The Subjective Climate Migrant: Climate Perceptions, Their Determinants, and Relationship to Migration in Cambodia

Abstract

This article examines the factors shaping the perception of climate change and the relationship between climate change perception and migration. Drawing on a 691-case survey of climate perceptions in Cambodia, it explores three dimensions of climate change perception. The first is the relationship of climate change perceptions to space, geography, and scale. Second is the influence of livelihoods to climate change perceptions, and third is the relationship of climate perceptions to migration. The results show that perceptions of climate change are not significantly influenced by spatial distance, meaning that divergent or even opposite climate perceptions might coexist within a relatively small geographical area. The data, however, show that climate perceptions are significantly influenced by both engagement in certain primary livelihoods and contextually specified socioeconomic marginality. Despite this subjectivity of climate perceptions, a strong, statistically significant relationship exists between climate change perception and the prevalence of migrants in the household. Overarchingly, the article challenges efforts to infer direct linkages between climate data and human behavior, arguing instead for a more subjectively attuned understanding of the impacts of climate change on migration, to account for the multiple factors that influence perceptions of and responses to climate change.

本文讨论了气候变化认知、气候变化认知与移民关系的塑造因素。通过柬埔寨691 个气候认知调查案例，本文探讨了气候变化认知的三个方面：气候变化认知与空间、地理和尺度的关系，谋生对气候变化认知的影响，气候认知与移民的关系。结果显示，气候变化认知受空间距离的影响不大，意味着不同甚至相反的气候认知可以在相对较小的地理范围内共存。然而，数据显示，某些谋生手段、特定社会经济条件下的边缘化程度，对气候认知的影响非常大。除了主观的气候认知，气候变化认知和家庭成员的移民比重之间有很强的统计相关性。总体来说，为了解释气候变化的认知及反应的影响因素，本文质疑了直接关联气候数据和人类行为的研究，认为应更主观地去理解气候变化对移民的影响。

Este artículo examina los factores que conforman la percepción del cambio climático, y la relación entre percepción del cambio climático y migración. Con base en un estudio de 69 casos de percepciones climáticas en Camboya, el artículo explora tres dimensione de percepción de cambio climático. La primera es la relación de las percepciones del cambio climático con el espacio, la geografía y la escala. La segunda es la influencia de los modos de sustento con las percepciones del cambio climático, y la tercera la relación de las percepciones del clima con la migración. Los resultados muestran que las percepciones del cambio climático no están significativamente influidas por la distancia espacial, lo cual significa que percepciones del clima divergentes o incluso opuestas podrían coexistir dentro de un área geográfica relativamente pequeña. Sin embargo, los datos muestran que las percepciones están influidas significativamente por el compromiso con ciertos medios de sustento primarios, y contextualmente con una marginalidad socioeconómica especificada. A pesar de la subjetividad de las percepciones climáticas, existe una fuerte y estadísticamente significativa relación entre la percepción del cambio climático y la prevalencia de migrantes en el hogar. Por sobre todo, el artículo reta los esfuerzos por inferir lazos directos entre los datos del clima y la conducta humana, respaldando por el contrario un entendimiento de los impactos del cambio climático sobre la migración más a tono con la subjetividad, para tomar en cuenta los múltiples factores que influyen las percepciones del cambio climático y las respuestas al mismo.

Key Words:

• adaptation
• climate change
• climate migration
• climate politics
• mobilities

关键词：:

• 适应
• 气候变化
• 气候移民
• 气候政治
• 流动性

Palabras clave:

• adaptación
• cambio climático
• migración por clima
• movilidades
• políticas climáticas

As declarations of a “climate emergency” by a succession of countries place climate change “at the centre of policy and planning decisions” (Turney 2019, 1), its human impacts have moved to the forefront of contemporary politics and policymaking. In particular, its impacts on human migration have been cast as a key global policy priority in recent years, with a growing list of governments, from the United Kingdom to Germany to the European Union, adopting a securitized lens on the issue (White 2012; Boas 2015). The result has been a “weaponising” of both migration (Misra 2018, 1) and adaptation (K. A. Thomas and Warner 2019), wherein the assumed necessity of solutions has come to shape the nature of the phenomena themselves, via a “transfiguration of threatened populations into threatening populations through climate change securitization” (K. A. Thomas and Warner 2019, 1).

