https://orcid.org/0000-0002-8466-4343
Figures
Abstract
Climate variability and climate change are among many interacting drivers of human migration, alongside social, geopolitical, and economic factors. Environmental stressors such as crop failures, rising sea levels, and water insecurity may contribute to mobility, but their influence is complex, indirect, and highly context-specific. Populations in Sub Saharan Africa are particularly vulnerable due to high exposure to climate change and limited adaptive capacity. Although migration patterns in Africa are increasingly well documented, empirical evidence directly linking long-term migration trends to specific climate impacts, such as crop yield variability, remains limited. In this study, we analyzed longitudinal data from 196,320 individuals in rural Burkina Faso, collected through a Health and Demographic Surveillance System from 1994 to 2016. We used Prentice-Williams-Peterson regression to assess the association between annual weather-induced crop yield variations and individual-level migration events. We found that reductions in crop yields were strongly associated with increased out-migration, particularly among male farmers, individuals with lower household wealth, and those with prior migration experience. These findings highlight the role of climate related livelihood impacts on shaping migration patterns and highlight the importance of effective climate adaptation strategies that account for migration dynamics in vulnerable settings.
Citation: Belesova K, Zabre P, Opata M, Sie A, Sauerborn R, Nayna Schwerdtle P, et al. (2026) Field to flight: Migration dynamics amidst climate driven crop yield fluctuations in Burkina Faso. PLOS Clim 5(3): e0000832. https://doi.org/10.1371/journal.pclm.0000832
Editor: Noureddine Benkeblia, University of the West Indies, JAMAICA
Received: March 3, 2025; Accepted: January 23, 2026; Published: March 5, 2026
Copyright: © 2026 Belesova et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Data Availability: The HDSS data underlying this study cannot be made publicly available due to ethical and confidentiality restrictions related to human participant data. The data are subject to oversight by the Centre de Recherche en Santé de Nouna (CSRN), Burkina Faso, which imposes these restrictions to protect participant privacy and comply with institutional and national ethical requirements. Data access requests may be directed to the Centre de Recherche en Santé de Nouna (CSRN) via the institutional email address: info@crsn-nouna.bf. Requests for data will be reviewed by the appropriate institutional body in accordance with CSRN data governance policies. Institution: Centre de Recherche en Santé de Nouna (CSRN), Burkina Faso Phone: +226 20 53 70 43 Website: Centre de Recherche en Santé – Burkina Faso.
Funding: JR, RS, PNS, MR, PS, AS are funded by the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation) – Project number 409670289. KB is supported by the NIHR Centre on Non-communicable Diseases and Environmental Change (NIHR203247).
Competing interests: The authors have declared that no competing interests exist.
Background and literature review
Both sudden- and slow-onset impacts of climate change—including rising temperatures, increased precipitation variability, flooding, and drought—are projected to intensify pressures on mobility patterns in many parts of the world [1]. Climate-related environmental change is described as one of the multiple interacting drivers of migration and displacement, [1] particularly affecting regions with high vulnerability and low adaptive capacity [2]. The Intergovernmental Panel on Climate Change (IPCC) emphasizes that climate change is rarely the sole driver of mobility and that migration decisions arise from the interaction of environmental stressor with economic, social, political and demographic factors, as well as individual and household agency [1]. Saharan Africa has been identified as a region where climate-related mobility pressures may intensify due to high exposure to climate hazards and limited scope for effective adaptation [2]. With high agreement and medium evidence, the IPCC projects that by 2050 there could be an additional 17–40 million internal migrants under 1.7°C warming and 56–86 million under 2.5°C warming in Sub-Saharan Africa alone [3]. Such increases may create socio-economic challenges for both sending and receiving communities [4,5], although migration can also function as an adaptive strategy that helps reduce climate-related risks, including adverse health outcomes. Existing research on climate change and migration in Africa highlights the complex and context-specific interactions between climate variability and change, food security, and human mobility, shaped by intersecting environmental, economic, and social factors that influence both exposure and response options.
In Burkina Faso, migration patterns are shaped by a combination of socio-demographic factors (such as age, gender, household size, income and wealth) and environmental factors (such as rainfall variability, temperature extremes, and declining crop yields). Economic opportunity and environmental stress play important roles in mobility decisions [6] with individuals weighing both push and pull factors under conditions of uncertainty. Agent-based models further emphasize that migration is mediated not solely by climate stressors but also by socio-political dynamics and governance structures [7]. Historical analyses of droughts in West Africa reveal that migration often functions as a livelihood diversification strategy, though its success depends on access to resources and is not without risks such as loss of land rights due to prolonged absence, reduced household labor availability, and increased vulnerability and exploitation or poor working conditions as destination sites [4]. These risks are closely linked to household resource access, as poorer households may lack the means to migrate safely or to benefit from migration outcomes [8]. Economic research indicates that climate variability is more likely to drive internal and regional migration than large-scale international migration [9], particularly in agriculturally dependent low- and middle-income countries in Latin America and Sub Saharan Africa [10].
