Deciphering climate-induced displacement in Somalia: A remote sensing perspective

comparing the matching seasonal aggregates LC. In addition, the change in the agricultural class was the only component of the drought LC Change, as drought should not directly destroy the settlement class as with flood. Fig 10 shows an example of the LC Change maps that were produced between the 2019 and 2018 Gu season showing the impact of a Drought in April 2019 around the Balcad region. Both Fig 10A and 10B depict the LC maps for April 2018 and 2019, respectively. Fig 10C displays the monthly drought map for April 2019. Taking into account the LC Change maps and drought maps, Fig 10D illustrates the impact of drought on agricultural fields. A significant portion of agricultural fields within drought regions (indicated in red) were either destroyed or damaged. Linking environmental factors and Internally Displaced People (IDPs) To start with, Fig 11 presents a map of displacements taking place within and between district levels. During drought events from 2016 to 2019, Doolow emerged as the most preferred desti- nation district for individuals displaced from their home districts. In total, nine migration movements were reported to Doolow, followed by six migration movements to Luuq. Con- versely, Qansax Dheere had the highest number of departures due to drought events within the same period. Similar displacement patterns are showcased in Fig 12B. Notably, flood- induced displacements predominantly occurred within districts, suggesting that flood events were less likely to prompt individuals to leave their districts.

A series of plots were generated for all districts within our AOI to visually demonstrate the impact of environmental threats such as drought and flood on internal displacement in Soma- lia. We visually compared the plots of the Qansax Dheere and Doolow districts. The Qansax Dheere district had the highest outgoing environment-related displaced people, while Doolow district, on the border with Ethiopia, had the highest incoming environment-related displaced people. ADI and loss of agricultural fields. In this study, agricultural fields (measured in hect- ares) are of significant interest because they provide food sources for humans and livestock and are the primary income source and means of survival in rural areas of Somalia. Fig 13 illustrates the impact of drought on variation of agricultural fields in the Qansax Dheere dis- trict from 2016 to 2019. The district’s percentage affected by three different drought levels is Fig 10. Land cover maps. a. Aggregated LC map in April 2018; b. Aggregated LC map in April 2019; c. ADI map of April 2019; d. LC Change map with the impact of drought on agricultural fields. https://doi.org/10.1371/journal.pone.0304202.g010 PLOS ONE Deciphering Climate-induced Displacement in Somalia: A Remote Sensing Perspective PLOS ONE | https://doi.org/10.1371/journal.pone.0304202 August 9, 2024 17 / 27

represented by pale orange (watch level), orange (warning level), and dark brown (alert). The purple curve indicates the total agricultural fields lost. As depicted in Fig 13A, the total agriculture fields started to decrease, shown by the upward movement of the purple curve, three months after the onset of drought in February 2016. In 2017, the agricultural fields began to disappear as a drought started in May, and despite an increase in drought intensity from May to July, the agricultural fields did not shrink. This minor reduction in the agricultural fields can be attributed to the natural increase in crop pro- duction that occurs during the agricultural growing season following the Gu growing season, Fig 11. Displacement maps. a. Displacement movements occurred at the district level during drought event between 2016 and 2019; b. Displacement movements occurred at the district level during flood event between 2016 and 2019. The extent coordinates for AOI: 41, 0.3 to 47, 5.4. https://doi.org/10.1371/journal.pone.0304202.g011 Fig 12. Diagram of the developed methodology for the Agricultural Drought Indicator (ADI). https://doi.org/10.1371/journal.pone.0304202.g012 PLOS ONE Deciphering Climate-induced Displacement in Somalia: A Remote Sensing Perspective PLOS ONE | https://doi.org/10.1371/journal.pone.0304202 August 9, 2024 18 / 27

