© 2024 The World Bank CLIMATE country note MIGRANTS dominican republic 1818 H Street NW, Washington DC 20433 Telephone: 202-473-1000; Internet: www.worldbank.org Some rights reserved This work is a product of the staff of The World Bank. The findings, interpretations, and conclusions expressed in this work do not necessarily reflect the views of the Executive Directors of The World Bank or the governments they represent. The World Bank does not guarantee the accuracy of the data included in this work. The boundaries, colors, denominations, and other information shown on any map in this work do not imply any judgment on the part of The World Bank concerning the legal status of any territory or the endorsement or acceptance of such boundaries. Rights and Permissions The material in this work is subject to copyright. This material should not be reproduced or distributed without the World Bank’s prior consent. Because The World Bank encourages dissemination of its knowledge, this work may be reproduced, in whole or in part, for noncommercial purposes as long as full attribution to this work is given. Attribution Please cite the work as follows: World Bank. 2024. Dominican Republic: Climate Migrants. Country Report. All queries on rights and licenses, including subsidiary rights, should be addressed to: World Bank Publications, The World Bank Group, 1818 H Street NW, Washington, DC 20433, USA; fax: 202-522-2625; e-mail: pubrights@worldbank.org. Table of Contents List of Figures and Tables iv Acknowledgements v Abbreviation List vi Key Definitions Relating to this Country Note vii Executive Summary 1 Background and Objectives 10 Context: Climate Change, Natural Degradation, and Migration 13 Climate Change and Natural Degradation Vulnerability 14 Migration in the Dominican Republic 17 Studies on Climate Migration 22 Part 1 Quantitative Modeling to Estimate Climate Migration in the Dominican Republic 24 Methodology 25 Results 32 Part 2: Qualitative Case Study of Restauración Municipality in Terms of Internal Emigration Caused by Climate Change and Natural Degradation 44 Methodology 45 Climatic and socio-economic context of Restauración 47 Agriculture and forestry in Restauración 49 Emigration and climate migration 52 Conclusions and Recommendations for a Climate and Migration Policy 61 Reference List 70 Appendix 75 Appendix 1. Map of regions and provinces of Dominican Republic. 76 Appendix 2: Internal migration between municipalities of the Dominican Republic. 77 Appendix 3: Methodological details of modelling exercise 78 Data 78 Scenario Framework 83 Modeling Methods 87 Appendix 4: Additional Tables and Figures Quantitative Modeling, Dominican Republic 99 Appendix 5: Prioritized Areas for the Emissions Reduction Program, Dominican Republic 102 iii List of Figures and Tables Figure 1: Precipitation variability in 2011-2015 compared to historical trend (1950 – 1999). 15 Figure 2: Proportion of Forest Lost (left) and Land that has Declined in Productivity (right) in the 2011-2015 period. 16 Figure 3: Internal out-migration and in-migration between 2005-2010, by province. 18 Figure 4: Internal in-migrants between 2005-2010 by municipality 19 Figure 5: Ratio of out-migration over people staying between 2005-2010, by municipality. 19 Figure 6: Scenario framework (from Groundswell Africa; Rigaud et al. 2021) 29 Figure 7: Total internal climate and development migrants by scenario; 2050 and 2100 34 Figure 8: Total internal migrants (climate + development) under each scenario; 2010-2100. 34 Figure 9: Projected hotspots of climate in-migration (reds) and out-migration (blues) by 2050. 37 Figure 10: Net internal climate migrants by livelihood zone and scenario; 2050 (2100 provided in appendix 6). 39 Figure 11: Projected climate in-migrants from Haiti (top maps) and expressed as a portion of the population (bottom maps) by municipality; 2050 and 2100). 42 Figure 12: Difference in portion of Haitian in-migrants 43 Figure 13: Location of Restauración (left - yellow area) and points within Restauración (right) 47 Figure 14: Deforestation (above) and Land Productivity category (below) in Restauración for period 2001-2015. 50 Figure 15: Number of interviews and emigration from Restauración in Census 2010. 53 Table 1: Datasets, variables, and sources. 30 Table 2: Projected internal climate migrants by scenario (total, as % of all migrants, % of pop); 2050 and 2100. 33 Table 3: Total estimated migrants and climate-migrants from Haiti to DR 41   iv acknowledgements The Task Team Leaders of this Country Note were Rosa María Martínez (Senior Social Development Specialist) and Ana I. Aguilera (Senior Social Development Specialist). Gonzalo Pons (Environmental and Social Development Specialist) led the technical design and coordinated the studies and report, with considerable support from Lilian Pedersen (Social Development Specialist) and Adria de la Cruz (Head of International Affairs, National Institute of Migration, Dominican Republic). Marcos Morales (World Bank Social Development Consultant and As- sociate Professor of the Autonomous University of Santo Domingo) designed and implemented the qualitative research, and Bryan Jones (World Bank Environment Consultant and Assistant Professor of the CUNY Institute for Demographic Re- search) undertook the quantitative modeling. Background notes were developed by Ingrid Schreuel (Social Development Specialist, World Bank) and Juan Manuel De La Cruz (Social Analyst, National Institute of Migration, Dominican Republic). The team extends its gratitude to Wilfredo Lozano (Executive Director, Nation- al Institute of Migration of the Dominican Republic), as well as the World Bank’s Michel Kerf (Country Director), Alexandra Valerio (Resident Representative), Maria Gonzalez de Asis (Practice Manager), Paola Guerra Guevara (Social Develop- ment Analyst), Kennan Rap (Senior Social Development Specialist), Katharina Siegmann (Senior Environmental Specialist), Viviane Wei Chen Clement (Senior Climate Change Specialist), Kanta Kumari Rigaud (Lead Environment Special- ist), Ana Gabriela Strand (Senior Program Assistant), Vanessa Marin Arbeláez (Program Assistant), Christian Camilo Gómez Canon (Poverty Consultant), Carol Franco Billini (Environment Specialist), Alejandro de la Fuente (Senior Poverty Economist), Alejandra De La Paz (External Affairs Associate), Carmen Amaro (Operations Officer), Nancy Lozano (Lead Economist), and Paola Ballon (Senior Economist), for their support and guidance across the development of this report. This Country Note and its related studies were supported with funding from the World Bank’s Forest Carbon Partnership Facility (FCPF) and its Country Climate and Development Reports (CCDR) for the Dominican Republic. Appreciation is further extended to Asyl Undeland (Senior Social Development Specialist), Bouke Berns (Social Specialist), and Siet Meijer (Senior Operations Officer) for their continuous support throughout the production of this Country Note. Additional thanks goes to Margie Peters-Fawcett for having edited the publication. v abbreviation list 1km 1 kilometer DR Dominican Republic ENI National Immigrant Survey ERP Emissions Reduction Program GCM General Circulation Model GDP gross domestic product GHG greenhouse gas GPW Gridded Population of the World HTI Haiti IPCC Intergovernmental Panel on Climate Change ISIMIP Inter-Sectoral Impact Model Intercomparison Project IUCN International Union for the Conservation of Nature km kilometer LAC Latin America and the Caribbean NDC Nationally Determined Contributions NPP Net Primary Productivity RCP Representative Concentration Pathways SLR sea level rise SSP Shared Socioeconomic Pathways vi KEY DEFINITIONS relating to this country note Climate migration Out-migration and In-migration The movement of people from one country, Out-migration relates to migrants who depart region, or district to settle in another, largely from a specific location (also called emigrants), as a result—directly or indirectly—of the slow while in-migration refers to those moving into onset of impacts on their livelihoods brought a particular destination (also called immi- about by climate change in the form of natural grants). The general term, “migrant,” defines degradation (such us shifts in water availabili- both immigrants and emigrants, depending on ty, crop productivity, ecosystem productivity) or the location taken into consideration. other factors (such as sea level rise). International or cross-border migrant Human mobility Any person who is outside a State of which he As an umbrella term, it covers all aspects or she is a citizen or national, or, in the case of relating to the movement of people, regard- a stateless person, his or her State of birth or less of why, whether or not the movement is habitual residence. The term includes migrants voluntary or forced. The term not only relates who intend to move permanently or temporari- to the involuntary internal and cross-border ly, and those who move in a regular or displacement of populations, but also to the documented manner as well as migrants in voluntary, the latter of which usually is planned irregular situations (IOM Glossary of Migration). and consented to. Slow-onset processes and Internal migration rapid-onset events Refers to the movement of people within a For the purpose of this Country Note, the State involving the establishment of a new definitions from Germanwatch (2021) are temporary or permanent residence. Internal applied: (i) Slow-onset processes are “(…) a migration movements can be temporary or phenomena caused or intensified by anthro- permanent and include those who have been pogenic climate change that take place over displaced from their habitual place of prolonged periods of time — typically years, residence such as internally displaced persons, decades, or even centuries—without a clear as well as persons who decide to move to a start or end point.” For example, changes in new place, such as in the case of rural–urban precipitation patterns or volume over the years migration. The term also covers both nationals or sea level rise. (ii) rapid-onset events are and non-nationals moving within a State, pro- “single, discrete events with a clearly identifi- vided that they move away from their place of able beginning and/or end and that occur or habitual residence (IOM Glossary of Migration). reoccur in a matter of days or even hours at a The operational definition for this study also local, national, or region scale.” For example, makes explicit that internal migrants are those cyclones or floods, which climate change can that change municipalities of residence. also modify its frequency or intensity. vii KEY MESSAGES 1 Climate-induced migration in the Dominican Republic (DR) as a result of changes in the availability of water, sea level rise, and crop and ecosystem productivity exhibits distinct patterns in internal versus cross-border movements. 2 Internal climate migration is strongly focused on the migration of people from rural to urban areas; and sometimes urban to urban migration, partly due to reduced agricultur- al outputs, decreased capacity for pastoralism, and water stress, particularly in irrigated cropland areas. 3 Santo Domingo, the DR’s capital, is projected to become a net receiver of internal cli- mate migrants by 2050. This may change by 2100, however, with the city’s center (Distrito Nacional) transforming, instead, into a net sender of climate migrants due to the foreseen increase in population density, combined with water stress, which would push people to move outward and into the city’s suburban districts. In Santiago, the country’s second largest city, it is anticipated that it will face increased water stress and is projected to be a net sender of climate migrants by 2050. viii 4 The cross-border flow of Haiti’s climate migrants into the DR is projected to settle in urban, agricultural, and border areas, with those in Dominican cities representing a higher abso- lute numer. In relative terms, however, they represent a higher share of the local population in border areas which are currently lagging in terms of infrastructure and human develop- ment, potentially elevating tensions among host and migrant communities who compete for already limited resources. 5 Many climate migrants leave their community of origin as a result of climate change and natural degradation impacts on their agricultural productivity, while others find ways to adapt within their communities by, for example, diversifying into forestry and agroforest- ry. Agroforestry allows forest conservation and land recovery; it also provides farmers an income source throughout the year. Nevertheless, those farmers on already low incomes may struggle to transition into forestry and agroforestry, since these require more time and investment than traditional cash crops. Some emigrants have considered returning to their original communities and would do so, were economic conditions more favorable. Certain policies have assisted farmers to overcome these barriers, such as the provision of plants and the support to establish plant nurseries through local associations. There are other policies, as well as climate adaptation strategies, which relate to the use of production technologies and technical support, although reportedly with less success. Some emigrants have considered returning to their original communities and would do so, were economic conditions are more favorable. 6 The findings of this Country Note call for the DR to ensure more effective integration of climate and migratory policies, as well as ensuring these are aligned with broader nation- al objectives for green, resilient and inclusive development. Policy recommendations may be framed in three main pillars: (i) Having a coherent legal and institutional framework for climate and migration policies; (ii) Supporting migration as a valid adaptation strategy to ensure a planned, safe, and dignified process; and (iii) Climate adaptation and mitigation focused on preparedness and anticipatory action. A territorial approach that contextual- izes policy solution and considers multiple sectors, such as agriculture and forestry, water management and urban planning, will also ensure the success of key policies in these areas. ix executive summary The Dominican Republic (DR) is vulnerable to pulsion. In terms of international migration— climate change and has a high rate of natural despite the larger number of people emigrat- degradation. The effects of climate change ing from the DR compared to those who have include not only rapid-onset events, such as immigrated—immigration flows are never- hurricanes and floods, but also slow-onset theless sizeable. According to the National processes, such as decreased water availabil- Immigration Survey 2017 (ENI for its Spanish ity, sea level rise (SLR), and land degradation. acronym; GoDR, 2018). International migrants Comparative studies demonstrate that the born outside the DR account for 5.6 percent country is ill prepared to confront these antic- of total population (one of the highest rates in ipated climate change effects. Deforestation the Latin America and the Caribbean [LAC] and land degradation also are significant in region) and approximately 8.1 percent of the the DR, especially in the northwest and south- total labor force. The largest migrant groups west regions, threatening ecological process- are those originating in Haiti (87.2 percent) es, resilience, emission mitigation, and the and Venezuela (4.5 percent). The most com- productivity of forestry and agriculture. mon places of arrival are Santo Domingo and Santiago, although Haitians represent a larg- er proportion of the population in the border The DR shows evidence of significant human provinces (8 percent). mobility flows of (i) internal migration, mainly rural to urban; and (ii) international cross-bor- Given this context, the DR is an important der migration, especially from Haiti. The most place to study migration induced by the im- recent data on internal migration from the pacts of climate change and natural degrada- DR’s 2010 National Population and Housing tion. In this report, climate migration refers to Census (GoDR, 2010)—referred to as the 2010 migration that can be attributed largely (di- Census going forward—shows that the rate rectly or indirectly) to the slow-onset impacts of immigration into large cities (Santo Domin- of climate change on livelihoods through go, Santiago, Higüey) is high. While internal natural degradation such us shifts in water out-migration trends are less clear spatially, availability, crop productivity, ecosystem pro- border areas appear to be places of high ex- ductivity, or to factors such as sea-level rise. 1 Given climate impacts on migration can be essential to their livelihoods and where direct or indirect (e.g., climate impacting live- natural degradation is a key issue. lihoods and that in turn triggering migration), Even though there are many climate migrants we do not conceptualize it as mutually ex- that are not related to agriculture or this lo- clusive to other types of migration, like labor cation, the case study concentrated in these migration. Furthermore, climate migration is because of the presence of poor and exclud- conceptualized as a valid adaptation strategy ed groups and their relevance for existing to the consequences of climate change and environmental programs. natural degradation. This Note builds upon previous studies under- The DR’s internal climate migration signifi- taken regarding climate migration in the DR, cantly relates to rural-to-urban migration and, and combines a quantitative modeling ap- in some cases, to substantial urban-to-urban proach with a qualitative case study. migration. In terms of the former, the The quantitative modeling applied new combination of reduced agricultural outputs, methods to predict the number of internal capacity for pastoralism, and water stress climate migrants in the DR by 2050 and 2100, (particularly in irrigated croplands) is driving as well as their place of origin and people into the urban areas. There are destination, showing trends and producing a differences, however, within the projected novel dataset of internal climate migration. It impacts of climate change on urban areas, also estimated the share of climate migrants many of which rely on nearby agricultural potentially coming from Haiti into the DR, lands (i.e., worsening agricultural conditions their projected destinations, and calculating will impact not only the rural economy but by how Haitians’ migration patterns could also urban). Cities such as Santiago, change as a result of various climate scenari- expected to experience increased water os. The study used variations from a modeling stress, is projected to be a net climate methodology originally developed for a World out-migration hotspot (although the city may Bank flagship publication, Groundswell: Pre- still grow its population regardless); on the paring for Internal Climate Migration (Rigaud other hand, the urban fringe of Santo et al. 2018).1 The qualitative case study relies Domingo continues to grow substantially on in-depth interviews held with internal cli- under all scenarios. Interestingly, estimates mate migrants who had been involved in indicate that the Distrito Nacional (central agriculture. While the quantitative method district) may become a net receiving point for provides a country-level view and estimates internal climate migrants by 2050, of climate migrants, the case study allows to originating from various points of departure deepen into the motivations behind their within the country. Toward 2100, however, the decision to migrate, the experiences during center may become a net climate their journeys, and their expectations. out-migration site as a result of increased The interviewed emigrants had originated population density combined with water from the border municipality of Restauración stress, thus driving climate migrants to move in Dajabón Province, a location where to the suburban strip of Santo Domingo. agriculture and forestry activities are 2 It was estimated that a large share of mi- The DR’s internal climate migration signifi- grants, both internal and cross-border from cantly relates to rural-to-urban migration and, Haiti into the DR, will be driven by climate in some cases, to substantial urban-to-urban change, and the number and share will contin- migration. In terms of the former, the com- ue to increase through the end of the century. bination of reduced agricultural outputs, During the period 2020-2050, depending capacity for pastoralism, and water stress on future scenarios, between 149,000 and (particularly in irrigated croplands) is driving 368,000 people could potentially migrate people into the urban areas. There are differ- internally as a result of climate change (1.2- ences, however, within the projected impacts 2.8 percent of the DR population and 30-74 of climate change on urban areas, many of percent of internal migrants during the period which rely on nearby agricultural lands (i.e., 2005-2010), and 234,000 to 473,000 addition- worsening agricultural conditions will impact al internal climate migrants in the second half not only the rural economy but also urban). of this century (2050 to 2100; 1.9-5.3 percent Cities such as Santiago, expected to experi- of the DR population). Similarly, depending ence increased water stress, is projected to on the scenario, between six and 20.1 percent be a net climate out-migration hotspot (al- of cross-border migrants from Haiti to the DR though the city may still grow its population will be driven by climate change by mid-cen- regardless); on the other hand, the urban tury, and between 11.2 and 28.1 percent in fringe of Santo Domingo continues to grow the second half of the century. Pessimistic substantially under all scenarios. Interestingly, scenarios (regarding both socio-economic estimates indicate that the Distrito Nacional and climate outcomes) would result in higher (central district) may become a net receiving numbers of climate and other migrants. While point for internal climate migrants by 2050, estimating these numbers accurately and pre- originating from various points of departure cisely is difficult and hold much uncertainty of within the country. Toward 2100, however, the the future (e.g., assuming there is no change center may become a net climate out-migra- in border policy or in Haiti’s fragile economic tion site as a result of increased population and political situation), these estimates nev- density combined with water stress, thus driv- ertheless do convey the message that cli- ing climate migrants to move to the suburban mate change, indeed, could displace a large strip of Santo Domingo. number of people into and within the DR, in the absence of adequate and timely policy measures, and brings important details on the spatial distribution of these migrants. These estimates should be used in conjunction with other data sources and knowledge to inform policymaking. 3 In contrast, Haitian climate in-migration into causes of these drivers, participants recog- the DR projects a different trend. The border nized how natural degradation and climate regions, El Cibao, and Higüey would receive change played a role in the intention to mi- proportionally more climate migrants from grate through loss of agricultural productivity, Haiti than from within country, while Santo showing an indirect impact of climate and Domingo province would receive proportion- natural degradation. Interviewees were able ally less (although high numbers in absolute to relate the changes in hydrological factors terms). Patterns of Haitian in-migration would (e.g., precipitation) to changes in the sowing also shift slightly due to climate change, with and harvest seasons, both of which can re- both El Cibao and the area of the Cordillera sult in an unpredictable loss of production on Central receiving proportionally more climate rain-fed farming systems. Likewise, low soil migrants than in the past. Meanwhile, Santo productivity, caused by deforestation, forestry Domingo continues to receive a smaller por- intensification, and water stress, is perceived tion over time. Together, the likely pattern is as the source of lower yields. Farmers have one of internal emigration from agricultural reported applying various strategies to adapt and pastoral regions by the Dominican popu- to productivity loss, such as rotating between lation fueling urban growth, while these same lands and soil treatment; these, however, have regions will presumably experience an influx proved less successful. Faced with these chal- of labor migrants from Haiti, many of whom lenges on their livelihoods, migration to seek may be driven by Haiti’s climate change better economic opportunities appears as the impacts. It is worth noting that general pat- main adaptation strategy to improve their terns of internal climate migration, on the living conditions. one hand, more strongly follow changes in environmental conditions within the DR. On While many emigrated as a result of low agri- the other hand, however, migrants from Haiti, cultural production and productivity, others while reacting slightly to the environmen- found ways to adapt within their communities, tal situation in the DR, seem more driven in for example, by diversifying activities into choosing their destination by existing social forestry and agroforestry. Agroforestry allows networks and the potential for rural labor forest conservation and land recovery; it also opportunities. provides the potential for farmers to be able to rely on a source of income throughout the year. Thus, some emigrants have considered Internal emigrants who were interviewed in returning to their original communities and Restauración, most of whom worked in the would do so, were economic conditions more agriculture, perceive that changes in precipi- favorable; from the interviews, in fact, some tation and land degradation are impacting ag- reportedly were returning to work in forestry. ricultural productivity—one of the main liveli- Despite the potential of forestry and agrofor- hoods in the area—causing people to migrate estry, there might also be additional challeng- as an adaptation strategy. The main reasons es that low-income farmers may experience to emigrate, they stated, were the lack of in their efforts to undertake these economic employment, low agricultural productivity, activities; as they require more time and re- and the pursuit of better education opportu- sources than for cash-crops. nities. However, when inquired about the root 4 Certain policies (e.g., provision of plants, sup- Other adaptation strategies and policies in- port to establish plant nurseries through local clude the use of production technologies and associations), however, have assisted farmers technical support, although these reportedly to overcome such obstacles. with less success. The above results call for a more integrated migration, climate, and development policy. As in many other countries and within various development organizations, climate change and human mobility are treated as distinct policy areas. As evidenced in this study, the two are inextricably related, with a growing body of literature calling for more cross-sectoral policies. On the one hand, recognizing migra- tion as a valid adaptation strategy could sup- port people to deal with the effects of climate change (especially to those trapped because of insufficient resources to even migrate), and support a planned, safe and dignified move- ment that increases the positive impacts while reducing the negative ones. On the other hand, migration policies that take into account the current relevance of Haitian workers for the DR labor market—particularly in terms of agricul- ture – could support the creation of programs that will incentivize large and small farmers to apply environmentally sustainable practices while, at the same time, ensure the supply of agricultural labor from which Dominicans seem to be exiting. Such integration, thought through an adaptation framework, will help to improve the effectiveness of two seemingly distinct yet interconnected sectoral policies. Furthermore, these should align with broader national objec- tives for green, resilient and inclusive develop- ment. A territorial approach considering mul- tiple sectors, such as agriculture and forestry, water management and urban planning, will allow a holistic response to climate migration. 