Global patterns of water‐driven human migration

Environmental change is growingly reported as an important driver of human migration. Among all environmental variables, water crises are the most critical factors. To date, patterns of interconnections between changes in water and migration are not yet clearly understood. Here, we explore these patterns through a systematic review that combined a quantitative text‐mining approach with qualitative thematic analysis. Our results generally concur with those of previous studies, which found that water‐driven migration usually occurs internally and that the population in low‐ and middle‐income countries and in dry regions are the most vulnerable and more likely to migrate or be displaced in the face of water‐related events. However, our causal network analysis highlights that water is not the only reason for migration: Its related problems could be major triggers driving people‐at‐risk to leave their original place. Based on observed evidence, water‐driven migration can be generally divided into four patterns: variability in water quantity, damaging water hazards and extremes, physical disturbances to water systems, and water pollution. These patterns are not independent but interconnected through multifaceted factors affecting people's livelihoods and their decisions to migrate. Understanding water‐migration dynamics requires systematic thinking of the interconnections between changes in water and in migration patterns, the investigation of interactions between fast and slow water variables and their dynamic link to other socioeconomic variables, an integrated water‐migration database to help identify early‐warning signals of damaging water hazards that may result in undesirable migration, and targeted water policies that focus on building the resilience of vulnerable regions and population to climate change.


| INTRODUCTION
Human migration is a complex social process with multiple definitions (BOX 1). Environmental change and human migration represent a two-way interaction between the environment and people. Although patterns of human migration vary in space and timing and are affected by complex environmental and social factors (Niva et al., 2021), there is a growing recognition that climate warming and many other types of environmental change are becoming important drivers of population mobility and human migration worldwide (Foresight., 2011;IPCC, 2022;Rigaud et al., 2018). The changing climate is accelerating human migration, especially in highly vulnerable regions (McLeman & Smit, 2006;Nawrotzki & DeWaard, 2018;Tacoli, 2011), such as in those low-and middle-income regions (Arnall et al., 2013;Gray & Mueller, 2012a, 2012bGray & Wise, 2016;McLeman, 2013). Reports have predicted that, by 2050, changing climate will likely lead to the movement of over 134 million people in the Sub-Sahara, South Asia, and Latin America (Rigaud et al., 2018). Migration will possibly be the only option for people in these regions to adapt to climate change; otherwise, they will become trapped populations (Foresight., 2011;Grecequet et al., 2017;Lustgarten, 2020a;McLeman, 2016). However, in some cases, migration may not be an appropriate adaptation to environmental change. As seen in Ethiopia, migrants in their new surroundings could be more vulnerable than they were at home because migration can accelerate environmental change through alterations to the landscape and overuse of the destination's natural resources (Hermans-Neumann et al., 2017;Rogers & Xue, 2015). This two-way interactive dynamic makes issues of climate migration challenging to understand, yet it is urgent that they be explored in-depth.
A rich body of literature examines the interaction between environmental change and migration, yet the knowledge remains fragmented because the observed evidence varies on different temporal and spatial scales, generating mixed results (Adger et al., 2015;Hunter et al., 2015;Koubi et al., 2016;Thiede & Gray, 2016). Specifically, fast-onset events, such as flooding and wildfires, and slow-onset events, such as drought and land degradation, are commonly viewed to be critical environmental factors in people's decisions to migrate. The influencing patterns, however, differ from place to place due to distinct demographic characteristics, environmental conditions, economic and socio-political contexts, and the degree of exposure to change (Fussell et al., 2014;Hoffmann et al., 2020;Niva et al., 2021). For example, research has found that in Bangladesh, floods may have a weaker impact on migration patterns than other events and factors such as rainfall shortage and related crop failure (Gray & Mueller, 2012a). Conversely, regular flooding was found to be a major trigger of population movement and migration in the Mekong Delta region (Dun, 2011). Additionally, some studies have found that changes in temperature correlate more strongly with migration in some regions than BOX 1 Definition of human migration-related terms Migration. The movement of persons away from their place of usual residence, either across an international border or within a State. Human mobility. Mobility is a generic term covering all the different forms of movements of persons. The term human mobility reflects a wider range of movements of persons than the term "migration." Displacement. The movement of persons who have been forced or obliged to flee or to leave their homes or places of habitual residence, in particular as a result of or in order to avoid the effects of armed conflict, situations of generalized violence, violations of human rights or natural or human-made disasters. This definition covers both internal and cross-border displacement. Climate migration. The movement of a person or groups of persons who, predominantly for reasons of sudden or progressive changes in the environment due to climate change, are obliged to leave their habitual place of residence, or choose to do so, either temporarily or permanently, within a State or across an international border. Climate migration is a subcategory of environmental migration; it defines a singular type of environmental migration. Internal and external migration. Internal migration is defined as the movement of people within a State involving the establishment of a new temporary or permanent residence. External migration, or international migration, or cross-border migration, refers to the movement of persons away from their place of usual residence and across an international border to a country of which they are not nationals. Source. International migration law by the International Organization for Migration (IOM, 2019). in others, especially when temperature changes cause crop failures, for example, the cross-border migration from Mexico to the United States and the increasing long-term migration in rural Pakistan (Feng et al., 2010;Mueller et al., 2014).
The research domain of climate migration has also been plagued by inadequate data and a lack of information transparency (McLeman, 2013). To date, questions of whether there is an explicit linkage between environmental change and human migration, and how they are interconnected, are not yet fully understood due to the lack of fine datasets on migration, particularly long-term data on migration trends and flux. Efforts have been made to create or integrate emerging datasets to explore the environment-migration nexus between climate or environmental change and human migration. For instance, Fang and Jawitz (2019) integrated census data into remote sensing data to map out human population distribution over many years in the United States. Lu et al. (2016) used mobile phone data to investigate patterns of human mobility in Bangladesh after Cyclone Mahasen in 2013. Some "person-year" datasets have been built to study population mobility and migration possibly caused by climate change in Ethiopia (Gray & Mueller, 2012b), Bangladesh (Gray & Mueller, 2012a), western Africa (Henry et al., 2004), Nepal (Massey et al., 2010), and Pakistan (Mueller et al., 2014). Globally, the Internal Displacement Monitoring Centre (IDMC) has created a novel database to provide comprehensive information on global internal displacement associated with conflict, violence, and sudden-onset natural disasters since the 2000s (IDMC, 2005). Niva et al. (2021) recently used a machine learning approach, based on gridded global net migration data, to specify the roles of environmental and social factors in migration. Aside from place-based studies, science-based evidence databases have been created, offering opportunities for comparative studies among different cases and over different scales. For example, a bibliographic database (CLIMIG) has been developed to include studies of migration and the environment and was used to map the global empirical studies and their common issues (Piguet et al., 2018).
Although previous empirical studies and data integration approaches have made significant contributions to climate migration research, a synthesis of fragmented evidence is limited, but would help to generalize migration patterns thereby contributing to the spectrum of the knowledge about interactions between environmental change and human actions. Starting with this motivation, this paper is set to generalize global patterns of water-driven human migration using an evidence-based synthesis. We focus on the role of water in migration as 74% of natural disasters and climate events between 2001 and 2018 were identified as water-related, placing millions of people at risk (UNICEF, 2021), and water factors, among all environmental variables, are regarded as key triggers for human migration or displacement worldwide (IDMC, 2022;IPCC, 2022;Lustgarten, 2020aLustgarten, , 2020bNagabhatla et al., 2020;Niva et al., 2021;Stoler et al., 2022). The objectives are to map knowledge of water-related migration studies, to generalize global patterns of human migration that are driven by water-related factors, and to synthesize studies in a way that quantitatively identifies knowledge gaps and future needs in the field.