as seen in Fig 13B. Despite a greater proportion of the Qansax Dheere district being affected by the lowest degree of drought between August and December 2018, the reduction in the agri- cultural fields was mitigated. This can be explained by the significance of the drought degree impacting agricultural fields. In other words, agricultural fields are primarily affected by the most severe level of drought and least affected by the warning level of drought. This trend is observable in Fig 13D where coverage of agricultural fields began to diminish from June to October 2019 as more areas of the Qansax Dheere district were affected by alert and warning drought levels. Examination of plots from other districts tells a similar story: broadly speaking, agricultural fields decreased due to increasing drought severity and extent. These plots help validate our choice of the ADI metric by demonstrating a clear link to agricultural devastation, thereby supporting the proposed cause-effect hypothesis. It can be inferred that the ADI, par- ticularly in severely affected regions, could indicate economic devastation, as seen in the reduction of agricultural fields. ADI and displacements. A series of plots were generated for each of the sixteen districts to visualise the variation in IDPs during drought events. Here, we represent only two IDP plots for Qansax Dheere and Doolow. Qansax Dheere signifies the district with the highest outgoing migration events between 2016 and 2019. However, we selected the Doolow district

due to its highest number of incoming displacements. Figs 14 and 15 provide a more compre- hensive picture of displacement movements due to drought over the four-year study period, taking into account other factors including flood events and growing seasons. The background of each figure indicates the percentage of the district (Qansax Dheere, in the case of Fig 14) affected by watch, warning, and alert levels, as shown on the left vertical axis. Following this, two vertical strips per year represent the Gu and Deyr growing seasons, respectively. It is expected that precipitation will increase at the commencement of these seasons, thereby reduc- ing the percentage and intensity of the drought through the growing seasons. Three distinct IDP values are presented in the subsequent figures: net IDP, internal IDP, and total IDP. The net IDPs depicts the overall movement for a given month, considering both those leaving due to environmental factors and those migrating to this district to escape devas- tation from other districts. A negative net IDP indicates that more people moved into the dis- trict rather than departing. The internal IDP values reflect only the number of individuals Fig 13. The impact of drought on agricultural fields in Qansax Dheere District during 2016 (a), 2017 (b), 2018 (c), and 2019 (d). https://doi.org/10.1371/journal.pone.0304202.g013 PLOS ONE Deciphering Climate-induced Displacement in Somalia: A Remote Sensing Perspective PLOS ONE | https://doi.org/10.1371/journal.pone.0304202 August 9, 2024 19 / 27

displaced from their homes who opted to remain within their home districts, while the total IDP numbers present the total number of individuals displaced in some way over the course of a month. As demonstrated in Fig 14, the initial displacement wave began two months after the onset of drought in November 2016 (the commencement of the dry season) and the majority of the displaced population returning to their homes in Qansax Dheere by May 2017. The remaining migrants took nearly a year to return to their homes. The second drought period spanned between June and September 2017, resulting in approximately 3,000 displaced people. This low number of displacements can be explained by the previous drought, during which many people had already left the district. Moreover, the drought that occurred during the agricul- tural growing season suggests there may have been enough food available for them to remain. The number of displaced people during the third drought period, from May to September 2019, did not exceed 3,000 people. This can be attributed to the availability of food during the agricultural growing season, much like the previous drought period. Two major drought periods were identified for the Doolow district between 2016 and 2019 in Fig 15. The initial drought began in June 2016 and concluded in January 2017, resulting in a significant external displacement from Doolow two months later. The second significant drought event occurred a month after the first migrants had returned home, lasting until

