Ghana, a lower-middle income country in West Africa, has a population of approx. 31 million inhabitants (GSS, 2021a). Ghana’s economy is mainly driven by the agricultural sector, which employs 33% of the workforce (62.9% when referring to the rural population), with a high dependency on rain-fed crops such as maize, millet, and cassava (GSS, 2022b; MoFA, 2021). The country faces challenges due to inadequate infrastructure and lacking access to essential services such as water, sanitation, and health care (GSS, 2021b). Cocoa is a key economic contributor, alongside other cash crops such as oil palm, cashew and rubber (Essegbey and MacCarthy, 2020). Ghana is divided into 16 regions (administrative level 1) and 261 districts (administrative level 2; see Figure 1).

The second country investigated is Nigeria, which is the continent’s largest economy and most populous country with 236.7 million people (The World Factbook, 2021). Nevertheless, the poverty rate exceeds 50%, inequality is increasing, and the economy is vulnerable to fluctuating oil prices. Agriculture employs about 35% of the workforce, and the sector is heavily dependent on rainfall (Alehile, 2023). Ongoing conflicts, such as the Boko Haram insurgency in the north-east and unrest in the Niger Delta, are exacerbated by governance challenges that threaten the overall stability of the country (Berger et al., 2021). Climate change increases challenges in key sectors such as agriculture and hydropower, and disrupting food and water security (The World Bank Group, 2021). Moreover, Nigeria’s rapid urbanization has led to the growth of informal settlements in cities such as Lagos and Abuja, where residents are exposed to various climate-related risks, including flooding and heat waves (Benjamin Obe et al., 2023; Ismail et al., 2024; Ndimele et al., 2024). Nigeria is divided into 36 states and 774 local government areas (LGAs). In both countries, the rural population’s reliance on agriculture amplifies vulnerability to impacts of climate risks like poor crop yields due to droughts.

In this study, the risk assessment framework proposed by Zebisch et al. (2023) based on the IPCC AR6 was adapted to identify areas where climate-induced hazards interact with pre-existing vulnerabilities potentially leading to migration. McLeman et al. (2021) expanded this framework by viewing migration as part of a continuum of agency, emphasizing that migration decisions are shaped by perceived risks and available options. When local adaptation measures are not sufficient to reduce risk and a certain threshold is crossed - such as resource depletion or a decline in livelihoods - households may choose to migrate as a response. As outlined in the IPCC AR6, the determinants of risk include hazard, vulnerability and exposure (IPCC, 2021). Hazards, such as droughts or floods, intensified by climate change, threaten agriculture and food security in the region (Zougmoré et al., 2016). Section 2.4 details the newly generated data related to hazard indicators. Vulnerability refers to the susceptibility of a population to harm due to various socio-economic and environmental factors, such as poverty, lack of infrastructure, and limited access to resources. In West Africa, high agricultural dependency, combined with socio-economic challenges, amplifies the vulnerability of rural communities to climate impacts (Sultan and Gaetani, 2016). Vulnerability indicators are classified into ‘sensitivity’ (socio-economic/ecological) and ‘capacity’, as proposed by Zebisch et al. (2023). The newly generated data related to vulnerability indicators is detailed in Section 2.5. Exposure refers to the presence of people, livelihoods, and assets in hazard-prone areas. In West Africa, large segments of the population reside in regions highly susceptible to climate hazards, increasing their risk of adverse impacts (Almar et al., 2023; Trisos et al., 2022). The data generated for exposure indicators is outlined in Section 2.7.

Migration is integrated into the risk assessment framework as a potential outcome resulting from the interaction of hazard, vulnerability and exposure. The overall risk is calculated using Equation 1:

Where the present or future hazard, vulnerability, and exposure component are combined for the risk indicator and w_H, w_V, and w_E are the respective weights assigned to each component. The indicator selection for each component, along with their weightings, is described in the Section 2.3. For all components and the overall risk, values ≥0.6 are assumed to have rather negative impacts on the population (GIZ and EURAC, 2017). This threshold is used to identify areas where migration is more likely to occur compared to areas below this threshold. Migration is conceptualized as one of the adaptive responses that individuals and communities may choose to reduce their exposure to climate risks, increase their resilience, or seek better opportunities (McLeman et al., 2021).