| WATER RESOURCES AND HUMAN MIGRATION: HYPOTHETICAL DESCRIPTIONS
Water-migration nexus represents one of the complex interactions between water and humans. Human settlements are highly dependent on water availability (Fang & Jawitz, 2019;LandScan Global, 2017) as water plays a crucial role in social welfare that directly relates to food and life quality. When water conditions are significantly altered that threaten people's livelihoods, effective adaptation is required to maintain harmonious human and water relations. Migration may be viewed as an effective way or an opportunity for people to adapt to the threatened water systems as it allows people to diversify their income and build resilience (Black et al., 2011). However, water-driven migration in many cases may be involuntary, especially when it is the last-resort solution, which could pose risks to people who are unable or unwilling to move to the hosting destinations. Therefore, it can both result from, and result in the degradation of hydrological conditions (both quality and quantity) in original places and destinations.
Water is a multiple role-player in triggering human migration (Stoler et al., 2022), but people's decision to move is usually determined by multidimensional environmental and socioeconomic factors (Niva et al., 2021). The interactive dynamics with other environmental factors such as sea-level rise (Hauer et al., 2020) and socioeconomic variables (e.g., heterogeneity in demographic characters and socio-political settings that differentiates the adaptive ability of countries and/or communities) need to be further explored. For example, water scarcity is likely to accelerate migration in the global dry regions, as observed in countries where people are forced to move from vulnerable to more viable regions for better lives (Foresight., 2011; Rigaud et al., 2018). These regions are dominated by low-or middle-income countries, in which the migration decision is significantly influenced by opportunities of the destination, as well the moving costs (Gray & Mueller, 2012a). In addition to water scarcity, changes in human water use behaviors could also affect migration decisions via a cascading effect. Specifically, the growing population and changing patterns of water use will amplify the magnitude and timing of water stresses, further affecting agricultural production, worsening water contamination, causing persistent conflicts over water, and putting human health and economic wealth at great risk. All these water stress-related risks will make vulnerable regions and their people more susceptible to changes in hydrological conditions, further driving population movement (IPCC, 2022;Wrathall et al., 2018). For example, injustice in water use for industries and for residents can cause conflicts over water among communities and cross sectors leading to mass migration, particularly in water-stressed regions like the Middle East, South Asia, and North Africa (Famiglietti et al., 2022;Gleick & Heberger, 2014). Evidence of this is the inter-states water conflict due to unequal water use rights between Israel and Palestine that has resulted in the movement of residents near the border (Baumgarten, 2009).
As water stresses increase under climate change, the number of water-driven human migrants is expected to increase in the near future. A large body of literature predicts that climate-induced water stresses will likely cause a larger proportion of the world's population to move from their original places, seeking better opportunities and life quality (e.g., IPCC, 2022;Rigaud et al., 2018). Many of these stresses stem from water-related factors, including the availability of and accessibility to freshwater, the quality of water to provide clean water and sanitation, and exposure to damaging water hazards (Nagabhatla et al., 2020;Stoler et al., 2022;Wrathall et al., 2018). It has been estimated that water deficits are linked to 10% of the rise in global migration (Damania et al., 2017). The livelihood of over 85% of the population in low-or middle-income countries and a quarter of the world's population will be threatened by the growing water stresses (Damania et al., 2017). While numerous empirical studies have reported the interconnection between water stresses and migration, the evidence-based knowledge still remains fragmented which can benefit from a generalization of patterns that help identify critical stresses.

| Literature retrieval
The data collected for this synthesis include both academic publications and grey literature retrieved from three sources: the Web of Science (WOS), Google Search, and an existing bibliographic database. Academic publications include journal articles, books, book chapters, and conference papers, all of which are indexed in the WOS core collection. Grey literature includes those materials produced by organizations external to commercial and academic publishing, for example, reports of the World Bank and the International Organization for Migration (IOM), government documents, technical reports, and working papers of institutions. Grey literature was collected using the Google Search engine and an existing bibliographic database for climate and migration (CLIMIG) (Piguet et al., 2018). Although we acknowledge the contributions of all literature in different languages to the field, we only included English documents in the dataset for this study. Further, we eliminated short articles-such as comments, notes, and editorials-and did not include published conference abstracts because these items rarely presented detailed content for the synthesis.

| Data processing
The procedure of data processing is illustrated in Figure 1. First, we collected academic publications through a keyword-based search in WOS for the years 1950-2022. We applied multiple combinations of words to the topic search for publications that accommodate issues in human migration and environmental or climate change or water problems. Although our study addresses water-related migration, we also included studies on environmental and climate change in the initial search because some studies may focus on water issues without using "water" as a keyword. For example, we found that some studies on drought problems were not included in our search using the combination of "water & human migration." However, when we applied "environmental or climate change & human migration" to the search, the query produced studies on drought-related human migration.
After multiple trials of different word combinations, we identified search terms and applied them to produce a representative corpus of the literature (Table 1): human or people NEAR/2 migrat* or mobil* or move* or resettle* or displace* AND water or drought or flood* or water NEAR/2 availa* (dated 31 July 2022). It should be noted that each combination of terms was applied for the search individually to reduce redundant records and duplications, and all combinations were then merged. Using the term "Human NEAR/2 migrat*," the search found publications where "human" and "migrat*" were separated by at most two words. It thus captured those articles containing combinations of words such as "migration of human" and "human decide to migrate" in their title, abstract, or keywords list. The search initially identified 1144 records.
As a keyword-based search relies heavily on the semantic contents used, it is not capable of aligning the precision of retrieval data to the needs of the search, and it thus usually generates digressive or noisy data (Bhagdev et al., 2008). To clean the data, we removed digressive studies from the raw dataset using the refined function of WOS. Specifically, studies categorized into non-environmental research fields were eliminated. For example, studies categorized into Engineering Electrical Electronic, Telecommunications, or Medicine were excluded as they digressed from the field of interest (i.e., environmental-related research). The second round of data cleaning was conducted through examining the title and abstract of each publication to keep the most relevant studies for further analysis. The final cleaning of the data generated 143 records with close research foci on the linkage between environment, water, and human migration. These data comprise sub-dataset (A).
Second, we used Google and CLIMIG to obtain grey literature and academic publications not covered in WOS. We applied keyword-based searching rules to Google with its advanced search function ( Table 1). Given that the search results were few (about 400 retrievals were returned per search with a total of 2334 records), we selected relevant records manually and excluded duplication through reading the title of each record. The records considered in Google Search included both academic articles and reports not captured in WOS or CLIMIG, but excluded newsletters, short commentaries, and editorial notes. As a result, the records retrieved from Google Search did not necessarily undergo the cleaning process by field (i.e., first cleaning), and all the selected results made up sub-dataset (B), which contains 121 records.
Meanwhile, we screened the data in CLIMIG (2300 records as of July 2019 when the bibliographic database was accessible for the public). The updated number of records in CLIMIG is 1412 after refinements. The biggest benefit of the existing CLIMIG database is its wide coverage of grey literature. Although this database has been improved, we found, at the time of our data collection (before its data refinements), that some of its records did not directly investigate the interactions between climate change and human migration. For example, we found that several studies on  Dataset 364 records 121 records F I G U R E 1 Schematic representation of data collection and processing.
T A B L E 1 Keyword combinations used to retrieve publications for the domains of human migration and environmental or climate change or water issues.

