Main

Climate change is undeniably one of the most pressing global challenges of the 21st century, with profound implications for human societies and the natural environment. Among its many effects, the impact on human mobility and international migration has garnered increasing attention from policymakers, researchers, and international organizations1,2,3,4. As environmental changes intensify and exacerbate existing vulnerabilities, many people in more vulnerable regions and communities will be expected to migrate in search of safer and more sustainable living conditions5,6,7. Recognizing the importance of addressing climate-induced migration, 37 countries had implemented 43 national policies linking climate change to migration as of December 31, 20198,9. This growing trend underscores the urgency of understanding the complex relationship between climate change and migration, as well as the need for adaptive policy responses.

As global greenhouse gas emissions continue to rise, the urgency of addressing climate change becomes increasingly apparent. Projections suggest that global warming is likely to exceed 1.5 °C in the 2020 s and could reach 2 °C before 2050 without sustained reductions in carbon emissions10. Recent studies have highlighted the multifaceted impacts of rising sea levels, focusing on human migration, displacement, and health outcomes9,11,12,13,14,15. Migration outcomes of slow-onset events differ from those of rapid-onset ones. While slow-onset impacts such as droughts, desertification, and land degradation are usually associated with progressive departures, rapid-onset events tend to be associated with sudden departures which are predominantly internal displacements16. However, sea level rise is not the sole climate-induced driver of migration. Extreme weather events—including hurricanes, floods, droughts, and heatwaves—are becoming more frequent and severe due to climate change, leading to widespread displacement and migration17,18,19,20. These events cause immediate destruction of homes and infrastructure and have lasting impacts on essential services such as healthcare, food security, water availability, and general livability. The strain on urban areas—where displaced populations often seek refuge—exacerbates existing social inequalities, intensifies tensions, and fosters conflicts over resources21,22,23,24.

The human toll of climate change is starkly illustrated by the rising incidence of heat-related mortality. Over the past two decades, heat-related deaths among people aged 65 and older have increased by 53.7%, with particularly severe impacts in regions experiencing more frequent and intense heatwaves25,26,27,28. The economic consequences are also significant. In 2019, an estimated 302 billion hours of potential labor capacity were lost due to extreme heat9,29,30. Countries like India and Indonesia are among the most affected, with losses in potential labor capacity equating to 4–6% of their annual GDP. The IPCC Sixth Assessment Report (AR6) emphasizes that human-induced factors will continue to increase the frequency and intensity of heatwaves, amplifying associated risks31. In response, migration is increasingly recognized as a key adaptation strategy for populations facing extreme heat19,32,33.

The empirical evidence on the linkage between climate change and migration is diverse, inconclusive, and often contradictory; and this is unsurprising given how complicated and contextual the drivers of mobility decision-making are. The complication is further exacerbated by the fact that the movement could be voluntary or involuntary, internal or international, instigated by slow-onset events such as droughts and land degradation or fast-onset ones such as floods and hurricanes, and the effects could materialize through direct or indirect impacts16. Much of the existing literature on the interactions between climate change and migration are regionally focused, due mainly to data challenges such that the evidence base is weak for international migration-climate change interactions. This is mainly due to the complex interactions among the multiplicity of drivers such that climatic factors interact with economic, political, social, cultural, and historical factors in myriad ways to influence international migration34. These complications often lead to complex and even contradictory findings.

Focusing on international migration between 1960 and 2000, Beine and Parsons find no direct evidence of long-run climatic influence on international migration35, beyond an indirect relationship through wage differentials. This somewhat contradicts the findings of Cattaneo and Peri who utilized bilateral migrant stock data for the same period but demonstrate that increases in average temperatures increased both internal5, rural-urban migration as well as international migration though the latter is more significant when source countries have middle-income status36. In poorer countries, however, the relationship tends to be inverse; rising temperatures were associated with decreased emigration, suggesting higher potentials of trapped populations5,6,16. This is, in turn, contradicted by Falco7, who relying on data on 108 countries between 1960 and 2010, finds that climate change, through negative shocks to agricultural productivity, significantly increased international migration from developing countries, with the effect being more pronounced in poorer countries and less so in middle-income countries.