This study focuses on the RCP4.5 scenario, a “middle of the road” pathway that projects a temperature rise of approximately 2.7°C by the end of the century (IPCC, 2023). This scenario aligns with the guidance from Zebisch et al. (2023) for the application of climate risk assessments. An overall workflow is illustrated in Figure 2.

In November and December 2023, two online workshops with experts from Ghana and Nigeria aimed to identify and weight key hazard, vulnerability and exposure factors that influence migration decisions in each country. The interactive tool Miro Board (Miro, 2024)¹ was used to introduce pre-defined factors from a literature review (based on Schürmann et al., 2022) to the participants. Four experts from Ghana and six from Nigeria reviewed these factors, with the option to modify, delete or add new factors, and to map connections between components. This process resulted in the identification of 11 hazard factors, 16 vulnerability factors and 10 exposure factors, although not all were relevant in both countries (e.g., fire events were important for Ghana but not for Nigeria). Some factors could not be included in the analysis due to unavailability of proxy indicators (see Supplementary Table S1).

Participants in these interviews were scientists from research institutes and NGOs with expertise in human migration and related research fields, particularly agricultural systems, rural and urban systems, food security, climate change risks, and adaptation strategies in their respective countries. The majority of the experts have more than 10 years of experience in their fields (Supplementary Table S2).

The “Budget Allocation” method (European Commission, 2023) was employed to weight these factors. Each expert was given a budget of 100 points to distribute among the factors for each composite category (hazard, vulnerability, and exposure) and for two time periods (the current situation as baseline and for 2050). This method enabled the experts to assign relative importance to each factor, allowing for equal weighting where necessary.

The median of the assigned budgets for each factor, for both countries and time periods, was normalized by dividing each value by the highest value across both the available present and future datasets. This approach allows to account for changes in the weightings over time. Table 1 lists the factors and respective proxy indicators used, and Figure 3 shows their normalized weightings.

While data from the current census are available for Ghana, an attempt was made to identify corresponding equivalents for Nigeria. The future vulnerability component was calculated using the present-state vulnerability data in conjunction with their future weightings, given the unavailability of gridded or subnational data for future periods for both countries. In the case of demographic pressure, however, future projections were accessible for both countries. This factor was calculated by summing the rural population for 2020 and 2050 per district or LGA, as it is assumed that a high rural population, in particular, reflects the potential scarcity of resources. The aggregated datasets were then normalized using the global minimum and maximum, which is further explained in Section 2.6.

To align with the method proposed by GIZ and EURAC, (2017) and Zebisch et al. (2023), each indicator was rescaled to a consistent range from 0.0 to 1.0, with higher values indicating more negative conditions for livelihoods. For the hazard component, first the indicators were aggregated to the administrative level 2 boundaries (districts in Ghana and LGAs in Nigeria). In case of the hazard component with future predications available, a global normalization approach was used to normalize the aggregated datasets (see Equation 4), utilizing the minimum and maximum values across both the observational and predictive model time layers. This “global” normalization ensures comparability between present and projected future conditions.

Where, X represents the input value, either from the observational dataset X o b s or the predicted dataset X pred , min X o b s , X pred denotes the minimum value across both datasets. The value max X o b s , X pred is the maximum value across both datasets.

Other hazard indicators, which could not be predicted to the future were normalized according to the formula:

In case of a negative direction (when a high indicator value represents a low risk, e.g., in case of available microfinance institutions; see Table 1), the X norm value was subtracted by 1.