Database
Keyword combinations WOS ((human OR people OR population) NEAR/2 (migrat* OR relocat* OR resettle* OR mobil* OR move* OR displace*)) & (water NEAR/2 availa* OR flood* OR drought OR disaster* OR hazard*)) Google ("human/people migration*" OR "movement*" OR "resettlement*" OR "relocation*" OR displacement*) & ("water*" OR "water disaster" OR "flood" OR "drought" OR "water hazard*") OR ("climate*" OR "environmental change") CLIMIG (migrat*) & (hazard OR disaster OR flood* OR drought* OR water) human adaptation to climate change and other non-environmental-related immigration studies (e.g., "Networks in the modern economy: Mexican migrants in the U.S. labor market" and "The first South Atlantic hurricane: Unprecedented blocking, low shear and climate change") are also included in the database. To improve the accuracy of the data for our analysis, we first applied "migrat," "flood," "drought," "hazard," "water," and "disaster" as key words to clean the data (Table 1). Next, we removed studies whose full text or abstract was not available or accessible. These search queries produced a total of 2170 records (1118, 233, 259, 128, 154, and 278 for the keywords listed above). All the retrievals were then merged, and duplications were removed, generating 1345 records for a further screen. The second round of data cleaning was conducted by screening the titles and abstracts of all records to exclude those that mention migration and environmental change but do not accommodate interactions between human migration and environments, or do not report any migration cases in their contexts. For example, some studies focus on how to facilitate or improve the wellbeing of environmental-induced migrants, from legal and human rights perspectives, rather than on the interactions, that is, environmental impacts and responses of people. After these records were removed from CLIMIG, 705 records remained matching our criteria and forming sub-dataset (C). Third, we merged the three sub-datasets and conducted the third round of data cleaning. The aim of this step was to keep those data specifically focused on water-related issues through text-mining approaches and full-text assessment. We excluded three types of documents from the dataset: studies of migration projection, studies of ancient human migration, and opinion papers that do not report observed cases. We excluded these studies for two reasons. First, studies that use data and modeling to project future population movement and migration, as well as studies that call for attention to climate migration, are usually not evidence-based and lack validation. Second, patterns of ancient human migration are very different from modern ones because socioeconomic and environmental conditions have vastly improved over the centuries. A total of 364 publications remained as the final dataset for the investigation.

| Methods
A combined quantitative content analysis and qualitative thematic analysis was employed. In the quantitative analysis, we tracked knowledge development of the field using a set of bibliometric techniques based on the data corpus obtained from the developed bibliographic dataset. This analysis aimed to identify trends in knowledge development and the intellectual structure of a research field or discipline, which enables investigators to generalize knowledge patterns from the past and recognize the potential for future development of a research domain (Di Stefano et al., 2010). A wide range of analyses can be performed, such as citation analysis (Huo et al., 2021;, scholarly network analysis (Janssen et al., 2006;, topic mapping and visualization (Fortunato et al., 2018;van Eck & Waltman, 2014), and co-occurrence analysis (Chen et al., 2016). In this paper, we presented general statistics on trends in research and perform spatial analysis of case studies. In addition, we employed a text mining of key terms following the natural language processing proposed in Frantzi et al. (2000). The mining approach helped to remove stop words, such as "is," "a," and "the," and to extract meaningful terms from the text.
Thematic analysis, as a complement to content analysis, moves beyond the quantitative identification of texts and focuses on linkages and patterns of themes in qualitative texts (Braun & Clarke, 2006). In this paper, thematic analysis was conducted based on the assessment of full texts, aiming to generate patterns of migration in a qualitative and comparative synthesis. To this end, we first sorted extracted key terms into categories based on the lexicon and ontology of words (Goyal et al., 2018;Hirst, 2009). These terms were then connected to construct a network based on their co-occurrence relations. Analysis of co-occurrence of words uses the co-occurrence matrix constructed by two words that occur together in the text to visualize the semantic relations of words. The cooccurrence of words in a network represents the frequency of words appearing together in the text. Two words that appear in the same text indicate the thematic connection between them, and a higher frequency means stronger relations (Chen et al., 2016). This method allows researchers to efficiently generalize and conceptualize content connections from a large text-based corpus, a process that would otherwise rely on an assessment of documents through comprehensive reading and summarization (Callon et al., 1991). Next, the network was clustered, using the Driver-Pressure-State-Impact-Response (DPSIR) framework as the guide (Famiglietti et al., 2022;Smeets & Weterings, 1999) to show the causal chain of how changes in water systems and the environment have affected human migration and social responses. The information was therefore synthesized to present patterns of water-driven migration.

| RESULTS
Results are presented in the following subsections: We first present overall trends in the research topic on water-driven human migration, followed by results from the spatial analysis of case studies worldwide, including geographical coverage and hotspots, and the direction of observed or reported human migration. Further, we present a network to illustrate the causal chain of how changes in water systems have affected human migration.

| Overview of research trends
Although research on climate migration has been steadily growing in the last few decades, the role of water in the dynamics of migration has received less attention. It should be noted that some studies in which water/climate is not reflected in their title, abstract, or keyword list were not included for analysis. They are viewed as digressive data, even though they have pointed out in the main text that water/climate is a multiplier that triggers migration, such as some studies on conflicts. However, as the scope of our paper is focused on water, this data-inclusive rule would not affect the exploration of key causal pathways of water-driven migration. Figure 2 illustrates general trends in studies on human migration that relate to water issues and climate change. Lines in the figure show the growth trends in the number of publications in each field over the past four to five decades (the blue line shows the publications on climate change and migration, and the black line shows the number of publications in water-related studies). The shadow area displays water-related studies as the percentage of studies in climate migration. According to our dataset, studies on water-related human migration started in the 1980s, a decade later than climate migration began in the 1970s (based on the CLIMIG database). As illustrated in the figure, studies on climate migration began to increase in 1990 while the research that specifically links water problems to migration received little attention until 2007. The annual number of publications on water-related migration has never exceeded 50% of those on climate migration. The pie chart in Figure 2 shows the percentage of diverse types of documents among the total 364 publications in studies on water-related migration. Journal articles and conference papers dominate the dataset, accounting for more than 60% of all publications. The second biggest group of documents is reports, which account for 22.4% of all publications. A relatively small percentage (12.7%) of documents comprises books, book chapters, working papers, and theses. The top academic outlets and organizations that have published water-related migration studies are listed in Table 2. These publishers have produced 40% of the total studies in the field. The journal Population and Environment has published the most academic articles, followed by the journals Climate and Development and Global Environmental Change. It is interesting that journals with interdisciplinary scopes or have more interests in the social dimensions of environmental sciences have produced more publications than those in water science. This may be a call for water science journals to include more social sciences into their scope so as to better contribute to the interdisciplinary development of water science. It is not surprising that IOM, with 23 reports, has produced the most grey literature given its mission and scope. The United Nations University has published a series of reports, making it the second most productive producer of grey literature. Other national and international organizations have also produced extensive grey literature, including the Institute for Sustainable Development and International Relations (IDDRI), the European Commission, and the Norwegian Refugee Council. These organizations have produced more than 60% of the existing reports.
In addition, we extracted funding information from the acknowledgments of publications to discover key financial supporters. Ninety-two publications do not report funding information. A total of 148 funding bodies were identified in the remaining publications. Among all the funding bodies, international organizations and developed countries have contributed the most to water-related migration studies, although the financial supporters are globally diverse. The top 10 funding bodies, major funded projects, and countries are listed in Table 3. Similar to the trends in report production, the International IOM has supported most water-migration publications, followed by the European Commission through the EACH-FOR Project launched in 2007. Other international and regional organizations, including the World Bank and the European Union, have also made great contributions to this research field. At the country level, there are fewer funding bodies from developing countries than from developed countries, with only two (i.e., China and Bangladesh) identified among the top 10. The United States, Germany, Canada, and the United Kingdom have financially supported more than 80 studies through various sources. Although there is no strong correlation between funding priority and studied countries and regions, studies on the least developed and developing regions obtained the most financial support, and international funding sources such as IOM and World Bank tend to support more large-scale studies than those country-based funding agencies. As climate migration is an interdisciplinary topic, avoiding undesirable migration patterns requires collaboration from multiple disciplines and co-developing solutions among different parties. More diverse and inclusive funding sources would be encouraged to achieve this interdisciplinary and transdisciplinary cooperation.