Climate variability and extreme weather are contributing to both displacement and reduced farm productivity in Burkina Faso. Recurrent droughts, decreasing rainfall, and rising heat have exacerbated land degradation and resource scarcity, which in turn drives internal migration and displacement [6]. For example, decades of desiccation (chronic or cumulative drying over time) in northern Burkina Faso have led to soil depletion and large-scale out-migration toward the south [6]. At the same time, acute-onset climate hazards – especially flooding – have caused significant sudden displacements across all regions of the country [11]. A so far underexplored climate-migration pathway is climate change impacts on agricultural yields. Rising temperatures are linked to lower sorghum yields, Burkina Faso’s staple cereal [12]. Projections further indicate that unmitigated climate change will depress crop production. One modeling study found sorghum yields could decline by ~5% nationally by late-century under a high-emissions scenario, with losses up to 35% in parts of the country [13], with implications for food security, agricultural livelihoods and migration patterns.
The relationship between environmental shocks and migration is complex and highly context-dependent. In Uganda research showed that weather shocks can reduce temporary migration among poor and smallholder farmers [14]. This decline in mobility is attributed to falling agricultural productivity and low farm revenues, which limit households’ capacity to finance migration as a coping strategy. These findings challenge the widespread assumption that environmental stress necessarily increases migration, highlighting instead how poverty and vulnerability can produce immobility or “trapped” populations. Similarly, a recent review found that while forced migration and environmental degradation are linked, their relationship is highly variable across different settings [15].
These empirical insights have important implications for policy. Maystadt et al. (2024) emphasize the potential for targeted interventions - such as social safety net programs - to enable migration as a proactive risk management strategy [15]. In addition, such measures can help discourage environmentally harmful practices in destination areas. Taken together, the evidence points to the need for context-sensitive policy responses that address both the drivers of mobility and the structural constraints that limit adaptive movement in response to climate-related stress.
Furthermore, other studies have identified non-linear migration responses to climate hazards, where adaptation strategies fail once critical thresholds or tipping points are crossed, resulting in abrupt displacement [16]. Climatic shocks - manifested as anomalies in temperature and precipitation - have been shown to negatively affect productivity, health outcomes, and economic growth, all of which play an important role in shaping migration decisions [17]. In response, agricultural adaptation strategies, including integrated soil fertility management, irrigation and agroforestry, have demonstrated potential to enhance resilience and mitigate climate-induced migration [18]. However, methodological challenges continue to limit our understanding of the interlinkages between climate, food systems, migration and health. Meta-analyses and systematic reviews identify persistent difficulties in measuring migration and climate variables, integrating diverse datasets, and isolating causal relationships [10].
This study builds on our previous research, which provided empirical evidence of the association between inter-annual crop yield variation and both child survival and nutritional status in a rural Burkinabe population [19]. In that work, we established that yield variation in this setting is largely driven by weather conditions [19]. Here we investigate how these same yield fluctuations influence out-migration that exceeds 2 months. Our study examines the nexus between climate change, crop yield, food security, and migration, framing both food security and migration as key social determinants of health. While the empirical focus is on the association between weather-induced crop yield variation and out-migration, the analysis is informed by an interdisciplinary perspective that integrates insights from agricultural science, demography, economics, epidemiology, and public health. This approach highlights how climate-related disruptions to agricultural livelihoods influence migration decisions and, in turn, affect population health and well-being. We adapted the conceptual framework of Tuholske et al. (2024) to analyze the climate-food-migration nexus, illustrating how local climate variability affects agricultural productivity and shapes rural household decisions to migrate in response to food insecurity and shifting economic opportunities [20] (See Fig 1). By leveraging a detailed, continuous individual panel dataset, we offer precise and temporally sensitive estimates of the climate-migration relationship and describe how the impacts are interacting with socio-economic and demographic conditions.
Adapted from Tuholske et al, 2024. Legend: Blue unbroken line: Confounders. Blue double link: Mediating factors.
Focusing on agriculture as a key pathway in linking climate variability to migration, this study contributes new insights into how environmental shocks drive mobility. We also examine how gender, household wealth, and prior migration experience modify these responses, enhancing our understanding of differential vulnerabilities. Our research aims to answer the question - How do weather-induced fluctuations in crop yields influence out-migration patterns in rural Burkina Faso, and how do these effects vary by gender, wealth, and prior migration experience?
Materials and methods
Study area and population
The study was conducted in the Nouna Health and Demographic Surveillance System (HDSS) site, located in Kossi Province in north-western Burkina Faso. This predominantly rural area, comprising 58 villages and one town, is reliant on rain-fed subsistence farming with a single agricultural season per year. Past studies in this setting suggest that about 90% of what households eat has been grown and harvested in their own fields (20A). The population has been under longitudinal surveillance since 1992, increasing from 26,626 to approximately 125,000 by 2022, due to natural growth, village incorporation (in 2000 and 2004), and in-migration. Over this period, no notable changes have been observed neither in the prevalent practice of manual, low input, low-intensity agriculture nor to the reliance of the local population on subsistence agriculture. The area has been affected by conflict after 2016 and mostly from 2019. Our study captures a conflict-free period.