As socio-economic factors may vary significantly between urban and rural areas, we first reduced outliers by identifying the 95th percentile. Any values exceeding the 95th percentile were replaced with the value below this threshold. Subsequently, data were aggregated at the administrative level 2 boundaries and normalized according to Equation 5. The normalized hazard and vulnerability factors are displayed in Supplementary Figures S4–S6. To aggregate individual indicators into composite indicators, Zebisch et al. (2023) recommend using a ‘weighted arithmetic aggregation’. This method involves multiplying each individual indicator by its respective weight, summing these products, and then dividing the sum by the total sum of the weights. This process is used to calculate the composite indicator (CI) of a risk component (see Equation 6).

Where CI is the composite indicator for the present or the future (e.g., hazard), I represents an individual indicator of a component, n is the number of indicators, and w is the weight assigned to the indicator. The weights used were identified in the expert consultation (see Section 2.3) and are shown in Figure 3. If no future dataset was available, the present-day data were utilized for the future risk components, with their weighting adjusted based on expert opinions.

A gridded projected population dataset published by Wang et al. (2022) with a 1 km resolution for the years 2020 and 2050 was employed to assess the exposure component. This dataset had been developed using the Shared Socioeconomic Pathways 2 (SSP2) “middle of the road” scenario. Six exposure layers were created for each year to represent different population groups identified by the expert consortium as being particularly vulnerable and more likely to migrate in response to hazards (see Table 1; Figure 3). Urban areas were defined on the basis of the spatial extent of settlements (CIESIN, 2021). Thus, in Ghana urban clusters are defined as grid cells with a minimum population of 5,000 and a density of 500 people per km². For Nigeria, urban areas are characterized by a minimum population of 10,000 and a population density of at least 1,500 people per km². Rural areas were defined as non-urban areas. Farmers were represented by overlaying rural population data with cropland areas, reflecting the assumption that smallholders reside near their farmland. The spatial distribution of the exposure layers is illustrated in Figure 5. After clipping each exposure layer with the population raster, population density was reclassified to a value of 1 for densities equal to or greater than 50 inhabitants per km². This reclassification emphasized population distribution while minimizing misinterpretations from urban density outliers during normalization. The detailed population distribution for each layer and time period is provided in Supplementary Figure S7. Subsequently, these distribution layers were used to calculate the exposure component according to Equation 6. The gridded exposure component was then aggregated to the administrative boundaries using the 99^th quantile, allowing to map the most at-risk population groups and recognize that certain areas, such as drylands or rural regions, may face higher migration pressures from hazards. In order to quantify population growth within a district or LGA, we calculated the percentage difference between 2020 and 2050 using data published by Wang et al. (2022) (see Supplementary Figure S8), enabling more accurate assessments of population exposure.

To assess whether high hazard and vulnerability scores align with migrants’ actual motivations, we analyzed data from national interviews developed and implemented in Ghana and Nigeria by research teams from the University of Cape Coast (Ghana) and the Federal University of Technology Minna (Nigeria). Interviews with migrants, non-migrants and potential migrants were conducted between May and September 2022 in Ghana and between June and October 2022 in Nigeria. For this study, we focused specifically on the responses of migrants. The questionnaire gathered information on socio-economic status, migration histories, and perceptions of climate change, among other factors.

In this study we used responses to the question: “What are the main reasons why you left your most recent place of origin/last destination?,” allowing multiple answers for migration motivations. This was combined with the question” Where was your place of origin before migrating to this current destination?” which provides information about the respondents’ places of origin (see Supplementary Tables S4, S5). For Ghana, 1,265 interviews were available while for Nigeria, 472 interviews were considered.

The spatial patterns of hazard and vulnerability scores was compared with the migrants’ actual motivation. We calculated the proportion of respondents identifying environmental factors as the primary reason for migration to compare with the hazard component. For the vulnerability component, we summed the proportion of respondents citing “job opportunities,” “access to markets,” and “education” as their main migration motivations, with “insecurity” also included for Nigeria. For each region (Ghana) or state (Nigeria), we identified the highest hazard and vulnerability scores and plotted them against the normalized motivation scores. This method enabled both visual and quantitative evaluations of how well perceived migration motivations aligned with the calculated risk components. In order to compare the risk scores with actual migration rates, we plotted the current net migration rates for Ghana from the Population and Housing Census (Ghana Statistical Service (GSS), 2023) against the maximum risk scores per region. As this data was not available for Nigeria, we have included the outcome for Ghana in Supplementary Figure S9.