November 2018. As before, the migrants departed from Doolow two months later. This sug- gests that people might need to gather resources before they can move, and the decision to relocate can follow with a necessary time lag to realise the “capabilities” [9, 41]. While higher precipitation during the Gu and Deyr growing seasons might seem to decrease the impact of droughts decreases during these periods, we observed instances that contradict this expectation. In the case of Qansax Dheere, displacements were often external with no incoming migration. However, Doolow had a high negative net IDP, meaning that large numbers of people moved into Doolow due to drought in other districts is Somalia. This is not unexpected as people who managed to arrive in Doolow are closer to humanitarian sup- port such as safety, food and healthcare provided at the nearby refugee reception centres and Fig 14. Internal, total, and net IDP for the Qansax Dheere district compared to the Drought ADI classification and flood events for this same period. Please note that the Gu and Deyr growing seasons are indicated on this figure as period of peak agricultural growth. Data for all indicators are generated on a monthly timescale. https://doi.org/10.1371/journal.pone.0304202.g014 PLOS ONE Deciphering Climate-induced Displacement in Somalia: A Remote Sensing Perspective PLOS ONE | https://doi.org/10.1371/journal.pone.0304202 August 9, 2024 20 / 27

camps. Doolow is one of the primary exit points from Somalia to Dolo Ado, a small Ethiopian border town, and often serves as a hub for cross-border flows of displaced people. Deyr Flood and displacements. Compared to drought, the impact of flood events is largely localised and confined to specific districts. Fig 16 demonstrates that the majority of displace- ments due to floods remained internal within a district, with Belet Weyne recording two exter- nal displacements in addition to internal ones. The flood events representing regular and extreme floods, respectively, are depicted as two distinct lines. Of the three flood events, only two (April 2018 and November 2019) led to displacements movements in Belet Weyne district. The effect of flood events on displacement is more immediate than that of drought. Internal displacement occurs shortly after the flood, followed by a swift return of the IDPs to their homes after the water recedes. In contrast, drought is more gradual, with people beginning to Fig 16. The relationship between flood events and IDP movements in Belet Weyne district between 2016 and 2019. https://doi.org/10.1371/journal.pone.0304202.g016 Fig 15. The relationship between drought events and IDPs movement in Doolow district between 2016 and 2019. https://doi.org/10.1371/journal.pone.0304202.g015 PLOS ONE Deciphering Climate-induced Displacement in Somalia: A Remote Sensing Perspective PLOS ONE | https://doi.org/10.1371/journal.pone.0304202 August 9, 2024 21 / 27

leave their homes with a delay after the drought starts. In addition, for those who opt to return home, it takes much longer, as it is likely that those migrating have decided that their previous living conditions were untenable. In conclusion, the impact of droughts on departure and return displacements is more com- plex than that of floods. However, reviewing the entire AOI period on drought and IDPs for all districts reveals that a district affected by a severe drought does not necessarily result in a high displacement of people. Discussion and concluding remarks In this study, we utilised remote sensing datasets to create various environmental indicators for Somalia including Agricultural Drought Index (ADI), flooding, and Land Cover (LC) changes. These were used to demonstrate the impact of extreme events on ground conditions in Somalia. Initially, the ADI methodology was developed around the cause-effect correlation found in agricultural droughts, where inadequate precipitation results in a soil moisture defi- cit, leading to decreased vegetation productivity. We generated a monthly time series of ADI, which corresponded to the different stages of the drought effects, categorised into three sever- ity levels: watch, warning, and alert. Drought monitoring precision could be improved by sup- plementing data input with in-situ measurements of higher spatial resolution or by consistently using remote sensing and model ensemble techniques. For flood mapping, a change detection and thresholding methodology was employed using,