Data sources and study design
We used Nouna HDSS data from 1994 to 2016, excluding 1992–1993 due to data incompleteness and post-2016 due to interruptions in surveillance continuity. HDSS longitudinal surveillance is structured as a dynamic open cohort, where all individuals and their households are assigned unique identification numbers that allow linking their demographic and health events over time. HDSS data collection included quarterly (later 4-monthly) house-to-house surveys registering all births, deaths, migrations, and other vital events (e.g., an individual moving to or forming another households), complemented by periodic full socio-economic censuses. Records of all events are time-stamped with an indication of the level of precision of the date of every event. For this study, we used monthly (or finer) precision of all records of event timing.
We used a time-to-event (survival) study design [21], which is well suited to analysing mobility outcomes in longitudinal cohort data. This approach accounts for the precise temporal ordering of exposures and outcomes, allows estimation of associations with recurrent events such as repeated migration episodes, and appropriately handles censoring, thereby avoiding survival bias (the exclusion of individuals who do not remain in the cohort for the full observation period). By applying this design to 23 years of HDSS data, we were able to align the timing of agricultural harvests and migration events with monthly precision, ensuring correct temporality and ruling out reverse ordering between yield variation and out-migration. The use of continuous measures for both exposure (FCPI) and outcome enabled us to examine association gradients and heterogeneity across occupational, demographic, and socio-economic subgroups while adjusting for potential confounders. Taken together, these design features strengthen the plausibility and interpretability of the observed associations.
Our primary outcome was out-migration for yield-sensitive reasons: work, crop cultivation, and livestock pasture. Presence of individuals leading up to migration events that are unrelated to yield (e.g., marriage, education) was included in the time at risk but censored upon migration. Events with unknown yield sensitivity (e.g., health reasons, following family) were excluded from main analyses. Migration was defined at the individual level as any absence from the HDSS area lasting more than two months.
Exposure variables
Exposure to agricultural productivity was measured using the Food Crop Productivity Index (FCPI), derived from province-level crop yield data (millet, sorghum, maize, fonio, rice) collected via field measurements of crop yield by Burkina Faso’s Agricultural Statistics Service (1994–2016) using the crop cut estimation method [22]. This is recognized as a near gold-standard method of yield estimation, offering greater accuracy than such proxies of crop yield as rainfall and vegetation indices derived from satellite imagery [23]. FCPI values express annual yield as a percentage of the 1992–2012 average. Lower FCPI values reflect yield deficits linked to climate variability. Two exposure windows were analyzed reflecting different lags between the exposure and outcome:
- Preceding harvest – FCPI in the agricultural season prior to each observation period.
- Cumulative FCPI – mean FCPI over the three preceding harvests, reflecting medium-term economic conditions.
We previously estimated that 72% of variation in FCPI can be explained by variation in weather parameters of physiological significance to crop growth [24].
Covariates
Covariates were derived from HDSS individual, household, and village-level data:
- Demographic and socioeconomic variables: age, sex, literacy, religion, ethnicity, occupational status.
- Occupational relevance to agriculture: categorized as (a) crop farming, (b) other agri-food work, (c) non-agricultural work, (d) unoccupied/unemployed, or (e) unknown.
- Household wealth index: based on 2009 data on housing and asset ownership, dichotomized above/below the median.
- Village infrastructure: an index constructed via principal component analysis (e.g., presence of health facility, market, road type).
- Migration history: order, permanence (return vs non-return), and duration (<1 year or ≥1 year) of out-migration events.
Statistical analysis
Out-migration was modeled as a recurrent event to reflect that individuals migrate multiple times during their lifetime. Subjects were at risk from birth but included from September 1, 1994. We applied Prentice-Williams-Peterson models – time-to-event (also referred to as survival) analyses models that allow modelling recurrent outcomes [25]. We used age as analysis time to estimate associations between FCPI and migration, which inherently controls the analyses for age, and adjusted for intra-subject correlation. Cumulative hazard plots were used for visual comparison of high vs low FCPI exposure. Models were adjusted for time-invariant individual and contextual confounders (ethnicity, religion, literacy, village type, market access, infrastructure index) and a linear time trend to account for secular changes as fixed effects. Effect modification, i.e., differences in the association between FCPI and out-migration, was assessed for sex, household wealth, agricultural occupation, and migration order, including two-, three-, and four-way interaction terms based on improvements in model fit (Akaike Information Criterion) and Wald test significance.
Sensitivity analyses
We conducted three sensitivity analyses:
- I. including all migration events regardless of reason;
- II. restricting to original HDSS villages (excluding those added in 2000/2004);
- III. restricting to individuals born after September 1, 1994 (with complete migration history).
Details are provided in Table A, Table B, Table C in S3 File. Statistical analyses were performed using Stata 17 (StataCorp LLC, College Station, TX, USA).
Ethics statement
This study involved human participants and received ethical approval from the Ethics Committee of the Medical Faculty of Heidelberg University, Germany, and the Comité Institutionnel d’Éthique du Centre de Recherche en Santé de Nouna, Burkina Faso. The approval reference number is 2022-11-240. Written informed consent was obtained from all participating households enrolled in the ongoing dynamic longitudinal cohort. All participants were adults over the age of 18; no children were involved in the study.
Results
Characteristics of out-migration
The average out-migration rate in our dataset over 1994 until 2016 was 45 out-migration events per 1,000 person-years of observation. The out-migration rate was highest in the age group of 18 - < 30 years, among those from the wealthiest households (measured using the wealth index of household assets and housing quality), and in villages with the highest level of infrastructural development (i.e., villages with highest presence of health-care facilities, drilled water wells, markets, and quality road connections).