A total of 36 individual factors were identified during the expert consultation, of which 24 are included in the analysis due to data availability (Figure 3). Among these, nine factors related to hazards were identified. The Ghanaian experts considered ‘Decrease in average rainfall’ to be similar to drought, leading to the inclusion of only the drought indicator for Ghana. High rainfall variability in Ghana and high rainfall events in Nigeria were considered to have a highest impact on the agriculture-dependent population; in the present and in the future. For vulnerability factors, ‘conflict prone areas’ was mentioned only by the Nigerian experts, while ‘limited preparedness for emergencies’ was only highlighted for Ghana. Conflict is seen as having the greatest impact on the decision to migrate in Nigeria at present and in future, while in Ghana the reliance on agriculture is attributed to a high vulnerability. In Nigeria, people living in coastal and flood-prone areas are expected to be more exposed in the future, while in Ghana, people living in rural and arid areas are expected to face higher levels of exposure.

For reasons of comprehensibility, the results are discussed at administrative level 1 (regional level for Ghana, Figure 6, and state level for Nigeria, Figure 7), although they have been visualized at administrative level 2 (district level for Ghana and LGA level for Nigeria). The highest hazard scores (≥0.6) in the current assessment are observed in districts located within the Upper East, Upper West, Northern East, and Northern Region. In contrast, the lowest hazard scores are concentrated in the central regions, including Ashanti, Eastern, and Central Region. Overall, hazard scores are predicted to increase across all districts in the future, except in some districts at the coast. The greatest increase is expected in the central regions, attributed to an increase of consecutive dry days and heavy rainfall events as well as a later onset of the rainy season. Highest vulnerability scores are observed in the Northern, Northern East, Savannah and Bono North regions, with values above 0.7, reflecting a high degree of negative impact of pre-existing adverse socio-economic conditions and low adaptive capacity of people. Factors contributing to this high vulnerability include a large agricultural workforce, limited access to education, and unfavorable soil conditions, specifically low organic carbon content. Conversely, the Greater Accra and Ashanti Regions exhibit the lowest vulnerability, or, more specifically, the highest adaptive capacity, due to better access to education and the presence of microfinance institutions. While most of the input data for both present and future vulnerability scenarios remain consistent, changes in the weighting of variables contribute to a shift in vulnerability patterns. Although the future vulnerability is distributed in a similar way to the current situation, a closer look at the differences reveals an upward trend, especially in the central regions. This is attributed to increased rural population.

Exposure is most pronounced in arid and rural areas, resulting to highest exposure levels in northern Ghana. These regions have a more arid climate and a higher proportion of arable land, which contributes to their higher exposure levels. Low exposure in the Savannah region is attributed to the low population distribution. While exposure in the northern regions is increasing, some coastal districts are experiencing a decline in exposure due to urban expansion, as urban areas are classified as less exposed compared to rural regions. An analysis of the percentage of difference in the population between 2020 and 2050 for Ghana and Nigeria (see Supplementary Figure S8) reveals a notable population increase in the central and southern regions of Ghana, particularly around districts that are already urbanized. In northern regions where an increase in risk has been calculated, the percentage difference is not as high as in the rest of the country, but still within a 20–40% range.

Finally, the risk component shows the combination of all three components, indicating high risk scores in the northern regions of Ghana, especially in the Upper East and Northern East Region. Those regions maintain their high-risk status in the future, while some districts within these regions are experiencing even higher scores. Risk scores are increasing in the majority of the regions, especially in the Upper West and Savannah regions.