Above all, the issue of climate migration is pervaded by a sense of exceptionalism: the undergirding assumption that migration promulgated by climate change is fundamentally distinct in character from migration in other forms. Not only does this contribute to its discursive construction as a “problem” but in a practical sense shapes the tools of its investigation. The assumed need to discern that part of mobility that is specifically and separately engendered by climate change has spawned a very particular set of methods. Above all, Ricardian analysis—wherein existing population data sets are cross-referenced with climate data—have come to dominate the analysis of climate migration, with global quantitative figures becoming characteristic of a field increasingly defined by its “quest for ever larger numbers” (Jakobeit and Methmann 2012, 301). Well-publicized studies over the years have invariably centered on headline figures, from the Overseas Development Institute estimation of 24 million displaced in 2016, via World Bank (2018) and academic (Jacobson 1988; Myers 1997, 2002; Stern 2007) projections of between 140 million and 1 billion climate migrants by 2050, to the current leading figure of 2 billion by 2100 (Geisler and Currens 2017).

Despite the range of these estimates, the aim in each case is to discern a more or less direct relationship between physical environmental and human population data, a goal to which much criticism has been addressed for its failure to consider the social contexts and conditions of migration (Piguet et al. 2011; Faist and Schade 2013; Gesing, Herbeck, and Klepp 2014; Nicholson 2014). Climate migration scholars have responded by introducing intermediary social and cultural variables (e.g., Adger et al. 2013; McCarthy et al. 2014). Yet the most intractable problem remains that of scale. Because climate migration research relies on existing climate data sets, the scale of analysis is determined not by the research question at hand but by the logistics of climate measurement infrastructure (Gubler et al. 2017). Not only are the “resolution,” coverage, and quality of these data variable—often problematically so in the Global South (Fussell, Curtis, and DeWaard 2014; Gubler et al. 2017; Johnson and Rampini 2017)—but the very nature of the data available has generally precluded quantitative analysis of climate migration at the community scale familiar to migration and adaptation scholars.

This scalar disjuncture has led in recent years to growing interest in subjective climate change perceptions, as a measure of resilience (Clare et al. 2017), risk perception (Funatsu et al. 2019; Zander, Richerzhagen, and Garnett 2019), and risk management (Carmenta et al. 2017). Nevertheless, this work has been limited in scope. Up to now, work examining climate perception in the Global South (Maddison 2007; Slegers 2008; Deressa, Hassan, and Ringler 2011; Ejembi and Alfa 2012; Vulturius et al. 2018; Funatsu et al. 2019) has tended to be directed toward testing the closeness of fit between perceived and recorded climate data, with relatively few studies (see Zander, Richerzhagen, and Garnett [2019]; De Longueville et al. [2020], for two recent exceptions), exploring the impact of climate perception on migration specifically.

This study seeks to extend this analysis, demonstrating not only that climate perceptions are significantly related to migration decisions but also the manner in which these perceptions are subjectively attuned to social and livelihoods factors. Decentering the question of fit between physical data and human perceptions of climate change, the article therefore explores the determinants and value of climate perception data, examining in depth the relationship between climate change perception and migration in three Cambodian communities defined by large internal socioeconomic differences between marginal and better-off households. A key point of the article is thus that climate change perceptions are not an effective proxy for physical climate data but should constitute an important sphere of research in their own right when wanting to understand the local impacts of climate change, as well as adaptation measures such as migration.

Connected to this, the article additionally challenges the frequently used contention that adaptation in situ is preferred among those with the resources to do so over migration (Bardsley and Hugo 2010; Zander, Petheram, and Garnett 2013; Laurice Jamero et al. 2017; Stojanov et al. 2017). In contrast to the recent literature on climate migration (Richards and Bradshaw 2017; Conway et al. 2019), our results indicate that socially and economically marginal households are not only less likely to migrate than better-off households—results that closely align with the migration literature in general (Stark and Bloom 1985; Skeldon 2002)—but less likely also to subjectively perceive changes in the climate.

After outlining the conceptual, methodological, and empirical framing of the study, the article addresses how perceptions are linked to adaptation measures focusing on migration. First, we outline the role of livelihoods in shaping climate perception; second, we highlight the relationship between climate perception and migration; and, finally, we interrogate the assumption that climate migration is most likely to occur among the most marginal members of communities. A discussion of the implications of these findings, focusing in particular on the role of climate perceptions within climate migration research and the need for a more nuanced and subjective approach to linking the two, will follow. A conclusion ends the article.

Climate Migration and the Subjectivity of Climate Perception

Among the myriad impacts of the changing climate, one of the most frequently and passionately discussed is the impact on human migration. In public discourse “we are told that such movement is unprecedented, that it represents a crisis, a flood, a disaster” (Hamid 2019, 20). Moreover, this is a position that pervades scientific analysis, encouraging a reliance on “topographies and maps of so-called hot spots of existing and expected climate-change migration,” rather than migration systems more broadly (Klepp 2017, 8). Nevertheless, even as the policy relevance of understanding and modeling such relationships increases, criticism of the methods employed to do so has risen in tandem. Climate migration models have been accused of arbitrary causal reasoning on the one hand (O’Brien 2013; Nicholson 2014; Meyer 2016) and binary and monofactoral thinking on the other; that is, does a given environmental variable impact aggregate migration or not—in others (Gómez 2013).