The vast majority of observed migration events were singular for each individual; only 8% of individuals experienced repeated out-migration (up to five episodes per person). Reasons classified as having ambiguous yield-sensitivity, (health reasons, returning to parents for reasons other than giving birth, following family or someone else, and reasons recorded as unknown or “other”) –were the most prevalent. Out-migration for yield-sensitive reasons such as work or cultivation accounted for 24% of events, while non–yield-sensitive reasons (marriage, divorce, study, household formation, and funerals) accounted for 15%. Most out-migration events were recorded as permanent (78%), although this likely reflects limitations in the HDSS system’s ability to identify returning migrants. Among temporary migrants, long-term episodes were more common than short-term episodes, which may also be influenced by data collection constraints. The median duration of temporary out-migration was 2 years (p10–p90: 1, 6). The median age at out-migration was 17 years (p10–p90: 4, 34). Tables 1 and 2 summarize characteristics of the study population and out-migration events.
Characteristics of crop yield variation
Inter-annual yield variability was expressed by the Food Crop Productivity Index (FCPI), with 100% FCPI representing the mean yield level over the study period [24]. The median FCPI values over the two exposure windows of interest were 103 (p10-90: 81, 118) % and 102 (p10-90: 85, 116) % for FCPI over the preceding years and cumulative exposure windows, respectively. Further details on FCPI variation in this area are reported elsewhere [24].
Association of out-migration with crop yield variation
Out-migration was consistently higher among those exposed to below-average FCPI than those exposed to above-average FCPI. This pattern applied to both – FCPI in the single preceding year and cumulative FCPI over the past three years (Fig 2).
The red solid line shows the cumulative hazard in those exposed to FCPI< 100%; the blue dashed line shows the cumulative hazard in those exposed to FCPI ≥ 100%. Red/blue shaded areas show the 95% confidence intervals, correspondingly.
For every combination of sex, relevance of occupation to agriculture, household wealth level, and order of migration event for the migrating individuals, we present hazard ratios (HRs) of out-migration in relation to reductions in FCPI from its 90th to 10th centile, i.e., the relative change in the rate of out-migration adjusted for potential confounders at the individual, household, village levels and across time.
We found evidence for association between FCPI and out-migration for some but not all combinations of these characteristics (See Table 3, and Table A, Table B, Table C, Table D, Table E in S2 File.
Of the occupational categories, we found the association among farmers, particularly male, and the unoccupied and no association among those in non-agricultural occupations. For those occupied in the wider agricultural and food sector, the association only held for migrants’ second out-migration events (See Table B in S2 File).
The magnitude of the association (assessed by the value of the central estimate of the hazard ratio) was consistently higher among men than women, among those with lower than higher household wealth, for migrants’ second and third than for their first, fourth or fifth out-migration (the latter two were rare, limiting statistical power), and in relation to the last years’ than cumulative FCPI.
The values of hazard ratios for groups of all combinations of these characteristics are reported in See Table 3, and Table A, Table B, Table C, Table D, Table E in S2 File. To illustrate the interpretation, for the first out-migration, the highest hazard ratio was found in relation to the preceding years’ FCPI among the poorer male farmers: 1.54 (95% CI 1.41, 1.69) and the unoccupied: 1.60 (95% CI 1.41,1.81) for a 90th to 10th centile reduction in FCPI. This translates into 84% (95% CI 62, 109%) increase in the rate of out-migration among poorer male farmers in the year of the lowest observed crop yield level (2000) as compared to the year with the highest observed yield level (2015), or 47% (95% CI 36, 59%) if the year 2000 is compared to the period average yield over the study period.
Discussion
This study provides empirical evidence linking weather-induced fluctuations in crop yields to out-migration patterns in rural Burkina Faso over a 22-year period. Our findings demonstrate a clear association between reductions in the FCPI and increased out-migration rates, particularly among poorer male farmers and individuals without stable employment. The differential effects observed across gender, wealth levels, and migration history highlight the complex interplay of socio-demographic factors in shaping climate-related migration decisions.
Interpretation of findings
In the Burkinabè subsistence-farming population relying on rain-fed agriculture, years of poor crop yield were significantly associated with increased rates of out-migration for work and farming elsewhere. This finding is consistent with some studies in other ecological contexts. For example, in Mexico, each additional month of drought increases the odds of rural-urban migration by 3.6% [26]. Similarly, a study in Bangladesh estimates that increased rainfall uncertainty would raise net out-migration rates by 20% in 2030, relative to 1990, assuming no adaptation measures are implemented [27].
However, previous studies have not consistently demonstrated a clear association between environmental stressors and increased out-migration. For example, one study in Senegal found that excessive precipitation was associated with increased migration, whereas heatwave were linked to decreased migration. Notably, both effects were more pronounced when environmental exposures were measured over the crop growing season, suggesting agricultural mediation [26]. Therefore, our findings highlight the likely context-specific and complex interplay between climate stressors and migration dynamics in rural Burkina Faso.