In Nigeria, there is, like in Ghana, a north–south gradient in hazard scores, with generally higher scores in the northern regions compared to the south (Figure 7). In the present scenario, Sokoto, Kebbi and Borno states display the highest hazard scores, which are projected to increase due to higher temperature and more heavy rainfall events (see Supplementary Figure S3). Both factors received a high weighting in the future by the experts. Some states in Nigeria show a low decrease in hazard scores in future. The observed decline in the map of difference for hazard in parts of the Sudano-Sahelian Zone can be attributed to the relatively low number of hot days and an accompanying lower increase, in comparison to other regions of the country. The relatively low values observed along the coast can be attributed to the fact that temperatures do not rise as high, and heavy rainfall events do not increase as much, as they do in the northern regions. Additionally, these areas typically experience higher precipitation levels than the northern regions, which are expected to increase slightly in the future. High vulnerability scores are most pronounced in Borno and Zamfara, with Borno being particularly affected by a high number of armed conflicts—and being a factor of socio-economic sensitivity. An increase in vulnerability is predominantly observed in the northern states, while the reduction in Borno’s vulnerability is primarily attributed to a lower weighting of conflicts for the future by the experts compared to the present (see Figure 3). In general, the majority of vulnerability factors are assigned with a higher weight, which leads to an overall increase of vulnerability scores in the future scenario. The northern regions exhibit the highest levels of exposure, which can be attributed to the high proportion of people living in arid or semi-arid areas and of population being engaged in agricultural activities. Furthermore, elevated levels of exposure are observed in certain coastal states like Lagos, Delta or Bayelsa, a pattern that is anticipated to intensify in the future due to population growth. On the contrary, in the central and northern part of Nigeria, the exposure score is predicted to decrease slightly. In Nigeria, the population growth is generally higher than in Ghana (see Supplementary Figure S8). It is notable that the north is experiencing a particularly strong increase in population, e.g., in Kano and Katsina state. This growth is accompanied by an overall increase in risk in these regions, driven by an increase in hazards and vulnerability. By 2050, the population of some southern LGAs is expected to double, contributing to higher exposure scores as more people inhabit areas with elevated exposure risks. The states with high overall risk in both the present and future remain largely unchanged, with further intensification in Yobe and Kebbi State. In contrast, some states in the southern part of Nigeria, such as Oyo, Ondo, and Cross River, are expected to experience a decrease in risk.

The relationship between present hazard and vulnerability scores and migrants’ motivations is illustrated in Figures 8, 9 for Ghana and Nigeria, respectively. It is important to note that, overall, relatively few migrants cited environmental factors as their primary reason for migration. In Ghana, the highest percentage of respondents citing environmental reasons within any region was 14.5%, while in Nigeria, the highest percentage within a state was 33.3%. In contrast, socio-economic factors related to vulnerability were cited more frequently, with the highest percentage being 74.5% in Ghana and 100% for Nigeria. This analysis aims to determine whether regions or states where migrants more frequently cited environmental or vulnerability related factors correspond to those with higher hazard or vulnerability scores.

In Ghana, migrants more frequently identified environmental factors as a primary reason for relocating from regions with high hazard scores (≥0.6). Conversely, migrants from regions with lower hazard scores less often attributed their migration to environmental conditions. Socio-economic factors are prevailing reasons for migration in regions characterized by both high vulnerability and hazard scores. This suggests that economic factors often play a crucial role in migration decisions, especially in regions facing environmental challenges.

In the case of Nigeria (Figure 9), states with high hazard scores, such as Borno, Sokoto, Zamfara, and Yobe, environmental factors did not exhibit a higher frequency of migrants citing environmental factors as their reason for migration. Instead, socio-economic factors were identified as key motivations for migration. In states such as Borno and Sokoto, where both high vulnerability scores and high hazard scores were calculated, socio-economic factors were identified as the primary motivation for migration, rather than environmental factors. This indicates that environmental-related stressors may not be perceived as direct threats by the population, but rather as contributors to deteriorating economic conditions, which subsequently drive migration. The indirect impact of environmental changes on livelihoods likely results in people prioritizing economic motivations over environmental ones when explaining their migration decisions. This underscores the complex interplay between environmental stressors and socio-economic conditions in shaping migration patterns.