Sentinel-1 imagery closely timed to flood incidents. Historical satellite data before the Sentinel era could potentially augment our analysis; however, our methodological constraints, along- side the discernible disparities in spatial and spectral resolutions between Landsat and Sentinel series, deterred the integration of Landsat data. Despite the absence of Sentinel data pre-2015, our study navigated these limitations by focusing on the robust capabilities and temporal alignment of available Sentinel-1 and Sentinel-2 data post-launch, ensuring consistency in our remote sensing analysis. The absence of pre-2015 Sentinel data and lack of driect ground truth data for validating the ADI and LC change map constitute significant limitations, reinforcing the need for future studies to explore complementary datasets where feasible, while maintain- ing analytical coherence. We developed a series of LC change maps to evaluate the repercussions of natural hazards, specifically drought and flood, on various land cover types, particularly those associated with agricultural productivity and human settlements. The LC methodology was constructed using the state-of-the-art U-Net technique, trained on the Sentinel-2 Level-2A dataset. The convolu- tional characteristic of U-Nets allows for the integration of information from surrounding pix- els into predictions, offering superior detection of large-scale structures within images, compared to the single-pixel detection of Random Forests. The U-Net training process was considerably expedited by collecting training data via poly-

gons, circumventing the laborious task of manually labelling pixels across an entire image. As a result, the final trained model outperformed conventional machine learning techniques typi- cally employed for land cover purposes, such as Random Forest. However, a notable limitation of LC change involves the inherent difficulty in disentangling the impact of drought from con- current events throughout the entire growing season. These concurrent events could include flooding and deliberate shifts in cultivation patterns, potentially complicating the interpreta- tion of the results. Our observations echo the findings of seminal works in this field [42–44], strengthening the validity of our assertions. Moreover, the nuances we’ve identified in the impact of concurrent events during the growing season, though briefly mentioned by [45, 46] PLOS ONE Deciphering Climate-induced Displacement in Somalia: A Remote Sensing Perspective PLOS ONE | https://doi.org/10.1371/journal.pone.0304202 August 9, 2024 22 / 27

are expanded upon in our study. These interconnections, along with our unique observations, signify the broader relevance of our research and its positioning amidst existing literature. Our analysis identified 2017 as the driest year for Somalia, with 2018 and 2019 being signifi- cantly impacted by floods. Regular rainfall periods, specifically the Gu and Dayer seasons, bring increased precipitation, which tends to mitigate drought impact during these growing seasons. In regions such as northern and central Somalia, where rapid-onset and slow-onset extreme climatic events occur either sequentially or concurrently, categorising environmental factors as beneficial or detrimental is intricate. This complexity is evident, for instance, when considering the mitigating effect of floods following severe droughts in agricultural areas. However, these environmental fluctuations may not always result in advantageous impacts on livelihoods. As we delve deeper into how these extreme weather phenomena affect people’s daily lives, it becomes clear that the consequences of rapid-onset events (like floods) and slow-onset events (like droughts) diverge, particularly in terms of individual or household response times. For instance, flood events prompt quick displacement, with minimal response time, indicating that people start evacuating shortly before floods occur, and often return to their homes soon after the floodwaters recede, contingent upon the area’s recovery. Our study confirms that the effects of floods are predominantly localised and confined to districts, and most flood-induced

displacement results in internal displacement within the same districts. Conversely, drought is a more progressive phenomenon, making adaptation to these more drastic changes challeng- ing. Displacement due to drought begins with a delay and subsequently sees a gradual increase in the number of IDPs. Additionally, returning to habitual residences post-drought is a consid- erably extended process. Our investigation into the underlying causes of displacement reveals a diverse and intersec- tional assortment of factors including climate-related threats, conflict, extreme poverty, and food insecurity. Given the profoundly inter-correlated nature of these factors and events, eval- uating their marginal effects proves unfeasible. When individuals face the repercussions of these threatening conditions on their livelihoods, well-being, and security, whether in the short, medium, or long term, the decision to relocate can be construed as a strategic response. However, the predictability of events in terms of timing and the magnitude of their impact var- ies, leading individuals to opt for immediate reactions to avoid imminent starvation and mate- rial deprivation, rather than employing strategic responses. Our research suggests that responses currently remain predominantly at an individual level, especially as mobility in response to slow-onset events significantly depends on individual "capabilities" [47]. Given projections of similar of greater magnitude climate change, develop- ing a robust assessment of current and future impacts solely through macro analyses of aggre-