In response, “a wave of new social science research into these hitherto under-emphasized cultural dimensions of climate change” (Adger et al. 2013, 112) has emerged, bringing much-needed nuance to the field. A key practical challenge facing climate modeling, however, is low-resolution climate data, especially in the Global South (Eklund et al. 2016), that leaves many studies with few practical options beyond upscaling and interpolating data in such a way as to elide issues of quality and distribution in certain areas (Eklund et al. 2016). Recognizing such shortcomings, efforts have been made to make greater use of alternative, smaller scale data on changing livelihoods, migrations, and adaptations. Nevertheless, a significant discrepancy remains “between the large scale at which biophysical impacts of climate change are generally studied, and the local scale of analysis typically adopted in bottom-up studies” (Conway et al. 2019, 504).

In recent years, interest in bridging this scalar lacuna—alongside a wider sense that “people-centred effects cannot be examined through meteorological observation alone” (Rankoana 2018, 367)—has resulted in growing attention toward the potential of climate perception as a proxy for climate impacts (see Karki et al. [2020] for a review of the climate perception literature). Studies exploring perceptions of climate change in both the Global North (Diggs 1991; Leiserowitz 2010; Semenza et al. 2008; Weber 2010; Akter and Bennett 2011; Hansen, Sato, and Ruedy 2012; Rankoana 2018; Whitmarsh and Capstick 2018) and Global South (Vedwan and Rhoades 2001; Hageback et al. 2005; Maddison 2007; D. S. Thomas et al. 2007; Ishaya and Abaje 2008; Slegers 2008; Gbetibouo 2009; Mertz et al. 2009; Deressa, Hassan, and Ringler 2011; Ejembi and Alfa 2012) show a majority of the study population having perceived climate change.

As interest in the value of climate perception data grows, however, it is increasingly recognized that the perception of climate change is shaped not only by physical changes in the climate but also by intermediary socioeconomic factors that frame the experience of climate change at community, household, and individual scales. The most recent literature acknowledges “both household internal dynamics and broader external factors, such as agro-ecological, climatic, and institutional economic and political frame conditions” (Žurovec and Vedeld 2019, 1). More specifically, acknowledgment is growing that environmental changes are perceived in large part through the lens of people’s livelihoods (Slegers 2008; Nielsen and Reenberg 2010; Ejembi and Alfa 2012; D’haen, Nielsen, and Lambin 2014; Macchi, Gurung, and Hoermann 2015; Rankoana 2018; Žurovec and Vedeld 2019; Funk et al. 2020) due to their impact on resource availability.

This is in some senses intuitive. Water scarcity—or the unreliability of water availability—for example, is both a key impact of climate change and strongly dependent on the use to which this water is put. Whereas paddy rice, for example, is estimated to require an average of 1,673 liters of water per kilogram to produce, cabbage requires only 237 (Mekonnen and Hoekstra 2011). Assuming that both access water from a common source, therefore, a rice farmer is likely to be considerably more sensitive to changes in the availability of water than a cabbage farmer.

From this perspective, the inherent difficulties in disentangling local environmental and broader climatic factors are rendered moot by the centrality of livelihoods to climate knowledge. Climate perception is very likely both a communally and individually mediated knowledge: the product of both received wisdom and subjective, personal experience. People experience risks themselves, but so, too, they invest time “communicating about risks” (Wahlberg and Schoberg 2000, 31), meaning that “interpersonal” and “personal” risk perception play a role in shaping conceptions of the climate (Wahlberg and Schoberg 2000, 31). Thus, climate perception exists at the nexus of these two framings. It is both tangibly felt and absorbed from the wider milieu, so that group membership, as well as livelihoods themselves, may influence climate perception (Warner 2010), especially in countries—such as the one explored in this article—where formal sources of climate data are generally lacking.

Moreover, the corollary here is of equal significance. Just as group membership shapes climate perception, so, too, then must group nonmembership, a principle with far-reaching implications for the much-discussed impact of climate change on marginal communities, or marginal members of communities. Despite nuanced analysis of the nexus of “marginality” and migration in a general sense (see, e.g., Kretsedemas, Capetillo-Ponce, and Jacobs 2013), this is a concept often left unpacked in the climate migration literature. Whereas vulnerability is essentially an externally relational state—that is, it describes the “exposure, system sensitivity, and adaptive capacity” (McLeman and Hunter 2010, 1) to something distinct from the community itself—marginality is internally relational: Somebody may be marginal regardless of the presence of a given threat. Thus, as a social and political construct, it is both highly variable and profoundly attuned to environmental and social context.