5 Recommendations to address current and future climate-driven migration in the DR are based on three pillars, in line with the key findings of the World Bank’s World Development Report 2023 (World Bank 2023) and the policy response framework outlined in the special focus on migration in Migra- tion and Development Brief No. 37 (Ratha et al. 2022). The following table includes specific recommendations, along with the three pillars: Table 1. Summary of Policy Recommendations to Address Climate Migration in the Dominican Republic PILLAR POLICY RECOMMENDATIONS Support technical roundtables among the National Climate (i) Change Council, the National Migration Council, Ministry of Agriculture, Ministry of Labor, Ministry of Environment, Ministry of Interior and Ministry of Economy, Planning and Development Coherent legal to better integrate environmental, agricultural, and migratory and institutional policies in the country. framework Provide technical advice and facilitate knowledge exchanges with small-island states in Latin America and in other regions, including through regional platforms. For instance, through the Greater Caribbean Climate Mobility Initiative (GCCMI) and sim- ilar regional platforms. Create clear and established legal frameworks to support cli- mate-induced migration pathways at the national level. Mobilize international technical, institutional, and financial support to provide regional responses to facilitate safe and planned cross-border climate migration. Support the professionalization and reduce turnover of mi- gratory civil servants with a dedicated module or seminar on climate-induced migration as part of the curriculum of the National Migration Institute´s Master’s degree in Migration and Development in the Caribbean. Promote further research, analytical work, and data generation and collection on these issues. 6 PILLAR POLICY RECOMMENDATIONS (ii) Develop early warning systems to enable prior planning for slow and rapid human mobility flows in response to climate change Supporting impacts and natural degradation. migration as a valid adaptation Planned relocation may imply the voluntary movement of strategy to people to other cities or areas within a country. In line with Law ensure a No. 368-22 on Territorial Development, Land Use, and Human planned, safe, Settlements, greater attention is needed to ensure that affected and dignified persons are involved fully in decisions regarding their reloca- process tion. Codesign and implement climate-induced relocation plans with local communities. Lessons learned from the DR’s Lake Enriquillo experience can be a starting point.1 Boost national and international resource mobilization to fi- nance subsidies and enable safe and planned migratory path- ways for internal and cross-border climate migrants. Support climate-informed territorial development and planning instruments and socio-environmentally sustainable investments that address rural-urban corridors and the unique challenges or border areas. Prepare and update local development and climate action plans to allow the mobility of people, identify their destination, and ensure appropriate preparations and key services for ar- rival at destination (e.g., health, education, and housing). For instance, support the continuity and adequacy of educational process for internal migrants. 7 PILLAR POLICY RECOMMENDATIONS Periodically update territorial planning instruments at the Do- minican Republic’s (DR) Ministry of Economy, Planning, and De- (iii) velopment (Ministerio de Economía, Planificación y Desarrollo, MEPYD) to take into account predicted climate migration flows. Climate Invest in the adaptation of rural and border communities and adaptation and other out-migration hotspots, to enable potential migrants to mitigation stay in place, where viable, are also important. National plan- focused on ning instruments at MEPYD, such as the regional development preparedness strategy for border areas “Mi Frontera RD” already integrate cli- and anticipatory mate and migration policy and can serve as reference for other planning instruments in the country. action Prioritize strategies aimed at supporting livelihoods and ad- dressing the risk of displacement caused by climate change in areas adjacent to protected zones. The Emissions Reduction Program (ERP) in the DR may consti- tute an avenue for future preparedness in green and sustain- able sectors. The program may provide technical and financial support in favor of national efforts. This program could be lev- eraged to reverse unsustainable practices in the forestry and agriculture sectors that are factors that influence the decision to migrate. Address the barriers to enter the agroforestry field through in- centivization and subsidy programs, and promote green jobs in the forestry sector. Provide technical and technological support to increase agricul- ture and land productivity, irrigation, as well as water manage- ment to address factors that influence the decision to migrate. Enhance waste management to reduce environmental pollu- tion and health risks, especially in water and sanitation sectors, ensuring equitable access to clean resources, and supporting public health and environmental preservation. 8 PILLAR POLICY RECOMMENDATIONS Minimize the use of harmful chemicals by large agricultural pro- (iii) ducers to reduce the speed of soil and water degradation and its subsequent impacts on climate-induced migration from rural areas. Climate adaptation and Improve urban-rural corridors and increase awareness of the mitigation importance of sustainable agricultural practices among rural and focused on urban youth. preparedness and anticipatory Strengthen technical assistance for preparedness as a critical action measure in cities expected to receive an influx of migrants over the next few decades. Invest in infrastructure and employment in cities (e.g., Santo Domingo, Santiago, Higüey), as well as in sustainable areas. Additional research is needed in this area. Likewise, given water stress and sea-level rise pressures on cities, climate resilient urban planning is needed to ensure cities are able to accommodate growing numbers of people under conditions of climate stress, including resilient infrastructure, housing, and ser- vices, with social inclusion and cohesion considerations. 1 The Government of the DR has relocated many affected by a rise in water levels in Lake Enriquillo to the new town of Boca de Cachón, a US$24 million community built by the government to house people on the verge of losing their homes to the lake. 9 background and objectives This report aims to introduce climate migra- Evidence, however, suggests that slow-onset tion in the policy debate of the DR by present- natural processes (e.g., sea level rise, land ing key findings based on quantitative and degradation, and changes in precipitation qualitative analyses. While the country has and water availability) may have larger im- undertaken many efforts to develop policies pacts on livelihoods and on migration (Ger- and institutional arrangements relating to manwatch 2021). At the same time, detri- migration and climate change, the interac- mental and unsustainable natural resource tion between the two in the policy discussion management and agriculture practices, as remains siloed. For example, the topic of well as deforestation, also are discussed as migration does not appear in the Dominican paths through which humans further con- Republic’s (DR) top climate change strategies, tribute to natural degradation and climate namely the Nationally Determined Contri- change; these also are recognized as areas butions (NDC) and the National Adaptation that could hold important mechanisms to Plan. Neither does climate change appear adapt and mitigate the consequences. prominently in any migration policy, except for that caused by natural hazards (e.g., earthquakes and hurricanes). Through this Country Note, the World Bank aims to contribute to a better understanding of climate-induced human mobility in the DR. While extreme weather events and disasters Taking advantage of the Emissions Reduction caused by natural hazards are the most visible Program (ERP) preparation,2 the World Bank’s climate change manifestations, other effects Inclusion Global Practice has carried out this of climate change and environmental issues, study to provide initial evidence on the rela- such as deforestation and slow-onset process- tionship between climate-induced migration, es (e.g., natural degradation, reduced water natural resource degradation, and agricul- availability), may have more profound con- ture practices with forestry impact. The study sequences. Hazards such as cyclones, floods, applied qualitative and quantitative research landslides, and droughts are more prominent- methods, detailed throughout this report. ly discussed in the context of climate change This study also created a new granular raster and its impacts on migration. dataset as a public good to support further research on climate migration. 10 This study discusses climate-induced migra- This study also uses a qualitative methodology tion, including the drivers as a result of natural to document a case study of the decision-mak- degradation. Following the World Bank flag- ing process of climate migrants related to the ship report, Groundswell: Preparing for Inter- agriculture sector from the municipality of nal Climate Migration (Rigaud et al. 2018)— Restauración in the Province of Dajabón. While referred forthwith as Groundswell—this note the quantitative modeling will provide a gen- will refer to climate change-induced migra- eral view of future migration flows caused by tion, or climate migration, to discuss migra- different drivers related to climate change and tion that is largely linkable to the impacts of natural degradation, the qualitative study will climate change (directly or indirectly), partic- delve into the decision-making and migration ularly through natural degradation and slow process, critical to formulate effective policies onset processes (such as water availability, and support these populations. Slow-onset crop productivity, ecosystem productivity, and processes and natural degradation are likely sea-level rise). to directly affect those people whose liveli- hoods depend on natural resources. As such, a World Bank team interviewed a number of emigrants who were involved in the agriculture The study applies quantitative modeling to sector; the objective was to learn how climate predict the number of climate migrants in and natural variables impacted the liveli- the DR by 2050 and 2100, as well their origin hoods of interviewees and their decision to and destination. The study uses a modeling migrate. While there are many other climate methodology that was initially developed for migrants that are not related to agriculture Groundswell, referred to above. Estimates of (e.g., urban-to-urban, driven by sea-level-rise, climate migrants and their origin and desti- etc.), the team selected this sector because nation are developed for different regions of of presence of poor and excluded groups and the world. Given the methodological chal- its relevance for existing environmental pro- lenge to include Small Island States in the grams. The team focused on emigrants from a modeling, however, the Caribbean region was single municipality, Restauración, to obtain a not included in these initial estimates; hence, more comprehensive understanding. The mu- this study produces estimates of internal nicipality not only has a population dedicated migration for the DR. Additionally, given the to forests and agricultural activities, but also prominence of international migration from holds a history of internal and international Haiti into the DR, this analytical work will migrations. The study delves into the decision include for the first time, additional modeling to migrate, the process of migration, and the to calculate the share of international climate adaptation efforts that may have supported migrants and their destinations, focusing spe- the decision. cifically on those originating in Haiti. 2 The ERP and the framework by the UNFCCC Conference of the Parties (Reduce Emissions from Degradation and Deforestation (REDD+), aim to improve the quality of life in rural communities in the DR and increase the resilience of ecosystems against the impacts of climate change, and promote activities such as agriculture and sustainable forest management, forestry, and coffee and cocoa agroforestry, among others. 11 The report is structured as follows: It (i) describes the context and relates to previous studies on climate change, natural degradation, migration, and climate migration in the DR; (ii) summarizes the methodology and results from the quantitative modeling exercise; (iii) documents the case study of Restauración, describing the context of the municipality and the methodology and re- sults of interviews; and (iv) discusses the implications for climate change and migration policy. 12 Context: Climate Change, Natural Degradation, and Migration 13 Climate Change and Natural Degradation Vulnerability The DR is vulnerable to natural hazards and From the historical data (WHO and UNF- climate change. It has a high risk of floods, CCC, 2021), future projections indicate that landslides, cyclones, and fires, and a consider- total annual precipitation by 2050 may able risk of earthquakes, droughts, and ex- decrease by 15 percent, on average, through- treme heat (GFDRR 2023). The DR is the tenth out the DR, with precipitation in the south- most impacted country across the globe that ern and western regions mostly decreasing. has suffered from climatic phenomena be- The start and end dates of the dry and rainy tween 1997 and 2016, based on the Global months may shift, with the weather pat- Climate Risk Index (CRI) 2018 (Eckstein, Kün- terns becoming more intense. Freshwater zel, and Schäfer 2018).3 In particular, climatic resources for human and agricultural use events and their impacts are expected to be are expected to decrease up to 25 percent aggravated by climate change, including ris- by 2050, as will decrease the overall wa- ing temperatures, changing precipitation pat- ter quality. These impacts are predicted to terns, and an increase in their frequency and be heterogeneously distributed across the intensity. Comparative studies demonstrate country. Figure 1 illustrates the uncertainty of that the country is not adequately prepared precipitation across the DR’s municipalities, to confront such effects (Notre Dame Global with the darker colors representing higher Adaptation Initiative 2023). variability between 2011 and 2015 compared to the historical trend (1950-1999). As such, it is increasingly more difficult to predict future These climate change impacts include rap- seasons and water availability. These are id-onset events (e.g., cyclones and extreme expected to impact vulnerable populations, rainfall) and slow-onset processes (i.e., related such as the farmers of rainfed agriculture in to sea-level rise, land degradation, reduction the border area of the DR. on water availability). Sea level rise is one of the most significant threats to small islands and atolls with low-lying areas. 3 Publisher: Germanwatch. Subsequent Germanwatch rankings position the DR as less impacted, since the relevant events in 1997 and 1998, such as Hurricane George, are no longer considered. 14 Figure 1: Precipitation Variability in 2011-2015. Compared to Historical Trend 1950–1999: Dominican Republic. Source: World Bank staff, using data from Terrestrial Precipitation 1900-2017 Gridded Monthly Time Series v 5.01. Notes: Comparison period is historical precipitation average from 1950–1999. Forest and land degradation is also high in the of the main environmental issues in the DR, DR, especially in some areas, threatening the given the fact that deforested areas are more ecological process, resilience, emission miti- susceptible to desertification and drought gation, and productivity of forestry and agri- which, in turn, reduces land productivity. Fig- culture activities. Forests, especially primary ure 2 illustrates two maps denoting the rate of forests, have the capacity to absorb carbon deforestation and the proportion of land that and contribute to meeting emission reduction are declining in terms of productivity in the goals. On the other hand, soil degradation, period 2011-2015; it also indicates that these deforestation, and land use changes gener- issues are more prominent in the southwest, ate emissions, compromise future economic northwest, and east of the DR. As such, Agri- activity, and increase vulnerability to extreme culture, Forestry and Other Land Use (AFO- natural hazards. Despite the country’s efforts LU) activities are extremely important in the to regain forest cover,4 deforestation contin- mitigation and adaptation country plans, as ues to be relatively high in comparison with recognized in the 2020 country’s NDC. other countries in the LAC region,5 and is one 4 Since the formulation in 1986 of the Tropical Forest Action Plan for the DR (PAFT-RD), the country has developed numerous strategies to increase the country’s forest cover, such as the Plan Nacional Quisqueya Verde (PNQV); Siste- ma Nacional de Áreas Protegidas (SINAP); Proyecto de Desarrollo Agroforestal de la Presidencia (PAP); Agroforestal system with coffee in shade (CAFÉ); and Programa Megaleche (silvopastoral system and conservation of forests on livestock farms). Many of these projects are part of the ERP. 15 Figure 2: Proportion of Tree Cover Lost (left) and Land that Has Declined in Productivity (right): Dominican Republic (2011-2015) Sources: World Bank staff, using (i) Hansen et al. 2013; and (ii) Land Productivity Dynamics LPD - MODIS (derived from NDVI product of MODIS/Terra Vegetation indices 16-Day L3 Global 250m SIN Grid V006). Notes: (Left) The grey municipality refers to San Pedro de Macorís, which had an unusually large tree cover loss rate of 0.48, possibly an error of measurement due to the small size of the forest cover. The denominator is the forest extent relating to 2000, which is the only available in the Hansen dataset. (Right) The proportion of land categorized as “De- clining” in terms of productivity. The other categories in the dataset are “Non vegetated area,” “Early sign of decline,” “Stable but stressed,” “Stable,” and “Increasing.” The denominator is the municipality’s total hectares. To prevent natural degradation, it is essential prone to water runoff, soil degradation, and a to comprehend the practices of the agriculture reduced ability to retain moisture. This, in turn, sector, and their causes, effects, and potential causes the surrounding areas, along with the alternatives. Along with illegal logging, one of agriculture practiced in those areas, to be- the main causes of forest loss is cattle ranch- come vulnerable, particularly where there are ing on a commercial scale, expansion of the steep slopes. Likewise, lack of land ownership agricultural frontier (GoDR, 2019), and unsus- of many farmers intensifies these practices tainable agricultural practices. One of these since reduces the incentives to invest in land practices is the recurrent burning of forest sustainability due to low opportunity cost. (tumba y quema) for agriculture and livestock However, these practices are also common use and/or for the production of charcoal, in large commercial agricultural producers,6 which leaves the surface of the land uncov- especially in their initial stages. ered for prolonged periods of time, becoming 5 Based on the data from Global Forest Watch (2023), World Bank staff calculates that the DR is 10th in the LAC re- gion to have faced the highest tree cover loss rate (as a percentage of forest cover in 2010) between 2011 and 2021. If calculated in terms of total area loss in lieu of a percentage rate, however, larger countries show having experienced greater loss, with the DR placed in this case as 17th. The calculation does not include reforestation. 16 Migration in the The most recent internal migration data Dominican Republic emerges from the 2010 Census (GoDR 2010), which reflects a continuing high arrival of people into cities such as Santiago or Santo Domingo. According to the Census, 497,470 The DR population has a long history of inter- people (5.3 percent of the population in 2010) nal migration. Notable patterns of territorial moved across municipalities between 2005 mobility in terms of volume and magnitude and 2010, and 404,854 (4.3 percent in 2010) occurred in the 1950s, with the number of mi- from province to province. Figure 3 reflects grants having tripled by the 1980s. The main the emigration flow between provinces regional attraction areas during this approx- around the country within this period (see imate 30-year period were the (i) north mac- Appendix 1 for a map of DR provinces and roregion of the country, given its successful Appendix 2 for a similar graph at the mu- agricultural production of export items such nicipality level). The most notable trends are as coffee, cocoa, tobacco, and rice; (ii) south- high internal immigration to the cities, espe- eastern region due to the sugarcane planta- cially within the Province of Santo Domingo tions; and (iii) capital city of Santo Domingo, (182,828), Distrito Nacional (48,882), and the the seat of political power (Ariza and Lozano Province of Santiago (44,700), particularly 1993). within the municipality of Santiago (31,051). Santo Domingo Province and the Distri- The 1960s and 1970s were years of high ru- to Nacional form part of Santo Domingo’s ral-urban mobility. The Cibao area, in par- metropolitan capital area, while the city of ticular, began to lose its attraction compared Santiago is the second largest in the country. with the capital’s National District. Other The three areas experienced 55 percent of provinces experienced a retreat in population internal mobility between 2005 and 2010. The as a result of lower agricultural production, migration map by municipality (Figure 4), also economic concentration of resources, and illustrates that in the East, the Higüey munici- rapid urbanization. During this period, 54 pality in Altagracia Province has high internal percent of urban growth reflected rural-ur- immigration (30,336), likely because of the at- ban displacement, slowing gradually in the traction of the third largest city and employ- 1970s and 1980s (Jiménez 1992). It was not ment opportunities in livestock production until the 1990s when the Northwest Cibao and tourism from the Punta Cana and Bávaro region (including Montecristi, Dajabón, San- beach areas. tiago Rodríguez, and Valverde) recorded the highest emigration rate. Likewise, the border zone provinces—the poorest parts of rural DR, 6 A study indicates that the DR has an extreme concen- with extreme poverty twice as high compared tration of land held by various companies and individ- with the rest of the country (UNEP 2013)—also uals. Only 50 major producers control more than 1,000 recorded a high emigration rate. Many Do- hectares each; 200 families control approximate- ly 600,000 hectares (equivalent to 50 percent of the minicans sold or leased their lands in these country’s arable land), and only 40 percent of privately provinces, ultimately moving to the cities. owned land is titled (Ovalles, 2011). 17 Figure 3: Internal Out-Migration and In-Migration between 2005 and 2010, by Province: Dominican Republic Sources: World Bank staff, based on data from the National Population and Housing Census 2010 of the Domin- ican Republic (GoDR 2010). Notes: The graph was created based on the 2010 Census survey (GoDR 2010) question regarding the name of the municipality in which the person (five years and older) was living in 2005. Each tick mark represents 6,000 people; the number represents those who, in 2005, were living in another municipality within the DR; and arrows in the same province represent those who have moved between municipalities within the same province. 18 Figure 4: Internal In-Migrants between 2005 and 2010, by Municipality: Dominican Republic Sources: World Bank staff, based on data from the 2010 National Population and Housing Census of the Dominican Republic (GoDR 2010). Notes: k = thousands. The graph was created based on a census question regarding the name of the municipal- ity in which the person (five years and older) was living in 2005. The number represents those who, in 2005, were living in another municipality within the Dominican Republic Figure 5: Ratio of Out-Migration over People Staying between 2005 and 2010, by Municipality: Dominican Republic Sources: World Bank staff, based on data from the 2010 National Population and Housing Census of the Do- minican Republic (GoDR 2010). Notes: The graph was created based on a Census question relating to the name of the municipality in which a person (5 years and older) was living in 2005. The ratio represents those who were living in a municipality in 2005 and had left it by 2010, against those who had remained from 2005 to 2010, inclusive. The data prevented the computation of the population size in 2005. 19 Internal out-migration trends are spatially less The DR also has a large inflow of international clear, although there are some notable trends. migrants, especially from its neighbor country, Of the four provinces with the highest internal Haiti. In 2022, the net migration rate for the immigration in 2010 (Santo Domingo, Distrito DR, however, was -2.673 per 1,000 population Nacional, Santiago and Altagracia), Santiago (own estimations based on UNDESA 2022); has a more balanced number of people who that is, a larger number of people emigrated emigrated in 2005 and immigrated in 2010; from the country, particularly to the United this suggests that, for many, it may have been States, compared with those who immigrat- an intermediate and temporary destination. ed to the DR. Nevertheless, the immigration Conversely, the Province of Santo Domingo flows are sizeable. Based on the 2017 National shows a clear growing trend of immigration Immigrant Survey (ENI) (GoDR 2018), 5.6 per- (more immigrants than emigrants), as well as cent of the DR’s population was born outside a large number of migrants who moved from the DR, with Haiti being most prevalent (87.2 municipality to municipality within the prov- percent of all immigrants or 4.9 percent of the ince or to Distrito Nacional. Aside from these population). Venezuela occupies second place four provinces, the others had more internal (4.5 percent of all immigrants). The number of emigrants in 2005 than immigrants in 2010. Venezuelan migrants has doubled in the last San Juan, for instance, had 17,135 more inter- six years (ENI 2012; ENI 2017), while the Hai- nal emigrants than immigrants, suggesting tian population grew by 8.6 percent from 2012 a considerable reduction in population. San to 2017(GoDR, 2018). Both immigrant popula- Cristóbal also has a large number of emi- tions are leaving their home countries follow- grants. With regard to the ratio of people em- ing outbreaks of conflict, violence, poverty, 7 igrating versus those staying in 2005, there is and political instability. Consequently, the no clear spatial pattern at the municipal level DR receives people in increasingly vulnerable 8 (Figure 5), except for a high ratio in a number conditions. An additional 2.7 percent of the of border municipalities. population were born in the DR of at least one immigrant parent. 7 Institutional deterioration, chronic poverty, unemployment, structural violence, and precarious health services are the main reasons for the exodus of Haitians from their country, while Venezuelans are escaping from a complex social, political, institutional, and economic crisis. Haitian migration is one of the most complex and challenging in the region (IOM DTM Feb 2021). 8 New trajectories; simultaneous flows of emigration, transit, and immigration; increased vulnerabilities due to the COVID-19 pandemic; border closures; and the collapse of informal economies (in which migrants are overrepresent- ed) have added additional layers of complexity to the challenges refugees and migrants face and have historically faced. (R4V 2022). 