The risk assessment framework is able to account for the interplay of environmentally related hazards, vulnerability (especially socio-ecological sensitivity and low adaptive capacity) together with exposed population groups. The study intends to quantify the impact of external factors influencing rural out-migration by integrating spatial data in order to identify areas where migration is more likely to occur. However, it is important to acknowledge that numerical values of thresholds are not absolute reflections of reality. Migration decisions may occur in regions independently from counted data or numerical thresholds, while some individuals may choose to stay in areas even though these areas are scientifically classified as high-risk areas. While adverse climate events can create conditions that lead individuals or communities to consider relocation, their decision to stay is often influenced by a range of factors. For example, many vulnerable communities use local coping strategies, such as diversifying livelihoods or adapting agricultural practices, to address climate-related challenges and reduce the need to migrate (van der Geest and Warner, 2015). In addition, strong community ties and deep cultural attachments may support a preference to remain in place despite certain risks (Kutor et al., 2025). Nevertheless, the method provides a systematic approach to linking spatial datasets and projecting shifts between current and future migration patterns. An increase in the risk score does not necessarily indicate that a greater number of people will migrate from these regions in the near future. Rather, it suggests that living conditions in these regions are deteriorating due to factors such as increased rainfall variability or greater resource pressure from a growing rural population, which is further compounded by existing socio-economic vulnerabilities. Nevertheless, this could also indicate that rural-to-urban migration may intensify in the future, placing greater strain on urban areas.

The results of our analysis for the present state are largely consistent with statements from migration studies, which indicate that areas in northern Nigeria and Ghana are the primary areas of internal out-migration (Alarima, 2019; Ango et al., 2014; GSS, 2023; Schürmann et al., 2024; van der Geest, 2011). Main destinations for internal migrants in Nigeria are often economically vibrant states such as Lagos and the Federal Capital Territory (FCT) around Abuja, which attract people seeking better job opportunities and living conditions (UN, 2023). For most LGAs in Lagos State, we computed an increase in vulnerability and exposure scores due to the growing rural and urban population, as well as the state’s coastal location, which will be more vulnerable to floods and rising sea level in the future (Adegun, 2023). Most internal migrants in Ghana move to Ashanti or Greater Accra Region (GSS, 2023), both of which have low risk scores. However, it should be noted that this study does not account for coastal flooding and erosion, which are expected to increase in the future and could impact coastal cities like Lagos or Accra (Rigaud et al., 2021).

The ongoing instability in northern Nigeria due to terrorist activities poses an additional layer of complexity. Although more rainfall is projected for the Sahel zone (Almazroui et al., 2020; Stanzel et al., 2018; Weber et al., 2023), which could improve yields for certain crops as cassava, groundnuts or rice, but at the same time potentially decrease yields for maize and millet (Tomalka et al., 2021). Even with potentially better agricultural conditions in the Sahel, the socio-political factors may continue to drive vulnerability and therefore rural out-migration. Studies have shown a correlation between increased rainfall and reduced conflict in communities. However, as rainfall becomes more variable and adaptation to these changes becomes more difficult, the potential for conflict may increase (Coulibaly and Managi, 2022; Nordkvelle et al., 2017). In addition, an increase of precipitation can lead to higher malaria transmission due to more occasions of open, stagnant water and higher moisture (Jambou et al., 2022). More frequent hot days and greater variability in precipitation challenges for water management and agricultural stability (Berger et al., 2021; Röhrig et al., 2019), exacerbating existing vulnerabilities in already economically weak regions in northern Ghana and Nigeria, which could influence migration dynamics. However, it is not possible in this study to make any statements regarding a potential improvement or deterioration in socio-economic sensitivity or adaptive capacity in future. An improvement of the adaptive capacity or reduction of socio-economic sensitivity could lower future risk values in our assessment. Furthermore, different weightings are applied to the future components which also affect the risk score. This highlights the inherent uncertainty in predicting local migration patterns (de Valk et al., 2022).