Despite the seeming lack of consensus as has been highlighted by recent reviews of the extant literature37,38,39,40 as well as empirical studies41,42,43, four broad insights can be gleaned. First, the challenge of untangling the complex relationship between climate change and migration has no end in sight37. Second, the relationship is non-linear and context-dependent7,36,43. Thirdly, much of the climate-related migration, especially in developing countries, which are also the most susceptible to its impacts, are at least in the initial stages, predominantly internal7,42,43. Fourthly, agriculture plays a central role as a transmission mechanism of the effects7,40,43,44, among other mechanisms, such as economic pull factors in destination countries, accessibility to adaptation and mitigation measures, information flows and educational levels in source countries, which all contribute in complex ways to influence perceptions and narratives of climate impacts37,38.

Our study contributes to this field by shedding further light on the relationship between climate change and global migration. The present contribution is relevant for three main reasons: 1) our analysis relies on an expanded 160 countries and regions, albeit over a relatively limited timeframe of 1980–2018; 2) it incorporates income levels of countries, proxied by GDP; and (3) it leverages the determined relation to then make projections of likely future migration trends given expected changes in climate baselines and GDP. There is a pressing need for more comprehensive evaluation of how climate change is reshaping international migration patterns globally. This study addresses this gap by investigating the primary climate-related drivers of international migration over the past 40 years, and project future trends. By developing a model that estimates migration sensitivity to climate-related factors—such as baseline climate conditions and economic drivers—this study leverages observational data and numerical simulations to provide a more nuanced understanding of the relationship between climate change and international migration than previous research.

Results

Linking global migration and climate change in the historical period

According to the definition in Section 2, the correlation between a specific climate index and migration—after detrending the data and applying a 13-year moving average—serves as an indicator of migration response to climate change (Fig. 1a and Supplementary Fig. 1). Using the regional divisions shown in Supplementary Fig. 2, we categorized 160 countries and regions into 18 subregions. Figure 1a illustrates the sensitivity of international migration to climate change (SIMC) for these subregions in response to six key climate indices, while individual country-level SIMC values are detailed in Supplementary Fig. 3.

Fig. 1: Sensitivity of International Migration to Climate and its relationship with national indicators.
figure 1

a Sensitivity of International Migration to Climate (SIMC), measured as the correlation between international migration and six selected climate indices—frost days (fd), ice days (id), heating degree days (hddheat18), cooling degree days (cddcold18), growing season length (gsl), and growing degree days (gddgrow10)—across 18 sub-regions. b Correlation between SIMC and national-level indicators for 160 countries and regions, including gross domestic product (GDP), electric power consumption (EPC), energy use (EU), life expectancy at birth (LEB), mortality rate under age 5 (MR5), arable land per capita (AL), crop production index (CPI), unemployment rate (UT), urban population (UP), rural population (RP), government expenditure on education (GEE), literacy rate (LRA), carbon dioxide emissions (CO₂), renewable internal freshwater resources (RFR), fossil fuel energy consumption (FFC), Human Development Index (HDI), Environmental Sustainability Index (ESI), and Happy Planet Index (HPI). Asterisks indicate statistically significant correlations at the 95% confidence level.

Full size image

Overall, the 160 countries and regions show contrasting responses to cold-related and warm-related climate indices (Fig. 1a), particularly in high-latitude regions (e.g., Northern Europe, Northern America, and Southern South America). These climate indices generally reflect the suitability of extreme low and high temperatures for biological perception (heating degree days (hdd) and cooling degree days (cdd)) and vegetation productivity (frost days (fd), ice days (id), growing season length (gsl), and growing degree days (gdd)) in certain regions. In these areas, an increase in cold-related days typically drives migration outflows, while an increase in warmer days is associated with migration inflows. Conversely, low-latitude and equatorial regions—especially in Southern Asia and Southeastern Asia—exhibit the opposite trend, where warmer days lead to greater migration outflows.