20 Migrants born in Haiti are commonly em- Although climate change or natural degra- ployed in the agriculture (representing 33 dation are not directly reported by Haitians percent of Haitian workers), construction (26.3 surveyed as a main reason for migrating, it percent), and retail and trade (16.3 percent) may be a latent factor influencing such deci- sectors (GoDR 2018). In the agriculture and sions. It is recognized that Haiti has a high construction sectors, Haitian-born workers rep- degradation of its natural resources 9 which resent approximately 30 percent of their labor will impact economic activities that make use force (own calculations based on data showed of them, such as agriculture, livestock and in GoDR, 2018). Haitian immigrant workers tourism. Haiti is also one of the most affected typically are employed in jobs that require little and vulnerable countries to climate risk and qualification, and are overrepresented in the climate change.10 In this sense, it is possi- informal job market, therefore leaving these ble that natural degradation has led to loss populations in a situation of vulnerability. Do- in economic productivity which has driven minican workers increasingly self-exclude them- migration. However, climate migration is not selves from such jobs, since (i) they are able to the most present reason to migrate for those access various social benefits that raise their surveyed, especially when climate degrada- basic wage; and (ii) there are cultural factors tion occurs over a long period of time, for ex- that prevent them from participating in these ample droughts or soil degradation. Likewise, kinds of jobs (ENI 2017, INM RD 2022). The ENI extreme natural events, such as the tropical 2017 indicates that 47.6 percent of Haitians live storm in August 2021 that further pushed mil- in provinces with a high population concen- lions of Haitians into food insecurity (Lutz and tration (e.g., Santo Domingo, Distrito Nacional, Yayboke, 2021), will likely continue to be im- and Santiago). Haitians, however, represent the portant factors in the decision to leave Haiti largest proportion of the population in provinc- towards the DR. es close to the Dominican Republic/Haiti border (8 percent) and in provinces that cultivate sugar cane (6.3 percent). 9 Haiti ranks 165th out of 180 countries in the Yale University Ecosystem Vitality Index. 10 Ranked as the second most vulnerable to climate risk between 1997 and 2016 in the Long-term Climate Risk Index for 1997-2016 (Germanwatch 2018) and third in the most recent (2000-2019). Ranked 168th out of 182 countries in the ND-GAIN 2020 Index (Notre Dame Global Adaptation Initiative 2022), which captures vulnerability to climate change and readiness to confront it. 21 Studies on Climate sources, however, is not linear. Poverty and the impacts of climate change could create a situ- Migration ation whereby the poorest are trapped and do not have the opportunity to even use migration as an adaptive strategy, thus trapping them in There is general agreement in the scientific places with degraded resources and plunging community of the connection between climate them further into poverty. Critics also observe change, natural degradation, and human mo- the simplistic evaluation that sometimes is bility, but more research on the subject is called given to the migration process, portraying it as for. Globally, climate change is emerging as a something that is either positive or negative. potent driver of human mobility. As document- Migration can have both benefits and negative ed by the Intergovernmental Panel on Climate consequences for the migrants, for the origin Change (IPCC) report of 2022, various studies community, and for the destination communi- have found links between climate change and ty. Hence, a more thoughtful discussion on how human mobility, such as migration or forced to increase the positive impacts of migration displacement. This will include international while managing the negative impacts is need- migration, but most climate-induced migra- ed. tion is expected to occur within country bor- ders. The report also states that “rapid-onset Empirical studies in the DR have applied quan- climatic events trigger involuntary migration titative and qualitative methods to document and short-term, short-distance mobilities (…) the links between migration, climate change, slow-onset climatic events (such as droughts and natural degradation. A study by Corde- and sea level rise) lead to long-distance inter- ro Ulate and Lathrop (2016) used a survey nal displacement, more so than local or inter- method to document the emigration process national migration (Dodman et al., 2022, IPCC relating to environmental events from the Chapter 6, p. 930).” The World Bank’s Ground- border municipality of Jimaní (Independencia swell report estimates that internal climate Province) toward a neighborhood located in migration may reach up to 216 million people Santo Domingo Norte. Extreme natural events, by 2050 due to slow-onset climate impacts. such as droughts and floods, were found to The IPCC notes the need for further research influence participant decisions to emigrate on this complex topic. from Jimaní, highlighting vivid flooding events in that area that called for relocation of the Research has conceptualized migration as an entire population. These were not reported, adaptation strategy to climate change and however, as the main reasons to migrate; socioeconomic challenges. Many studies sum- instead, migration was reported as an adap- marized in the IPCC (2022) and other reports tation strategy to face economic and social have found that migration is used as an adap- disadvantages, with these environmental tation strategy to escape degraded resources, events only contributing to them but not being consequences of extreme climate events, or perceived as the main triggers. Additionally, to find better socioeconomic opportunities Wooding and Morales (2014a and 2014b) have (sometimes viewed as a last resource). The studied, qualitatively, the interactions between relationship between migration and low re- migration, natural degradation, and conserva- 22 tion policies in two communities around the they are missing a thoughtful consideration natural park of Nalga de Maco, located in the of migration. Among the recommendations Northwest and close to the border with Haiti. for further research, the report indicates the The study documents how waves of internal importance of gathering meteorological data and Haitian migration to the area, coupled for the study of climate migration, the need with unsustainable agricultural practices, have for adequate methods to capture localized contributed to natural degradation in the area movements, and a review of land exploitation through agricultural expansion. The authors systems and their impact on natural degrada- also discuss that natural degradation can be tion and migration. seen as a second-order reason for migrating, by influencing economic hardship and pover- ty, which is commonly referred as some of the This report will define climate-migration in main reasons for migrating, both by internal line with recent evidence. In this report, we immigrants and those of Haitian origin. Other use a similar definition as the one used on reports estimate the amount of migration that Groundswell WB Report, referring to migra- has occurred as a result of extreme natural tion that can be attributed largely (directly or events. For example, IFRC (2021) reports that indirectly) to slow-onset impacts of climate in 2017, 69,000 people were classified as being change on livelihoods through natural deg- internally displaced in the DR as a result of radation such us shifts in water availability, disasters caused by natural hazards. crop productivity, ecosystem productivity, or to factors such as sea-level rise. Nevertheless, it’s important to highlight a caveat with this Additionally, a recent report indicates the definition. Climate change can have, at the theoretical and contextual considerations to same time, direct impacts in migration deci- further study this topic in the DR. The report, sion making (such as when a storm destroys “Diagnostic of Information for Public Policies: someone’s place of living) or indirect (such Migration, Environment, and Climate Change, as reducing income of farmers because of a Dominican Republic,” includes a compre- drought and pushing them to find other eco- hensive review of national and international nomic activities) (Cissé, 2022, IPCC Chapter evidence on migration, climate change, and 7). Additionally, various studies point out to climate-induced migration, with the propos- the difficulty of determining the exact reason al for a study of climate-induced migration for migrating, given the complexity of the specifically in the DR (Wooding and Morales, process and how push factors may be inter- 2016). The report highlights how migration twined. Both methodologies try to control within national borders is a strategy to escape for it in different ways, but will ultimately still the effects of extreme natural events (e.g., hold error and uncertainty because of this. In hurricanes, earthquakes), slow-onset impacts line with this, when studying climate migra- of climate change, and the consequences of tion using self-reports from people, it should overexploitation of natural resources (e.g., soil not be expected that people could perfectly degradation). It also outlines climate change identify impacts from climate change from and municipal policies in the DR (e.g., ad- those caused by other factors. This will be aptation strategies in agriculture), and how further discussed when describing the results from each method. 23 part 1: quantitative modeling to estimate climate migration in the dominican republic 24 Methodology This study uses a methodological approach to estimate climate migrants that was first used in World Bank’s Groundswell report (2018). Analyses for the Groundswell report were the result of col- laboration between World Bank Group staff and researchers from the Center for International Earth Science Information Network, Columbia University Earth Institute, City University of New York (CUNY) Institute for Demographic Research, and Potsdam Institute for Climate Impact Research. The novel methodological approach to estimating climate change impact on internal migration patterns was built from spatial methods to project future population distributions that can be calibrated to reflect the impact of various potential drivers of change. This study builds upon the Groundswell method- ology, incorporating spatial refinements and integrating an international mobility component. The results and a summarized version of the methodology are described below. A more detailed explana- tion of the models, scenarios, data, and limitations is provided in Appendix 3. Models The model, a gravity-based spatial alloca- influenced by socioeconomic and demograph- tion-type framework (hereafter referred to as ic characteristics of populations, economic the INCLUDE model (Jones and O’Neil 2013; conditions and livelihoods, historical and exist- 2016)), produces scenario-based estimates of ing connections, political systems and stability, internal migration. The scenarios that were geographic characteristics and, most impor- used are combinations of socioeconomic tantly for this work, climate impacts. With (development) and climate (emission) pro- regard to climate and degradation impacts in jections. On the one hand, the development particular, the model inputs changes in water projections provide estimates of national-lev- availability/stress, crop yield/productivity, pro- el population change and urbanization rates, ductivity of the ecosystem/natural biome, and among other narratives and estimates. On seal-level rise. The INCLUDE model estimates the other hand, projections of greenhouse gas the number of climate-induced migrants and (GHG) emission concentration drive changes their future locations by comparing population in environmental conditions which, in turn, distributions that incorporate climate impacts may alter socioeconomic conditions through with a scenario based on only a development impacts on various socioeconomic livelihoods trajectory (constant climate). Additionally, the (further details on the scenarios are provided calculations disaggregate between climate below). Future changes in the spatial distri- and “other” migrants11 to further investigate bution of the population, driven by migration, the impacts of climate on spatial patterns of occur in the model as a function of the rel- population change. To be considered an inter- ative attractiveness of various subnational nal migrant in this model, a person must move 12 points in space. Relative attractiveness is across municipal boundaries. 25 The international model is derived from a variation of the INCLUDE model, whereby it first esti- mates the bilateral country-level migrant flows between Haiti and the DR, and then identifies the portion of climate migrants and distribution of these migrants between locations. This work draws on a method to estimate origin-destination international movement that could intensify or de- cline as a result of environmental change. The method was used to produce migration projec- tions across Central America and Mexico for the New York Times Magazine/ProPublica report on climate migration in the Americas (Jones, 2020). The international model operates based on two steps. First, total migrants (climate and other) for each time period are estimated at the country level from Haiti to DR. Haiti is considered as a single unit, so out-migrants are modeled for the country on aggregate. Second, migrants are distributed to municipalities within the DR based on the estimated relative attractiveness of each—using a gravity-type function that accounts for environmental conditions, economic conditions (through an agglomeration effect), and the exist- ing Haitian population (considered a proxy for social networks). Similar to the INCLUDE model, climate migrants are distinguished from other migrants by comparing outcomes to a counterfac- tual scenario in which no climate change is assumed (e.g., climate is held constant at 2020 con- ditions), although there is much more uncertainty in these estimates than for the internal model. Advantage is taken of existing bilateral flow data (taken from the ENI 2017) to train and project future migration from Haiti to the DR as a function of aggregate climate impacts at the national level (i.e., crops, water, ecosystem productivity13 ), and the existing data on the distribution of Hai- tian migrants across the DR (from the 2010 Census), to estimate potential changes in origin-desti- nation flows from Haiti to municipalities within the DR under the four alternative future scenarios detailed below. The internal and international models are loosely coupled, in that the distribution of Haitian immi- grants at each time period influences (slightly) internal migration in the subsequent time period, and internal migration within the country influences how migrants from Haiti chose a destination. When estimating migrants for each five-year period (in a sequential step-by-step manner), the internal model was applied first, followed by the international model that was used to distribute projected migrants from Haiti across the DR. 11 Climate migrants are those driven by factors, either directly or indirectly related to climate change, while “other” migrants are those who are driven by factors other than climate change. The model is able to distinguish between the two by assessing the counterfactual scenario in which no climate change occurs against future scenarios that include climate change. Hence, “other” migrants are simply the number of migrants derived from the scenario without climate change, and climate migrants are the difference between this scenario and the one pertaining to climate change. 12 While this definition is debatable, a decision was taken for the purpose of this study so that estimates would be more meaningful, actionable, and consistent with historical estimates of internal migration. It should be acknowl- edged that smaller or larger geographic definitions would influence the number of estimated migrants. 13 Note that sea level rise is not included in this model. 26 Scenario Framework This work adopts the extended scenario-based Two RCPs (2.6 and 8.5) are included in this approach from the Groundswell Africa series modeling work as drivers of climate impacts. (Rigaud et al. 2021), in which four plausible RCP2.6 is a low emission scenario (Low Emis- future internal climate migration scenario sion scenario), while RCP8.5 is characterized combinations are examined based on com- by increasing GHG emissions over time, lead- binations of the Representative Concentra- ing to high atmospheric concentration (High tion Pathways (RCP) (van Vuuren et al. 2014) Emission scenario). RCP8.5 implies little to no and Shared Socioeconomic Pathways (SSP) climate policy, and is characterized by signifi- (O’Neill et al. 2014). For each scenario, the cant increases in CO2 and CH4 emissions. projection represents an ensemble of mod- el runs using combinations of crop, water, On the other hand, SSP scenarios span a wide ecosystem, and Sea Level Rise (SLR) impact range of potential future development path- models from the Inter-Sectoral Impact Model ways, and describe trends in demographics, Intercomparison Project (ISIMIP—more details human development, economy and lifestyle, on this data below). The scenario approach policies and institutions, technology, and has several advantages. First, exploring mi- environmental and natural resources. Broadly, gratory outcomes across alternative physical they are organized according to their respec- and demographic/socioeconomic futures tive challenge to adaptation and mitigation allows researchers to begin to characterize in each future world. Importantly, climate the size and sources of uncertainty associated change impacts are not directly included in with projections of climate-induced migra- these scenarios. They can be thought, how- tion. Second, varying both climate and demo- ever, as consistent with broad assumptions graphic/socioeconomic pathways provides a regarding the primary factors driving chal- means to evaluate different policy options in lenges to adaptation and mitigation, namely terms of the impacts on spatial population population and emissions, respectively. Na- outcomes and the potentially avoided climate tional-level estimates of population, urban- impacts of achieving more advantageous ization, and gross domestic product (GDP) climate and societal outcomes (e.g., Oleson et are publicly available and are used for this al. 2015; Martinich et al. 2018). project. Population estimates include as- sumptions regarding international migration; however, as mentioned, these assumptions On the one hand, RCPs are global scenarios are made in the absence of any information representing trajectories of GHG concentra- regarding climate change, exposure, and tions (and other pollutants) resulting from vulnerability. No assumptions are made in human activity and corresponding to a specif- these scenarios regarding internal migration. ic level of radiative force in 2100. These were In this work, two different SSPs are proposed developed in advance of the IPCC 5th Assess- for consideration: (i) SSP2 (Middle of the Road ment Report (IPCC, 2014). The scenarios simu- scenario) describes a world with development late not only the volume of emissions, but also that occurs at rates consistent with historical their corresponding climate change impacts. 27 patterns, and therefore has moderate levels mitigative capacity in places where it matters of investment in human capital, technolog- most in terms of global emissions. In other re- ical change, and economic growth. (ii) SSP4 gions, however, development progresses slow- (Inequality scenario) describes a mixed world, ly, inequality remains high, and economies with relatively rapid technological develop- are relatively isolated, leaving them highly ment of low-carbon energy sources in key vulnerable to climate change with limited emitting regions, leading to relatively large adaptation capacity. Following the Groundswell approach, four plausible socioeconomic and climate futures are pro- posed by combining these SSP and RCP scenarios (Figure 6). The scenarios allow for an examina- tion of the relative importance of climate futures in driving potential migration outcomes. The scenarios can be characterized as follows: 1 A Pessimistic reference scenario (SSP4 Inequality and RCP8.5 High Emissions), in which global emissions remain high and development is disparate. Population growth in middle-income countries generally slows over the next few decades and the population declines after mid-century due to significant economic un- certainty. Urbanization rates are high due to rural economic decline, and GDP growth and education levels remain stagnant. Urban growth is poorly planned and high emissions drive greater climate impacts. This scenario poses high bar- riers to adaptation because of the slow pace of development and isolation of regional economies. The other three scenarios are defined in comparison to this reference scenario. 2 3 4 A more Climate-Friendly A more Inclusive Development sce- An Optimistic scenar- scenario (SSP4 Inequality nario (SSP2 Middle of the Road and io (SSP2 Middle of the and RCP2.6 Low Emis- RCP8.5 High Emissions), which retains Road and RCP2.6 Low sions) with lower emis- high emissions from the Pessimistic Emissions), which com- sions that reduce climate scenario, but provides a development bines the lower emission impacts, but holds the scenario that is more optimistic and scenario that reduces development scenario the potential for adaptation is higher climate impacts and consistent with the pessi- than under SSP4. Population growth provides a development mistic scenario. is higher, but urbanization rates are scenario that is more lower than in SSP4, while progress in optimistic. education and GDP are higher than in SSP4. 28 Figure 6: Scenario Framework Source: Framework drawn from Rigaud et al. 2021. For each scenario, a small model ensemble consisting of four models was applied. Each ensem- ble member considers a different combination of General Circulation Model (GCM)—a global “climate model”—and sectoral impact models (which use the data produced by the GCMs to project changes in crop yield, water stress, and ecosystem productivity). The ensemble approach accounts for uncertainty in climate outcomes by considering a range of projected futures under each RCP. Data The data used in the modeling was drawn from the original Groundswell project, updated to reflect more recent updates where possible, and modified slightly for this project. Table 1 includes a list of the data employed in the modeling analysis. Detailed information about each data source and how they were included in the model can be found in Appendix 3. 29 Table 1: Datasets, Variables, and Sources Variable Source Resolution Time Series Time Step Indicator Climate Driven Water Availability ISIMIP 0.5° 1970-2100 5-year Deviation from baseline Yes Agriculture/Crop Yields ISIMIP 0.5° 1970-2100 5-year Deviation from baseline Yes Biomes/Ecosystem ISIMIP 5-year Deviation from baseline Yes 0.5° 1970-2100 Productivity Sea Level Rise Changes in coastline Yes NASA 30m 1990-2150 5-year (inundation) Bilateral Internal N/A ONE Municipal 2005-2010 5-year Count of migrants Migration Count of migrants Bilateral International ONE 5-year (Haiti to Dominican N/A Municipal 1990-2010 Migration Republic municipalities) Municipal Population ONE Municipal 2010 N/A Count of people N/A Spatial Population 1-year Population counts WorldPop 100m 2000-2020 No (Totals, Age, and Sex) by age and sex Elevation SEDAC 30m 2015 N/A Corrected elevation No Slope SEDAC 30m 2015 N/A Avarage slope No Water Bodies ESRI Vector 2019 N/A Surface water No World Database on IUCN Vector 2019 N/A Mandate for protection No Protected Areas Source: World Bank staff Notes: ISIMIP = Inter-Sectoral Impact Model Intercomparison Project; ONE = National Office of Statistics (Ofi- cina Nacional de Estadística) of the Dominican Republic; Worldpop = University of Southhampton’s database; SEDAC = NASA’s Socioeconomic Data and Applications Center; and IUCN = International Union for Conserva- tion of Nature. The impacts of climate change act through the a viability that will figure into the migration first four variables on the list: water availabil- decision. Crop yields are modeled only for ity/stress, crop yields, ecosystem productivity, maize, wheat, rice, and soybeans, which are and SLR. For the purpose of this project, out- the main global staple crops selected for the puts are used from the ISIMIP modeling effort Groundswell study. 15 The ecosystem models for crop production, water availability, and act in areas not covered by crop productivity ecosystem impacts, which cover the historical data and simulate the natural growth of sev- period 1970–2010 and projections for 2010– eral different plant functional types, includ- 2100. The future sectoral impact models are ing grasses. Hence, Net Primary Productivity driven by a range of general circulation mod- (NPP), simulated by these models, serves as els.14 The crop, water, and ecosystem simu- an estimate of the productivity of a location’s lations—at a relatively coarse spatial scale natural biome, including grassland biomes (0.5°)—represent indicators that capture the that may potentially support pastoral liveli- impact climate may have on specific types of hoods. livelihoods, 30 Other variables that capture aspects not re- mask used to restrict the land deemed suit- lated to climate change or natural degrada- able for human habitation. The population tion also are included in the model. Bilateral variable (WorldPop) serves as the base year internal and international migration data, population distribution (total population the latter of which consists only of flows count by grid cell), as well as two variables from Haiti to the DR, are from the National in the migration model (age and sex struc- Statistics Office (Oficina Nacional de Es- ture). Each of these variables were tested and tadística) of the DR, as are the municipal validated, and previous applications of the population data. These data were used to model found them to be statistically and/or help train the model and are unique to this practically significant in driving positive or INCLUDE model application. The final four negative impacts on the relative attractive- variables on the list are not drivers of mi- ness of various locations. (For more method- gration but, instead, constitute the spatial ological details, see Appendix 3.) 14 Applied here are data that are driven by two general circulation models that provide a good spread for the tem- perature and precipitation parameters of interest: the HadGEM2-ES and IPSL-CM5A-LR climate models (more details in the methodological appendix. 15 While other important crops are missing, the assumption made for modeling is that the deviation of these crops from historical productivity (which is what is actually used in the model) should correlate with that of other missing crops. 31 RESULTS This section reviews results and key findings from the modeling work. For purposes of clarity, the results are organized into separate subsections pertaining to (i) internal migration and (ii) migration from Haiti to the DR. Projected Internal Climate Migration 2020-2100 Depending on the scenario, the average esti- The most Optimistic scenario (RCP2.6/SSP2) mated number of internal climate migrants yields roughly 170,000 internal climate mi- range between 149,000 to 368,000 for the peri- grants by mid-century, rising to 234,000 by od 2020-2050 (1.2- 2.8 percent of the country’s the end of the century and comprising 1.3 population and 30-74 percent of 2005-2010 percent and 1.9 percent of the population at internal migrants), and between 234,000 and that time, respectively. While these represent 473,000 additional climate migrants for the a relatively small proportion of total popula- period 2051-2100 (Table 2). Under each of the tion in the Optimistic scenario, climate mi- four scenarios, the model produces estimates grants nevertheless comprise over 27 percent of internal climate migrants, those driven to to 31 percent of all internal migrants under move due to some climate-related impact both scenarios. For reference, according to (water availability, crop yield, ecosystem pro- the 2010 Census (GoDR, 2010), 497,470 people ductivity, and sea-level rise), and “other” inter- (5.3 percent of the population in 2010) moved nal migrants, which encompasses any migra- across municipalities between 2005 and tion not related to climate change but, rather, 2010. It should be noted that, reflecting the to development issues. The most Pessimistic difficulty to predict an uncertain future, the scenario (RCP8.5/SSP4) yields roughly 346,500 reported estimates for each scenario repre- internal climate migrants by mid-century, sent, in themselves, averages of a wider range rising to 473,000 by the end of the century, (as reported in Table 2), estimated using the comprising 2.9 percent and 5.3 percent of the ensembled models (composed by four models population at that time, respectively. each). Hence, estimates should not be consid- ered a certain or exact number. 