gate socio-economic statistics of vulnerable regions and existing climate products seems unattainable. Specifically, detangling the intertwined aspects of economy, security, and climate appears to be an insurmountable challenge. Climate change and extreme weather events not only trigger migratory movements but also exacerbate other factors such as economic deprivation and food insecurity. In the context of Somalia, a country grappling with both environmental hazards and armed conflict for decades, the urgent need for sustainable solutions to its humanitarian crisis is evident. Addressing inter- nal displacement and migration in Somalia (and similarly afflicted countries) affectively will require resources beyond the Somalia’s capacity and call for multinational and multilateral col- laboration. The primary long-term strategy should aim at minimising the factors exacerbating climate change. Considering climate-induced migration, understanding the intricate and multifaceted rela- tionship between environmental variables and human migration goes beyond mere conceptual PLOS ONE Deciphering Climate-induced Displacement in Somalia: A Remote Sensing Perspective PLOS ONE | https://doi.org/10.1371/journal.pone.0304202 August 9, 2024 23 / 27

discourse. A growing body of research highlights the inadequacy of treating climate factors as isolated and peripheral drivers of human mobility; a more nuanced comprehension is neces- sary to decipher the interactive dynamics intrinsic to this nexus (i.e., [9, 11]). Especially in a context like Somalia, which is susceptible to severe climatic conditions and persistent conflict, the fundamental causes of displacement prove challenging to explicate due to the relentless interconnections between conflict and environmental degradation. Regarding the technical limitations of our study, the most significant hurdle was procuring a dataset - whether satellite or ground data - that aligns with the timescale and resolution of our AOI. For instance, to generate the flood product map, an image from Sentinel-1 immedi- ately after the flood event is crucial to capture the flood’s full extent. Nevertheless, the pro- longed revisit time of the Sentinel-1 sensor sometimes resulted in unavailable images of the flooded region until several days after the flood began. This could lead to an inaccurate repre- sentation of the flood’s full scope. Moreover, for Land Cover products, the quality of ground truth data is crucial, as it directly affects the quality of the predictions. This includes a lack of ground truth data for validating past drought occurrences. While datasets like the Land Cover Net exist, they are often too expansive, encompassing the entire African continent, or they possess an unspecific classifica- tion scheme that doesn’t match ours.

For other products, such as drought, our only alternative was to validate the predicted drought regions using reliable local drought reports. Future research on drought mapping should aim to improve the portrayal of drought status. This could be achieved either by enhancing data input through the use of in-situ measurements with finer spatial resolution, or by persisting with remote sensing and model ensemble techniques. In addition, considering Synthetic Aperture Radar (SAR) data, full-polarisation techniques, particularly "target decomposition", can offer enhanced capabilities over raw polarised data such as VV, HH, and HV/VH when it comes to detecting specific objects. Although integrat- ing such techniques was beyond the scope of the present study, it is indeed an avenue worth exploring. Future research endeavours with similar applications could greatly benefit from incorporating these advanced polarimetric techniques to achieve more accurate results. Compounding the challenges of this study were the limitations we faced in accessing high- quality, micro, longitudinal socio-economic data, especially on a more granular level such as district-based data. The issue of data quality and availability remains a significant hurdle, espe- cially for economically disadvantaged African nations [31, 32]. A key contributing factor in our case study of Somalia is the protracted armed conflict and war the country has been experiencing for decades. This has resulted in a dearth of govern- mental reports due to the inherent difficulties in collecting ground data, with only limited

information available from international observers. Despite these challenges, our study offers an innovative assessment of detailed environmen- tal indicators and IDPs in Somalia. This approach could potentially be expanded to cover more extended time periods and wider geographical areas. Future research could examine cross-border mobility and environmental factors. However, given our findings and existing data challenges, we maintain that one of the first steps should be to improve data quality and availability at both the national and international levels for regions most affected by climate change. Finally, we propose that triangulating quantitative and qualitative information can aid in illuminating such complex phenomena. This approach could underpin future research aimed at enhancing our understanding of climate-induced migration. PLOS ONE Deciphering Climate-induced Displacement in Somalia: A Remote Sensing Perspective PLOS ONE | https://doi.org/10.1371/journal.pone.0304202 August 9, 2024 24 / 27