Our results show that increases in out-migration were somewhat higher in response to the preceding single year’s FCPI than cumulative FCPI. This suggests that socio-economic processes and coping strategies, such as selling disposable assets to cope with low yields, may mitigate out-migration over the long term. A systematic review found that migration decisions are closely linked to access to agricultural means for climate change adaptation and vary by landholdings [28]. Other research demonstrated that climate shocks might lead to immediate migration responses, but migration may also be delayed until in-situ adaptive strategies are exhausted [29].
The association between FCPI and out-migration was stronger in poorer households. ‘Distress migration’ is more often pursued by socioeconomically vulnerable individuals [30] and is less planned compared to economic or investment migration, which is more strategic. ‘Distress migration’ can undermine livelihoods and exacerbate vulnerability, leading to negative outcomes such as withdrawing children from school and eroding resilience against future shocks [31,32]. This is supported by a study in Burkina Faso, documenting first time migration in the absence of other options, despite its erosive effects [4]. However, other studies show that households require some degree of wealth to migrate and that poorer households can become immobile or trapped [29]. This highlights the importance of addressing socio- economic inequalities in climate change adaptation to ensure equitable outcomes for high- risk populations. Implementing weather-based crop insurance targeted at poor farmers provides financial support during climate shocks [33], reducing inequities. Targeted social support measures, such as grants and cash for lower-income groups, can protect against forced migration and negative health effects from climate shocks, aligning with forecast- based financing strategies to minimize disaster displacement [33].
Our study reveals gender disparities in the migration response to declining crop yields. Men exhibit a stronger association between lower yields and out-migration compared to women. This may be attributed to traditional gender roles in rural Burkina Faso, where men are often primary income earners and thus more likely to seek employment opportunities elsewhere when local livelihoods are threatened. Additionally, the finding that individuals with prior migration experience are more responsive to yield fluctuations suggests the role of social networks and migration pathways in facilitating mobility, consistent with migration systems theory [34]. This aligns with other research indicating that climate change-related migration is highly gendered [35]. For example, a study in Pakistan found that heat stress increased long-term migration among male but not female farmers [36]. These findings highlight the need for gender-specific policies to address the distinct capability of migration in response to climate stressors and ensure equity in benefits brought by such responses across genders.
The results suggest a cumulative impact of migration experience on migration behavior, with individuals who have previously migrated being more likely to respond to crop yield reductions by migrating again. The influence of past migration on future migration behavior is poorly understood due to the lack of longitudinal studies in climate migration research. Yet, some evidence indicates that migration is a learned behavior. For instance, a study of 15 European countries found that decisions to migrate are embedded within a longer migration history and the influence of past moves diminishes as individuals progress in their migration “careers” [37]. This suggests that past migration experiences shape individuals’ adaptive strategies, emphasizing the need for longitudinal studies to capture the evolving dynamics of migration in response to climate change.
We observed that not only farmers but also those employed in the wider agricultural and food sector may out-migrate more in years of poor yield, if they have a general predisposition to migration, such as prior experience of out-migration. While many studies on migration and slow-onset events focus on populations with directly resource-dependent livelihoods, such as farmers, there is a growing body of research examining the risk of environment-related migration and displacement among other occupational groups, which constitutes a novel aspect of this study.
Policy implications
Our findings point to several policy priorities for addressing mobility pressures arising from climate-related yield variability in rural Burkina Faso. Because we observe that poorer male farmers and individuals with prior migration experience are most responsive to yield shocks, interventions should prioritize these groups. Strengthening agricultural resilience through measures that directly buffer households against yield fluctuations, such as integrated soil fertility management, agroforestry inputs, and expanded agricultural extension services, may help stabilize production and reduce the need to migrate during poor harvest years [18]. Given the heightened sensitivity of poorer farmers to income shortfalls, targeted cash transfers or input subsidies during drought or low-yield seasons could help maintain short-term consumption and reduce migration undertaken in response to livelihood pressures. Investments in small-scale irrigation, water harvesting systems, and improved rural road networks would also help reduce crop losses and improve market access, particularly for remote communities where FCPI variability is greatest.
Additionally, recognizing that migration can serve as a legitimate and beneficial adaptation strategy, policies should support safe and orderly mobility. This includes improving civil documentation processes, enabling the portability of social protection entitlements, and ensuring that young migrants, who represent a large share of those leaving, have access to skills training that enhances their safety and economic prospects. Vocational training in climate-resilient agricultural practices and small enterprise development can help diversify income sources for both migrants and non-migrants, reducing vulnerability to climate shocks. Finally, because climate impacts and mobility flows in West Africa are strongly transboundary, coordinated regional planning and data-sharing mechanisms can help to anticipate and manage migration dynamics linked to agricultural volatility.
Limitations
This is an observational study, where the possibility of residual confounding cannot be excluded. As the study was designed to compare the entire population against itself in one year versus another in a longitudinal 22 year data set, cross-sectional factors at the individual and household level are less likely to confound the associations of interest than time-varying factors, such as government policies, large-scale land ownership changes, economic development, and other changes in population’s general mobility. Time-varying factors can be sudden and short-term or gradual and long-term. To adjust for any unmeasured long-term changes, we included in our analyses an adjustment for linear time trend. Sudden and short-term time-varying factors were discussed with representatives of the local research center to identify any events that could have confounded our analyses. None of the events that were identified in these discussions met the criteria of confounding, i.e., events that were not on the causal pathway between FCPI and out-migration but are independently associated with FCPI and out-migration. For example, droughts and floods were on the causal pathway, there were no relevant government policy, large-scale land ownership changes, or changes in the agricultural practices or in the population’s reliance on subsistence farming over the study period. The armed conflict in the area started only after the end of the study period in the year.