This divergence can be attributed to baseline temperatures: low-latitude regions already experience high temperatures, and additional warm days further reduce climate livability, making migration a predictable response. Meanwhile, in high-latitude regions, which are generally cooler, warming trends improve living conditions, attracting migration inflows. However, not all countries conform to these patterns. For example, migrants from lower-latitude countries in Asia and Africa demonstrate relatively high uncertainty regarding temperature changes, suggesting that the relationship between climate change and migration varies significantly based on latitude. This latitude-dependent trend is consistent with the notion that baseline climate conditions, which shape natural living environments, are heavily influenced by a region’s distance from the equator45. Given the strong correlation between latitude and climate conditions (Supplementary Fig. 4), latitude can serve as a convenient proxy for describing migration responses to climate under current patterns. However, we argue that climate baselines—dynamic and variable by nature—offer a more precise lens for understanding such responses than the fixed geographic attribute of latitude. This distinction becomes particularly critical under scenarios of substantial future climate change, where regional climate norms may shift dramatically, fundamentally reshaping climate conditions. Latitude alone cannot adequately reflect these changes. Therefore, our analysis focuses on climate baselines rather than latitude to more accurately capture the dynamics of climate-driven migration.

The role of climate baseline and economic conditions

We also analyzed the relationship between SIMC and 18 national indicators across 160 countries (Fig. 1b) and found that SIMC is primarily correlated with single-variable national indicators. While these indicators capture various social factors—such as the economy, education, health, population composition, and the environment—they are not entirely independent of one another. Most single-variable indicators are significantly correlated with economic indicators (Supplementary Fig. 5), as economic development tends to strengthen other social systems, including education and healthcare. Although multivariable indicators are more independent, SIMC does not show a strong response to them. Given the strong correlation between GDP and most national indicators, as well as its consistently significant association with SIMC across the six climate indices, we therefore focus on the response of SIMC to GDP in subsequent analyses, which serves as a representative economic indicator.

The evidence indicates that international migration flows in response to climate change are influenced by a country’s climate baseline (geographical location) and social conditions, particularly economic factors. Moreover, the combined influence of these factors on SIMC is not merely additive (Fig. 2, Supplementary Figs. 6 and 7). Specifically, in countries with relatively low baseline temperatures, warming-related shifts in cold-associated climate indices tend to suppress population outflows. In contrast, in countries with higher baseline temperatures, warming-related shifts in heat-associated climate indices tend to promote population outflows. A further clarification is that, in countries with a GDP below the global median, SIMC does not exhibit a strong dependence on the climate baseline. This implies that while climate change may have made living conditions in these countries less favorable, it has not resulted in significant migration outflows. In contrast, in countries with stronger economic conditions, the negative (or positive) impacts of climate change on livability are associated with substantial migration outflows (or inflows). This finding aligns with previous research, which suggests that only relatively wealthy individuals and households can afford to migrate across borders or internationally in response to climate impacts46,47. It highlights that economic conditions can constrain the extent to which SIMC depends on a country’s climate baseline. Notably, this constraint operates in one direction: while SIMC’s relationship with economic conditions is moderated by the climate baseline, the reverse is not true.

Fig. 2: Relationship between the Sensitivity of International Migration to Climate (SIMC) and climate baseline and economic conditions.
figure 2

a–f Distribution of Sensitivity of International Migration to Climate (SIMC) for a frost days (fd), b ice days (id), c heating degree days (hddheat18), d cooling degree days (cddcold18), e growing degree days (gddgrow10), and f growing season length (gsl) by climatological annual average temperature. Red bars represent countries in the top 50% of GDP, while blue bars represent those in the bottom 50%, and White bars denote the results of total countries. The height of the bars indicates the correlation coefficient.

Full size image

We developed an idealized climate baseline-economic dependency model to estimate SIMC and compared these estimates with observed SIMC (Fig. 3). Details of the model are provided in the Methods section. The results reveal significant discrepancies between the estimated and observed SIMC in nearly all African countries and parts of northern Latin America. These findings suggest that migration responses to climate change in these regions are more unpredictable, making it difficult to forecast migration patterns based solely on geographical location and economic conditions.

Fig. 3: The combined impact of climate baseline and economic conditions on the Sensitivity of International Migration to Climate (SIMC).
figure 3

The impact is determined by the difference between the real SIMC and the estimated SIMC from the idealized model. An opposite sign between the estimated and observed SIMC indicates that climate baseline or economic factors have minimal influence on SIMC in that country. The bar plot in the left lower corner denotes the percentage for each category. A matching sign, with the estimated SIMC being greater (smaller) than the observed value, suggests the model overestimates (underestimates) the impact of these factors on international migration. For calculation details, refer to the Methods section.