Acknowledgments We would like to express our sincere gratitude to our colleagues Haodong Qi and Miko Stanek for their critical review and insightful suggestions to improve our article. Author Contributions Conceptualization: Tuba Bircan. Data curation: Rahman Momeni, Tuba Bircan, Robert King. Formal analysis: Rahman Momeni, Tuba Bircan, Eloy Zafra Santos. Funding acquisition: Tuba Bircan. Methodology: Rahman Momeni, Tuba Bircan. Project administration: Tuba Bircan. Supervision: Tuba Bircan. Visualization: Rahman Momeni, Robert King, Eloy Zafra Santos. Writing – original draft: Rahman Momeni, Tuba Bircan. Writing – review & editing: Tuba Bircan. References 1. Popa V, Cocos O. Syria, an agonizing country. The socio-economic, political and environmental impli- cations of the conflict. Romanian Review on Political Geography/Revista Romaˆ na Geografie Politica, 2015 May; 17(1): 29–37. 2. Aburas R, Najeeb A, Baageel L, Mackey TK. The Syrian conflict: a case study of the challenges and acute need for medical humanitarian operations for women and children internally displaced persons. BMC Medicine. 2018 Dec; 16(1): 1–6. https://doi.org/10.1186/s12916-018-1041-7 PMID: 29747641 3. Huynh BQ, Basu S. Forecasting internally displaced population migration patterns in Syria and Yemen. Disaster Medicine and Public Health Preparedness. 2020 Jun; 14(3): 302–7. https://doi.org/10.1017/ dmp.2019.73 PMID: 31452495 4. Podesta J. The climate crisis, migration, and refugees. [Internet]. 2018. Brookings Institution. [cited

2022 Sep 5]. Available from: https://www.brookings.edu/research/the-climate-crisis-migration-and- refugees. 5. Roth BJ. Temporary shelter: Venezuelan migrants and the uncertainty of waiting in Colombia. Journal of Immigrant & Refugee Studies. 2021 Aug 28: 1–3. 6. IDMC. Global Report on Internal Displacement 2021. [Internet]. IDMC. 2022. [cited 2023 Mar 8]. Avail- able from: https://www.internal-displacement.org/global-report/grid2021/ 7. Laczko F, Aghazarm C. Migration, Environment and Climate Change: Assessing the Evidence. Interna- tional Organization for Migration (IOM); 2009. 8. Piguet E, Pe´ coud A, De Guchteneire P. Migration and climate change: An overview. Refugee Survey Quarterly. 2011 Sep 1; 30(3): 1–23. 9. Abel GJ, Brottrager M, Cuaresma JC, Muttarak R. Climate, conflict and forced migration. Global Envi- ronmental Change. 2019 Jan 1; 54: 239–49. 10. Boas I, Farbotko C, Adams H, Sterly H, Bush S, Van der Geest K, et al. Climate migration myths. Nature Climate Change. 2019 Dec; 9(12): 901–3. 11. DeFries RS, Edenhofer O, Halliday AN, Heal GM, Lenton T, Puma M, et al. The missing economic risks in assessments of climate change impacts. Paper, London Sch. Econ. 12. Hoffmann R, Sˇ edova´ B, Vinke K. Improving the evidence base: A methodological review of the quantita- tive climate migration literature. Global Environmental Change. 2021 Nov 1; 71: 102367. 13. Sıˆrbu A, Andrienko G, Andrienko N, Boldrini C, Conti M, Giannotti F, et al. Human migration: the big data perspective. International Journal of Data Science and Analytics. 2021 May; 11(4): 341–60.

