Understanding the global subnational migration patterns driven by hydrological intrusion exposure

Climate change is intensifying hydrological risks globally, significantly influencing human settlement patterns. While previous research has largely focused on national-level migration, this study addresses the gap by examining subnational migration patterns driven by hydrological intrusion (including floods, sea-level rise, and storm surges). The increasing frequency and intensity of these events are displacing millions, underscoring the need to understand the complex interplay of factors influencing migration decisions at a granular level. The study acknowledges the multifaceted nature of migration, recognizing the influence of socioeconomic factors, cultural nuances, political policies, and other elements alongside environmental pressures. Some literature suggests that hydrological intrusion may not always be the decisive factor, with adaptation strategies sometimes chosen over relocation. Resource scarcity, particularly in economically disadvantaged regions, can also trap populations in high-risk areas. The study adopts a holistic framework considering hazard, exposure, and vulnerability to provide a comprehensive analysis of climate-related migration risks. Previous hazard-related assessments have limitations, focusing on precipitation levels, flood severity, and floodplain area, without fully accounting for the heterogeneity of population distribution and the varying resources in different areas. While country-level studies on climate change-induced migration exist, they often overlook the crucial subnational scale, where most migration occurs. Linear models used in previous studies may also fail to capture the nuanced non-linear relationship between hydrological risk and migration. This study uses an innovative approach involving population-weighted hydrological intrusion computed via remote sensing and interpretable nonlinear ensemble learning models to address these limitations.

Existing research acknowledges the link between climate-related risks and human migration, but much of it focuses on national-level trends or specific case studies. There is considerable debate about the relative importance of climate change versus other factors like socio-economic conditions in driving migration. Some studies have shown a strong correlation between climate-related disasters and migration, while others find weaker or more complex relationships. The existing literature highlights the importance of considering the vulnerability of different populations to climate change impacts. For instance, studies point to the disproportionate impact of climate change on marginalized groups, who may lack the resources to relocate or adapt to changing conditions. The challenge of measuring the contribution of climate change to migration is also frequently discussed, due to the complexity of the phenomenon and data limitations. Methodological issues surrounding the definition of climate migration, the choice of indicators, and the modelling techniques used complicate comparisons across different studies. Some studies point to the importance of considering the nonlinear nature of the relationship between climate change and migration, emphasizing potential thresholds beyond which climate change becomes a significant driver of displacement. This paper aims to build on this existing research by applying a more comprehensive framework and utilizing advanced analytical techniques to investigate the subnational dynamics of climate-driven migration.

This study uses a novel approach to analyze global subnational-level population migration driven by hydrological intrusion risk. First, it estimates subnational migration rates (Mig R) by analyzing inter-annual variations in population distribution data from WorldPop, adjusting for natural population growth using data from the World Bank. The accuracy of this subnational migration estimation method was validated by comparing aggregated subnational estimates with country-level estimates from the World Population Prospects 2022 (WPP22), resulting in an R-squared value of 0.89. The study acknowledges that the yearly stock data used limits its ability to analyze specific characteristics of population flow, such as origin-destination patterns or temporal aspects. Hydrological intrusion exposure (HIE) is quantified using a population-weighted index, leveraging high-resolution dynamic surface water data from remote sensing (Landsat and other sources), processed and resampled to 100m resolution using ArcGIS Pro. The HIE calculation considers both the extent of surface water and the population density in affected areas. Additionally, the study calculates the divergence of HIE with neighboring units (HIE_D) to assess the influence of regional disparities in risk exposure on migration decisions. To investigate the relationship between HIE, HIE_D, and Mig R, the researchers employ an interpretable machine ensemble learning framework combining Light Gradient Boosting Machine (LightGBM) and Shapley Additive exPlanations (SHAP). A country-year fixed-effects model is used to mitigate the impact of unobserved variables on the results. The model incorporates various socioeconomic, geographic, and climate variables to account for multivariate drivers of migration. Vulnerability is assessed using indicators such as GDP per area, nighttime light intensity, urbanization rate, average education years, temperature, elevation, and distance from water. Finally, the study analyzes the heterogeneous effects of HIE and HIE_D on different income groups (low, middle, and high) and population subgroups (age groups, gender) using separate models for each group. This approach allows for a more nuanced understanding of the drivers of climate-related migration.