32 Table 2: Projected Internal Climate Migrants by Scenario, by 2050 and 2100: Dominican Republic (total, as percentage of all migrants, and as percentage of population). Optimistic 2020 - 2050 2051 - 2100 SSP2 - RCP2.6 Min Avg Max Min Avg Max Climate migrants 148,787 169,654 195,611 183,941 234,021 288,782 % of internal migrants 24.35% 26.85% 29.74% 25.68% 30.54% 35.17% % of population 1.13% 1.29% 1.49% 1.48% 1.89% 2.33% More Inclusive Development 2020 - 2050 2051 - 2100 SSP2 - RCP8.5 Min Avg Max Min Avg Max Climate migrants 311,148 368,222 440,025 362,190 463,158 587,748 % of internal migrants 40.24% 44.35% 48.78% 40.49% 46.53% 52.47% % of population 2.37% 2.80% 3.35% 2.92% 3.74% 4.74% More Climate Friendly 2020 - 2050 2051 - 2100 SSP4 - RCP2.6 Min Avg Max Min Avg Max Climate migrants 124,718 148,651 177,935 193,062 266,293 337,127 % of internal migrants 13.20% 15.34% 17.82% 15.58% 20.29% 24.37% % of population 1.04% 1.89% 1.48% 2.18% 3.00% 3.80% Pessimistic / Reference 2020 - 2050 2051 - 2100 SSP4 - RCP8.5 Min Avg Max Min Avg Max Climate migrants 267,496 346,497 428,617 336,448 473,204 618,478 % of internal migrants 24.59% 29.69% 34.32% 24.34% 31.15% 37.16% % of population 2.23% 2.89% 3.57% 3.79% 5.33% 6.97% Source: World Bank staff analysis Note: Within each scenario, and ensembled model was composed using four models (GCM) each. These create a range of possible results, represented by the min and max, and the average of these. While the average is reported mostly in the text, the reader should keep in mind the wider range of possibilities and the uncertainty that it represents. Unsurprisingly, the total number of projected RCPs are held constant, SSP4 (Inequality) climate migrants are higher in the two more future leads to slightly fewer climate migrants pessimistic climate scenarios (those based on at mid-century, followed by a more rapid RCP8.5), with over 473,000 and 463,000 peo- increase in climate migration after mid-centu- ple in 2100 driven to move by climate-related ry. Even in the reference Pessimistic scenario factors under SSP4-8.5 (Pessimistic reference) (SSP4/RCP8.5), the number of climate migrants and SSP2-8.5 (More Inclusive Development), is lower than in the scenario that holds high respectively. Hence, these estimates suggest emissions but the SSP2 (“middle of the road”) that, regardless of development projection, a development projections. Likewise, the number future with high emissions would lead to more of projected climate migrants by 2050 under climate migrants than one with lower emis- SSP4/RCP2.6 (more Climate-Friendly)—the sions. scenario with the optimistic climate scenario but coupled with the more pessimistic so- Interestingly, the lowest number of climate cio-economic future, are the lowest. What is migrants by 2050 did not come from the Op- discovered is that in the earlier decades of this timistic scenario but, rather, from the sce- projection, socioeconomic concerns dominat- nario that projects low emissions with an ed (reflected in the portion of total migrants unequal development (SSP4/RCP2.6 - “More defined as “others” and in Figure 7), while the climate-friendly”), revealing a trend in which impacts of climate change on migration in- the relative importance of development issues crease in the latter half of this century will initially crowds out the climate impacts on mi- become continuously more important, particu- gration, but later takes larger relevance. When larly in the form of rural-to-urban migration. 33 The SSP2-8.5 future (“More inclusive development”) is the scenario in which internal climate mi- grants are closer to the number of other internal migrants, particularly by end-century (Figure 7). As Table 2 and Figure 7 show, this scenario has the highest percentage of climate migrants rel- ative to all internal migrants (44 percent in 2050 and 47 in 2100), and even by 2100, in the max estimate of the potential range, climate migrants slightly outnumber other migrants. The out- come is reasonable given that the SSP2 future (“Middle of the Road) represents a more positive socioeconomic future for a larger proportion of the population (e.g., continued slow convergence in educational attainment and income across all segments of the population) than the SSP4 future (Inequality), which would lead to less migrants driven by economic factors, particularly if the diffusion of wealth reinforces the viability of the agriculture sector. At the same time, because SSP2 is coupled with the RCP8.5 future (high emissions) in this scenario, climate impacts are severe, and thus the number of people forced to move due to climate-based disruption is high- er. This scenario should be considered indicative of a likely outcome if climate change continues unabated but socioeconomic conditions improve. Conversely, the reference Pessimistic (SSP4-8.5) scenario, which also projects a similarly high number of climate migrants—although substantially more internal migrants driven by development factors (and hence, more migrants in total)—is a more likely outcome if a future of unabated climate change is coupled with increased inequality within the country. Figure 7: Total Internal Climate and Other Migrants by Scenario, 2050 and 2100: Dominican Republic Source: World Bank staff analysis Notes: RCP = Representative Concentration Pathways; SSP = Shared Socioeconomic Pathways. Whiskers represent min and max values from ensembled model (uncertainty in the estimate). Other migrants are calculated using the climate migrants estimates, and so the uncertainty is the same. It is important to note that the population size Under SSP2, the population reaches 13 million of the DR changes between SSPs. As one might in 2050, peaks around 2070, and then declines deduce by looking at the proportion of total to just over 12 million in 2100. As such, and population comprised of climate migrants, the despite projecting a similar total number of cli- SSP4 (Inequality) future is one where the DR mate migrants in the SSP4 scenarios (relative population declines post mid-century, de- to SSP2 scenarios), climate migrants represent creasing from 12 million in 2050 to 8.9 million a substantial larger proportion of population in in 2100. 2100 under both SSP4 scenarios. 34 Estimates of total internal migrants (climate much higher rates of internal mobility than plus others) show expected trends in which the SSP2 future. Second, regardless of SSP, the pessimistic development and climate scenari- RCP8.5 scenario (High Emission) leads to more os lead to larger numbers of internal migrants internal climate migrants than the correspond- (Figure 8). Three significant trends emerge. ing RCP2.6 scenario, an expected result given First, both of the SSP4-based (Inequality) sce- the more severe climate impacts projected un- narios project a higher number of total internal der RCP8.5. Finally, under the Optimistic (SSP2- migrants than either SSP2-based scenario. The 2.6) scenario, domestic migration is projected SSP4 scenario also projects a much smaller to decline slightly over the next few decades total population (over 3 million fewer people) from observed 2010 numbers, before ticking for the DR by the end of the century, meaning back up after mid-century. Figure 8: Total Internal Migrants (Climate plus Other) under each Scenario, 2010- 2100: Dominican Republic Source: Numbers for 2050 and 2100 come from World Bank staff analysis. The number for 2010 represents the total number of people moving between districts between 2005-2010, calculated by World Bank staff based on the 2010 National Population and Housing Census of the Dominican Republic (GoDR 2010). Notes: RCP = Representative Concentration Pathways; SSP = Shared Socioeconomic Pathways. Dotted lines represent min and max values from ensembled model. There is a fair degree of stability across scenar- The largest net loss in population to cli- ios in terms of the municipalities with the big- mate-related factors over all scenarios occurs gest gain and loss of populations as a result of in the municipalities of Azúa, San Cristóbal, climate change. The largest gains from climate Baní, La Romana, Barahona, San Pedro de migrants are consistently projected to occur Macorís, Santiago, and San Juan. With the in the urban municipalities of Santo Domingo exception of Santiago and San Cristóbal, Norte, Higüey, Santo Domingo Este, Los Al- the projected climate out-migration of these carrizos, Villa Hermosa, and Santo Domingo locations is considerable compared to their Oeste, in that order. When net total internal population size. Most notably, the municipal- migrants are considered, the largest gainers ities with the largest out-migration in relative are mostly the same locations, with the addi- terms to their population are the southern tion of Distrito Nacional. coastal municipalities of Sabana Grande de 35 Palenque, Barahona, and Azúa, the border Climate out-migration hotspots include a municipalities of Jimaní, Comendador, and few cities in El Cibao, including La Vega and Pedernales (which is also on the southern Moca, as well as a few urban locations along coast), and the municipalities of Tamayo, Nei- the South Coast (e.g., San Pedro de Macorís, ba, and San Juan. La Romana, and San Cristóbal). Santo Do- mingo itself yields an interesting pattern over Maps with hotspots of climate in- and out-mi- time. The urban fringes of Santo Domingo gration at the native 1-kilometer (1km) resolu- (Norte, Oeste, and Este) consistently attracts tion show these trends in more detail (Figure climate in-migrants over the century. In con- 9). For practical purposes, hotspots are the trast, central Santo Domingo attracts them locations that exhibit the largest (top or bot- from across the country during the first half of tom 10 percent of all grid cells) differences in the century, but then becomes a net exporter population totals across the four scenarios of climate-induced migrants to its suburban with climate change when compared to their fringe in the second half of century, as the corresponding “development-only” migration densely settled urban core faces increased scenarios,16 and for which there is agreement water stress because of reduced water avail- between three or four of the ensemble mem- ability and increased population density, and bers in each scenario regarding the outcome. deteriorating conditions along the coastline. From the top panel, the primary hotspots of Finally, in El Cibao, a distinct pattern is clear climate in-migration are the suburbs/urban over time, with climate impacts generally fringe of Santo Domingo, Villa Hermosa, and nudging people to the Northwest, away from the city of Higüey in the Southeast, and the the larger cities and toward the current small- area along the main highway to Santiago er urban areas further north, where conditions (region of El Cibao North and Northwest). are projected to be—in relative terms from climate change—better. 16 Simply the SSP2 and SSP4 futures projected under the assumption that climate remains constant. 36 Figure 9: Projected Hotspots of Climate In-Migration and Out-Migration by 2050: Dominican Republic Source: World Bank staff analysis Notes: Reds are hotspots of climate in-migration and blues of out-migration. Hotspots are the locations that exhibit the largest (top or bottom 10 percent of all grid cells) differences in population totals across the four scenarios with climate change when compared to their corresponding “development-only” migration scenarios. Darker shades of each color indicate areas where all four ensemble models project high levels of in/out migra- tion across all scenarios, and lighter shades indicate that three of the four models agree. Places with larger populations are more likely to drive large numbers of migration and to appear as hotspots. A complete map for 2100 is provided in Appendix 4. 37 It is essential, when considering the hotspot maps, not to confuse areas of consistently projected high levels of climate-induced in-/out-migrants with a total population gain and loss. Hotpots of climate in-migration, for example, are regions where the number of total in-migrants is increased by the impacts of climate change when compared to a no-climate change scenario. This can occur even if total population is declining (as a result of natural decrease or out-migration for non-climate reasons). Similarly, the population can be increasing in a place characterized as a climate out-migration hotspot. Hotspots should be interpreted only as representative of the impact of climate change on total internal migration. For example, when “other” migrant move- ments —unrelated to climate change—are included in the projections, larger cities, such as San- tiago or Distrito Nacional, continue to represent a net increase, even if they have out-climate migrant hotspots. The largest numbers of internal climate mi- in urban areas—so-called urban-to-urban mi- grants will arrive to dense settlements (i.e., grants—as well as rural-to-urban migrants). cities), and come from irrigated croplands, pas- The more dire climate scenario (RCP8.5) toral lands, and rainfed croplands. This study produces more than twice the number of mi- assesses the impact of climate change on grants into urban areas than the RCP2.6 (Low internal migration by livelihood zone, dividing Emission) scenario under both SSPs. Irrigated the country into regions dominated by urban croplands and pastoral/rangelands (largely in landscapes, irrigated croplands, rainfed crop- El Cibao and adjacent regions of the Cordille- lands, pastoral/rangelands, and woodlands; ra Central), are generally the largest net send- and tally the projected number of net climate ers of climate migrants. Rainfed croplands migrants over each. Under all scenarios (Figure generally fair better (although are still net 10), dense settlements are significant net re- senders under all but the SSP4-8.5 Pessimistic ceivers of climate-induced migrants, suggest- scenario in 2100 (See Appendix 4). The low ing that the vast majority of people who move population in woodland areas renders these as a result of climate impacts seek out urban regions somewhat neutral. areas (this includes migrants already located 38 Figure 10: 1 Net Internal Climate Migrants by Livelihood Zone and Scenario, 2050: Dominican Republic. Source: World Bank staff analysis 1 Data for 2100 is provided in Appendix 4. Note: SSP = Shared Socioeconomic Pathways. These numbers include a small and conservative estimate of SLR’s effects on human mobility. SLR as a driver on its own accounted for a small portion of projected climate migrants (less than half a percentage point by 2050 and less than 2.5 percent by 2100 in the SSP2 scenarios), but this could be due to limitations on modeling the diverse factors of this phenomenon beyond inunda- tion and storm surge and because movements due to SLR may be within the same municipality (see limitations section in Appendix 4 for more information). Projected International Climate Migration 2020-2100 For this project, the INCLUDE model was com- which hold climate constant (SSP-only). How- bined with an econometric model for project- ever, whereas the internal INCLUDE mod- ing international migration from Haiti to the el operates at a 1 kilometer resolution and DR, which reports slightly different outputs. results are aggregated up to the municipal Like the internal model, the impacts of cli- level, the international model projects a single mate change are inferred by comparing the estimate of total migrants from Haiti to the scenarios in which emission pathways and DR at each time step, and those migrants are 17 climate change impacts are included (e.g., then distributed across the 155 municipali- RCP2.6 and RCP8.5), as well as scenarios ties of the DR. Climate migrants originating 39 in Haiti are derived by municipality within the Likewise, the wide range of outcomes across DR and at the national level. Also similar to scenarios reflects a significant level of uncer- the internal model, other and total migrants tainty in the relationship between nationwide are projected and reported for each time- climate drivers in Haiti and of movement step/scenario. Nevertheless, the uncertainty into the DR. Additionally, it should be noted in this model is more pronounced than in the that these figures reflect no change in border internal model, given the less-granular origin policy (on either side) over the course of the data and the high correlation of climate and century—an unlikely outcome if the number economic drivers in Haiti. of migrants changes substantially, if con- ditions deteriorate substantially, or both. It also reflects a future in which Haiti’s fragility, The range of outcomes across and within conflict, and vulnerability conditions continue scenarios is fairly wide, reflecting wide uncer- the trends of the last few decades (which are tainty in the exact number of migrants. For reflected in the historical data used for this example, under the Pessimistic scenario, if model). Given the increased uncertainty in each ensemble member is considered individ- this international model, the estimate table ually, the maximum number of the scenario is (Table 3) and the description of results focus more than three times the minimum and with on relative changes and trends instead of differences of many hundreds of thousands. estimated numbers. The range of outcomes On the one hand, the wide range of climate for each scenario in Table 3 reflects the mini- migrants reflects uncertainty in the climate mum and maximum share of climate migrants inputs from the ISIMIP models (data source projected by each individual model ensemble of the climate inputs). On the other hand, member, and the average reflects the ensem- uncertainty in the total number of migrants ble mean. 18 The reported trends should be is driven by not only climate uncertainty but taken cautiously and used in conjunction with also by the interaction between climatic and other data sources and knowledge to inform socioeconomic conditions. policymaking. 17 This was the number of municipalities in the 2010 Census, which was the basis for this analysis. This number has changed since then. 18 As mentioned in the methodological section, for each scenario, a small model ensemble consisting of four models was applied. Each ensemble member considers a different combination of GCM (a global “climate model”) and sectoral impact models (which use the data produced by the GCMs to project changes in crop yield, water stress, and ecosystem productivity). The ensemble approach accounts for uncertainty in climate outcomes by considering a range of projected futures under each RCP. 40 Climate migrants coming from Haiti into the DR are expected to represent six to 20.1 percent of all migrants by 2050, with the highest number and share of climate migrants in the Pessimistic scenario. Under the Optimistic SSP2/RCP2.6 scenario, the average projected share of climate migrants from Haiti to the DR is around 6 percent by mid-century. This proportion rises to 11.2 percent by the end of the century. The scenario SSP2/RCP8.5 (more Inclusive Development) that holds development pathways inclusive but uses the Pessimistic climate scenario, has an average projection of 19 percent by mid-century, owing to the worsening environmental conditions in this scenario. This percentage is expected to rise to 26 percent by the end of the century. Unlike the middle-income DR, the population under SSP4 (Inequality) in Haiti, a low-income country, is projected to be substantially higher than under SSP2 (Middle of the Road), a result of a much slower decline in fertility rates projected under SSP4 (relating to stagnant education attainment levels under the more Inequality scenario). As such, the average total number of projected arriv- als from Haiti is expected to be much greater under the two SSP4 scenarios. Climate migrants correspond to over 20 and 28 percent of total migrants from Haiti under the Pessimistic scenario for mid and end of century, respectively (the highest share among scenarios). Two trends domi- nate these projections: (i) the SSP4 scenarios produce more total migrants as a function of wors- ening socioeconomic conditions in Haiti relative to the DR, and (ii) the RCP8.5 scenarios project a much larger number of climate migrants than the corresponding RCP2.6, in absolute terms, as well as a proportion of total migrants. Table 3: Percentage Climate Migrants from Haiti to the Dominican Republic % of climate migrants 2020 - 2050 2051 - 2100 relative to total migrants Min Avg Max Min Avg Max Optimistic RCP2.6 / SSP2 4.9% 6.0% 7.0% 9.6% 11.2% 12.8% More Inclusive Develompent RCP8.5 / SSP2 14.5% 118.8% 20.3% 18.0% 26.0% 27.6% More Climate Friendly RCP2.6 / SSP4 6.9% 7.8% 8.5% 10.5% 13.3% 14.4% Pessimistic / Reference RCP8.5 / SSP4 16.3% 20.1% 22.5% 22.2% 28.1% 31.1% Source: World Bank staff analysis Notes: RCP = Representative Concentration Pathways; SSP = Shared Socioeconomic Pathways 41 The estimated distribution of these future climate migrants into municipalities seems largely similar to current trends, but with some changes in the local intensity of in-migrants. In terms of absolute number of climate in-migrants, the capital region, urban areas of El Cibao (especially Santiago, Guayubín, and Mao), the livestock and agricultural region of San Juan, and the eastern municipality of Higüey, are the largest receivers of Haitian immigrants. Many of these munici- palities, however, are dominated by urban areas and are thus heavily populated. Expressed as a proportion of municipal population (Figure 11), municipalities along the Haitian border (e.g., Mon- te Cristi, Pepillo Salcedo, Pedro Santana, Jimaní, and La Descubierta), and important agricultural locations in the Southwest such as Sabana Yegua, Polo, and Rancho Arriba a bit further to the Northeast, disproportionately receive Haitian migrants. If this trend develops at the intensity pro- jected by the model, these municipalities may find certain public services stressed by the influx of migrants. With adequate preparation and planning, however, the influx of potential agricultural workers could also provide a boom to the local economies. It therefore is important to consider any potential large-scale migration for the opportunities it offers, as opposed to assuming mobili- ty is an inherent threat. Figure 11: Projected Climate In-Migrants from Haiti (top maps), Expressed as a Proportion of Pop- ulation (bottom maps) by Municipality, 2050 and 2100: Dominican Republic 2050 2100 Source: World Bank staff analysis 42 Migrants coming from Haiti, who are driven by most of El Cibao, including Santiago, the latter the impacts of climate change, could change a city projected to be a net sender of domestic the distribution of Haitian migrants across climate migrants. Elsewhere, a corridor running municipalities. When projected Haitian mi- through most of the Cordillera Central and grants are disaggregated into climate/other south through Azúa Province is projected to by municipality, Figure 12 illustrate how these experience a relative increase in the proportion climate migrants may shift the country’s pro- of Haitian in-migrants, while Santo Domin- portional distribution (relative to all Haitian go and areas along the border with Haiti are migrants).19 The model projects an increasing expected to experience slight declines in the proportion of Haitian migrants arriving across relative proportion of Haitian arrivals. Figure 12: Difference in Proportion of Haitian In-Migrants: Dominican Republic 2050 2100 Source: World Bank staff analysis. Notes: To calculate this, the analysis takes the 2050 and 2100 distribution of all migrants and compares it against a hypothetical scenario in which Haitian climate migrants follow the same pattern of municipal residence as non-climate migrants - the difference can be attributed to the municipal pattern driven by climate impacts. 19 To calculate this, the analysis takes the 2050 and 2100 distribution of all migrants and compares it against a hypothetical scenario in which Haitian climate migrants follow the same pattern of municipal residence as non-climate migrants - the difference can be attributed to the municipal pattern driven by climate impacts. 43 part 2: qualitative case study of restauracion municipality in terms of Internal emigration caused by climate change and natural degradation 44 Methodology This study used a qualitative in-depth semi structured interview method to explore the deci- sion-making of internal migrants and its relationship to climate and environmental variables. Because of the method, the study was performed in a delimited area—Municipality of Restaura- ción—so that responses and experiences were able to be better contextualized. Restauración was selected based on consul- tation with local experts, since it interlinks a target area for reducing carbon emissions and because of its relevance on migration, agri- culture and forestry. It is not only part of the influence area of ERP (Appendix 5), but also is a municipality that has historical relevance to the process of migration, internally and inter- nationally, due to its location on the border with Haiti and registering previous significant internal migration waves. Furthermore, for- estry and agriculture are important economic activities and, theoretically, they constitute a mechanism through which climate change and natural degradation may impact internal migration. The characteristics of Restauración are described in the following section. Interviewees were Dominicans who had emigrated from localities within Restauración to other localities in the DR and who had been involved with agriculture in their community of origin. While the main target subjects were internal migrants who emigrated because of climate change, this characteristic was not possible to identify, ex-ante, due to various challenges. Instead, it was decided to prioritize those who had (i) emigrated, (ii) who they themselves or their families had worked in agriculture within their community of origin, and (iii) who were originally from com- munities which, according to key informants, suffered significantly from degradation in terms of natural resources compared to other vicinities within the municipality. The expectation was that by focusing on these characteristics, the sample would include emigrants who had been impact- ed by climate change and natural degradation. Nevertheless, while the driver may have been climate change, it was difficult to accurately establish a distinction between those environmental impacts caused specifically by climate change or other causes. The methodology therefore cen- ters on a discussion of mechanisms that may have been caused by climate change (e.g., reduced water availability or land productivity). It is important to note, however, that the influence of direct human actions, such as deforestation or the overexploitation of resources, as well as other climatic variables, may also have influenced these mechanisms. 45 Because of the challenges identifying mi- ditional interviews were not providing new grants, which by definition do not live any- information, which is a recommended criteria more in the specific area of interest, the study to determine sample size in qualitative meth- used a non-probabilistic sampling method ods, called “reaching saturation” (Skovdal and called “snowball”. The research team started Cornish, 2015). The limitation of this method with interviews to key informants (e.g., local is that the subjects may be connected, poten- public workers, local leaders, etc.) and asked tially reducing the variability of experiences. if they could connect the team with people However, it was assessed that this was the that emigrated from the two localities and best available method for this research topic. had been involved with agriculture; as the Because of the logistical challenges of com- team interviewed new participants, more con- muting to diverse locations in the country tacts were requested from new interviewees. where participants emigrated, most interviews The team continued this process until ad- had to be performed remotely via cellphone. Ultimately, 17 emigrants and 10 key informants were interviewed, reaching an acceptable sample size and representation of subgroups to be able to carry out a rigorous qualitative study. While both men and women tend to emigrate from Restauración, despite multiple efforts to find times and venues suitable for women, most of those the team was able to interview were male (13 out of 17). It was also shared by key informants that most of those who emigrate are the young —a characteristic reflected in the sample with a median age of 22 when migrated (minimum: 17; max- imum: 49). In the sample, some had migrated as recently as one year ago, while others had mi- grated as far as 40 years ago (median: 12; mean: 17). By design, every interviewee either had been involved in agriculture prior to emigrating or had helped her/his family in this activity. In particu- lar, most of the women did not consider themselves as farmers, preferring to focus on education or household tasks, despite alluding to assisting their spouses or parents in agricultural tasks, such as harvesting. The final sample size involved those who had emigrated from Las Rosas (11 out of 17) and Cruz de Cabrera (3), with the remaining coming one each from Berenjena, Cruce de Mariano, and the urban area of Restauración. 