In our study area the highest resolution of crop yield data was at the province level, due to which we had only one exposure measure per year for the whole study population. Therefore, we could not use year as an additional indicator variable to control for any potentially confounding temporal factors other than time trend (which we controlled for). This may have introduced some ecological error, as true yield shocks are likely to vary within the province due to local agro-ecological conditions and infrastructure. The time-to-event framework used in our analysis partially mitigates these concerns by exploiting temporal variation in exposure rather than cross-sectional spatial variation. Furthermore, we controlled for a time-invariant village-level infrastructural development index - a principal component of village infrastructural characteristics: presence of health-care facilities, drilled water wells, markets, and the quality of road connections. Although individuals are still compared within risk sets, the analysis does not rely on spatial contrasts in FCPI across villages. As a result, bias from fixed within-province heterogeneity is reduced relative to designs based on cross-sectional exposure differences, and any remaining ecological measurement error is expected mainly to attenuate estimated effects rather than generate spurious associations.
Household wealth data were only available for 2009, therefore, we had to assume household wealth stability over the study period.
Recommendations for future research
While this study advances understanding of the climate-migration nexus, several areas warrant further investigation. Longitudinal studies incorporating more granular climate data and diverse socio-economic indicators could elucidate causal pathways more precisely. Additionally, qualitative research exploring the experiences of migrants can provide deeper insights into decision-making processes and the role of non-economic factors in migration. Future research should also examine the health impacts of climate-induced migration, considering both the risks associated with displacement and the potential health benefits of reduced exposure to environmental hazards. Integrating health metrics into migration studies would offer a more comprehensive view of the climate-food-migration-health nexus, informing holistic adaptation strategies.
The variation in crop yields in this setting is largely attributable to weather. Poor crop yields are projected to worsen in the future, even under the aspirational target of 1.5°C global warming. Consequently, out-migration from the study area is likely to increase under any climate scenario. This future will necessitate increased local and national adaptation measures.
Development of adequate adaptation measures requires robust and context-specific projections of migration flows under different climate change and socio-economic scenarios based on high-quality empirical evidence. Estimating the number of people projected to migrate or be displaced due to climate change presents significant challenges. The variability of climate impacts across regions complicates uniform predictions, while models must account for complex interactions between environmental, economic, and social factors, which are often nonlinear, unpredictable, and context-specific. Migration data, particularly in developing regions, is scarce, fragmented, often outdated, leading to gaps and biases in projections and since most environmental displacements are internal, they are harder to track than cross-border movements [38].
Building a more comprehensive evidence base for attribution, projections, and development of suitable policies requires further research on similar associations in other areas. There are more than 60 HDSS sites globally, which follow a standardized methodology and pass quality-assurance. All HDSS sites are in low-income countries vulnerable to climate change [38–40]. Hence, our approach is scalable and, if expanded, can be used to substantially improve the empirical evidence for the attribution and projections of climate migration.
If adequately managed, migration can constitute an adaptation strategy [41–43]. Migration decisions are influenced by individual characteristics, household composition, social networks, and broader historical, political, and economic factors. Migration can be part of household strategies to diversify risk [39,44,45], which may be what our results reflect given poorer households are more likely to migrate in response to weather-related crop loss. Understanding whether, to what extent and how out-migration in response to low crop yield currently constitutes an effective adaptation strategy and what factors influence its effectiveness, is important. Further research is needed to determine what strategies, policies, and other interventions could benefit sustainable management of migration flows in both sending and host locations, ensuring the health and wellbeing of migrants.
A recent policy synthesis suggested key principles for stewarding safe, orderly and regular migration in the context of climate change, including: avoiding the universal promotion of migration as an adaptive response to climate risk; preserving cultural and social ties of mobile populations; enabling the participation of migrants in decision-making in sites of relocation and resettlement; strengthening health systems and reduce barriers for migrant access to health care; and optimizing of social determinants of migrant health with attention to immobile and trapped populations [46]. However, effective climate change mitigation action should remain the priority to help minimize the stress that climate change imposes on requiring people to leave their homes.
Conclusion
In the rural population of Burkina Faso studied, out-migration appears to increase with crop yield reductions, particularly among farmers, and especially among poorer male farmers with prior experience of migration. There is also some evidence that those occupied in the wider agricultural and food sector and those not in paid employment (unoccupied), out-migrate more during the years of poor yield. These findings are particularly important in the context of the projected reductions and increased unpredictability of crop yields with future climate change in this and similar settings. Our findings provide an important basis for further research and projections of possible out-migration flows under future climate change and socio-economic scenarios. It is important to understand whether, to what extent, and how out-migration in response to poor crop yields functions as an effective adaptation strategy across different settings, and which policies and programs are needed to protect migrant health and the health of both sending and host communities. Finally, this research emphasizes the need for urgent action on climate change mitigation to prevent further exacerbation of the climate change-driven stressors that necessitate migration in subsistence farming communities in Africa.