The study analyzed data from 46,776 subnational units across 249 countries and regions. Significant regional disparities in migration patterns were observed, with Oceania showing significant out-migration and Europe and America showing significant in-migration. The average global Mig R was -0.59%, indicating net out-migration in a substantial portion of the world. The analysis confirmed that increasing hydrological risks lead to population displacement. However, the relationship is non-linear: the effect of HIE on out-migration diminishes as exposure increases, exhibiting an S-shaped curve. Below a certain HIE threshold (6.74%), increased HIE leads to increased out-migration. Above this threshold, the effect plateaus or even reverses (inward migration in some cases), possibly due to resource constraints limiting migration capacity. HIE_D (the difference between a region's HIE and its neighbors') significantly amplifies the migration effect. Greater disparities in risk exposure between a region and its neighbors increased the likelihood of out-migration. Socioeconomic factors also played a role, but their impact was less significant than HIE_D. Analyzing income groups, the study revealed that high-income regions showed resilience to hydrological risks, with even high-HIE areas attracting population due to economic opportunities. Middle-income regions showed a pattern similar to the global model, while low-income regions exhibited migration primarily to nearby areas due to resource limitations. Examining population heterogeneity, the study found that younger people exhibited a stronger outflow trend compared to older people, potentially linked to labor mobility and employment opportunities. The elderly population showed a significantly lower tendency to migrate even in high-risk areas, suggesting vulnerability and limited adaptive capacity. Gender differences were less pronounced, with men showing slightly greater sensitivity to HIE and HIE_D than women.

This study's findings highlight the significant but non-linear influence of hydrological intrusion risk on subnational migration patterns globally. The non-linear relationship, captured by the S-shaped curve, demonstrates a threshold effect: migration only occurs after exceeding a critical level of exposure. This suggests that policies focused solely on responding to large-scale migration events might be insufficient; proactive interventions are crucial to prevent displacement before thresholds are crossed. The importance of HIE_D underscores the role of spatial inequalities in amplifying migration pressures. Policy interventions should therefore not only focus on reducing hydrological risks but also on addressing regional disparities. The study's findings challenge the widespread reliance on linear models in migration research, demonstrating the need for a more nuanced understanding of the complex and context-dependent relationship between environmental risks and migration. The heterogeneous effects across income groups and demographics emphasize the importance of tailored policies considering the specific vulnerabilities of different populations.

This study provides a comprehensive assessment of the subnational-level impact of hydrological intrusion on global migration patterns. It highlights the non-linearity of this relationship, indicating a threshold beyond which migration is triggered. The significant impact of HIE_D underscores the role of regional disparities in driving migration. These findings support the need for context-specific policies that consider both risk reduction and equitable resource distribution. Future research could explore the use of higher-resolution data (e.g., cellular phone data) to improve understanding of migration flows and investigate the long-term impacts of hydrological changes on migration dynamics. The Libyan flood serves as a stark reminder of the urgency of developing early warning systems and community-based adaptation strategies.

This study has several limitations. The reliance on yearly stock data prevents the analysis of detailed origin-destination bilateral migration patterns and temporal dynamics. Data biases related to disaster reporting in less developed regions affect the accuracy of the hazard variable used. The temporal scope is limited; a longer time series could provide a more robust analysis of migration trends. The study also assumes consistent natural population growth at the national or regional level, which may not be entirely accurate. Finally, the focus is on micro-migration within national borders; further research is needed to analyze international migration driven by hydrological intrusion.