46 Climatic and Socioeconomic Context of Restauración Restauración is located in the northwest of the country and in the southern part of Dajabón Prov- ince. 20 It is demarcated to the north by the Municipality of Loma de Cabrera, to the south by Elías Piña Province, to the east by Santiago Rodríguez Province, and to the west with the border of Haiti (Figure 13). Restauración was created as a municipality in 1892 and is 283.6 km2 in area; it has 42 vicinities, including an urban area of the same name. Figure 13: Location of Restauración1 and Vicinities within Restauración,1 Dajabón Province, Dominican Republic Source: World Bank staff using Open Street Maps locations and roads and Census 2010 borders. 1 The yellow area in the map on the left represents Restauración, red dots represent main cities, and red and brown lines represent key highways; the map on the right shows vicinities of interest in Restauración, with the main highways and roads drawn in dark brown and light brown, respectively. 20 The province is formed of four municipalities: Loma de Cabrera, Restauración, Partido, and El Pino 47 According to data from the 2010 Census (GoDR Restauración has one main highway that 2010), Restauración had a population of 7,274 connects its vicinities with Haiti and other people. Fifty-four percent were men and 39 provinces of the DR. The main road to Restau- percent lived in the urban area. Within Da- ración, Highway 45, connects the municipality jabón Province, Restauración is the munici- and its urban area to Lomas de Cabrera to pality representing the lowest average age the Northeast, and provides access to not (26.4 years), and has the highest average only the provincial capital of Dajabón but number of children per family (2.5) (GoDR also to other roads that reach Santiago city 2017). It has poverty and extreme poverty via Duarte Highway, one of the main high- levels of 79 and 38 percent, respectively. Only ways in the DR. Trade, as well as most oth- 9.9 percent of the population is supplied with er transportation, head in this direction to 21 drinking water from a tap in their homes Dajabón. In the other direction, highway and 23 percent of the water is sourced from 45 also connects the Southwest to the inter- springs. In terms of fuel, the most prevalent national highway that traverses the DR to are firewood (52 percent) and coal (5.4 per- Haití and back to Elias Piña. Transport in this cent) (FUNGLODE n.d.). Restauración’s illit- direction is more difficult due to the border eracy rate is at 25.2 percent, higher than the crossings. Other local roads of lower quality national rate of 12.8 (GoDR 2017). emerge on Highway 45, such as one heading southeast in the direction of Elias Piña, con- necting many communities (e.g., Las Rosas and Los Cerezos) (Figure 13). 21 Annual operational development plan of the municipality of Restauración for 2020-2024 (Restauración Municipal Government 2020) 48 Agriculture and Forestry in Restauración Agricultural and forestry production constitutes the main source of income in the communities of Restauración. The municipal annual development plan reflects agro-industrial wood processing as the main economic activity, followed by agriculture and livestock. Agriculture is the sector with the highest employment (37 percent). The most common type of agriculture in the area is rainfed, which depends on the rainy and dry seasons and is concentrated on short-cycle products such as beans, peas, corn, and cassava, among others. Forest plantations are mainly of the pine varieties: Pinus occidentalis and Pinus caribaea, as well as Honduran mahogany, caoba (Swietenia mac- rophylla). For some key informants, the management and exploitation of forests have expanded significantly in the area during the last decade. Some interviewees complement their exploitation of wood with agroforestry products (e.g., coffee or cacao). In some cases, farmers compartmen- talize their land to enable multiple cash-crops, trees, and perennial crops. While livestock is also an important economic activity in Restauración, it is not as predominant as in the neighboring municipalities of Dajabón Province, and the neighboring province of Santiago Rodriguez Province where it is the main activity. Nevertheless, the intensity of these activities in the region has had environmental, trade, and economic implications for the municipality, particularly in the form of deforestation and land degradation. These communities have experienced various Nevertheless, many continue to use such un- changes in agricultural practices and agrarian sustainable practices, and the slash-and-burn systems. Changes in agricultural practices agriculture continues predominantly in the are the result of awareness raising by experts area. In terms of agricultural systems, these from the Ministry of Agriculture (Ministerio have changed due to natural and human de Agricultura), nongovernmental organiza- factors that have influenced crop substitution. tions, and international cooperation agencies. These include a decrease in peanut produc- These institutions encourage farmers to cease tion because of reduced demand and orien- the practice of burning and the use of agro- tation of buyers, the substitution of coffee chemicals and, instead, advocate replacing plantations for varieties resistant to diseases them with sustainable practices that include common in the area (roya del café; (Hemileia the use of live barriers, crop rotation, and or- vastatrix)); and its consequent impact on the ganic fertilizers. recovery of the forests. 49 Deforestation is a core issue in the municipality and province. According to the Hansen Global For- est Change v1.9 (2000-2021) dataset (based on Hansen et al. 2013), Dajabón is the third province with the highest forest lost rate between 2001 and 2020 with 7.6 percent; the second if we exclude the Capital that only has 8.7 percent of forest cover in 2000. Restauración was the 32nd (out of 155) district with more forest cover in 2000 due to 60.2 percent of its territory being covered by forest. Similarly, between 2001 and 2020, it had a deforestation rate of 22.8 percent, ranking 56th in the country, albeit 31st if considered as hectares lost (Figure 14 (upper image) shows the spatial distribution of deforestation in the municipality). Deforestation in the municipality is attributed to three main factors: (i) the presence of sawmills to take advantage of forest resources; (ii) slash- and-burn agriculture; and (iii) deforestation to create grassland for livestock. The presence of sawmills has led to the cutting of timber trees, mainly in mountainous areas, and the consequent loss of forest. Slash-and-burn agriculture also has contributed to deforestation, with the cutting of trees to make way for short-cycle agricultural crops or livestock production. The predominant importance of livestock as an economic activity in neighboring areas of Dajabón Province and Santiago Rodriguez Province also affects Restauración’s forests. Figure 14: Deforestation and Land Productivity Category in Restauración, Dominican Republic, 2001‒2015. Forest Cover Land degradation Sources: (i) Hansen Global Forest Change v1.9, based on Hansen et al 2013; (ii) Land Productivity Dynamics LPD-MO- DIS (derived from NDVI product of MODIS/Terra Vegetation indices 16-Day L3 Global 250m SIN Grid V006). 50 According to key informants, deforestation has had various impacts on water availability and soil productivity. Participants pointed to deforestation as the main reason that water streams, rivers, and rainfall are decreasing. Declining rainfall directly affects agriculture production, since crops depend on it. Further, rivers have raised in political importance as an issue of national security. Soil erosion, and its consequent reduced productivity, also is linked to deforestation; with the loss of forest cover, soil becomes devoid of protection from rain and wind. Torrential rains tend to wash away the soil and increase the erosive process—somewhat more pronounced in the mountainside areas. According to the Land Productivity global datasets, about half of the land in Restauración declined in productivity between 2011 and 2015. Figure 14’s lower image illustrates how this has generalized across the municipality in the period 2001-2015, including in the areas of Lomas de Cabrera and Las Rosas. “Well, the [trees] that have disappeared, the ones that we farmers have cut down. The upper part had a lot of mahogany, cedar, almond, cabirma, cigua. And that has disappeared, most of the rivers have been ceded to the farmers due to the lack of resources, if you are the father of a family and you do not have resources, you are going to depredate a forest to produce food (...) number of farmers manage the ground with fire (…), knocking down trees and the land has been degraded.” (Key interviewee, male, and a member of the Local Forest Association). When discussing Restauración’s environmen- workers to practice agriculture in their lands, tal issues, key informants often perceive that a with landowners receiving an (often large) contributing factor is the intensive unsustain- share of the production. A problem expressed able agricultural practices carried out by immi- by the interviewees is that these schemes often grant workers. Many key informants comment- use unsustainable practices, such as slash and ed on the large number of Haitian workers burn and indiscriminate deforesting. It was employed and living in the area, who occupy also recognized that Dominicans landowners or use land that has been abandoned or which sometimes have a shared responsibility in this is shared by Dominican farmers working under issue by allowing (or even incentivizing) these sharecrop schemes. Under these schemes, Do- practices. In some cases, key informants re- minican landowners allow Haitian immigrant ported that Haitian immigrants use lands that Dominicans left after emigrating. 51 Emigration and Climate Migration Emigration from Restauración According to interviews, the main reasons for people to emigrate from Restauración’s communities (emigration drivers) are lack of employment, low agricultural productivity, and the need and wish to pursue better educational opportunities. The first two are related as they represent barriers to meeting people´s needs and motivating them to find other options. In this sense, migration be- comes an adaptation strategy to low agriculture productivity and lack of alternative jobs. This strategy is most attractive to young people, who see the challenges that their parents confront with agriculture and aspire to different education and opportunities. “(I used to work) in agriculture, I used to work with peo- ple, and they gave me something (to do)… picking peas, planting pine trees in forests… It’s not easy for young people to be involved in agriculture… I don’t like agriculture.” (Interviewee No. 16, age 25, when migrated, female, who migrated to Santiago) The main destinations for emigrants from these communities are the urban centers of Dajabón, Loma de Cabrera, Montecristi, San- tiago, and Santo Domingo. The economic dynamism of urban centers makes these the preferred places of destination, since employ- Social networks, family, and/or friends are key ability is easier. In these locations, instead resources to facilitate the migration process of agriculture, participants are engaged in in any given destination. The migratory pro- activities of the tertiary or secondary sector, cess usually initiates as a result of contact such as mechanics, operating machinery, or with family or friends who serve as the link to having small businesses. Santiago and Santo migration; they are considered a key to facil- Domingo present advantages over Dajabón, itating the process. In the observed cases of Montecristi, and Loma de Cabrera in terms of families, usually the first to leave the commu- offering more diverse labor opportunities. The nity is the father; once established at the des- advantage of Dajabón and Loma de Cabrera tination, the other family members will follow. is their proximity to the communities of origin, The main help that friends or family provide which allows emigrants to maintain ties with is the provision of a place to live while the the communities from where they emigrat- now emigrant gets a job and can live inde- ed and in some cases continue carrying out pendently. Absent of these links could be seen productive work. These main destinations are as a barrier to use this adaptation strategy. also evident when looking at data from the 2010 Census (Figure 15). 52 Figure 15: Number of Interviews Held for the Study, and the Emigration from Restauración, Dominican Republic Sources: World Bank staff based on data from the 2010 National Population and Housing Census of the Dominican Republic (GoDR 2010) and from the number of interviews carried out for this study. Notes: The graph was created based on a question included in the 2010 Census regarding the municipality in which the person (5 years and older) was living in 2005. Emigrants on this map are those participating in the Census outside Restauración in 2010 and who reported as having been residing in Restauración five years pre- viously, in 2005. The data did not report emigrants from Restauración Municipality to their own district, and so are represented as NA. 53 Migration Caused by Climate and Natural Degradation Even though the participants’ most present reasons for emigrating are related to education or work, when inquired, they recognize how natural degradation and changes in climate played a role through loss of agricultural productivity. Participants were aware of the link between environ- mental degradation and migration, given the fragility of agriculture livelihoods due to hydrome- teorological factors (i.e., too much or too little rain), deforestation, and the loss of soil production, all of which translates into a loss of agricultural production. The following comments exemplify these links: “… What I cultivated was lost; when it was not because of one thing it was because of another. When I saw that what I grew resulted in many (economic) commitments, I got out of it, but I had to pay for those commitments [loans] because everything was lost.“ (Interviewee No. 8, 41, male, migrated to Montecristi). “… At the time there, people mistreated the trees, and noth- ing was achieved, and that is why many people decided to move from there. At the time, several of us emigrated be- cause the situation was terrible.” (Interviewee No. 11, 57, female, migrated to Santo Domingo). 54 Participants referred to the hydrological changes, particularly regarding precipitation and the effects on the sowing and harvesting seasons, which led to unpredictability and production loss in what had been a rainfed farming system. Since the area relies on rainfed agriculture, the reduc- tion or loss of agriculture production is identified by emigrants as the result of extremes (i.e., from dry to too much rain during the rainy season), thus influencing their decision to leave and seek employment elsewhere. Interviewees added that because of these extremes, it was difficult to maintain their traditional knowledge and practices. Furthermore, one person perceives that the reduced formation of rain clouds is directly the result of deforestation. “There in the countryside, there is no irrigation system… and (if) I sowed and it did not rain, it was spoiled… and if it rained, the rain was spoiling it. The agricultural system is based on the rain that makes it wet …. When the weather was favorable, the land produced peanuts, beans, corn… Many farmers from the countryside have emigrated.” (Interviewee No. 1, 60, male, migrated to Santiago). “All of that is needed. When there are no forests or any- thing like that, where do the clouds go to provide the fall of water?” (Interviewee No. 5, 41, male, migrated to Montecristi). “… I joined forces with a boy … We planted about two loads of beans, and they were lost [because] no water fell, and the harvest was a burden... I was left with a lot of trouble. I went to Dajabón and he went to Santiago. Agriculture is an adventure, of course, because if you plant and it rains, it could happen that you get production, but if it doesn’t rain, what can happen?” (Interviewee No. 6, 30, male, migrated to Santo Domingo). 55 Likewise, issues relating to low soil productivity are perceived as the cause of lower yields. For many interviewees, the land “is tired” or “it is denied to man.” The low productivity of soil relates to erosive processes, whereby when it is washed away by excessive water or wind, it no longer has the capacity to sustain agricultural crops. This fact is validated by the following comments, one of which is as follows: “As long as one helped the land, it was productive, but if he did not help it, it was not. If you repeat it three times it is productive, but the fourth time the products are not given, they give ‘aplejíos’ [crops that are not developed]. “ (Interviewee No. 5, 41, male, migrated to Montecristi). Farmers have attempted applying different strategies in order to adapt to the reduction in produc- tivity. Among the strategies used by the interviewees to deal with reduced production are the use of fertilizers, herbicides, and pesticides to improve crop productivity. Farmers also have moved to plots that are less degraded or that have been unused for some time; in other words, a migration of the land instead of migration to another place of residence. These strategies did not always achieved expectations and are often compounded by the challenges and sometimes it was diffi- cult of finding more productive lands (for example, one that could be irrigated): “But, if one sows with fertilizer there, he has a little strength, but if he doesn’t find how to put fertilizer or how to help her (the land), it doesn’t matter. In the times that I worked, there were years that came good and others that came bad, but when there was water everything was fine; but not when it was dry.” (Interviewee No. 5, 41, male, migrated to Montecristi). 56 “People sow in different places ... They say (one “(…) One lived changing (land). This year he sowed should) sow in the first and last months. The land is here and the other year elsewhere. (…) (because) one denied … If one had control of the water it would be thought that the land was bad and changed, seek- better. If it were irrigated land it would be better, but ing improvement.” if you sowed and it doesn’t rain you lose it.” (Interviewee No. 9, 27, male, (Interviewee No. 6, 30, male, migated to Santiago). migrated to Santo Domingo). Faced with such livelihood challenges, migration and the search for employment have become the main strategy to improve living conditions. As explained by one interviewee: “Las Rosas is difficult, because it is difficult to live without a job. Over there, people live off agriculture and things are lost when it rains or it is dry, and the land has been denied to man there. If it is not one thing, it is the other. To live there, you must have a job. “ (Interviewee No. 17, 41, female, migrated to Dajabón). 57 Staying, Despite Climate Change and Natural Degradation While many end up migrating influenced by the climatic and environmental conditions and low agricultural productivity, others stay and continue to work in agriculture and forestry. While our study design was meant to capture perceptions of people who migrated, key informants in the area also gave us the view of people who remain in the area and are able to make a living with agriculture and agroforestry. These cases contrast with those who have emigrated and serve to observe a resilience to these changes and strategies used. One of the most commented adaptation strat- One key informant even mentioned that the egies was the diversification of activities, development of forestry and agroforestry in mainly in the agroforestry and forestry sectors. Restauración has significantly improved the Agroforestry was referred to as one of the economic prospects of the municipality and main farming diversifications, particularly in has motivated people to return, including he, relation to the cultivation of coffee and co- himself. Likewise, other interviewees indicated coa, products that are encouraged by local that communities such as Baúl, Los Cerezos, branches of the Ministry of Agriculture (Min- and La Berenjena also have seen a return of isterio de Agricultura). Agroforestry not only migrants. While this excitement is likely war- promotes the conservation of forestry and re- ranted, there also may be difficulties to en- sulting land recovery, it also allows farmers a ter these economic activities for low-income source of annual income throughout the year farmers with few resources or land to invest, when the forests are cultivated with perennial as they require more time and investment crops (e.g., oranges, avocados). The cultiva- than “cash-crops”. Some policies are cana- tion of coffee and cocoa are reminiscent of lized through local associations, such as the traditional agricultural activities, practiced provision of plants and support to set up plant within areas that were originally affected by nurseries, and have help farmers overcome pests, where new varieties now have been these barriers (Figure 16). Nevertheless, this did introduced that are resistant to the roya del not seem widespread even across households café disease. of the same community where it was imple- mented. 58 Figure 16: A Plant Nursery and Cocoa Crop in La Berenjena, Dominican Republic Source: World Bank staff fieldwork Other adaptations strategies and policies, such as the use of production technologies or technical support, were also mentioned, although with less success. Production technologies, such as the use of herbicides and chemical fertilizers, are reported as allowing farmers to increase their pro- ductivity. Likewise, interviews mentioned adaptation policies provided by the government. For example, they mentioned technical support (“extensionismo agrícola”) and the use of machinery. Even though there are problems with technical support, the workshops and training sessions with farmers allow them to implement strategies that improve crop productivity. This support is provided not only through government institutions such as the Ministry of Agriculture, but also through non-governmental organizations that provide occasional aid to farmers. Additionally, the Ministry occasionally supports farmers with their tractor to help prepare the land. While these strategies and policies were mentioned occasionally, it was clear that they needed improvement to effectively support farmers. “The use of chemical products burning are very harmful. But that is what is used in practice. It’s always about rais- ing awareness so they don’t use it. There are agricultural plants, which, for example, they use to try to take care of the soil, for example, live barriers... That is one way. (...) For example, organic fertilizer. There are many practices that producers are always taught to take care of the soil and try to eliminate those (unsustainable) practices that they use.” (Key informant, male, employed by the local branch of the Ministry of Agriculture). 59 Interestingly, factor that was mentioned by a few emigrants and key informants is the desire of some to return to their communities of origin if labor conditions improved. Mentioned previously, the im- provement and economic potential of the forestry and agroforestry sectors prompted some people to return to Restauración. Other key informants also mentioned some people that returned once they realized the difficulties of the urban life, including security issues, difficulties finding jobs, and the longing for a calmer lifestyle. Others also returned after finishing studies if they had a profes- sion suitable for the area. Return migration was also seen as a viable possibility by some inter- viewed emigrants due to changes in their communities, such as electric power service, construction of aqueducts, improvement of neighborhood roads, and new jobs such as reforestation brigades. Nevertheless, some key informants mentioned cases of people returning after a negative experi- ence in the cities and having difficulties being reintegrated as they had sold their lands and had no place to practice agriculture again. Hence, keeping their lands after migrating (or having their family still living in the area) and having a viable economic alternative seems to be a pre-condition for such return. “… If one has something to survive on, one lives calmly in the field. I even have a little house in the country. Country life is much calmer than in town, because in town one goes to work afraid of being mugged or someone doing something to you on the way to work. In the country, one can leave the house open without fear that thieves will steal the little things one has there.” (Interviewee No. 16, 25, female, migrant to Santiago). 60 conclusions and recomendations for a climate and migration policy 61 Results of this study show how climate change and natural degradation may cause increased cross-border and internal migration in the Dominican Republic, even if indirectly. Without signifi- cant adaptation policies, significant flows will continue into the near future. Agriculture emigrants from Restauración perceived that changes in precipitation patterns and land degradation con- tributed to their low agricultural productivity, making it difficult to sustain a life in these rural communities, and ultimately influencing their decision to migrate to urban areas. Estimates of the quantitative model in fact show that the slow onset of events, such as sea-level rise, reduc- tion of water availability, and crop and ecosystem productivity, will cause approximately between 170,000 to 368,000 additional internal climate migrants in the period 2020-2050 and between 234,000 and 473,000 during the latter half of the century (2050-2100). Similarly, by 2050, between six and 20.1 percent of migrants from Haiti may arrive to the DR driven by climate change, with the percentage and total number increasing in the second half of the century. This study there- fore indicates that natural degradation and climate change will continue to cause migration within and to the DR. Internal climate migration will come from Climate migration should not be conceptu- several locations, but will primarily arrive to alized as an exclusive category distinct from a few spots, contributing to the existing ru- other types of migration. Even if people report ral-to-urban flow, changes in the labor struc- that the main reason to migrate is “to find a ture, and demographic pressures. Results show better job,” understanding their rationale for that internal out-migration/emigration will not being able to make a dignified living in come from many locations, such as coastal their place of origin is nevertheless important. cities, irrigated and rainfed cropland, and This study documents how climate change pastureland. In-migration will continue to ar- and natural degradation influence this pat- rive to many cities, especially the urban strip tern; for example, through lower productivity of Santo Domingo Province. This will likely in agriculture—one of the main economic contribute to people—especially the young— activities in the Dominican rural countryside. leaving rural communities and agriculture Even though the modeling approach has sector jobs to seek better opportunities in the been able to distinguish climate migrants cities. Cross-border climate migration from from other types of migrants, climate mi- Haiti also will follow similar patterns, arriving grants may nevertheless report the economic in larger numbers; this also will continue to reasons as the most pressing, given that the supply labor in the agriculture sector, espe- impacts of climate change indirectly affect cially in the border regions from where young their livelihoods and provide them with a Dominicans are leaving. Understanding these reason to migrate. The qualitative and quan- trends will help inform policymaking to max- titative modeling approach in this study has imize the positive development impacts of provided an avenue to explore the subjacent migration while, at the same time, to minimize motives and a comprehensive view on the the costs. issues. 