Full size image

In contrast, the estimated SIMC for European countries, Australia, parts of Asia, and certain North American countries aligns closely with the observed SIMC. This alignment indicates that latitude and economic conditions play a substantial role in shaping SIMC in these regions. For instance, in Nordic countries, global warming has led to a warmer and more hospitable climate (e.g., longer growing seasons and fewer fd), combined with strong economic conditions, a well-developed welfare state, and historically liberal immigration policies. These factors have attracted significant inflows of international migrants. While this correlation between climate improvements, economic strength, and migration inflows in Nordic countries is intriguing, it does not necessarily imply a causal relationship. However, the significant statistical correlation provides a valuable basis for further exploration into potential causality.

Outlook of international migration under future warming

Using the idealized model, this study projected potential changes in SIMC for the mid and late 21st century under different climate scenarios, based on future climate and GDP projection data (Fig. 4). Compared to observed results, SIMC is expected to increase significantly in mid-to-high latitude countries in the Northern Hemisphere under both the SSP2-4.5 and SSP5-8.5 scenarios. In contrast, changes in SIMC are projected to be less pronounced in the Southern Hemisphere and low-latitude countries. These projections suggest that international migration in regions such as northern North America, eastern and northern Europe, and northern Asia will become more sensitive to future warming. However, this increased sensitivity under future scenarios is primarily driven by economic factors, particularly GDP growth, rather than shifts in climate baselines. This finding highlights the dynamic, rather than static, nature of how climate baselines and economic factors regulate international migration.

Fig. 4: Projections and attributions of Sensitivity of International Migration to Climate (SIMC) in the 21st century.
figure 4

a Mean variation in SIMC across the six climate indices for the mid-21st century under the SSP2-4.5 scenario. b the relative contributions of climate baseline and GDP changes calculated by the Ridge Regression Model. c,d SIMC c projections, and d attributions for the end-21st century under the SSP2-4.5 scenario. e,f SIMC e projections, and f attributions for the mid-21st century under the SSP5-8.5 scenario. g,h SIMC g projections, and h attributions for the end-21st century under the SSP5-8.5 scenario.

Full size image

As global warming raises average annual temperatures in high-latitude regions, the cooler climate advantage that once existed in these areas diminishes. Nevertheless, these regions will remain more livable than low-latitude areas, whose economies are projected to suffer greater devastation under these climate scenarios. In this context, the relatively stronger economies in high-latitude regions will exert a greater influence on international net migration than average temperatures. This explains why the projected increase in SIMC for mid-to-high latitude countries is smaller at the end of the 21st century compared to the mid-21st century. It also clarifies why SIMC growth under SSP5-8.5 is less pronounced than under SSP2-4.5: despite more severe warming under SSP5-8.5, the relative advantage of cooler climates in these regions continues to decline.

However, certain limitations in the projections should be noted. The model does not fully account for relative changes in the climate baseline. Specifically, while mid-to-high latitude regions will experience warming, they will remain relatively cooler compared to low-latitude areas. Despite these limitations, the study highlights that climate change is likely to drive significant increases in international migration flows into mid-to-high latitude developed countries.

Discussion

This study examines the relationship between international net migration rates and 37 climate indices across 160 countries and regions over the past four decades to assess the SIMC. The findings reveal a strong correlation between international migration and temperature-related climate indices at the decadal scale. In Europe, North America, and South America, warmer climates are positively correlated with attracting more international migrants, whereas in South Asia and Southeast Asia, warming is linked to increased migration outflows. The SIMC in these countries and regions is shaped by both climate baselines and economic conditions. The climate baseline directly influences livability, while economic conditions determine the extent to which SIMC responds to climate baselines. In lower-GDP countries, SIMC shows limited correlation with climate baselines.

The weaker relationship between international migration and temperature changes in low-latitude regions of Africa and Asia may initially appear to result from geographic location. However, deeper analysis highlights the role of income levels and immobility challenges. Previous studies48,49 describe involuntary immobility as the inability to migrate despite aspirations to do so, a condition exacerbated by low economic capital and heightened vulnerability to environmental threats. This dual disadvantage undermines the viability of migration as a climate adaptation strategy for the poorest populations50. The Foresight Report emphasizes that this category of individuals faces significant barriers, raising questions about the applicability of the migration-as-adaptation thesis and underscoring the need for nuanced, multi-scalar perspectives51.