Nonetheless, it is important to acknowledge that individual aspirations and capabilities influence migrants’ decisions, shaping their responses to environmental and socio-economic factors (Adger et al., 2024; De Haas, 2021). In addition to these personal drivers, the role of government policies is also critical, as highlighted by the experts in Nigeria. Despite its potential to improve the lives of vulnerable populations, few West African governments incorporate migration into climate adaptation plans. Due to limited capacity to manage urban growth and infrastructure, policies often discourage rural-to-urban migration (Farrell, 2018; Teye and Nikoi, 2022). In order to address the challenges of rural–urban migration, policies should aim to create rural employment opportunities and reduce pressure on urban infrastructure. Actions to promote sustainable rural development and thus sustainable rural–rural and urban–rural migration should focus on agricultural resilience, land accessibility and diversification of the rural economy sector. Governments could expand sustainable agriculture programs to provide farmers with climate-smart techniques, access to credit and improved irrigation systems to reduce the current and future risks associated with climate-related hazards and socio-economic vulnerabilities.

The factors identified as highly relevant by the experts, including ‘high rainfall variability’, ‘drought’, ‘dependence on agriculture’, ‘limited access to microcredit’ in Ghana, and ‘heavy rainfall’ and ‘conflict’ in Nigeria, are also identified in recent literature as important determinants of migration decisions (Teye and Nikoi, 2022). These stressors are of particular importance in agricultural communities where livelihoods are closely intertwined with environmental conditions. For example, Rigaud et al. (2021) demonstrate that climate-induced changes, including variability in rainfall and increased drought frequency, play a direct role in migration, as these conditions lead to a reduction in agricultural output and economic instability. Additionally, heavy rainfall events and subsequent flooding have been demonstrated to displace large populations, thereby highlighting the critical role of extreme weather events in migration decisions in Nigeria (Ibrahim and Mensah, 2022; Rigaud et al., 2021).

Although environmental factors are often highlighted as drivers of migration, it is essential to recognize their interaction with pre-existing socio-economic vulnerabilities. Kaczan and Orgill-Meyer (2020) conducted a systematic review and propose that environmental factors should be regarded as contextual rather than primary drivers of migration. This is also evident in our analysis, which shows that in Ghana the reasons for environmental migration are consistent with the characteristics of the region’s geography, suggesting that external factors such as environmental hazards may influence the decision to migrate. In Nigeria, by contrast, socio-economic factors appear to dominate, even in areas that are highly vulnerable to environmental hazards.

Conflicts in Nigeria not only drive rural–urban migration but also exacerbate food insecurity (Ayuba et al., 2023), which is also identified by the experts as having major influence on migration decisions in the future. In addition, financial capacity plays a crucial role in migration decisions. Research indicates that it is often wealthier households that have the means to migrate in response to climate change (Duijndam et al., 2022; Hirvonen, 2016; Kaczan and Orgill-Meyer, 2020). Limited access to microcredit, e.g., to buy necessary inputs for agricultural production on loan basis, further constrains adaptive capacities (Twumasi et al., 2020).

Regarding different population groups, rural populations, particularly farmers, are more exposed to these stressors than urban populations. This is undermined by the ongoing trend of rural–urban migration in Ghana and Nigeria (Abbass, 2012; Dick and Schraven, 2021). This migration trend is not only a response to immediate environmental stressors but also a reflection of deeper structural inequalities in access to resources and opportunities between rural and urban areas.