Supporting information
S2 File. Extended Results.
Table A: Full version of Table 3. First out-migration. Table B: Association of out-migration, undertaken for “yield-sensitive” reasons, with food crop yield. Second out-migration. Table C: Association of out-migration for reasons, undertaken for “yield-sensitive” reasons, with food crop yield. Third out-migration. Table D: Association of out-migration for reasons, undertaken for “yield-sensitive” reasons, with food crop yield. Fourth out-migration. Table E: Association of out-migration for reasons, undertaken for “yield-sensitive” reasons, with food crop yield. Fifth out-migration.
https://doi.org/10.1371/journal.pclm.0000832.s002
(DOCX)
S3 File. Sensitivity Analysis Results.
Table A: Results of Sensitivity Analysis 1. Table B: Results of Sensitivity Analysis 2. Table C: Results of Sensitivity Analysis 3.
https://doi.org/10.1371/journal.pclm.0000832.s003
(DOCX)
Acknowledgments
We would like to acknowledge Cassia Rocha Pompeu for her assistance with graphic design.
References
- 1.
Pörtner HO, Roberts DC, Poloczanska ES, Mintenbeck K, Tignor M, Alegría A. IPCC, 2022: Summary for policymakers. 2022.
- 2. Rigaud KK, De Sherbinin A, Jones B, Bergmann J, Clement V, Ober K, et al. Groundswell. Washington, DC: World Bank. 2018;10(2):94
- 3.
Trisos C, Adelekan I, Totin E, Ayanlade A, Efitre J, Gemeda A, et al. Africa. Climate Change 2022: Impacts, Adaptation and Vulnerability. Contribution of Working Group II to the Sixth Assessment Report of the Intergovernmental Panel on Climate Change. 2022.
- 4. Vinke K, Rottmann S, Gornott C, Zabre P, Nayna Schwerdtle PF, Sauerborn R. Is migration an effective adaptation to climate-related agricultural distress in sub-Saharan Africa?. Population and Environment. 2022;:1–27.
- 5. Torres JM, Casey JA. The centrality of social ties to climate migration and mental health. BMC Public Health. 2017;17(1):600. pmid:28679398
- 6. Henry S, Schoumaker B, Beauchemin C. The Impact of Rainfall on the First Out-Migration: A Multi-level Event-History Analysis in Burkina Faso. Population and Environment. 2004;25(5):423–60.
- 7. Kniveton D, Smith C, Wood S. Agent-based model simulations of future changes in migration flows for Burkina Faso. Global Environmental Change. 2011;21:S34–40.
- 8. Roncoli C, Ingram K, Kirshen P. The costs and risks of coping with drought: livelihood impacts and farmers’ responses in Burkina Faso. Clim Res. 2001;19:119–32.
- 9. Millock K. Migration and environment. Annu Rev Resour Econ. 2015;7(1):35–60.
- 10. Hoffmann R, Dimitrova A, Muttarak R, Crespo Cuaresma J, Peisker J. A meta-analysis of country-level studies on environmental change and migration. Nat Clim Chang. 2020;10(10):904–12.
- 11.
Opitz Stapleton S, Nadin R, Watson C, Kellett J. Climate change, migration and displacement: The need for a risk-informed and coherent approach. London: Overseas Development Institute (ODI). 2017.
- 12. Sultan B, Defrance D, Iizumi T. Evidence of crop production losses in West Africa due to historical global warming in two crop models. Sci Rep. 2019;9(1):12834. pmid:31492929
- 13. Sultan B, Roudier P, Quirion P, Alhassane A, Muller B, Dingkuhn M, et al. Assessing climate change impacts on sorghum and millet yields in the Sudanian and Sahelian savannas of West Africa. Environ Res Lett. 2013;8(1):014040.
- 14.
Kafle K, Wang Y, Kiiza B. Too poor to migrate? Weather shocks reduce temporary migration among smallscale farmers in Uganda. 2024.
- 15. Maystadt J-F, Mishra AK, Mueller V, Smoldt M. The Causes and Policy Responses to Forced Migration and Environmental Degradation in Africa. Annual Review of Resource Economics. 2024;16(1):301–22.
- 16. McLeman R. Thresholds in climate migration. Popul Environ. 2017;39(4):319–38.
- 17. Dell M, Jones BF, Olken BA. What Do We Learn from the Weather? The New Climate-Economy Literature. Journal of Economic Literature. 2014;52(3):740–98.
- 18. Arumugam P, Chemura A, Aschenbrenner P, Schauberger B, Gornott C. Climate change impacts and adaptation strategies: An assessment on sorghum for Burkina Faso. European Journal of Agronomy. 2023;142:126655.
- 19. Belesova K, Gornott C, Milner J, Sié A, Sauerborn R, Wilkinson P. Mortality impact of low annual crop yields in a subsistence farming population of Burkina Faso under the current and a 1.5°C warmer climate in 2100. Sci Total Environ. 2019;691:538–48. pmid:31325854
- 20. Tuholske C, Di Landro MA, Anderson W, van Duijne RJ, de Sherbinin A. A framework to link climate change, food security, and migration: unpacking the agricultural pathway. Popul Environ. 2024;46(1).