While the climate baseline is influenced by latitude, economic development disparities are shaped by global asymmetrical power relations, including historical colonialism and unequal exchanges between the Global North and South52. Attributing economic disparities solely to geography risks environmental determinism53. This study’s reliance on historical climate and GDP data to predict SIMC assumes continuity of past trends, but future migration patterns may be shaped by evolving socio-economic, political, and technological factors, such as policy changes, conflicts, and innovations. Moreover, using GDP as the primary economic indicator oversimplifies complex migration drivers, potentially neglecting inequalities, resource access, and social networks.

Beyond temperature changes, precipitation variability also significantly affects migration through its impact on agriculture54. Existing precipitation indices fail to capture critical migration patterns, likely because migration is more sensitive to specific precipitation events, such as timing shifts in the rainy season, rather than annual totals. Future analyses should incorporate tailored precipitation indices aligned with local agricultural calendars to better understand these dynamics. Warmer climates may reduce moisture availability, prolong droughts, and foster pests and diseases, leading to declining crop yields in vulnerable regions like sub-Saharan Africa, South America, and South Asia55,56.

Migration from low-latitude regions is gaining attention as an adaptation strategy amid these challenges. Yet, this phenomenon presents both opportunities and challenges. Migrant-sending regions may benefit from remittances, while receiving regions experience labor force growth. However, brain drain and infrastructure pressures could create significant issues for both. High-latitude countries, with their stronger economies, are better positioned to adapt to increased migration inflows, but these dynamics also raise climate justice concerns. Global North countries bear greater responsibility for greenhouse gas emissions, yet disproportionately benefit from the economic advantages of migration, while low-income countries face severe climate impacts. At COP29 (2024), while developing nations requested $1.3 trillion annually for climate mitigation and adaptation by 2035, only $300 billion was pledged, reflecting a persistent gap between commitments and action57. Policymakers should address these disparities, prioritizing climate justice when designing migration and adaptation policies.

Our approach is limited in a number of ways, due largely to data availability. Future studies could explore, data permitting, the extent of immobility—both voluntary and involuntary—which have not garnered deserved attention. For mobile populations, future studies could analyse how the relationship works among different income groups within countries, as present studies, including the present one, focus on income at the national levels. Furthermore, considering the central role that agriculture plays as a transmission mechanism between climate and migration7,43,44, future studies could test this relationship further by focusing on only agriculture-dominant economies.

Further focus on the non-linear effects40,43 of climate change could provide a more comprehensive understanding of migration mechanisms, aiding policymakers in crafting effective strategies. Policymakers must also recognize that those least responsible for greenhouse gas emissions and least benefiting from their causes are often the most vulnerable and reliant on migration as an adaptation strategy58,59. Climate justice considerations should guide international migration policies to ensure equitable outcomes.

Methods

Data and preprocessing

To objectively and quantitatively assess climate change, we selected 37 indices from the HadEX3-ETSCI Gridded Land Surface Extremes Indices dataset60, with a horizontal resolution of 1.25° x 1.875°. This dataset, produced under the coordination of the WMO CCl/WCRP/JCOMM Expert Team on Climate Change Detection and Indices (ETCCDI) and the WMO Expert Team on Sector-Specific Indices (ET-SCI), encompasses 27 ETCCDI indices and ten ET-SCI indices. These indices comprehensively capture changes in both mean and extreme states of temperature, precipitation, and drought (see Supplementary Table 1 for detailed descriptions).

Net international migration rates (per 1000 population) for 1980–2020 were sourced from the United Nations Department of Economic and Social Affairs Population Division via the Migration Data Portal (United Nations, World Population Prospects 2024). Positive values indicate population inflow, while negative values reflect outflow. This dataset provides complete time series for population size (by age and sex) and components of population change, including international migration, for 236 countries or regions with a minimum population of 1000 in 2023. Further details on statistics and estimation methods are available in the official technical documentation (https://population.un.org/wpp/Methodology/).

This study integrates key indicators to evaluate population, socio-economic development, and environmental factors, providing a comprehensive profile of each administrative entity. We selected 18 indicators, categorized into Economic, Health, Agricultural, Social, Educational, Environmental domains, and Comprehensive Indicators. Supplementary Table 2 provides detailed descriptions of these indicators.