14. Momeni R, Bircan T, King R. Environment-induced internal displacement. The case of Somalia (Deliver- able 4.3). 2022. Leuven: HumMingBird project 870661 –H2020. PLOS ONE Deciphering Climate-induced Displacement in Somalia: A Remote Sensing Perspective PLOS ONE | https://doi.org/10.1371/journal.pone.0304202 August 9, 2024 25 / 27

15. Monteiro J, Martins B, Pires JM. A hybrid approach for the spatial disaggregation of socio-economic indicators. International Journal of Data Science and Analytics. 2018 Mar; 5(2): 189–211. 16. UNFPA S. Population Survey of Federal Republic of Somalia for the 18 Pre-War Regions of Somalia. 2014, Nairobi: UBON, Publishing Service Section. 17. Achour ME, Lacan NI. Drought in Somalia: a migration crisis. The State of Environmental Migration. 2012; 6:12. 18. Jayawardhan S. Vulnerability and climate change induced human displacement. Consilience. 2017 Jan 1(17): 103–42. 19. Black R, Bennett SRG, Thomas SM, Beddington JR. Climate change: Migration as adaptation. Nature. 2011; 478, 447–449. https://doi.org/10.1038/478477a PMID: 22012304 20. Hogeweg M. The power of remote sensing in predicting climate change impact. Remote Sensing. 2021; 13(5), 872. 21. Liu J, Kuang W, Zhang Z, Xu X, Qin Y, Ning J, et al. Spatiotemporal characteristics, patterns, and causes of land-use changes in China since the late 1980s. Journal of Geographical Sciences. 2014; 24 (2), 195–210. 22. Gutmann MP, Field V. Katrina in historical context: Environment and migration in the U.S. Population and Environment. 2010; 31, 3–19. https://doi.org/10.1007/s11111-009-0088-y PMID: 20436951 23. Adamowski J, Fung Chan H, Prasher SO, Ozga-Zielinski B, Sliusarieva A. Comparison of multiple linear and nonlinear regression, autoregressive integrated moving average, artificial neural network, and wavelet artificial neural network methods for urban water demand forecasting in Montreal, Canada.

Water Resources Research. 2018; 49, 178–95. 24. Lindvall K, Kinsman J, Abraha A, Dalmar A, Abdullahi MF, Godefay H, et al. Health status and health care needs of drought-related migrants in the Horn of Africa—a qualitative investigation. International Journal of Environmental Research and Public Health. 2020 Aug; 17(16): 5917. https://doi.org/10. 3390/ijerph17165917 PMID: 32824046 25. Van Baalen S, Mobjo¨rk M. Climate change and violent conflict in East Africa: Integrating qualitative and quantitative research to probe the mechanisms. International Studies Review. 2018 Dec 1; 20(4): 547–75. 26. Warner K, Afifi T. Where the rain falls: Evidence from 8 countries on how vulnerable households use migration to manage the risk of rainfall variability and food insecurity. Climate and Development. 2014 Jan 2; 6(1): 1–7. 27. European Space Agency (ESA). ‘Sen2cor’ [dataset]. [Internet]. European Space Agency. 2020 - [cited 2023 Feb 20]. Available from: https://step.esa.int/main/third-party-plugins-2/sen2cor/ 28. NASA. NASADEM_HGT v001 [dataset]. NASA. [Internet]. 2000. NASA. [cited 2022 Nov 14]. Available from: https://lpdaac.usgs.gov/products/nasadem_hgtv001/ 29. Beguerı´a S, Latorre B, Reig F, Vicente-Serrano S. SPEI Global Drought Monitor, CSIC [data set]. Evanston (IL): CSIC. 2000 - [updated 2020; cited 2022 Mar 10]. Available from: https://spei.csic.es/ map/maps.html#months=1#month=6#year=2020 30. Devarajan S. Africa’s statistical tragedy. Review of Income and Wealth. 2013 Oct; 59: 9–15.