62 The findings of this study call for more inte- Moreover, addressing climate migration gration between migration and climate policy should be further integrated in far-sighted within an adaptation framework, with a view green, resilient, and inclusive development to improving the effectiveness of both types planning, following a territorial approach. of policies. As it currently stands, neither the Following the framework of the World Bank’s NDCs nor the National Adaptation Plan of Country Climate and Development Reports the DR treats migration as an essential policy (CCDRs), the best climate policy is one that is, area. A similar case is true for climate change at the same time, good development policy. and natural degradation as a subpolicy area Climate migration should not be addressed within the migration policy, other than forced independently; instead, it should be further displacement caused by natural hazards, integrated with other areas of development such as hurricanes, floods, or earthquakes. policy. Some sectors that seem especially As discussed in this report, the inclusion of important are social inclusion, urban planning the drivers and impacts of migration caused and resilience, water management, and ag- by climate change and natural degradation riculture and forestry, although these should within an adaptation framework would be an not be considered an exhaustive list. Likewise, important step toward effective climate and given the different climate migration paths migration policy. and drivers in each region, a territorial ap- proach is needed to design policy solutions suited for each region, in addition to national policies. Policy recommendations can be framed along three main pillars. These will fall in line with the key findings of the 2023 World Development Report (World Bank 2023) and the policy response framework outlined in the 2022 KNOMAD/World Bank migration and development brief (KNO- MAD 2022). These frameworks and the results of this study highlight three pillars of policy recom- mendations (summarized in Table 4): (i) Coherent legal and institutional framework; (ii) Support- ing migration as a valid adaptation strategy to ensure a planned, safe, and dignified process; and (iii) Climate adaptation and mitigation that focus on preparedness and anticipatory action. 63 (i) Policy coherence on the legal and institutional framework Policies must be supported by a coherent legal and institutional framework. National policies and strategies relating to climate-driven migration must be integrated into climate policy and process- es, particularly in the context of adaptation. Support is necessary ensuring lower turnaround rates and the professionalization of migratory civil servants, with a dedicated module or seminar on cli- mate-induced migration as part of the curriculum of the Master’s degree on Migration and Develop- ment in the Caribbean at the DR’s National Migration Institute (Instituto Nacional de Migración). At the same time, issues relating to climate change must be integrated into migration policies and pro- cesses in a noticeably clear manner, including the promotion of continuous policymaking dialogue among the two disciplines—through the National Climate Change Council, the National Migration Council , Ministry of Agriculture, Ministry of Labor, Ministry of Environment, Ministry of Interior, and the Ministry of Economy, Planning, and Development, —and strengthening the necessary capacities and awareness among civil servants. Likewise, the institutional framework must consider the pro- duction of more research, analytical work, and geographically sensitive data generation and collec- tion, to further understand these issues and ways to address it. (ii) Supporting migration as a valid adaptation strategy to ensure a planned, safe, and dignified process Migration policy in the Dominican Republic ing their potential and planned relocation. can transform into effective climate policy when accepted as a valid adaptation strate- There are different alternatives to support this gy. Migration is often conceptualized as an climate migration process. These may include adaptation strategy to deal with degraded the identification of areas for voluntary re- resources, climate change, or other harmful location based in early warning systems of situations in communities of origin. Recogniz- unviable situations, developing action plans ing migration as a valid adaptation option in codesigned with local communities, and mo- climate policy could support people in dealing bilizing technical and financial support. This with the effects of climate change (especial- could be the case, for example, of coastal ly those trapped because of insufficient re- cities and towns affected by sea level rise, ap- sources to even migrate), as well as support a plying lessons learned from the DR’s Sierra de planned, safe, and dignified movement that Bahoruco, Sierra de Neiba, and Lake Enriquillo will increase positive impacts while reducing 22 experience. Other ways that appeared rele- the negative. Planned relocation includes vant in the context of the study are, for exam- measures to minimize the human costs arising ple, reskilling programs for people seeking jobs from climate change (Ferris and Weerasinghe in the cities and incentives for educational op- 2020; Bergmann 2021). In this sense, greater portunities to be tailored for those wishing to attention is needed to ensure that affected return with occupations that match the needs persons are involved fully in decisions regard- of their communities. 22 The Government of the DR has relocated many affected by a rise in water levels in Lake Enriquillo to the new town of Boca de Cachón, a US$24 million community built by the government to house people on the verge of losing their homes to the lake. 64 Migration policy can also support designing solutions to safe and orderly cross-border climate-in- duced migration, in a manner that heightens the positive social and environmental impacts. Mi- gration policies that recognize the current importance of Haitians workers for the labor market of DR, particularly in agriculture, and their vulnerability to social and climate drivers, could help create programs that incentivize them to utilize sustainable practices that take care of the envi- ronment, while helping supply the labor in agriculture that Dominicans appear to be exiting. (iii) Climate adaptation and mitigation focused on preparedness and anticipatory Recognizing that climate change and natural Improving waste management planning and degradation may cause large flows of human practices is crucial, particularly in the water mobility, and that climate and environmental and sanitation sectors and in precarious settle- policies may influence the size of these flows, ments. Inadequate waste disposal and man- becomes relevant in considering a more holis- agement not only exacerbate environmental tic adaptation framework. Climate adaptation pollution but also pose severe health risks, and mitigation policies, such as those imple- contaminating water sources, and hindering mented through the agriculture, water, and access to clean water and sanitation facili- forestry sectors, could have significant influ- ties. It also directly affects maritime, freshwa- ence on the need for migration. For example, ter, estuaries, and groundwater ecosystems, helping farmers adapt to the effects of cli- which are vital to the DR´s economy and mate change, supporting improvement in pro- environmental conservation. Improved waste ductivity, and reducing the drivers of natural management is especially critical in under- degradation and water stress may have sig- privileged communities, where the lack of nificant effects on the number of people that infrastructure and resources can lead to dire will need to migrate, as well as the potential consequences on living standards and public for return migration, for those who were pre- health outcomes. By focusing on enhancing viously displaced and their origin community waste management strategies, including the becomes viable again. Supporting people development of sustainable disposal methods to overcome barriers to enter agroforestry, and systems, the DR can mitigate these risks, promoting green jobs in the forestry sector; 23 protect livelihoods and ecosystems. This is not providing technical and technological support just an environmental imperative but a fun- to increase agriculture and land productivity, damental aspect of advancing public health, as well as irrigation and water management, economic growth, and social sustainability. were all discussed in the qualitative study as policies that will improve people’s livelihoods (albeit with mixed success) in their communi- ties of origin—all of which influence migra- tion. These could leverage existing programs at the country level, such as the ERP in the DR. 23 While not mentioned in the interviews, other sectors also are able to produce green jobs. The ERP includes projects to generate green jobs even in the livestock and agriculture sectors. 65 Adaptation policies in the form of preparedness migration and adaptation, based in localized and anticipatory actions are not only need- results and knowledge, and avoid imposing a ed in origin rural communities, but also in the common solution and plan nationally. adaptation of cities that are projected to be both in and out-migration hotspots. Results The Law No. 368-22 on Territorial Develop- from this study highlight the large influx of ment, Land Use, and Human Settlements, climate migrants that will arrive to cities, as establishes guidelines for zoning, land use, well as the underlying motivations to migrate environmental protection, and urban devel- to these locations, mainly the search for good opment, aiming to balance development jobs, services and education. Urban planning needs with environmental conservation and will need to take into account estimates like the welfare of the population. It emphasizes those produced by this study to prepare resil- the importance of participatory planning, ient services and labor markets. Moreover, the involving communities and stakeholders in results also highlight large numbers of ur- decision-making processes to ensure that ban-to-urban migration driven by water stress development projects meet the needs and caused by growing populations and sea level aspirations of the local population. As such, rise. In this sense, resilient infrastructure and national planning instruments at the Minis- water management systems will be required try of Economy, Planning, and Development to support the growing population pressure (MEPYD), such as the regional development and climate stressors. Further quantitative and strategy for border areas “Mi Frontera RD” al- qualitative research is needed to validate these ready integrate climate and migration policy projections and recommendations with local and can serve as reference for other planning communities, leaders, and experts and more instruments in the country. localized data sources. An integrated territorial development lens can help integrate adaptation policies across rural, border, and urban areas. The highly regionalized climatic and economic situa- tions of origin and destination communities, and therefore of the drivers and journeys of climate migrants, evidence the need to con- sider local and territorial approaches. Results from this study show that drivers and jour- neys of migrants are deeply related to their place of origin, which influence their social networks across the country and the viable intermediate and final destinations they have as options. This evidence the importance of geographically and regionally contextualizing climate migration and the policies to address it. In this sense, it is important that national policies facilitate regional plans for climate 66 Table 4. Summary of Policy Recommendations to Address Climate Migration in the Dominican Republic PILLAR POLICY RECOMMENDATIONS Support technical roundtables among the National Climate (i) Change Council, the National Migration Council, Ministry of Agriculture, Ministry of Labor, Ministry of Environment, Ministry of Interior, and Ministry of Economy, Planning, and Development, Coherent legal to better integrate environmental, agricultural, and migratory and institutional policies in the country. framework Provide technical advice and facilitate knowledge exchanges with small-island states in Latin America and in other regions, including through regional platforms. For instance, through the Greater Caribbean Climate Mobility Initiative (GCCMI) and simi- lar regional platforms. Create clear and established legal frameworks to support cli- mate-induced migration pathways at the national level. Mobilize international technical, institutional, and financial sup- port to provide regional responses to facilitate safe and planned cross-border climate migration. Support the professionalization and reduce the turnover of migratory civil servants with a dedicated module or seminar on climate-induced migration as part of the curriculum of the National Migration Institute´s Master’s degree in Migration and Development in the Caribbean. Promote further research, analytical work, and data generation and collection on these issues. (ii) Develop early warning systems to enable prior planning for slow and rapid human mobility flows in response to climate change Supporting impacts and natural degradation. migration as a valid adaptation Boost national and international resource mobilization to fi- strategy to nance subsidies and enable safe and planned migratory path- ensure a planned, ways for internal and cross-border climate migrants. safe, and dignified process 67 PILLAR POLICY RECOMMENDATIONS (ii) Planned relocation may imply the voluntary movement of peo- ple to other cities or areas within a country. Greater attention is needed to ensure that affected persons are involved fully Supporting in decisions regarding their relocation. Codesign and imple- migration as a ment climate-induced relocation plans with local communities. valid adaptation Lessons learned from the DR’s Sierra de Bahoruco, Sierra de strategy to Neiba, and Lake Enriquillo experience can be a starting point1 . ensure a planned, safe, Support climate-informed territorial development, planning in- and dignified struments and socio-environmentally sustainable investments process that address rural-urban corridors and the unique challenges or border areas. Prepare and update local development and climate action plans to allow the mobility of people, identify their destina- tion, and ensure appropriate preparations and key services for arrival at destination (e.g., health, education, and housing). For instance, support the continuity and adequacy of educational process for internal migrants. (iii) Periodically update territorial planning instruments at the Ministry of Economy, Planning, and Development (Ministerio de Economía, Planificación y Desarrollo, MEPYD) to take into Climate account predicted climate migration flows. adaptation and mitigation Invest in the adaptation of rural and border communities and focused on other out-migration hotspots, to enable potential migrants to preparedness stay in place, where viable, are also important. National plan- and anticipatory ning instruments at MEPYD, such as the regional development action strategy for border areas “Mi Frontera RD” already integrate climate and migration policy and can serve as reference for other planning instruments in the country. Prioritize strategies aimed at supporting livelihoods and ad- dressing the risk of displacement caused by climate change in areas adjacent to protected zones. 68 PILLAR POLICY RECOMMENDATIONS (iii) The Emissions Reduction Program (ERP) in the DR may consti- tute an avenue for future preparedness in green and sustainable sectors. The program may provide technical and financial sup- port in favor of national efforts. This program could be leveraged Climate to reverse unsustainable practices in the forestry and agriculture adaptation and sectors that are factors that influence the decision to migrate. mitigation focused on Address the barriers to enter the agroforestry field through in- preparedness centivization and subsidy programs, and promote green jobs in and anticipatory the forestry sector. action Provide technical and technological support to increase agricul- ture and land productivity, irrigation, as well as water manage- ment to address factors that influence the decision to migrate. Enhance waste management to reduce environmental pollu- tion and health risks, especially in water and sanitation sectors, ensuring equitable access to clean resources, and supporting public health and environmental preservation. Minimize the use of harmful chemicals by large agricultural pro- ducers to reduce the speed of soil and water degradation and its subsequent impacts on climate-induced migration from rural areas. Improve urban-rural corridors and increase awareness of the im- portance of sustainable agricultural practices among rural and urban youth. Strengthen technical assistance for preparedness as a critical measure in cities expected to receive an influx of migrants over the next few decades. Invest in infrastructure and employment in cities (e.g., Santo Domingo, Santiago, Higüey), as well as in sus- tainable areas. Likewise, given water stress and sea-level rise pressures on cities, climate resilient urban planning is needed to ensure cities are able to accommodate growing numbers of people under condi- tions of climate stress, including resilient infrastructure, housing, and services, with social inclusion and cohesion considerations. 1 The Government of the DR has relocated many affected by a rise in water levels in Lake Enriquillo to the new town of Boca de Cachón, a US$24 million community built by the government to house people on the verge of losing their homes to the lake. 69 Reference List Ariza and Lozano. 1993. Tendencias de la migración interna en República Dominicana: 1981-1991: movilidad territorial, patrones de desarrollo y mercados de trabajo. Santo Domingo. Bergmann. 2021. Bergmann, J., Vinke, K., Fernández Palomino, C.A., Gornott, C., Gleixner, S., Laudien, R., Lobanova, A., Ludescher, J., Schellnhuber, H.J., 2021. Assessing the Evidence: Climate Change and Migration in Peru. Bertaud A. and S. Malpezzi. 2003. The Spatial Distribution of Population in 48 World Cities: Implications for Economies in Transition. Working Paper. Cissé et al. 2022. Health, Wellbeing, and the Changing Structure of Communities. In: Climate Change 2022: Impacts, Adaptation and Vulnerability. Contribution of Working Group II to the Sixth Assessment Report of the Intergovernmental Panel on Climate Change [H.-O. Pörtner et al. (eds.)].. Cambridge University Press, Cambridge, UK and New York, NY, USA, pp. 1041–1170, doi:10.1017/9781009325844.009. Clark, C. 1951. Urban population densities. Journal of the Royal Statistical Society 114(4): 490–496. doi:10.2307/2981088. Cordero Ulate, A, and G., Lathrop. 2016. Relations between Environment and Migration in the Dominican Republic: Evidence, Interpretations and Policies from a Case Study. Geneva, Switzerland: International Organization for Migration (OIM) Danielson, J.J., and Gesch, D.B. 2011, Global multi-resolution terrain elevation data 2010 (GMTED2010): U.S. Geological Survey Open-File Report 2011–1073, 26 p. DeLorme Publishing Company. Environmental Systems Research Institute (Redlands, Calif.). 2014. World Water Bodies, 2013. [Shapefile]. Environmental Systems Research Institute (Redlands, Calif.). Retrieved from https://earthworks.stanford.edu/catalog/stanford-jb896tp9580 Dodman et al. 2022. Cities, Settlements and Key Infrastructure. In: Climate Change 2022: Impacts, Adaptation and Vulnerability. Contribution of Working GroupII to the Sixth Assessment Report of the Intergovernmental Panel on Climate Change [H.-O. Pörtner et al. (eds.)].. Cambridge University Press, Cambridge, UK and New York, NY, USA, pp. 907–1040, doi:10.1017/9781009325844.008. Eckstein, D., V. Künzel, and L. Schäfer. eds. 2018. Global Climate Risk Index 2018: Who Suffers Most From Extreme Weather Events? Weather-related Loss Events in 2016 and 1997 to 2016. Briefing Paper. Germanwatch. www.germanwatch.org/sites/default/files/publication/20432.pdf. 70 Ferris, E., & Weerasinghe, S. (2020). Promoting Human Security: Planned Relocation as a Protection Tool in a Time of Climate Change. Journal on Migration and Human Security, 8(2), 134–149. https://doi.org/10.1177/2331502420909305 Fox-Kemper et al. 2021. Ocean, Cryosphere and Sea Level Change. In: Climate Change 2021: The Physical Science Basis. Contribution of Working Group I to the Sixth Assessment Report of the Intergovernmental Panel on Climate Change [MassonDelmotte, et al.(eds.)]. Cambridge University Press. In Press. FUNGLODE. n.d. Dajabón Province, Infrastructure for sustainable development. Dajabon Provincial Vision 2044. http://rd2044.org/Provinciashtml/dajabon/ Germanwatch. 2021. Slow-Onset Processes and Resulting Loss and Damage: An Introduction. GermanWatch Publication Series. www.germanwatch.org/en/19796. GFDRR (Global Facility for Disaster Reduction and Recovery). 2022. “Thinkhazard: Dominican Republic.” Washington, DC: GFDRR. https://thinkhazard.org/en. Global Forest Watch. 2023. “Dominican Republic.” www.globalforestwatch.org/dashboards/country/DOM (accessed February 2023). GoDR (Government of the Dominican Republic). 2010. “IX Censo Nacional de Población y Vivienda 2010.” Santo Domingo: GoDR. GoDR. 2018. “Segunda Encuesta Nacional de Immigración 2017”. Santo Domingo: National Statistics Office, Government of the Dominican Republic.] GoDR. 2018. Segunda Encuesta Nacional de Inmigrantes en la Republica Dominicana: ENI-2017 Informe General. Santo Domingo: National Statistics Office, Government of the Dominican Republic. GoDR. 2019. Marco de Gestión Ambiental y Social (MGAS) REDD+. Santo Domingo, República Dominicana: Ministry of Environment and Natural Resources, Government of Dominican Republic, World Bank, Forest Carbon Partnership Facility (FCPF). GoDR. 2020. Nationally Determined Commitments (NDC). Santo Domingo: Government of the Dominican Republic. Hansen et al. 2013. "High-Resolution Global Maps of 21st-Century Forest Cover Change." Science Vol. 342(6160): 850-853. www.science.org/doi/10.1126/science.1244693. Houck, R. 2005. A Study of Ridgeline and Steep Slope Regulations in Mountain Communities throughout the United States (Asheville, North Carolina: Land of Sky Regional Council) 71 IFRC (The International Federation of Red Cross and Red Crescent Societies). 2021. Dominican Republic IFRC Country Office Plan. Santo Domingo: IFRC https://www.ifrc.org/sites/default/files/2021- 08/Do minican_Republic_Plan_2021.pdf IOM (International Organization for Migration). 2019. International Migration Law: Glossary on Migration. Geneva: UN IOM. IOM. 2021. “Displacement Tracking Matrix: Haiti.” Le Grand-Saconnex, Switzerland: UN International Organization for Migration. IPCC (Intergovernmental Panel on Climate Change). 2014. Climate Change 2014. Fifth Assessment Report. Geneva: IPCC. www.ipcc.ch/report/ar5/syr. IPCC. 2022. Climate Change 2022: Impacts, Adaptation and Vulnerability. Contribution of Working Group II to the Sixth Assessment Report of the Intergovernmental Panel on Climate Change [H.-O. Pörtner et al. (eds.)]. Cambridge University Press. Cambridge University Press, Cambridge, UK and New York, NY, USA, www.ipcc.ch/report/ar6/wg2. IUCN (International Union for Conservation of Nature). 2015. World Database on Protected Areas. Jiménez, M. 1992. Migraciones internas, 1986-1991, República Dominicana. Santiago, Chile: CEPAL/CELADE Jones. B. 2020. Modeling Climate Change-Induced Migration in Central America & Mexico. Methodological Report. Retrieved from the ProPublica website https://assetsc3.propublica.org/Cli mate-Migration-Modeling-Methodology.pdf Jones, B. and O’Neill, B.C., 2013. “Historically grounded spatial population projections for the continental United States.” Environmental Research Letters, 8(4), p.044021. Jones, B. and O’Neill, B.C., 2016. “Spatially explicit global population scenarios consistent with the Shared Socioeconomic Pathways.” Environmental Research Letters, 11(8), p.084003. Lutz, J., and Yayboke, E. 2021, November 22. “Haitian Migration: Food Insecurity, Fragility, and a Better Way Forward”. CSIS (Center for Strategic and International Studies). Published in: https://www.csis.org/analysis/haitian-migration-food-insecurity-fragility-and-better-way-forward Martinich et al. 2018: Reducing Risks Through Emissions Mitigation. In Impacts, Risks, and Adaptation in the United States: Fourth National Climate Assessment, Volume II [Reidmiller et al. (eds.)]. U.S. Global Change Research Program, Washington, DC, USA, pp. 1346–1386. doi: 10.7930/NCA4.2018.CH29 72 Notre Dame Global Adaptation Initiative. 2023. “The ND-GAIN Country Matrix: Dominican Republic.” Indiana, United States: University of Notre Dame. https://gain-new.crc.nd.edu/country/dominican-re public. Oleson et al. “Avoided climate impacts of urban and rural heat and cold waves over the US using large climate model ensembles for RCP8. 5 and RCP4. 5.” Climatic change, 146(3-4), pp.377-392. O’Neil et al. 2014. “A New Scenario Framework for Climate Change Research: The Concept of Shared Socioeconomic Pathways.” Climatic Change, 122: 387–400. https://link.springer.com/article/10.1007/s10584-013-0905-2. Ovalles U., P. (2011). Identificación de las causas de la deforestación y la degradación de los bosques en la República Dominicana: Informe final. Santo Domingo: Programa REDDCCAD/GIZ en Centroamérica y República Dominicana. R4V. 2022. RMRP 2022. Regional Refugee and Migrant Response Plan (RMRP). Inter-Agency Coodination Platform for Refugees and Migrants from Venezuela. Le Grand-Saconnex and Geneva, Switzerland: UN International Organization for Migration and United Nations High Commissioner for Refugees. Ratha, D. et al. eds. 2022. Migration and Development Brief 37: Remittances Brave Global Headwinds. Special Focus: Climate Migration. Washington, DC: KNOMAD-World Bank. Rich, D. C. 1980. Potential models in human geography. Concepts and techniques in modern geography 26. Norwich: University of East Anglia. Rigaud et al., eds. 2018. Groundswell: Preparing for Internal Climate Migration. Washington, DC: World Bank. Rigaud et al., eds. 2021. Groundswell Africa: Internal Climate Migration in West African Countries. Washington, DC: World Bank. https://openknowledge.worldbank.org/handle/10986/36404. Restauración Municipal Government. 2020. Plan de Desarrollo Municipal, Restauración, 2020-2024. Accessed through SISMAP (Sistema de Monitoreo de la Administración Pública) Skovdal, M. and F. Cornish. 2015. Qualitative Research for Development: A Guide for Practitioners. Rugby, UK: Practical Action Publishing. doi:10.3362/9781780448534. Stevens et al. 2015. Disaggregating Census Data for Population Mapping Using Random Forests with Remotely-Sensed and Ancillary Data. PLoS ONE 10(2): e0107042. https://doi.org/10.1371/journal. pone.0107042 Theilacker, J. and Anderson, K. 2010. Steep Slope Ordinance: A Guide. Pennsylvania Land Trust Association UNDESA (United Nations Department of Economic and Social Affairs). 2022. 2022 Revision of World Population Prospects. https://population.un.org/wpp/ 73 UNEP (United Nations Environment Programme). 2013. Haiti – Dominican Republic: Environmental challenges in the border zone. Nairobi, Kenya: UNEP. van Vuuren et al. 2014. “A new scenario framework for climate change research: scenario matrix architecture”. Climatic Change, 122(3), pp.373-386. Warntz, W., & Wolff, P. (1971). Breakthroughs in geography. New York: Plume Warszawski et al. 2014. “The Inter-Sectoral Impact Model Intercomparison Project (ISI–MIP): Project framework.” Proceedings of the National Academy of Sciences (PNAS) 111 (9), pp. 3228-3232. World Health Organization (WHO) and the United Nations Framework Convention on Climate Change (UNFCCC). 2021. Dominican Republic. Health & climate change country profile 2021. Small Island Developing States Initiative. Wooding B. and M. A. Morales. 