- 21.
Kleinbaum DG, Klein M. Survival analysis a self-learning text. Springer. 1996.
- 22. Belesova K, Gasparrini A, Sié A, Sauerborn R, Wilkinson P. Annual Crop-Yield Variation, Child Survival, and Nutrition Among Subsistence Farmers in Burkina Faso. Am J Epidemiol. 2018;187(2):242–50. pmid:28641367
- 23. Kosmowski F, Chamberlin J, Ayalew H, Sida T, Abay K, Craufurd P. How accurate are yield estimates from crop cuts? Evidence from smallholder maize farms in Ethiopia. Food Policy. 2021;102:102122. pmid:34898811
- 24. Belesova K, Gasparrini A, Sié A, Sauerborn R, Wilkinson P. Household cereal crop harvest and children’s nutritional status in rural Burkina Faso. Environmental Health. 2017;16:1–11.
- 25. Amorim LDAF, Cai J. Modelling recurrent events: a tutorial for analysis in epidemiology. Int J Epidemiol. 2015;44(1):324–33. pmid:25501468
- 26. Nawrotzki RJ, DeWaard J, Bakhtsiyarava M, Ha JT. Climate shocks and rural-urban migration in Mexico: Exploring nonlinearities and thresholds. Clim Change. 2017;140(2):243–58. pmid:28435176
- 27. Iqbal K, Roy PK. Climate change, agriculture and migration: evidence from Bangladesh. Clim Change Econ. 2015;06(02):1550006.
- 28. Mukherjee M, Fransen S. Exploring migration decision-making and agricultural adaptation in the context of climate change: A systematic review. World Development. 2024;179:106600.
- 29. Nawrotzki RJ, DeWaard J. Climate Shocks and the Timing of Migration from Mexico. Popul Environ. 2016;38(1):72–100. pmid:27795604
- 30. Parida JK. MGNREGS, distress migration and livelihood conditions: a study in Odisha. J Soc Econ Dev. 2016;18(1–2):17–39.
- 31. Cundill G, Singh C, Adger WN, Safra de Campos R, Vincent K, Tebboth M, et al. Toward a climate mobilities research agenda: Intersectionality, immobility, and policy responses. Global Environmental Change. 2021;69:102315.
- 32. Walelign SZ, Cutter SL, Lujala P. Resettlement capacity assessments for climate induced displacements: Evidence from Ethiopia. Climate Risk Management. 2021;33:100347.
- 33. Thalheimer L, Simperingham E, Jjemba EW. The role of anticipatory humanitarian action to reduce disaster displacement. Environ Res Lett. 2022;17(1):014043.
- 34. Massey DS, Arango J, Hugo G, Kouaouci A, Pellegrino A, Taylor JE. Theories of International Migration: A Review and Appraisal. Population and Development Review. 1993;19(3):431.
- 35. Chindarkar N. Gender and climate change-induced migration: proposing a framework for analysis. Environ Res Lett. 2012;7(2):025601.
- 36. Mueller V, Gray C, Kosec K. Heat stress increases long-term human migration in rural Pakistan. Nature Climate Change. 2014;4(3):182–5.
- 37. Bernard A, Perales F. Is Migration a Learned Behavior? Understanding the Impact of Past Migration on Future Migration. Population & Development Rev. 2021;47(2):449–74.
- 38. Gemenne F. Why the numbers don’t add up: A review of estimates and predictions of people displaced by environmental changes. Global Environmental Change. 2011;21:S41–9.
- 39.
Carmin J, Tierney K, Chu E, Hunter LM, Roberts JT, Shi L. Adaptation to climate change. Climate change and society: Sociological perspectives. 2015. p. 164–98.
- 40. Jia P, Sankoh O, Tatem AJ. Mapping the environmental and socioeconomic coverage of the INDEPTH international health and demographic surveillance system network. Health Place. 2015;36:88–96. pmid:26454610
- 41. Jamero MaL, Onuki M, Esteban M, Chadwick C, Tan N, Valenzuela VP, et al. In-situ adaptation against climate change can enable relocation of impoverished small islands. Marine Policy. 2019;108:103614.
- 42. Esteban M, Takagi H, Jamero L, Chadwick C, Avelino JE, Mikami T. Adaptation to sea level rise: Learning from present examples of land subsidence. Ocean & Coastal Management. 2020;189:104852.
- 43. Bettini G. Climate migration as an adaption strategy: de-securitizing climate-induced migration or making the unruly governable?. Critical Studies on Security. 2014;2(2):180–95.
- 44. Hunter LM, Luna JK, Norton RM. The Environmental Dimensions of Migration. Annu Rev Sociol. 2015;41:377–97. pmid:29861536
- 45. Hayward B, Salili DH, Tupuana’i LL, Tualamali’i’ J. It’s not “too late”: Learning from Pacific small island developing states in a warming world. Wiley Interdisciplinary Reviews: Climate Change. 2020;11(1):e612.
- 46. Nayna Schwerdtle P, Stockemer J, Bowen KJ, Sauerborn R, McMichael C, Danquah I. A Meta-Synthesis of Policy Recommendations Regarding Human Mobility in the Context of Climate Change. Int J Environ Res Public Health. 2020;17(24):9342. pmid:33327439