| Spatial analysis of water-migration patterns
In identifying spatial patterns, we examined study scales with a detailed investigation of contents in countrylevel studies. For the spatial information, we used a spaCy library in Python (Al Omran & Treude, 2017) to extract studied locations from their abstract. In Figure 3, we present the scales of studies, while hotspots and migration directions of studied countries are mapped out in Figures 4 and 5.

| Study scales
As can be seen in Figure 3, studies have been conducted at various scales from local to global levels. Studies at the country and lower levels dominate the whole research field in water-related migration, followed by studies at the global scale. Research has also used continental, regional, and basin scales. Among global studies, there have been more reports published by international and regional organizations than there have been academic articles published in journals. No matter the source, most of these publications present overall arguments or comparisons of observed countries to explore global patterns of human migration and possible linkages between this migration and water-related issues. For example, a report published by the Stockholm Environment Institute examines linkages between water scarcity and population movements by combining global models with local observations (Dow et al., 2005). Another report published by the Food and Agriculture Organization (FAO) reviews global empirical studies, showing that water stress, along with other environmental factors such as temperature, plays an important role in determining migration through its impact on food security (Wrathall et al., 2018). The focus of these studies is the Global South and least developed regions, as they tend to be more susceptible to changes in water variability (Nagabhatla et al., 2020;Warner et al., 2013). For continental and regional studies, the Sahel Desert region in Africa and low-lying islands in the Pacific and Caribbean have received the most attention, as they are confronting chronic water scarcity due to climate warming and increasing inundation resulting from sea-level rise. For studies at the basin scale and water in surrounding areas,

| Global hotspots attracting research attention
The map shown in Figure 4 illustrates global hotspots for case studies at the country level. All countries are identified if they are discussed in documents as case studies or as observed examples. The study hotspots are the most vulnerable regions to climate change. As can be seen in the figure, case studies have been undertaken all over the world, yet the majority were conducted in countries in Africa, Asia, and North and South America, accounting for 39%, 37%, 14%, and 7% of the total number of studies, respectively. Research in Africa is geographically diversified and examines most countries in the east and northwest parts of the continent, mostly in sub-Saharan Africa. This region becomes a hotspot not only because of the growingly observed climate-induced events there that have affected people's livelihoods, but also the weak adaptability of countries and communities to these events resulting in millions of internally displaced populations and cross-border migration. The IDMC report estimated that 2.6 million people were displaced in sub-Saharan Africa in 2021 due to climate-related disasters, such as ongoing droughts and severe flooding due to erratic rainfall patterns in parts of East Africa and the Horn countries (IDMC, 2022). The number of internal and cross-border migration and displacement in this region is projected to increase to 85.7 million by 2050 as the adverse effects of climate change worsen (Rigaud et al., 2018). Countries in southern and western Asia, as well as Latin America & the Caribbean, are also highlighted as hotspots. Countries in these regions are susceptible to climate-induced disasters like hurricanes and damaging hazards like drought that has caused repeated crop failure and other physical damage (Cantor, 2018;Jägerskog & Swain, 2016). A total of 96 countries or territories were identified, in which the populations have more or less suffered from waterrelated disturbances, causing either forced or voluntary migration. Bangladesh, Ethiopia, the United States, and Mexico are highlighted as the top three hotspots for studies among all countries that face water-related migration problems. Studies in the United States are more related to displacement or temporary movements due to damaging hurricane hazards, especially Hurricane Katrina which caused over half of its pre-Katrina population to move out of New Orleans after 5 months of the hurricane (Kates et al., 2006), and also reflect the permanent migration events during the Great Plain drought of the 1930s (McLeman & Ploeger, 2012). Except for studies of the United States, and Canada, as well as those of several countries in Europe (e.g., Netherlands, Portugal, and Spain), most are concentrated in the developing and least developed countries. On the continent of Africa, Ethiopia, Ghana, Senegal, Burkina Faso, Nigeria, Tanzania, Kenya, F I G U R E 3 Spatial scales of studies on water-related migration (As some studies address multiple countries or regions, the total number of studies containing location information is larger than the number of publications in the dataset, i.e., more than 364).
Sudan, and Morocco are widely studied countries, with main focuses on water scarcity, erratic rainfall events and drought hazards that caused crop failure, land and ecosystem degradation, and physical damage. All these countries have appeared in more than seven studies. Ethiopia is the most frequently studied country in Africa, appearing in 20 publications. Apart from Bangladesh, other widely studied Asian countries include India, Vietnam, Thailand, and Syria. Widely studied countries in South America include Mexico, Ecuador, Brazil, Colombia, Guatemala, and Peru.