In addition to these structural factors, historical gender roles have influenced migration patterns in West Africa, with men more likely to migrate for employment opportunities and women more likely to migrate for family reunification or marriage. However, this trend is changing as more women are choosing to migrate independently (Setrana and Kleist, 2022). Age is also an important factor in migration decisions, with younger people generally more likely to migrate in search of better education and employment opportunities (Alarima, 2019; Ango et al., 2014). Furthermore, historical pre-colonial trade networks, ethnic ties and shared official languages continue to shape current migration cultures and cross-border migration patterns (Teye, 2022).

The presented framework integrates the aforementioned environmental factors, socio-economic conditions, and the vulnerability of exposed population groups, addressing the complex and ongoing debate about the interplay between environmental stressors and socio-economic vulnerabilities in shaping migration patterns, particularly in rural areas. This approach enables a more nuanced analysis of migration dynamics and helps identifying areas where multiple risk factors coincide.

The study primarily focuses on external factors, such as environmental hazards, socio-economic vulnerabilities, and population exposure, with only limited consideration of individual perceptions. It does not account for personal factors such as social capital or cultural attitudes. Additionally, a more differentiated analysis of especially highly exposed demographic groups, including women and youth, could not be conducted due to the lack of data. This omission may underrepresent the social dynamics that influence migration decisions. Besides, the study focuses on rural migration because many factors, particularly those related to environmental conditions, are most relevant to rural populations dependent on agriculture. As such, migration from urban areas, which is also prevalent in Ghana and Nigeria, was not examined in detail.

The weighting process with experts introduces a certain degree of bias, reflecting the subjective judgments and potential limitations of the experts’ perspectives even though expert consultation proves a higher reliability in statements than nonexperts (Han and Dunning, 2024). Such bias may influence the relative importance assigned to different indicators, potentially skewing the results toward certain vulnerabilities while underrepresenting others. Furthermore, the identification of factors and their proxy indicators have an impact on the results. However, the weights assigned mainly correspond to recent literature (see Section 4.1) and the selection of proxy indicators was restricted to the data available.

The normalization of data, required to integrate multiple datasets with varying scales, introduces an additional layer of uncertainty. Different normalization techniques, or the use of alternative quantiles to address outliers, particularly for vulnerability factors, could alter the data range and subsequently influence the overall results. Conversely, normalization facilitates interpretation and enables a wide range of data to be combined. In particular, the use of global minimum and maximum values permits the illustration of changes between the present and the future.

For the estimations of the future risk component, the study relies on climate and population projections under the RCP4.5 scenario, which can vary depending on the models and assumptions used. In particular, rainfall projections are more challenging to model accurately than, for instance, temperature data (Tomalka et al., 2021). To reduce the degree of uncertainty of the climate projections, ensemble means were employed. Although climate models predict increased rainfall in certain regions, particularly in the Sahel of Nigeria, uncertainty remains regarding the distribution of this precipitation throughout the year and its impact on rain-fed agriculture. It is also evident that the selected CMIP6 models underestimate heavy rainfall events and overestimate consecutive dry days in Ghana. Conversely, they demonstrate comparable patterns for the climate indices in Nigeria. On the other hand, the CMIP6 models seem to be able to estimate annual precipitation and extreme temperatures. Additionally, population data for Nigeria are highly uncertain, as there has not been an official population census for almost two decades.

The risk framework employed in this study is a simplified representation of the complex interactions between environmental hazards, vulnerabilities, and exposure. While simplification is necessary for modeling and analysis, it also means that certain feedback loops and dynamic interactions are not fully captured. To illustrate, the framework does not incorporate potential feedback mechanisms where increased migration could either mitigate or exacerbate local vulnerabilities, according to the context. A distinction between different types of migration may also be crucial. This is particularly true for Ghana and Nigeria, where temporal and seasonal migration is common alongside permanent migration. To identify potentially vulnerable areas, we used the responses of individuals who had already left their place of origin, without further distinguishing between different types. This could be an important consideration for more in-depth analysis. Furthermore, although adaptive capacity is included as a component of vulnerability, the framework may not fully represent the range of adaptive strategies or the potential for communities to develop new adaptive capacities in response to changing conditions.