To project climate change impacts on international migration for the mid-21st century (2040–2060) and late 21st century (2080–2100), we utilized the GDP dataset from ref. 61. This dataset, based on the population-development-environment (PDE) model and Cobb-Douglas production function, incorporates historical data on fertility, mortality, migration, education, capital stock, total factor productivity, and labor force for parameterization and validation. The GDP values, normalized to 2010 prices, are available under three shared socioeconomic pathways (SSPs). SSP1 represents a sustainable development trajectory with reduced reliance on natural resources and fossil fuels; SSP2 follows a business-as-usual trajectory, continuing recent trends while gradually achieving development goals; SSP5 emphasizes fossil-fuel-driven economic growth to address socio-economic challenges62.

For 21st-century climate simulations, we used data from 27 models in the CMIP6 project, selected based on the availability of daily surface air temperature data63. Future climate projections were analyzed under three major emission scenarios: SSP1-2.6, SSP2-4.5, and SSP5-8.5, corresponding to the global forcing pathways of RCP2.6, RCP4.5, and RCP8.5 under the socio-economic contexts of SSP1, SSP2, and SSP5, respectively. For basic information on these models, refer to Supplementary Table 3.

To ensure consistent data availability throughout the study period, we implemented a rigorous quality control process for the immigration, climate indices, and national indicator datasets. As the climate data were derived from interpolated observations from station data, some countries and regions lacked complete climate change information. Additionally, due to limited availability of immigration and national indicator data prior to the 1980s, our analysis focused on 160 countries and regions from 1980 to 2018.

Technological and social advancements, along with global warming, have introduced significant monotonic trends in both the climate indices and national indicators. To avoid spurious correlations caused by these trends, we compared the correlations between immigration and climate indices before and after removing linear trends. Furthermore, acknowledging the lagged response of international migration to climate change, we applied a moving average to the data (see Supplementary Fig. 1).

As the sliding window size increased, the observed response of immigration to climate change initially decreased, then increased. When the sliding window exceeded 11 years, the results with linear trend removal converged with those without trend removal, providing more robust insights into the relationship between climate change and migration.

Based on these findings, we adopted the approach of using data with linear trends removed and a 13-year moving average applied for the subsequent analyses. To further assess the significance of correlations between two autocorrelated time series, we used the effective number of degrees of freedom (\({N}^{e{ff}}\)) as described by refs. 64,65. The \({N}^{e{ff}}\) is calculated as:

$$\frac{1}{{N}^{{{\rm{e}}}{ff}}}\approx \frac{1}{N}+\frac{2}{N}{\sum }_{j=1}^{N}\frac{N-j}{N}{\rho }_{{XX}}(j){\rho }_{{YY}}(j)$$
(1)

where N denotes the sample size, \({\rho }_{{XX}}(j)\) and \({\rho }_{{YY}}(j)\) are the autocorrelations of two sampled time series at the time lag of \(j\).

Selection of key indices/indicators

This study defines the correlation between net migration and climate indices as the SIMC. A higher absolute SIMC value indicates a stronger influence of climate change on international migration. Specifically, an SIMC value greater than 0 implies that an increase in a climate index corresponds to higher migration inflows, while an SIMC value less than 0 signifies increased migration outflows.

International migration decisions are significantly influenced by the overall conditions of the destination country, resulting in notable variations in SIMC across countries and regions. To determine, which aspects of climate change international migration is most sensitive to and how national conditions influence this sensitivity, we analyzed the correlation between SIMC values for 160 countries and 18 national indicators (see Supplementary Fig. 2).

The findings show that migration linked to six climate indices—fd, id, hdd, cdd, gsl, and gdd—is most significantly affected by national indicators. These indices share a common focus: they are temperature-related and primarily reflect average climatic conditions over a year rather than extreme events. Consequently, this study narrows its focus to these six indices.

These indices can be broadly categorized into two groups: those associated with cold conditions (id, fd, hdd) and those linked to warm conditions (cdd, gsl, gdd).