| Direction of reported migration
As extracted geophysical information does not disclose problems and migration directions found in these places, we conducted a content-based assessment of the full text of studies to investigate reported places of origin and destinations. Data on the exact numbers of migrants are rarely estimated in publications. We thus used times of studied countries or places and their observed destinations in all publications as a proxy to represent directional information of human migration. Figure 5 shows the various times a country was mentioned in publications illustrating the hotspots of original places and destinations. The nodes in the figure denote the countries and the band demonstrates the direction of migration from the place of origin to the destination. Those nodes without a direction to other nodes illustrate the internal migration.
As can be seen from the figure, the water-related migration in most observed or studied countries is dominated by internal migration, that is, within the country (either from rural to rural, from rural to urban, or from urban to urban). Internal short-distance migration usually occurred within lower-income countries, with a few cases observed in developed countries that are related to specific events. Examples include the rural-urban migration in Bangladesh due to flood-related damage and declined water availability for agriculture (Adri & Simon, 2017;Rayhan & Grote, 2007), drought-related migration in rural areas of Ethiopia (Gray & Mueller, 2012b), and the 1930s Dust Bowl mega-drought in the North American Great Plains (McLeman & Smit, 2006). External (or cross-border) migration more often occurred between low and higher countries or among middle-and higher-income countries. For example, the increasing number of migrants and refugees, surpassed the number from Mexico in 2018, move legally and/or illegally to the United States from Guatemalan, El Salvador, and Honduras in recent years, and over half of them have come from the highlands region where people's livelihoods are significantly affected by climate-induced water events and variabilities such as hurricanes and unpredictable rainfall (Blitzer, 2019). Other external migration events have more often been observed from Mexico to the United States, between Bangladesh and India, and between Syria and Jordan, with fewer cases of migration from Tuvalu to New Zealand and between Somalia and Ethiopia. These external migration cases are different from each other. In the case between Mexico and the United States, the changing precipitation patterns is believed to F I G U R E 4 Global hotspots for studies on water-related migration by country. be a driving force of the increasing number of Mexicans migrating to the United States as it results in a significant reduction in crop yields and threatens food security (Feng et al., 2010;Rigaud et al., 2018). Bangladesh is one of the most climate-sensitive countries with over 10 million people living in poverty in the coastal regions. The geographical location and poor economic circumstances make it a vulnerable region to natural disasters including storm surges, heavy regional rainfall, drought, and river bank erosion. These water-related disasters have become key drivers of migration flows in Bangladesh, including flows to urban centers, to foreign countries, especially to India and even more distant places (Bedarff & Jakobeit, 2017). Tuvalu emigration represents a typical case of those people on low-lying F I G U R E 5 Directions between countries of origin and destinations. The number of occurrences in publications was used as a proxy of human migration directions. The band shows the direction of migration from the place of origin to the destination, with the narrower side of the band indicating the destination. The nodes (countries) without a direction to other nodes illustrate the internal migration, for example, Canada and Chile. The bigger the band, the more times the migration phenomenon was reported in documents.
islands suffering from the impacts of repeated storm surges and hurricanes and chronic freshwater shortage and sealevel rise. Due to strong place attachment and identity (Adger et al., 2013;McNamara & Gibson, 2009), migration is usually the last-resort solution of people in these low-lying nations, especially in the Pacific and Caribbean (Mortreux & Barnett, 2009). External water-driven migration cases reported in Syria and Jordan, as well some other countries in the Middle East (e.g., Afghanistan, Israel, and Lebanon) and Africa (e.g., Somalia and Ethiopia) are usually the complex dynamics among water security issues (especially freshwater scarcity), land degradation, crop failure, interstate and inter-communal conflicts, violence, and the failure of governance and management (Findlay, 2011;Reuveny, 2007).

| Causal network of water-related human migration
To identify possible patterns of migration in relation to water, we first extracted frequent terms from both the title and abstract of all documents. We selected those which appear more than five times and excluded any stop words as well as words that are not identical to any research topics, such as "and," "we," and "but". These identified terms were then sorted into categories according to the lexicon and ontology of words. The mining process generated 495 frequent terms and phrases that are frequently mentioned in texts. We used these terms and phrases to acquire the symmetrical cooccurrence matrix to construct a co-occurrence network that helps to represent the closeness between frequent terms.
The connected network of terms was clustered into the DPSIR framework ( Figure 6). In the framework, Driver (D) defines factors that motivate human activities such as agricultural production (Bradley & Yee, 2015), but in many cases, it also defines elements that influence the system formed in the long run and those that are beyond direct control or management, such as climate change and population growth (Ehara et al., 2018;Famiglietti et al., 2022;Foden et al., 2019;UN, 2016). These drivers result in Pressures (P) that change the State (S) of natural and social systems, for example flood events and other natural disasters (i.e., P) that affect water access and the livelihood of communities (i.e., S). Changes in states of the environment produce direct and indirect consequences of water and human systems. These consequences are defined as Impacts (I), for example, asset loss, crop failure, poverty and famine in this study. In Responses (R) to these changes and impacts, actions or measures are implemented to avoid larger negative impacts or to explore new opportunities, such as migration, managed resettlement programs and the financial support of adaptations.
A final number of 191 key terms and phrases were identified which are linked by over 2000 lines to represent the relative important relations in the network ( Figure 6). In the figure, nodes illustrate extracted terms or phrases, and the size of nodes demonstrates the frequency with which corresponding key terms occurred in the collection of documents, meaning the relative importance of terms. All nodes are connected according to their co-occurrence relations. Two terms are connected by a line meaning that they appear at the same time in the text. The frequent terms mentioned in documents refer to data, factors, rural areas, social, disasters, and adaptive policies. For water problems, disasters and drought events are the top two terms. When all terms are connected in the network, causal chains become clearer. Along with environmental and climate change being prominent as major drivers, agricultural development, economic development, population growth and urbanization also appear to be other important drivers of migration. These key drivers exert pressures on the Earth's water system that are exemplified by the intensification of rainfall variability, increase of the frequency and uncertainty of water-related disasters (e.g., drought and flood), and decline of water resources, and thereby increasing the water-related risks to social systems such as food production and health issues. Specifically, drought events act as a critical hazard, resulting in human migration in the observed case areas, which is followed by flooding. These damaging water-related hazards and changes in water conditions as identified at the pressure level are major causes of increases in vulnerability and reductions in livelihoods of households and communities, giving rise to further risks, including economic loss, that trigger the displacement and movement of affected populations as the response. For example, agricultural production is closely linked to food security and farmers' livelihood (S), which is affected by severe drought (P) due to pressures exerted by climate change (D). Intensive agriculture irrigation (D) has resulted in the decline of water storage (S) which combined with drought events (P) to affect farmers' water accessibility (S) and cause crop failure and the loss of income and livelihoods (I), motivating or forcing farmers to migrate to other places for a better life (R). These casual chains have been observed especially in low-and middle-income regions in Mali, Niger, Burkina Faso, and Sudan in the Sahel region in Africa, Bangladesh in Asia, Mexico in America, and some small low-lying islands such as Kiribati and Tuvalu in the Pacific Ocean (Figures 5 and 6).