2014a. Migración y medio ambiente. Una reflexión pertinente. Migration Policy Brief. Santo Domingo. Written for Observatorio Migrantes del Caribe y CIES-UNIBE. http://obmica.org/images/Publicaciones/MigrationPolicyBrief/Migration_Policy_Brief%20espaol web.pdf Wooding B. and M. A. Morales. eds. 2014b. Migración y sostenibilidad ambiental en Hispaniola. Migration Policy Brief. Santo Domingo: Observatorio Migrantes del Caribe y CIES-UNIBE. http://obmica.org/index.php/publicaciones/libros/94-migracion-y-sostenibilidad-ambiental-en-his paniola. Wooding B. and A. Morales. (2016) Diagnosis of information for public policies: Migration, environment, and climate change in the Dominican Republic. Geneva, Switzerland: International Organization for Migration. Available at https://publications.iom.int/system/files/pdf/assessing_the_evidence_dominican_republic_sp.pdf World Bank. 2021. Project Appraisal Document - Dominican Republic Emission Reductions Program. World Bank. 2022. Remittances Brave Global Headwinds—Special Focus: Climate Migration. Migration and Development Brief 37. November. Washington, DC: World Bank, in partnership with the Secretariat of The Global Knowledge Partnership on Migration and Development. www.knomad.org/publication/migration-and-development-brief-37. World Bank. 2023. World Development Report 2023: Migrants, Refugees, and Societies. Concept Note. Washington, DC: World Bank. https://www.worldbank.org/en/publication/wdr2023. 74 appendix 75 Appendix 1. Map of Regions and Provinces of the Dominican Republic Source: World Bank staff based on data from the 2010 National Population and Housing Census of the Dominican Republic (GoDR 2010). 76 Appendix 2: Internal Migration between Municipalities of the Dominican Republic Source: World Bank staff based on data from the 2010 National Population and Housing Census of the Dominican Republic (GoDR 2010). Notes: Tick marks represent 12,000 people. The graph was created based on a question in the 2010 Census regarding the municipality in which a person (5 years and older) was living in 2005. The number represents those who, in 2005, were living in another municipality of the Dominican Republic. 77 Appendix 3: Methodological Details of Modeling Exercise Data The data used in the modeling was drawn from the original Groundswell (Rigaud et al. 2018) proj- ect, updated to reflect more recent revisions, where possible, and modified slightly for this proj- ect. Table A3.1 includes a list of the data employed in the modeling analysis. Table A3.1. Datasets, Variables, and Sources Variable Source Resolution Time Series Time Step Indicator Climate Driven Water Availability ISIMIP 0.5° 1970-2100 5-year Deviation from baseline Yes Agriculture/Crop Yields ISIMIP 0.5° 1970-2100 5-year Deviation from baseline Yes Biomes/Ecosystem ISIMIP 5-year Deviation from baseline Yes 0.5° 1970-2100 Productivity Sea Level Rise Changes in coastline Yes NASA 30m 1990-2150 5-year (inundation) Bilateral Internal N/A ONE Municipal 2005-2010 5-year Count of migrants Migration Count of migrants Bilateral International ONE 5-year (Haiti to Dominican N/A Municipal 1990-2010 Migration Republic municipalities) Municipal Population ONE Municipal 2010 N/A Count of people N/A Spatial Population 1-year Population counts WorldPop 100m 2000-2020 No (Totals, Age, and Sex) by age and sex Elevation SEDAC 30m 2015 N/A Corrected elevation No Slope SEDAC 30m 2015 N/A Avarage slope No Water Bodies ESRI Vector 2019 N/A Surface water No World Database on IUCN Vector 2019 N/A Mandate for protection No Protected Areas Source: World Bank staff The impacts of climate change takes place constitute the spatial mask used to restrict the through the first four variables on the list: land deemed suitable for human habitation. water availability/stress, crop yields, ecosystem The population variable (WorldPop) serves as productivity, and sea level rise (SLR). Bilateral the base year population distribution (total internal and international migration data, the population count by grid cell), as well as the latter of which consist only of flows from Haiti two variables in the migration model (age and to the Dominican Republic (DR), are from the sex structure). Each of these variables was DR’s National Statistics Office (Oficina Nacion- tested and validated, and previous applica- al de Estadística) (as are the municipal popula- tions of the model found them to be statisti- tion data). These data were used to help train cally and/or practically significant in driving the model, and are unique to this application positive or negative impacts on the relative of the INCLUDE model. The final four vari- attractiveness of various locations. ables are not drivers of migration; instead, they 78 Inter-Sectoral Model Intercomparison Project This project uses outputs from the ISIMIP modeling effort for crop production, water availability, and ecosystem impacts, which The Inter-Sectoral Impact Model Intercom- cover the historical period 1970–2010, and parison Project (ISIMIP) is an ongoing com- projections for 2010–2100. The future sec- munity-driven modeling effort organized by toral impact models are driven by a range the Potsdam Institute for Climate Impact of general circulation models. Here, the data Research, and is designed to provide a frame- is driven by two general circulation models work to compare multiscale, cross-sectoral that provide a good spread for the tempera- climate impact projections (Warszawski et al. ture and precipitation parameters of interest: 2014). Based on the Representative Concen- the HadGEM2-ES climate model developed tration Pathways (RCP) and Shared Socio- by the Met Office Hadley Centre for Climate economic Pathways (SSP), ISIMIP facilitates Science and Services (United Kingdom) and a quantitative assessment of impacts across the IPSL-CM5A-LR climate model developed multiple sectors and models, based on com- by the Institut Pierre Simon Laplace (France) mon climate and socioeconomic background climate modeling center. The crop, water, and scenarios and climate model inputs. A ma- ecosystem simulations—at a relatively coarse jor goal of the project is the development of spatial scale (0.5°)—represent indicators that policy-relevant metrics. Over the course of capture the impact climate may have on spe- this century, policymakers will be tasked with cific types of livelihoods, the viability of which assessing the costs associated with mitiga- will figure into the migration decision (e.g., tion efforts against those for adapting to a climate acting through other mechanisms). warmer world. In that spirit, the project is These climate impacts were selected because motivated by and organized around a central the literature shows that water scarcity, de- question: how do impacts vary between 2°C, clining crop yield, and the decline in pastur- 3°C, and 4°C of global warming? Research is age are among the major potential climate designed to isolate “tipping points,” the lev- impacts facing lower- and middle-income el of environmental change associated with countries, and these impacts also will be sig- rapid increase in negative sectoral impacts. nificantly important drivers of migration. Water and Crop Models The primary ISIMIP drivers used in this analysis are water availability and crop yields. Output from the water sector model is representative of river discharge, measured in cubic meters per second in daily/monthly time increments, and is influenced by rainfall and changing temperature. Crop sector model outputs estimate the annual crop yield of four staple crops (maize, wheat, rice, and soybeans) in tons per hectare at a 0.5° x 0.5° grid cell resolution, and they are a function of rainfall, temperature, CO2 concentrations, irrigation, and other management practices. Because the impact of climate change on local conditions (i.e., the deviation from historic local norms) is more indicative of potentially disruptive change than absolute yields, the approach adopted here is from the Groundswell report (Rigaud et al. 2018) in which the data are transformed to reflect periodic deviation from the 40-year historical baseline. 79 The data are converted to five-year average water availability and crop production (in tons) per grid cell, and an index is then calculated that compares those values with the 40-year average for water availability and crop production for 1970–2010: Eq. 1 where Davg is the five-year average crop production/water availability and Bavg is the baseline average crop production/water availability for the 40-year period 1970–2010. The indexes for wa- ter availability and crop production represent deviations from the long-term averages. To account for uncertainty in future climate of the two water models and two crop mod- outcomes, also adopted is the Groundswell els: the LPJmL water and crop models, the approach to select ISIMIP crop and water WaterGAP2 water model, and the GEPIC model outputs based on different combina- crop model (Table A3.2). The crop and wa- tions of climate, crop, and water models. Ap- ter models were selected by experts at the plying the combinations—two global climate Potsdam Institute for Climate Impact Re- models driven by two different emission sce- search, based on several criteria, including narios which, in turn, drive two sets of sectoral model performance over the historical peri- impact models (described below)—provides a od, diversity of model structure, diversity of range of plausible population projections while signals of future change, and availability of also indicating regions where the models tend not only observationally driven historical but toward agreement (Rigaud et al. 2018). The also global climate model-driven historical modeling for this assessment employed two and future simulations. Table A3.2 presents General Circulation Model (GCMs) – global the combinations of models used. “climate models” -- which drive combinations Table A3.2. Matrix of Global Climate Models and Crop/Net Primary Productivity and Water Model Combinations Crop/NPP Simulation HadGEM2-ES, HadGEM2-ES, IPSL-CM5A-LR, IPSL-CM5A-LR, LPJmL (crop) GEPIC (crop) LPJmL (crop) GEPIC (crop) Water simulation LPJml (NPP) Visit (NPP) LPJml (NPP) Visit (NPP) HadGEM2-ES,LPJmL (water) Model 1 HadGEM2-ES, WaterGAP2 Model 2 IPSL-CM5A-LR, LPJmL (water) Model 3 IPSL-CM5A-LR, WaterGAP2 Model 4 Source: Adopted from Rigaud et al. 2018. Notes: NPP = Net primary productivity. 80 Ecosystem productivity is estimated in terms Ecosystem Productivity of NPP. The ecosystem models simulate the natural growth of several different plant func- tional types, including grasses; thus, NPP sim- In the same way that crop production is an ulated by these models serves as an estimate important metric of farm-based livelihoods, of the productivity of a location’s natural bi- ecosystem productivity is an important mea- ome, including grassland biomes that may sure for pastoral livelihoods. In this project, potentially support pastoral livelihoods. Like ecosystem productivity is used as a potential the water and crop metric, NPP is transformed driver in nonurban areas where the crop data to represent local periodic deviation from the indicate that agriculture is not taking place historical baseline. The NPP sectoral models (e.g., those places likely suitable for pasto- proposed in this work are the LPJmL and Visit ralism). Using ecosystem productivity only models—the former to be used with the LPJml in areas lacking crop productivity, data is crop production and water availability models, deemed preferable to including an overlay of while the latter is to be used with the GEPIC net primary productivity (NPP) on top of the crop and WaterGap water models—and the crop production, since there is high spatial models are also driven by the same GCMs as collinearity between the crop and ecosystem the water and crop models. metrics. Sea Level Rise The analysis also will consider SLR projections from the IPCC Sixth Assessment Report (AR6) (IPCC 2022), available through the Physical Oceanography Distributed Active Archive Center at the NASA Jet Propulsion Laboratory (Fox-Kemper et al. 2021). Data cover the period 2020-2150 and are provided for all future scenarios covered in the AR6. Projections for individual processes that cause sea level to change also are included in the dataset. Globally averaged projections, regional projections on a regular global grid, and local projections at individual tide gauge lo- cations are all provided. Following the Groundswell methodology, sea-level rise data are used to represent the loss of habitable land due to the SLR of each coastal grid cell. Internal/International Migration and Municipal Population Bilateral international migration flows between Haiti and the DR are derived from two sources; the DR’s 2010 National Population Census (referred to as 2010 Census going forward) and the 2nd National Immigration Survey (ENI 2017). These data can be accessed through the DR’s National Statistics Office (www.one.gob.do). Five-year immigration flows are reported at the municipal level over the period 2005-2010, and are disaggregated by age and sex. 81 Internal migration flows (at the municipal Subnational population totals at the munici- level) are from the 2010 Census. Five-year bi- pal level are from the 2010 Census, and were lateral flow data are compiled based on each used to cross-validate the gridded spatial household’s current place of residence and population data. They also provide a histor- place of residence five years prior. The data ical point of reference for projected future were provided by the World Bank team, but municipal level population. also can be accessed through the DR’s Na- tional Statistics Office. Spatial Population Age and Sex Structure The WorldPop organization (University of Southampton; Stevens et al. 2015) produces popu- lation estimates with age/sex breakdowns for each 100 meter x 100 meter grid square on the planet. The intended function of WorldPop data is to serve as default, open access datasets for United Nations agencies planning humanitarian and development interventions, and to help governments fill census gaps. As such, the mission of the WorldPop group overlaps with the goals of the predictive analytics project. Estimates of the spatially explicit distribution of the population are derived from census data and produced using a semi-automated dasymetric modeling approach that incorporates cen- sus and a wide range of open access ancillary datasets with a flexible, “random forest” estima- tion technique. A combination of widely available, remotely-sensed, and geospatial datasets (e.g., settlement locations, settlement extents, land cover, roads, building maps, health facility locations, satellite nightlights, vegetation, topography, refugee camps) contribute to the mod- eled dasymetric weights, and then the random forest model is used to generate a gridded prediction of population density at ~100m spatial resolution. This prediction layer is then used as the weighting surface to perform dasymetric redistribution of census counts at a country level. Assessment of the product indicates improvements in mapping accuracies and other population mapping approaches (Stevens et al. 2015), particularly the (i) Gridded Population of the World (GPW), a data center of NASA’s Earth Observing System Data and Information Sys- tem; and (ii) Global Human Settlement Layer Population Estimates (GHS-POP), supported by the European Commission, Joint Research Centre, and Directorate-General for Regional and Urban Policy, with the first (i.e., (i)) tending to over-allocate population across rural areas, and the second (i.e., (ii)) over-concentrating population in urban areas and, crucially, often missing settlements in West Africa that are difficult to detect via satellite imagery. For this work, WorldPop serves two purposes. First, the 2020 WorldPop data serve as the base year distribution upon which the model is applied at its native 1 kilometer (km) resolution. Sec- ond, the data are used to cross-reference the census data, specifically the municipal level popu- lation, to ensure that the spatial model represents the underlying census data correctly. 82 Elevation, Slope, Surface Water, and Finally, the International Union for the Con- Mandate for Protection servation of Nature (IUCN) World Database on Protected Areas is used to mask land as The spatial population model includes a a function of mandate for protection (IUCN geospatial mask that acts as a weight, pro- 2015). Specifically, any area classified under portionally scaling the population potential IUCN categories (Ia) (strict nature reserve), (Ib) (attractiveness) for each grid cell as a func- (wilderness area), (II) (national park), (III) (na- tion of the area within each cell deemed tional monument or feature), or (IV) (habitat/ suitable for human habitation. A mask is species management area) is masked as not constructed from four geospatial data layers: suitable for development/habitation. surface water, elevation, slope, and protect- ed land. These data are overlaid to produce In some cases, existing base year popula- a single mask from which is extracted the tion was found in cells otherwise completely portion of each cell suitable for habitation. masked as a function of mandate for pro- The ESRI World Water Bodies 2013 (DeLorme tection. There are two possible explanations. 2014) dataset is applied to mask global sur- First, the algorithm used to distribute the ex- face water. Elevation and slope data are from isting population across grid cells in the GPW the Global Multi-Resolution Terrain Elevation data base year does not specifically account Data 2010 (GMTED2010; Danielson and Gesch for protected land and, as such, population 2011). The elevation of the highest perma- and protected land may overlap in the base nently populated settlement in each conti- year. Second, in many cases, new mandates nent is applied as a ceiling to exclude land for protection grandfather existing popula- from future habitation as a function of high tions (e.g., people living in newly designated elevation. In general, development costs in- national forest land in the United States). For crease substantially on land exhibiting a slope modeling purposes, both cases are treated greater than 15 percent, which is also the identically. For example, cells that are 100 point at which many municipalities impose percent masked as a function of mandate for development regulations (e.g., Theilacker and protection are not eligible to receive any pro- Anderson 2010). Here, account is made for jected future population growth. These cells, the likelihood that improved technology will however, are allowed to lose people during reduce the costs associated with excess slope periods of population decline. This decision and, instead, impose a threshold of 25 per- reflects perception of this study of real-world cent, an oft-cited “no-development” threshold population change in areas with not only ex- in municipal regulations across the United isting population but also prohibitions on new States (Houck 2005). development. Scenario Framework This work adopts the extended scenario-based approach from the Groundswell Africa report (Rigaud et al. 2021), in which four plausible future internal climate migration scenario combina- tions are examined (Figure A3.1) based on combinations of the RCPs (van Vuuren et al. 2014) and SSPs (O’Neill et al. 2014). For each scenario, the projection represents an ensemble of model runs using combinations of crop, water, ecosystem, and flood impact models from the ISIMIP. The scenario approach has several advantages. First, exploring migratory outcomes across alterna- 83 tive physical and demographic/socio-economic futures allows researchers to begin to character- ize the size and sources of uncertainty associated with projections of climate-induced migration. Second, by varying climate and demographic/socio-economic pathways provides a means for considering and evaluating different policy options in terms of the impacts on spatial population outcomes as well as the potential avoided climate impacts of achieving more advantageous cli- mate and societal outcomes (e.g., Oleson et al. 2015; Martinich et al. 2018). Representative Concentration Pathways (RCPs) Developed in advance of the IPCC 5th As- Climate output, consistent with two RCPs (2.6 sessment Report, RCPs represent the latest and 8.5), are included in this work as drivers of generation of global scenarios for climate the climate impacts considered in the model- change research (IPCC 2014). The RCPs are ing. The two RCP scenarios considered in this trajectories of greenhouse gas (GHG) (and work are now discussed in more detail. other pollutants) concentrations resulting from human activity corresponding to a RCP2.6 is a Low Emission scenario. GHG specific level of radiative forcing in 2100. For emissions begin to decline by 2020, and ra- example, RCP4.5 implies a future where radi- diative forcing peaks by mid-century before ative forcing of 4.5 W/m is realized by the end declining to near current levels by 2100. This of the century. An important characteristic of scenario is consistent with the extremely rapid the RCPs is that they do not rely on a fixed adoption of new, cleaner technologies, slower set of scenario-specific assumptions regard- population growth, and strong environmental ing economic development, technological policy. To achieve an RCP2.6 future, new tech- change, or population growth. Instead, there nologies would need to be widely employed in are many different socioeconomic futures or the next 5 to 10 years. The extended RCP2.6 pathways that may lead to the same level scenario assumes “negative emissions” by of radiative forcing. This framework allows 2070, meaning humans are removing more researchers to consider alternative policy CO2 and CH4 from the atmosphere than they decisions with combinations of societal, eco- are releasing. nomic, and technological change. As such, a future with high population but rapid de- RCP8.5 is characterized by increasing GHG velopment of clean technology may achieve emissions over time, leading to high atmo- the same level of radiative forcing as a world spheric concentration. It is a future consistent characterized by low population growth but with scenarios of energy intense development, continued reliance on fossil fuels. This frame- continued reliance on fossil fuels, and a slow work is exceptionally useful from a policy rate of technological development. Alterna- analysis perspective, as it allows researchers tively, pathways characterized by rapid pop- to specify specific levels of global tempera- ulation growth and land use intensification ture change (e.g., 1.5°C) and then explore (croplands and grasslands) are also consis- alternative policy options to achieve emission tent. RCP8.5 implies little to no climate policy, levels consistent with the goal. Previous sce- and it is characterized by significant increases narios, by contrast, specified the socioeco- in CO2 and CH4 emissions. nomic conditions from which climate change/ impacts were then calculated. 84 Shared Socioeconomic Pathways (SSPs) The five SSPs, described in more detail in O’Neill et al. (2015), span a wide range of potential future development pathways, and describe trends in demographics, human development, econ- omy and lifestyle, policies and institutions, technology, and environment and natural resources. Broadly, they are organized according to the respective challenges to adaptation and mitigation in each future world. Importantly, climate change impacts are not directly included in these sce- narios; however, they can be thought of as consistent with broad assumptions regarding the pri- mary factors driving challenges to adaptation and mitigation, namely population and emissions, respectively. National-level estimates of population, urbanization, and GDP have been released for each SSP and are available through the SSP database (https://tntcat.iiasa.ac.at/SspDb/ds- d?Action=htmlpage&page=about). Population estimates include assumptions regarding international migration; however, once again, these assumptions are made in the absence of any information regarding climate change, exposure, and vulnerability. This work will attempt to model the potential impacts of climate change on international migration. No assumptions are made regarding internal migration. In this work, it is proposed to consider two different SSPs, described here in more detail. SSP2 (Middle of the Road) describes a world with development that occurs at rates consistent with historical patterns, and therefore has moderate levels of investment in human capital, tech- nological change, and economic growth. Demographic outcomes are consistent with Middle of the Road expectations about population growth, urbanization, and spatial patterns of develop- ment. Trends vary across regions and over time but, on average, they fall near the center of ex- pectations about future outcomes rather than toward the upper or lower bounds of possibilities. SSP4 (Inequality) describes a mixed world, with relatively rapid technological development in low-carbon energy sources in key emitting regions, leading to relatively large mitigative capacity in places where it matters most to global emissions. In other regions, however, development pro- ceeds slowly, inequality remains high, and economies are relatively isolated, leaving them highly vulnerable to climate change with limited adaptive capacity. 85 Scenario Combinations Proposed for the Modeling Following the Groundswell approach, four plausible socioeconomic and climate futures are pro- posed by combining these SSP and RCP scenarios (Figure 6). The scenarios allow for an examina- tion of the relative importance of climate futures in driving potential migration outcomes. The scenarios can be characterized as follows: 1 A Pessimistic reference scenario (SSP4 Inequality and RCP8.5 High Emissions), in which global emissions remain high and development is disparate. Population growth in middle-income countries generally slows over the next few decades and the population declines after mid-century due to significant economic un- certainty. Urbanization rates are high due to rural economic decline, and GDP growth and education levels remain stagnant. Urban growth is poorly planned and high emissions drive greater climate impacts. This scenario poses high bar- riers to adaptation because of the slow pace of development and isolation of regional economies. The other three scenarios are defined in comparison to this reference scenario. 2 3 4 A more Climate-Friendly A more Inclusive Development sce- An Optimistic scenar- scenario (SSP4 Inequality nario (SSP2 Middle of the Road and io (SSP2 Middle of the and RCP2.6 Low Emis- RCP8.5 High Emissions), which retains Road and RCP2.6 Low sions) with lower emis- high emissions from the Pessimistic Emissions), which com- sions that reduce climate scenario, but provides a development bines the lower emission impacts, but holds the scenario that is more optimistic and scenario that reduces development scenario the potential for adaptation is higher climate impacts and consistent with the pessi- than under SSP4. Population growth provides a development mistic scenario. is higher, but urbanization rates are scenario that is more lower than in SSP4, while progress in optimistic. education and GDP are higher than in SSP4. 86 Figure A3.1: Scenario Framework adopted from Groundswell Africa; Rigaud et al. 2021) Source: Rigaud et al. 2021. Modeling Methods The modeling approach is divided into two sections. The first details the model used to project international migrants from Haiti to the DR, while the second describes the INCLUDE model for projecting internal climate migration. For each decadal projection, the internal model was ap- plied first, and then the international model was used to distribute projected migrants from Haiti across the DR. The models are loosely coupled in that the distribution of Haitian immigrants at each time step influences (slightly) internal migration in the subsequent time step, and internal migration within the country influences how migrants from Haiti chose their destination. The International Model This work draws on a method for estimating origin-destination international movement that could potentially intensify or decline as a result of environmental change used to produce projections of migration across Central America and Mexico for the New Times Magazine/ ProPublica report on climate migration in the Americas (Lustgarden 2020). Advantage is tak- en of existing bilateral flow data (from the ENI 2017) to train and project future migration from 87 Haiti to the DR as a function of aggregate climate impacts at the national level (crops, water, ecosystem productivity24 ), and the existing data on the distribution of Haitian migrants across the DR (from the 2010 Census), to estimate potential changes in origin-destination flows from Haiti to municipalities within the DR25 under the four alternative future scenarios detailed below. The international model operates in two steps; (i) at the country level, total migrants for each time period are estimated; and (ii) migrants are distributed to municipalities within the DR. From the historical migration data, out-migration rates from Haiti to the DR are calculated over each of the historical periods. National-level outflows to the DR are then modeled as a function of the theorized drivers of movement, using a Poisson log-linear regression model: Eq. 1 where (mij,t) is the count of migrants from country HTI (Haiti) to DR at time t; A is the five-year deviation of crop yield/NPP from historical baseline; W is the five-year deviation in water avail- ability from the historical baseline; D is the size of the population born in HTI currently residing in the DR; and (Pi,t) is the population of country i at time t. To distribute the projected migrants from HTI to municipalities within the DR, a probability distri- bution is calculated, where the likelihood that a migrant chooses to reside within a municipality is computed as: Eq. 2 where vi,t is the potential attractiveness of each municipality i to an migrant from HTI, and migrants are distributed according to: Eq.3 where Ii,t is the number of immigrants arriving in municipality i at time j, and the denominator is the sum of vi,t. As such, migrants are distributed to municipalities proportional to the relative attractiveness of each location. 