31. Hoogeveen J, Nguyen NT. Statistics reform in Africa: aligning incentives with results. The Journal of Development Studies. 2019 Apr 3; 55(4): 702–19. 32. Naude´ W, Vinuesa R. Data deprivations, data gaps and digital divides: Lessons from the COVID-19 pandemic. Big Data & Society. 2021 Aug; 8(2): 20539517211025545. 33. Be´ de´ carrats F, Cling JP, Roubaud F. The data revolution and statistical challenges in Africa: Introduc- tion to the special report. Afrique Contemporaine. 2016; 258(2): 9–23. 34. Clement MA, Kilsby CG, Moore P. Multi-temporal synthetic aperture radar flood mapping using change detection. Journal of Flood Risk Management. 2018 Jun; 1(2): 152–68 35. Twele A, Cao W, Plank S, Martinis S. Sentinel-1-based flood mapping: A fully automated processing chain. International Journal of Remote Sensing. 2016 Jul 2; 37(13): 2990–3004. 36. Long S, Fatoyinbo TE, Policelli F. Flood extent mapping for Namibia using change detection and thresh- olding with SAR. Environmental Research Letters. 2014 Mar 11; 9(3): 035002 37. Oliver C, Quegan S. Understanding Synthetic Aperture Radar Images. SciTech Publishing; 2004. 38. Renno´ C. D. et al. (2008). HAND, a new terrain descriptor using SRTM-DEM: Mapping terra-firme rain- forest environments in Amazonia. Remote Sensing of Environment. 2008, 112(9), 3469–3481. https:// doi.org/10.1016/j.rse.2008.03.018 39. van der Walt S, Scho¨ nberger J, Nunez-Iglesias J, Boulogne F, Warner J, Yager N, et al. (2014). scikit- image: Image processing in python. PLOS ONE

40. Laban N, Abdellatif B, Ebeid HM, Shedeed HA, Tolba MF. Sparse pixel training of convolutional neural networks for land cover classification. IEEE Access. 2021 Mar 30; 9: 52067–78. 41. Cammalleri C, Arias-Muñoz C, Barbosa P, de Jager A, Magni D, Masante D, et al. A revision of the Combined Drought Indicator (CDI) as part of the European Drought Observatory (EDO). Natural Haz- ards and Earth System Sciences Discussions. 2020 Aug 7: 1–34. 42. Ronneberger O, Fischer P, Brox T. U-net: Convolutional networks for biomedical image segmentation. In International Conference on Medical Image Computing and Computer-Assisted Intervention; 2015. p. 234–41. Springer, Cham. 43. Zhang Z, Geng X. Deep learning for remote sensing data: A technical tutorial on the state of the art. IEEE Geoscience and Remote Sensing Magazine. 2017; 4(2): 22–40. 44. Audebert N, Le Saux B, Lefèvre S. Deep learning for classification of hyperspectral data: A comparative review. IEEE Geoscience and Remote Sensing Magazine. 2019; 7(2): 159–173. 45. Houborg R, McCabe MF. High-resolution NDVI from Planet’s constellation of Earth observing nano-satel- lites: A new data source for precision agriculture. Remote Sensing of Environment. 2018; 215: 91–103. 46. Chen J, Ban Y, Li S. China: Open access to Earth land-cover map. Nature. 2014; 514(7523): 434–434. 47. De Haas H. A theory of migration: the aspirations-capabilities framework. Comparative Migration Stud- ies. 2021 Dec; 9(1): 1–35. https://doi.org/10.1186/s40878-020-00210-4 PMID: 33680858 PLOS ONE

Deciphering Climate-induced Displacement in Somalia: A Remote Sensing Perspective PLOS ONE | https://doi.org/10.1371/journal.pone.0304202 August 9, 2024 27 / 27