| GLOBAL PATTERNS OF WATER-DRIVEN HUMAN MIGRATION: EVIDENCE SYNTHESIS
In this section, we generalize patterns of water-driven human migration. The generalization describes the synergistic effects changes in water have on human migration that help to establish a roadmap for future exploration of the interaction between water and human migration. We used the casual chain network ( Figure 6) and previous environmental migration frameworks (Hauer et al., 2020;McDowell, 2011) as guides and compared evidence from the literature to tease out four patterns: variability in water quantity, damaging water hazards and extremes, physical disturbances, and pollution in water systems. These patterns are not independent but are interconnected through multifaceted factors affecting people's decisions to migrate in dynamic ways (Figure 7).
Human migration is a complex social process, which is influenced by multifaceted factors interacting within and across societal institutions, economic structures, political and cultural paradigms, and environmental conditions (Figure 7). Socioeconomic activities produce excessive pressure on water, causing disruptions to global freshwater systems. These disturbances can alter hydrological regimes and conditions, resulting in water variabilities at various levels F I G U R E 6 Co-occurrence networks of key terms extracted from titles and abstracts. illustrated as "Water systems" in the figure. Climate change, as a multiplier, is exacerbating these variabilities and amplifying water risks to society. Changes in features of water systems have negative impacts on the associated life support systems, such as land and soil for food production and ecosystems that provide valuable services to humans and other creatures. Degraded life support systems cascades to result in the increase of the vulnerability of the entire society (i.e., social, economic, political, and demographical vulnerability), which further decreases people's livelihood and wellbeing exemplified by the increase of damage costs, loss of income and livelihoods, food insecurity, water competitions and conflicts. In response to these synergistic changes, both individuals and governments take action, generating either positive or negative feedback to water systems, for example, increasing conflicts and environmental degradation in destinations (Hermans-Neumann et al., 2017;Pricope et al., 2013;Rigaud et al., 2018). Individuals may be well-prepared for these changes or might choose for unexpected migration in reaction to these changes. When people consider migrating, their decisions are driven not only by changes in water and its associated system states (highlighted in the left cycle in the figure) but also by their assessments of resources (e.g., situations exemplified in the upper right box of the figure) and their perception of risks and other obstacles they may face now and in the future (e.g., exemplified in the bottom right box of the figure).

| Variability in water quantity drives human migration
Variability in water quantity, in particular water scarcity, is a driving force of human migration as it not only affects water availability for associated life support systems, including land, soil, and natural habitat, which are indispensable for the existence of all life on Earth, but also increases the vulnerability of social-ecological systems and leads to the loss of people's livelihood. Sustainable water availability maintains the fertility of land and soil for agricultural production, ensuring farmers' livelihood and food security for the rest of the population. Variability in precipitation, for example, affects soil moisture and cultivability of land which will further impact farming patterns and crop production. The impact of water variability on people's life quality and safety is more significant in the world's dry regions where freshwater resources are substantially depleted (Rodell et al., 2018). The increasing water variability in these regions has already been observed as a significant multiplying driver of migration, which is likely to be accelerated in the decades to come (Foresight., 2011;Rigaud et al., 2018).
In the Sahel and Sahara regions of Africa, groundwater, much of which is found at depths greater than 500 m, is the only source of water for most local uses (Hamed et al., 2018). Livelihoods are dependent on the availability and storage of freshwater and are very sensitive to rainfall variability, heightening the importance of a reliable groundwater source in these regions. However, water shortages lead to crop failure and increase water-driven conflicts, which are exacerbated by continued groundwater depletion (Calow et al., 2010 Stafford Smith et al., 2011), Sudan (de Ramaix, 2011), Mali, Senegal, and Mauritania (de Bruijn & van Dijk, 2003Liehr et al., 2016). In Morocco, changes in rainfall patterns have substantially affected regional agricultural production patterns. Precipitation in this area started to decline in the mid-1970s and is expected to decrease by up to 30% by 2050, intensifying the water scarcity impacts on irrigation for agricultural food production in many river basins, especially the Oum Er-Rbia river basin, which contains half of Morocco's irrigated agriculture (Tangermann & Bennani, 2014). The increasing water shortage has driven a large number of people to migrate from agricultural areas along the Oum Er-Rbia river basin to urban areas (Ouassou et al., 2007). In South Africa, the chronic water scarcity in Cape Town and the surrounding areas with the well-known "day zero" water crisis has been a critical factor of increased conflicts and migration (Adaawen et al., 2019). The emerging drying trend in this region (Reager et al., 2016) will likely make people more vulnerable to the variability in the water supply and accelerate environmental migration. Similar patterns are also found in dry regions in Asia. Long-term over-exploitation of water resources for irrigation has reduced water storage in the Ganges-Brahmaputra basin (which includes one of the major aquifers in Asia) and forced permanent internal migration in northwest Bangladesh (Kabir et al., 2018). Although the exact number of migrants has not been documented, an increase in local conflicts between Bangladesh and India since the 1990s due to water scarcity in the Ganges basin has long been believed to be the main reason for internal and cross-border migration in this region (Swain, 1996). According to Ahmed (2009) and Baten and Titumir (2016), the lack of environmental security (mainly water security, i.e., the Farakka-Barrage water dispute and the drying Ganges River), among other factors, would have motivated nearly 67% of people to leave the Khulna-Sathkhira, Rajshahi, Dinajpur, and Nilphamari regions in Bangladesh. An empirical study of a west coast province in India showed that the increase of 100 feet (30.5 m) in the depth of groundwater wells could be associated with about a 2.5% increase in the probability of a household having at least one younger migrant (Fishman et al., 2013). In the region near Suruç in eastern Turkey, the depletion of groundwater reservoirs for cotton production has been observed since the 1950s. A steadily declining groundwater supply supported cotton production for decades, but when the groundwater became severely depleted in the 1990s, the majority of the population was forced to leave the region, with some permanently moving to urban areas and others becoming seasonal migrants (Kadirbeyoglu, 2008).

| Damaging water hazards and extremes force people to migrate
Water-related hazards result from complex interactions in the ocean atmosphere-land process cascade (UNESCO, 2016), which are represented as rapid and slow events (or fast-onset and slow-onset variables). Climate change, as the multiplier, is generating significant impacts on variability in freshwater systems and increasing the likelihood and damage of extreme water hazards. The surge of these types of water hazards threatens ecosystems' health and vulnerable livelihoods through hydrological processes, causes the loss of human lives, increases material damage, and escalates costs.
People's decisions to migrate are critically determined by either rapid or slow disastrous events, or the combination of both. Rapid events, such as floods and cyclones, directly damage infrastructure and possessions and threaten safety. They usually lead to internal population displacement and temporary movement. In many cases, these temporary migrants are involuntary and eventually return to their place of origin. For example, in many African countries such as Mozambique, young adults who are the main household breadwinners (Osbahr et al., 2008) move to other places for external wage opportunities to rebuild housing and lost assets caused by flooding, resulting in the growing rates of movement from rural to urban regions. On the other hand, slow events, which occur gradually such as droughts, bring chronic changes to water systems that push humans' livelihoods to a tipping point that forces the decision to migrate. Although this tipping point is challenging to identify as thresholds are highly context-specific (Hoffmann et al., 2020), the dependence on and cascading effects of human-water systems are seen in many places (Rocha et al., 2018;Xu et al., 2021). In global south countries such as Ethiopia, Bangladesh, and Mexico, the impacts of such gradual changes on people's livelihoods have been well documented, showing the potential upward trends in the number of climate migrants through 2050, which will increase dramatically afterwards (Rigaud et al., 2018).
Slow-onset drought impacts have significantly affected short-term temporary migration internally and externally, as evidenced in the African rural highlands and hinterland (e.g., Ethiopia and its neighbors) and in dry regions in Asia (e.g., the Middle East region). A multivariate assessment of rural Ethiopia has reported that drought impacts have doubled labor-related migration, with 10% of adult men eventually migrating each year (Gray & Mueller, 2012b). A severe drought from 2004 to 2005 in the pastoral areas on the Ethiopia-Kenya-Somalia border has resulted in livestock losses of up to 70% and forced mass migration of pastoralists in search of water, food, jobs, and relief aid (ODI, 2006). In 2009 and 2012, the average number of people displaced by drought in Somalia was 83,043 per year (Glinnetti & Frank, 2014). In west Asia and the Middle East, the severe drought in Syria between 2006 and 2015 resulted in multiyear crop failures that created food insecurity for more than 1 million people (Gleick, 2014), contributing to increased asylum-seeking (Abel et al., 2019) and a mass migration of an estimated 1.5 million people into cities (Abel et al., 2019;Kelley et al., 2015) and to its neighboring country Lebanon (Jägerskog & Swain, 2016).
Although gradual slow events seem to have more significant impacts, both slow and rapid water hazards are believed to have a role in forcing people to move internally and externally, in particular when these hazards generate physical damage to socioeconomic capital. The combined impact of both types of events on migration patterns is typically evidenced in Bangladesh where internal rural-urban migration (Chen & Mueller, 2018;Gray & Mueller, 2012a) and external migration to India have been observed (Quencez, 2012). Further, more frequent occurrences of rapid events could force people in coastal regions and on small islands to permanently migrate, but their migration decision is often constrained by the ability to move and social institutions. For example, social, demographic and physical drivers are perceived to drive migration in areas of Bangladesh where rapid events, such as coastal floods and cyclones, are dominating hazards as they often result in crop failures (Chen & Mueller, 2018;Chumky et al., 2022;Gray & Mueller, 2012a). Extreme rainfall events are not always the causes of migration in Bangladesh, but recurrent events such as coastal flooding have triggered massive permanent population mobility in the country's coastal provinces (Carrico & Donato, 2019), as similarly evidenced in Vietnam (Jäger et al., 2009). Like populations in many coastal regions, those on low-lying islands suffer from a high risk of disastrous water hazards (e.g., inundation, rainfall variation, and coastal erosion) due to warming temperatures and sea-level rise. The people who live there are one of the most vulnerable and least resilient populations to climate-induced damaging water hazards because of their inability and unwillingness to migrate (i.e., limited financial resources, a strong sense of place, and bonding relations) (Adger et al., 2013;Hauer et al., 2020;McMichael et al., 2021). While there is little strong evidence of the link between migration and water hazards, as climate change continues, people on these small islands will most likely have to be relocated using external financial support.