24 Note that sea level rise is not included in this model. 25 In the model, Haiti is considered as a single unit; that is, out-migrants are modeled for the country on aggregate. Migrants are then distributed across municipalities within the DR using a gravity-type function that accounts for environmental conditions, economic conditions (through an agglomeration effect), and the existing Haitian population (considered a proxy for social networks). 88 Historically, population potential is often con- Internal Model (INCLUDE) sidered as a proxy for attractiveness, under the assumption that agglomeration is indicative of the various socioeconomic, geographic, polit- The INCLUDE model downscales national ical, and physical characteristics that make a population projections to subnational raster place attractive. grids as a function of geographic, socioeco- nomic, and demographic characteristics of the For this assessment, the calculation of po- landscape and existing population distribu- tential is modified by adding variables that tion. Gravity-type approaches, commonly used describe local/regional conditions, including in geographic models of spatial allocation and climate impacts on economic livelihoods and accessibility, take advantage of spatial regu- weighting the attractiveness of each location larities in the relationship between population (grid cell) as a function of the historical rela- agglomeration and patterns of population tionship between these variables and observed change. These relationships can then be char- population change. Population potential is, acterized as a function of the variables known conceptually, a relative measure of agglomer- to correlate with spatial patterns of population ation, indicating the degree to which amenities change. and services are available. In the INCLUDE model, this value shifts over time as a function The INCLUDE model uses a modified form of the population distribution, assumptions of population potential, a distance-weighted regarding spatial development patterns (e.g., measure of the population taken at any point sprawl versus concentration), and of certain in space that represents the relative accessi- geographic characteristics of the landscape. bility of that point (e.g., higher values indicate The Groundswell approach expanded the a point more easily accessible by a larger model by considering the local impact of number of people). Population potential can climate on certain key sectors. In this further be interpreted as a measure of the influence expanded version of the model, the agglom- that the population at one point in space ex- eration effect is enhanced or muted as a func- erts on another point. Summed over all points tion of the influence of Haitian in-migrants. within an area, population potential represents Furthermore, the version of the model applied an index of the relative influence that the here operates at higher spatial and temporal population at a point within a region exerts on resolution (1 km and five-year time intervals, each point within that region (Rich 1980), and respectively). it can be considered an indicator of the po- tential for interaction between the population Beginning with the 2020 gridded population at a given point in space and all other popu- distribution for each country, the model es- lations. Population potential will typically be timates changes in the spatial population higher at points that contain, or are close to distribution (including the impact of climate large populations; thus, it is also an indicator change) in five-year time steps by (i) calcu- of the relative proximity of the existing pop- lating a population potential surface (a distri- ulation to each point within an area (Warntz bution of values reflecting the relative attrac- and Wolff 1971). tiveness of each grid cell); and (ii) allocating 89 population change to grid cells proportional- (crop, water, NPP, drought likelihood, and sea ly, based on potential. To generate estimates level). The differences in the spatial popula- of internal migration under climate change, tion distribution between the two scenarios scenarios are then run for each of the rele- that include climate drivers and this “no-cli- vant SSPs that exclude the impacts of climate mate” scenario are attributed to migration change. That is, the values for all variables induced by changing conditions, since the that are influenced by climate change are only variables that have changed are those held constant at current day values impacted by a shifting climate. In this version of the INCLUDE model, population potential (Vi) is calculated as a parametrized negative exponential function: Eq. 4 where spatial mask (l) prevents population from being allocated to areas that are protected from development or unsuitable for human habitation, including areas that will likely be affected by sea level rise between 2020 and 2050. Pj is the population of grid cell j, and d is the straight-line distance between two grid cells. The population and distance parameters (α and β) are estimated from observed patterns of historical population change. The β parameter is indicative of the fric- tion of distance or the cost of travel that generally determines the shape of the distance–density gradient in and around urban areas (e.g., sprawl versus concentration). The α parameter captures returns on agglomeration externality, interpreted as an indicator of the characteristics that make a place more or less attractive. Importantly, the SSPs include no climate im- tion effect that drives changes in the spatial pacts on aggregate total population, urban- distribution of the population. All of the data ization, or the subnational spatial distribution are incorporated into the model as 1 km grid- of the population. The INCLUDE approach ded spatial layers. The value Ai is calculated was modified by incorporating additional as a function of these indicators. Numerically, spatial data, including the ISIMIP sectoral im- it represents an adjustment to the relative pacts, and projections of the mean sea-level attractiveness of (or aversion to) specific rise. The index Ai is a weight on population locations (grid cells), reflecting current water potential that is calibrated to represent the stress, crop yields, ecosystem services, and influence of these factors on the agglomera- the likelihood of drought relative to “normal” conditions. 90 Calibrating the Model The spatial model is calibrated over two an adjustment to relative attractiveness. decadal periods (2000-2010 and 2010–2020) In order to carry out the procedure, model es- of observed population change relative to ob- timates of the α and β parameters are neces- served conditions. As noted above, the value sary, and Ai must be calibrated. Two separate Ai is calculated as a function of these different procedures are employed. climatic/socioeconomic indicators and acts as The α and β parameters are designed to capture broad-scale patterns of change found in the distance-density gradient, which is represented by the shape/slope of the distance decay func- tion from Equation 2. The negative exponential function described by Equation 4 is significantly similar to Clark’s (1951) negative exponential function, which has been shown to accurately cap- ture observed density gradients throughout the world (Bertaud and Malpezzi 2003). To estimate α and β, the model in Equation 2 is fitted to the 2000-2010 and 2010-2020 population change from WorldPop, and the values of α and β that minimize the sum of absolute deviations are calculated as: Eq. 5 where and are the modeled and observed populations in cell i, and S is the sum of absolute error across all cells. The model is fitted for two time steps (2000-2010 and 2010-2020) and the average of the α and β estimates is taken. In this modified version of the population potential model, index Ai is a cell-specific metric that weights the relative attractiveness of a location (population potential) as a function of environ- mental and/or socioeconomic conditions. The modeling approach requires that the relationship between Ai and the different local indicators is estimated, which are hypothesized to impact pop- ulation change. When α and β are estimated from historical data (e.g., observed change between 2000 and 2010), a predicted population surface is produced that reflects optimized values of α and β, such that absolute error is minimized. Figure A3.2 includes a cross section (one dimension) of grid cells illustrating observed and predicted population for 10 cells. Each cell contains an error term that reflects the error in the population change projected for each cell over a five-year time step. 91 It is hypothesized that this error can be explained, at least partially, by a set of omitted variables, including environmental/sectoral impacts. To incorporate these effects, the value of Ai is first cal- culated, such as to eliminate εi (Figure A3.2) for each individual cell (which is labeled “observed Ai”): Eq. 6 where ∆ and ∆ are the observed and modeled population change for each cell i and Ai is the factor necessary to equate the two. The second step is to estimate the relationship between observed index Ai and the different poten- tial drivers of spatial population metrics by fitting a spatial lag model: Eq. 7 where C,H and N are the five-year deviations from the historical baseline on crop yield, water availability, and net primary production, respectively. Together these variables and their respec- tive coefficients constitute the set of explanatory variables that go into producing index Ai. Note that for any grid cell in which C (crop yield) is a nonzero value, the value of N (net primary pro- duction) is automatically set to zero, so that only one of the two variables is contributing to index Ai. Finally, ρ is the spatial autocorrelation coefficient and W is a spatial weight matrix. From this procedure, a set of cell-specific A values is estimated. 92 Figure A3.2. Cross Section of Grid Cells Illustrating Observed and Projected Population Distributions Source: World Bank staff analysis Note: The error term is used to calibrate index A(i). For future projections, projected values of each independent variable are used along with their respective coefficient estimates from Equation 7 to estimate spatially and temporally explicit val- ues of A_i. To produce a spatially explicit population projection for each time step, estimates of α and β from the historical data (which reflect the business-as-usual nature of SSP2) are applied to produce estimates of the agglomeration effect, to which the spatio-temporally variant estimates of A_i for the SSP4 (described above) are applied and, finally, exogenous projections of nation- al urban and rural population change are incorporated and the model is applied as specified above. Past testing of the model indicates that cells meeting certain criteria should be excluded from the calibration procedure. First, cells that are 100 percent restricted from future population growth by the spatial mask (l, Equation 2) are excluded, as the value of v_i in these cells (0) renders the observed value of A_i inconsequential. Second, the rural and urban distributions of observed A_i were found to include significant outliers that skewed coefficient estimates in Equation 7. In most cases, these values were found to correspond with very lightly populated cells, where a small over/under prediction of the population in absolute terms (e.g., 100 persons) is actually quite large relative to total population within the cell (e.g., large percent error). The value of A_i (the weight on potential), necessary to eliminate these errors, is often proportional to the size of the error in percentage terms, and thus can be quite large, even though a significantly small portion of the total population is affected. Including these large values in Equation 7 would have a substantial impact on coefficient estimates. To combat this problem, the most extreme 2.5 percent of obser- vations are eliminated on either end of the distribution. 93 Estimating Internal Climate Migrants Gravity models do not directly model internal is fair to assume that differential population migration. Instead, internal migration is as- change between the climate impact scenarios sumed to be the primary driver of deviations and the development-only scenarios occur between population distributions in model as a function of migration. In this work, to be runs that include climate impacts and the considered an internal migrant, a person must development-only (the “no climate” models) move across municipal boundaries. Thus, that include the nonclimate related drivers. for each municipality, the impact of climate Migration is a “fast” demographic variable change is considered to be the difference be- compared with fertility and mortality; it is tween the Climate and No-Climate scenario responsible for much of the decadal-scale (e.g., SSP2/RCP8.5 versus SSP2/No-Climate). redistributions of population (Rigaud et al. To estimate total internal migration under 2018). Without significant variation in fertil- any scenario, the positive differences at the ity/mortality rates between climate-migrant municipal level between any scenario are populations and nonmigrant populations, it summed with its corresponding No-Climate scenario to derive total climate migrants. Sources of uncertainty and Limitations There are several sources of uncertainty in the climate migration modeling results presented here. Additionally, it is important to acknowledge how choices regarding data inputs and scenarios impact the outcomes. Uncertainty will impact the range of estimated number of climate migrants within and across each scenario, including the projected difference between the four socioeco- nomic/climate change scenarios and the corresponding respective development-only (No Cli- mate-Change) scenarios. These are described here. The following is a description of some of the sources of uncertainty in the model: Variation in ISIMIP impacts across ensemble members (uncertainty in sectoral change). In 1 many areas of the world, the combination of the LPJmL water and crop models project different impacts than the WaterGap-GEPIC combination of water and crop models. This results in different projected regional effects in the INCLUDE model—with some model ensemble members projecting improving conditions and net climate in-migra- tion, while others may project net climate out-migration in the same region. Variations between the two General Circulation Models (GCM) can amplify the ISIMIP 2 differences (climate uncertainty). The GCMs were selected in part because their future precipitation trends differ substantially in magnitude and, partly, even in sign across different parts of the world. This variance in precipitation across the GCMs, in turn, impacts the water and climate models, which then drives different patterns of projected future migration. 94 Time-space arc in the exogenous population projections and the drivers of spatial popu- 3 lation change (temporal uncertainty). There is a temporal component to the modeling which can influence population distribution trajectories. For example, stronger sectoral impacts early in the 80-year time period of the projections will have greater influence than the same impacts later in the period. This is because those early impacts affect the gravitational pull of locations in the future. This creates a “temporal” momentum over which later Climate (and No-Climate) impacts may have less influence. Similarly, the timing of population change (growth or decline) projected by the SSPs, relative to the development of sectoral impacts, can influence outcomes. For example, for most locations in the study, projected population growth is greatest over the next decade; if conditions also are predicted to deteriorate somewhat severely during that period, the impact on migration will be greater than if the deterioration had taken place during a more demographic stable period. Spatial trends in climate drivers that run concurrent to socioeconomic drivers will have a 4 larger impact than those that run counter to socioeconomic drivers (spatial uncertainty). For example, if sectoral impacts occur in ways that reinforce trends in places where the No-Climate scenarios suggest there will be a population gain or loss, that impact will be magnified (multiplicative effect) relative to impacts that counteract the agglomera- tion effect embedded in the SSP-only (No-Climate) model. In other words, if the No-Cli- mate model finds a place is relatively attractive, and the sectoral climate impacts are positive or neutral (relative to other areas that see negative impacts), then this will tend to reinforce the attractiveness of that area. Conversely, in remote areas experiencing population decline and negative climate impacts, the “push” factors will be reinforced. This creates a “spatial” momentum that gains traction over time (and which interacts temporal momentum (see Item 3 above). The combined impact of these two factors can increase the range of outcomes within scenarios (across ensemble members), thereby increasing uncertainty. Model parameterization (uncertainty in the historical data). The model was calibrated 5 using observed population changes in association with observed climate impacts (e.g., as represented by ISIMIP model outputs) for two time periods: 2000-2010 and 2010- 2020. This was done using the two separate sets of model combinations: LPjML water and crop models, and WaterGAP water and GEPIC crop models. There are different parameters that correspond to the different models. If the parameter estimates are close together across the different crop/water model, then there will be less variation in the population distribution projected by each of the models. As such, the uncertainty around the ensemble mean (measured using the coefficient of variation) will be low- er and parameter estimates will be higher. Conversely, if parameter estimates are not similar, the parameter estimates will tend to be lower because the signal of the climate impact on population distribution is diluted; thus there will be greater uncertainty around the ensemble mean. 95 Limited historic time series for model calibration (uncertainty in the relationship between 6 the drivers of migration and changes in the population distribution). Limited historic data of the type necessary to fit the model (e.g., spatially explicit data pertaining to the de- pendent and independent variables in the model) leads to a relatively short period over which to train the model. As noted above, the model was calibrated using two historic decadal periods; 2000-2010 and 2010-2020. Thus, the empirical relationship between changes in local conditions and spatial population dynamics (the population response to changing conditions), is based on a limited set of data, which leads to uncertainty in projected outcomes arises as result of several factors. First, the model parameters may describe a historic period that is not representative of the true, longer term relationship between drivers and population outcomes. While not necessarily the case, it is import- ant to consider whether the period 2000-2020 deviates in any way from the longer his- toric baseline and, importantly, whether there is any reason to believe the period is not representative of the likely relationship between drivers and outcomes moving forward. Second, because the calibration period is short, it may not be representative of the full range of outcomes that have been observed over time, and thus the model parameters, which themselves are averages of the empirical relationship over the two decadal peri- ods, may not truly be measures of central tendency. Finally, the projections in this work cover a long time-horizon (80 years). Basing long-term projections on a short historic time-series is potential problematic, as it is highly likely that there will be some devia- tion from the observed historic relationship. Challenges associated with the cross-border component of the modeling (uncertainty in 7 the projections of Haitian in-migration to the DR). The international component of the spatial population model is a form of econometric model commonly used to project bilateral flows of people (as well as goods and services, etc.) between nations/regions. While this approach is rooted in the current “state-of-the-art”, it is important to consid- er the limitations associated with the approach, particularly within the context of the two countries in this study. Most notably, the model is trained based on migration data originating from a single census period, a result of the lack of existing historic data. Whether the data are truly representative of historic trends is difficult to ascertain, par- ticularly in a country where migration patterns have been, at least anecdotally, known to shift substantially over time as a function of conditions and extreme events. In this model, the migratory response to conditions is based on a single snapshot, and thus may not be representative of the true relationship(s), or more likely, the true range of potential outcomes. Related, the model is also based on a relatively small set of driv- ers, and instead relies heavily on the size and location of the historic diaspora in driving projections (in this model mitigated or enhanced as a function of changing environmen- tal conditions). Once again this is not an unusual approach, given the relative dearth of historic data on migration and the drivers of migration in most countries. However, here we have a small set of drivers drawn from a single historic period, it must be noted that significant uncertainty in projected future migration may arise as a result. 96 Additionally, it is notoriously difficult to project “shocks” that may drive significant changes in migration. While the modeling work here includes projections related to slow-onset environmental change, it does not include projections of rapid-onset ex- treme events like hurricanes (and the associated flooding/mudslides) that have been known to affect the region. Furthermore, uncertainty in the total number of migrants is driven by not only climate uncertainty but also by the interaction between climatic and socioeconomic conditions. Likewise, the wide range of outcomes across scenarios reflects a significant level of uncertainty in the relationship between nationwide climate drivers in Haiti and of movement into the DR (in comparison to the 1-km resolution that was available for the internal INCLUDE model). Additionally, it should be noted that these figures reflect no change in border policy (on either side) over the course of the century—an unlikely outcome if the number of migrants changes substantially, if con- ditions deteriorate substantially, or both. It also reflects a future in which Haiti’s fragil- ity, conflict, and vulnerability conditions continue the trends of the last few decades (which are reflected in the historical data used for this model). Future socio economic and political shocks are not considered in the model, although it can be argued that the data on Haitian migration used to train the model may already reflect, at least partially, socio economic and political conditions of the country. However, because the timing of such events is important in projecting the temporal component of any impact on migra- tion, there remains substantial uncertainty in the timing of mobility that may be related.  The base-year distribution (uncertainty in the current population distribution). Through 8 past research, it was found that the choice of spatial population product (e.g., GPW, WorldPop, GHS-Pop) for the base year (2020) had a significant impact on both aggre- gate and spatially explicit projections of future climate-induced displacement and mi- gration. In some cases, the variation between projections produced for the same coun- try/scenario, using different data products in the base year (e.g., GPW versus WorldPop), was larger than the variation observed across different scenarios using the same base year product (e.g., SSP2/RCP2.6 versus SSP4/RCP8.5, using WorldPop as the initial pop- ulation distribution). This suggests that many of the modeling outcomes might reflect the base year distribution more so than the actual drivers of change. It is important to consider the implications of the choice of base year product, and to consider how the base year distribution is likely to impact outcomes (e.g., using GPW will lead to larger numbers of migrants into cities than World-Pop, owing to the known bias toward rural areas in the former). 97 The crop models exclude several locally important crops. The ISIMIP models project 9 change in crop yields for four major staple crops: maize, wheat, rice, and soybeans. This choice reflects the global importance of these crops. In the DR, however, sugar, banan- as and, to a lesser extent, coffee and cocoa, are of significant importance. At this time, there are no data pertaining to these crops over the time period necessary for appli- cation in this model. As such, the analysis relies on the staple crop data to indicate the degree to which conditions can be expected to vary at the local level, relative to his- torical conditions. A reduced capacity to produce maize or wheat could correspond to declines in the capacity to produce sugar or bananas (depending on what is driving the decline). Theoretically, however, the opposite also could be true. The results presented here should be consumed with the knowledge that local variations may differ from pro- jections in the ISIMIP data. The modelling approach might miss human mobility impacts from SLR. The complexity 10 of SLR impacts pose a number of challenges from a modeling perspective. First, while inundation and storm surge are factors that can be projected and are modelled in this study, other impacts such as salt-water intrusion, loss of drinking water supply, or the perceptions regarding SLR and its impact on residential choice are very difficult to model at present. This is due to both a lack of data and the uncertainty surrounding the impact of SLR on household-level perceptions regarding coastal threats. Also, in this study, a migrant is defined as someone that moves between municipalities. However, many of those likely to be impacted by SLR may not choose to leave their current mu- nicipality, but only to relocate within that community. In our modeling approach, these families would not be considered migrants. Even assessing the movement patterns using the 1km -resolution data (highest resolution available) might miss some people relocating because of SLR. 98 Appendix 4: Additional Tables and Figures from Quantitative Modeling, Dominican Republic Figure A4.1. Hotspots in the Dominican Republic, 2051-2100 Source: World Bank staff analysis 99 Figure A4.2. Net Climate Migrants by Livelihood, 2051-2100: Dominican Republic Source: World Bank staff analysis 100 Figure A4.3. Livelihood Zones: Dominican Republic 1 Data is used only to differentiate results by livelihood zone. It was not used to produce estimates during modeling. Figure A4.4. Projected Population by Livelihood Zone, 2050 and 2100: Dominican Republic Source: World Bank staff analysis 101 Appendix 5: Prioritized Areas for the Emissions Reduction Program, Dominican Republic Source: World Bank (2021) Project Appraisal Document - Dominican Republic Emission Reductions Program 102