Impacts of climate baseline and economic conditions on SIMC

Modelling climate migration is a challenging endeavor due, largely, to the multiplicity of migration drivers34, the continuum of slow- and rapid-onset events having both direct and indirect effects16, and the often contradictory role of climate as both a driver and inhibitor of migration43. Added to these are challenges relating to spatial and temporal data limitations, especially relating to immobility, both voluntary and involuntary37, which are, in turn, influenced by more subjective sentiments such as place attachment66,67 and negative incentives to migrate such as the increasingly restrictive and dynamic migration policy milieu68. Given that migration is usually neither the only option nor the first choice for climate-impacted populations43, any long-term modelling of climate-related migration is inexorably riddled with uncertainties and are, at best, indications rather than predictions.

We propose an idealized model for quantitatively assessing whether the SIMC is correlated to a country’s climate baseline and economic conditions. To achieve this, we first constructed a Ridge Regression model using the climatological annual mean temperature and Gross Domestic Product (GDP) of each country or region. Ridge Regression is a regularization technique that effectively addresses multicollinearity among predictor variables, helping to produce more stable and reliable estimates69. The algorithm minimizes the following cost function:

$${Minimize}{\sum }_{i=1}^{n}{\left({y}_{i}-{\hat{y}}_{i}\right)}^{2}+\alpha {\sum }_{j=1}^{p}{\beta }_{j}^{2}$$
(2)

Here, \({y}_{i}\) represents the observed SIMC values, \({\hat{y}}_{i}\) represents the estimated values, \({\beta }_{j}\) denotes the coefficients of the model, and \(\alpha\) is the regularization parameter controlling the strength of the penalty on the coefficients. The value of \(\alpha\) was determined using cross-validation techniques to identify the optimal level of regularization that minimized prediction error. Although our observational analysis includes 18 national-level indicators to assess migration responses, the model incorporates only GDP as a non-climatic factor. This choice is primarily based on the widespread availability and acceptance of GDP projections under different Shared SSPs, whereas many of the other indicators lack future data, making them unsuitable for migration projections. Moreover, our statistical analysis shows that most of these additional indicators are highly linearly correlated with GDP, suggesting that GDP can partially capture their variability. While including more variables could enhance the model’s comprehensiveness and interpretability and help reduce omitted variable bias, it would also introduce multicollinearity among predictors, thereby weakening the robustness of the regression model. Using this ridge regression model, we further calculated the relative contributions of changes in the climate baseline and GDP to changes in SIMC. Based on these, the system of regression equations can be expressed as:

$${SIM}{C}_{i}=\alpha +{\beta }_{1} \, {\cdot} \, {Temperature}+{\beta }_{2} \, {\cdot} \, {GDP}+\varepsilon$$
(3)

In the regression equation, the term \(\varepsilon\) denotes the error term, which captures the unexplained variation in the dependent variable that cannot be accounted for by the independent variables included in the model. It reflects random noise, omitted factors, or measurement errors. Since the error term is not directly estimated as a coefficient, it is not reported in the Table 1. Instead, the Table 1 summarizes only the estimated parameters (intercept and regression coefficients) together with their corresponding significance levels.

Table 1 Estimated regression coefficients and their significance levels for climate indicators
Full size table

If the estimated SIMC has an opposite sign to the observed SIMC, this indicates that the SIMC is not significantly influenced by the climate baseline or economic factors in that country. Such cases are assigned a value of -1. Conversely, when the estimated and observed SIMC share the same sign, further differentiation is made based on the magnitude of the estimated SIMC relative to the observed SIMC. If the estimated SIMC is greater than the observed value, the model overestimates the influence of climate baseline and economic factors, implying a weak response. These cases are assigned a value of 1. If the estimated SIMC is less than the observed value, the model underestimates the influence, suggesting a strong response. These cases are assigned a value of 2.

To provide a comprehensive assessment, we averaged the performance across the six climate indices for each country. The resulting scores, ranging from -6 (indicating insensitivity to all six indices) to 12 (indicating sensitivity to all six indices), in intervals of 3, were categorized as follows: extremely insensitive, insensitive, neutral, weakly sensitive, sensitive, strongly sensitive, and extremely sensitive. Overall, we use a system of Ridge Regression equations to estimate SIMC values associated with six climate indicators. By comparing the estimated SIMC with the observed SIMC, we apply a scoring system to derive a composite measure of migration sensitivity in response to both climate conditions and economic levels.