| Physical disturbances to water systems cause human migration
In this study, physical disturbances to water systems are defined as changes in the shape and structure of water bodies and the physical degradation of their associated landscape and ecosystems. Changes in water storage and drainage directions due to climate change and human activities can modify the landscape and fragment joint ecosystems. For example, pastoralism is the livelihood of the majority of people living in the drylands of East Africa where accessibility to water has become a prominent risk to local people for farming. The availability of water is an important structuring component in retaining rangelands that sustain the pastoralism. Declining water storage on the land surface and subsurface can dry up the rangelands on which livestock relies, further affecting the livelihood of pastoralists (IOM, 2016). Pastoralists living near the borders of Ethiopia, Kenya, Somalia, Sudan, and Uganda are viewed as a group at high risk of being displaced by the loss of rangelands, and cross-border movement and displacement are frequently observed in this region (Glinnetti & Frank, 2014). In 2007 and 2009, reduced pastoralism in Somalia resulted in the resettlement of over 20,000 pastoralists in search of alternative sources of food and income (Findlay, 2011).
Major dams and water infrastructure constructions cause mass migration because many of these projects dry up wetlands and alter hydrological conditions that sustain people's lives along the river. These major projects have resulted in a wave of environmental migration or refugees, especially in the arid Sahel region of Africa and many transboundary river basins (Pearce, 2017). The water engineering-related human migration can be well managed or preplanned through relocation programs by governments such as the relocation of over 1 million population in China due to the construction of the Three-Gorges Dam (Heming & Rees, 2000). However, this type of migration can also be unmanaged or unexpected. For example, the ongoing Ganges water competition and conflicts have been escalated by the construction and operation of the Farakka Barrage in India, which is exacerbating the dry season and increasing the likelihood of flooding, thus intensifying the water stress in Bangladesh (Climate Diplomacy, 2020;Hassan, 2019). Since the dam became fully operable in the 1970s, it has generated significant adverse impacts on Bangladesh, including reducing river flow, modifying the river bed and river morphometry, degrading surface and groundwater and aquatic systems, and increasing salinity, all of which have started to affect public health and people's access to water downstream (Hassan, 2019;Thakur et al., 2012). The hydrological conditions affected by the Farakka Barrage have caused the mass influx of environmental migrants into India every year, further creating secondary conflict in the destination, particularly in the region of Assam and villages near the river (Iqbal, 2010).
The example of the Ganges River represents a common water security issue that could become the driving force of human migration around transboundary river basins in the world. In the region around Lake Chad, the Maga Dam located in the southern area of Lake Chad near the border of Chad and Cameroon has diverted 70% of the flow of the Logone River to rice farms. This diversion has dried up parts of the floodplain pastures that once supported 130,000 people and has dramatically reduced the inflow to the lake (Pearce, 2017). The shrinkage of Lake Chad and upstream water impounding have put the local population at high water risk and have increased competition and conflicts, forcing more than 2.6 million people (both internally displaced and external refugees) to leave the lake region (i.e., across Nigeria, Niger, Cameroon, and Chad) from 2013 to 2016 (IOM, 2016). Further, the construction of hydropower dams in the Mekong River basin has influenced flood patterns that triggered household and individual migration in the Delta area of Vietnam (Dun, 2011). With temperatures rising and rainfall becoming more erratic in many parts of the world, tensions will likely become increasingly strained over the management of river flows across international borders.

| Pollution to water systems pushes human migration
Water pollution stems from accumulative biochemical disturbances generated by long-term human activities or by natural disturbances such as floodwater. Human-induced water pollution and contamination have been reported in almost all water bodies across the globe, including rivers, oceans, lakes, reservoirs, and aquifers (Famiglietti et al., 2022), affecting people's access to drinking water and impacting public health. Although the evidence of the impacts of water pollution on patterns of human migration is not as strong as that of disturbances to water quantity, populations affected by degraded water quality can also be forced to migrate in search of better and safer water resources. For example, accident-caused water pollution has been documented as a critical causality of population displacement, such as the Flint (Michigan, United States) water crisis that is caused by lead contamination in drinking water sources, triggering mass emigration of Flint residents (Morckel & Rybarczyk, 2018). Polluted water has motivated the migration decisions of households experiencing increased travel times and costs to fetch or purchase freshwater, such as women in the Sonora State of Mexico (Nagabhatla et al., 2020).
Excessive groundwater pumping along coastlines has contributed to saline contamination of many aquifers, land subsidence, and inundation, changing patterns of internal migration in many coastal cities, such as Semarang, Pekalongan, and Jakarta in Indonesia (Buchori et al., 2018;Khairulbahri, 2022) and along the southwest coast of Bangladesh (Rakib et al., 2019). The situation will worsen as sea levels rise under climate change. Saltwater intrusion into groundwater systems has already become a major issue, threatening millions of people's access to water and their livelihoods (Hauer et al., 2020). In global coastal zones, the influencing force of salinization and erosion on migration decisions is believed to outweigh that of flood inundation (Duijndam et al., 2022). For example, in the vulnerable coastal communities in Bangladesh, exposure to salinity contamination has increased drinking water risks and resulted in the spread of human diseases and high treatment costs. Strong risks for migration risk have been identified in areas where drinking water sources are threatened by salinity contamination, particularly when adaptive capacities are low, including in communities in Gabura, Munshigonj, Atulia, Burigoaliny, and Padmapukur in the southwest of Bangladesh (Rakib et al., 2019).
Just as water quality decline can trigger emigration, it can also result from immigration. Migrants may contribute to or cause water pollution in hosting destinations, further triggering conflicts and the surge of new migration. Waterdriven conflicts and migration in Middle Eastern countries provide a good example. The mass migration and exodus of refugees from Syria following its civil war in 2011 exerted great pressure on the receiving neighboring countries of Lebanon and Jordan (Weinthal et al., 2015). Demand for water grew across Lebanon from 0.1 to 7.4 million cubic meters per year, 30% and 24% of which came from public water supplies and groundwater wells, respectively (Jägerskog & Swain, 2016). This greater demand increased sewage generation, which remains untreated and is disposed of in surface and groundwater systems. The increase of untreated sewage has rendered more freshwater in Lebanon unusable (Jägerskog & Swain, 2016). With declining water quality and rising global water demands, migration triggered by water pollution is expected to grow in decades to come.

| WATER-DRIVEN HUMAN MIGRATION: CHALLENGES AND PERSPECTIVES
Although not the only reason, water has a pivotal role to play in shaping new patterns of human migration and movement locally and internationally. Water-related fast and slow events or variables could be major triggers forcing people at risk to emigrate from their place of origin, particularly when the event has caused damage to properties and reduced capital. As demonstrated in the evidence, for example, in Bangladesh, Burkina Faso, Sudan, and Senegal, water may have played a crucial role in determining human-environment interactions by pushing food systems and human livelihoods toward critical tipping points. Rapid water events are likely the causes of displacement and short-term migration, while slow events and the reoccurrence of rapid events may be the main causes of long-term/permanent and large-scale migration in many places, for example, Suruç in Turkey and the Khabur River in Syria. Migration patterns can be even more complicated when both fast and slow variables interact over time and across space, as evidenced by the case in Bangladesh where storm surge, flood, drought, and river erosion are combined to cause the majority of temporary population displacement, labor migration, and permanent migration. Moreover, water-related threats undermine food security and people's wellbeing, which together have great potential to drive human displacement (especially of vulnerable individuals and communities) and change global migration patterns. Further investigation of the dynamic interactions between fast and slow variables, and their link to social systems would be critical.
Understanding water-driven (or environmental) migration requires systematic thinking of the interactions between environmental change and social responses. Previous studies have asked questions of where, why, and who may be moving in the face of environmental changes and what factors may affect their decisions to move (Wrathall, 2012). Yet, the temporal dynamics of environmental migration are less understood, for example, the timing of people's decisions to move or stay as the response to the changed environmental conditions. Rarely addressed in the literature are critical hydrological thresholds and their interactions with thresholds of joint system variables, which might determine the timing of migration. What is the minimum amount of water required to sustain farmers' livelihoods in an area, past which food production would collapse even with the support of external sources? When food production ceases, what decisions on migration or movement have to be made as an adaptation to this situation? Answers to these questions are unclear because few studies have addressed them. This knowledge gap calls for investigations to determine the critical timing of interventions to avoid undesirable demographic shifts, causing locked-in paths of development (e.g., the competing trap for resources in in-migration destinations). Further, the hysteresis effect of regime shifts has been observed as a common phenomenon in interconnected systems due to interactions among critical system variables (Scheffer et al., 2002). This phenomenon has been observed in environmental migration studies, such as the case of Santa Rosa de Agu an in Honduras (Wrathall, 2012). Understanding the temporal dynamics of water-driven migration needs to consider such hysteresis effects among variables, as this knowledge will help develop informed decision-making before undesired patterns of migration are established.
It should be highlighted that interactions between water and human migration are bidirectional, but such bidirectional impacts may vary case by case depending on migration patterns and whether they are internal or cross-border and rural or urban. Specifically, water can be the driver of human migration. The mobility of large numbers of human migrants can, however, create a positive feedback loop within water systems in places experiencing a large influx of displaced peoples, accelerating water insecurity in these hosting destinations. News reports during the Syrian war underscored that water in Jordan was already overconsumed and depleted, making it extremely difficult to meet the increasing water needs of Syrian refugees and rising domestic water-related conflicts (Sweis, 2012). Severe droughts and rainfall variability in the 1970s and 1980s led to an influx of people in the Lake Chad region seeking fertile land and water on the shores. The increased population has led to tensions between resource users and the decline of water storage in the region. The drying lake has forced many families who relied on its water to migrate to other areas in search of a reliable water supply (Mwiturubani & van Wyk, 2010;Usigbe, 2020). As a result, unmanaged environmental migration could lead to maladaptation. For policy makers, failure to secure water resources for people will worsen conflicts and lead to millions of people moving out of their homes (Abel et al., 2019). Targeted water policies are recommended to focus on those vulnerable regions based on real-time monitoring data on hydrological conditions, such as remote sensing and more frequent population surveys.

| CONCLUSION
Previous studies have reported a large number of migration cases that are closely linked to water-related events. Yet, a comprehensive synthesis to understand how different dimensions of physical water systems affect human migration is limited. This is due partly to the fact that a large body of evidence is qualitatively described and the knowledge remains fragmented and buried in the mass of textual big data. This paper, using a text-mining approach and thematic analysis, sheds some light on the knowledge of water-related migration research to generalize global patterns of water-driven human migration.
This overview discovered that attentions on general topics of climate migration are far more than those on the specific link between water and migration. To advance the understanding of the water-migration interaction and explore effective solutions for undesirable migration patterns, the wider participation of water scientists and engagement with communities are needed which would benefit from interdisciplinary collaborations and more diverse and inclusive funding sources for co-development projects. The quantitative analysis also supported previous findings that waterdriven migration usually occurs internally and that the population in low-and middle-income countries and in dry regions are the most vulnerable and more likely to migrate or be displaced. Although the casual chain of multifaceted factors indicated the complicated influencing mechanism of people's migration decisions, some patterns can be drawn from existing evidence that are linked to four interconnected interruptions to physical water systems: variability in water quantity, damaging water hazards and extremes, physical disturbances to water systems, and water pollution.
In summary, interactions between water and human migration are bidirectional. Global human migration is a systematic social process affected by multiple external driving forces and internal adaptive capacities, as well as their interactions. Water-related issues do not stand out as the single trigger but can play a crucial role in determining people's migration decisions. Understanding bidirectional water-migration dynamics would benefit from (1) systematic thinking of the interconnections between changes in water and in migration patterns, especially the spatial and temporal distinction (e.g., rural vs. urban, local vs. international, and temporary vs. permanent); (2) further investigations of the interactions between fast and slow variables in water systems, and their dynamic link and hysteresis effects to other socioeconomic variables (e.g., intersectional relations with other issues such as violence and conflicts for a better estimate of water's contribution to migration); (3) an integrated water-migration database to help identify early-warning signals of damaging water hazards that may result in undesirable migration; and (4) targeted water policies that focus on building the resilience of vulnerable regions and population to climate change.