From this standpoint, therefore, this article is directed toward two interlinked objectives. Leaving the discussion of climate perception as a proxy for physical climate data aside, it aims first to clarify the nature of the factors that influence climate perception, the relation of these factors to livelihoods, and the role of marginality in shaping climate perception. Second, the article seeks to demonstrate the utility of climate perception data as a tool within climate migration analysis. In this way, the approaches outlined here serve to advocate a more person-centered, subjective analytical approach to the relationship between climate change and human mobility.

Methodology

Data were collected across nine village sites (see Figure 1) located in some of the most climate-vulnerable regions of Cambodia. The nine sites were selected on the basis of (1) high levels of environmental vulnerability, (2) high levels of migration, and (3) high levels of socioeconomic marginality. They are grouped into three agro-ecological zones that cross-cut provincial boundaries to better reflect the environmental characteristics of the country. The first zone, the Tonle Sap and lakeside zone—incorporating three floating villages in Kampong Chhnang, Pursat, and Kampong Thom—was selected because of the vulnerability of its fisheries to high temperatures and prolonged droughts (Oudry, Pak, and Chea 2016; Uk et al. 2018). The second zone, the flood-prone riverside zone—incorporating three villages in Kampong Cham and Prey Veng provinces—is distributed across the alluvial regions adjacent to the Mekong River, a zone with a historic propensity to both floods and droughts (Assessment Capacities Project 2011). The final zone, the drought-prone east zone—incorporating three villages contained within Prey Veng province—is located in the drought-prone regions close to the eastern border of the country (European Space Agency 2019), which have been identified as highly vulnerable to climate change due to the impact of both chronic and severe droughts on heavily rice-dependent monocropping (Assessment Capacities Project 2011).

Figure 1 Study site locations, flood and drought in Cambodia (Source: Open Development Cambodia) and Prey Veng province (inset; Source: European Space Agency 2019).

Migration also played a crucial role for site selection, with all nine village sites located in provinces where high proportions of the population work outside of their home villages. Of Cambodia’s twenty-five provinces, for example, Prey Veng, in which five of the nine sites are located, has the highest proportion nationally of internal migrants, and Kampong Cham, in which a single study site is located, has the third highest level. Of the three provinces containing the three sites in the Tonle Sap and lakeside zone, Kampong Thom has the fourth highest level of internal migration, Pursat has the seventh highest, and Kampong Chhnang has the eleventh highest. As such, being both climatically vulnerable and associated with high levels of migration, the nine sites are ideal loci to examine the intersection of climate change and migration dynamics.

Finally, to capture the role of marginality in relation to climate migration—and in particular how socioeconomic marginality plays a role in structuring the perceptions and response to climate change (Chu and Michael 2019)—the nine sites were additionally selected on the basis of the presence of marginalized groups. Three socioeconomically marginalized groups, brick workers, beggars, and ethnic Vietnamese populations, were selected. Brick workers and their households are characterized by multidimensional marginality rooted in their poverty, lack of assets, and the social stigma attached to their work. Beggars suffer social exclusion and poverty (Beazley and Miller 2015; Springer 2017; Parsons and Lawreniuk 2018; Parsons 2019), and Vietnamese communities are characterized by physical and social marginality, as well as political exclusion (Berman 1996; Amer 2013; Sperfeldt 2017; Parsons and Lawreniuk 2018; Canzutti 2019). From the perspective that marginality is coconstituted by natural environmental, economic, and social factors (Von Braun and Gatzweiler 2014; Chu and Michael 2019), the three groups were also selected on the basis of their geographical concentration in (and historical association with) parts of Cambodia characterized by ecological and environmental precarity. Thus, three of the target villages had high proportions of brick workers relative to the national average, three had high concentrations of beggars, and three had high concentrations of ethnic Vietnamese households.

In each of the nine villages, every household belonging to one of the three marginalized groups was presampled before the arrival of enumerators. A randomly selected sample of the remaining households of the village, equal in size to the targeted group members, was then selected to provide a comparative data set. Both the marginalized households and randomly selected cohort were then given a survey. The total amount of surveyed households across the nine villages was 691. The survey was designed to collect data on social and demographic indicators, perception of climate change perceptions, livelihoods, and migration patterns associated with each household.

Household characteristics such as composition, income, assets, and social and demographic indicators were collected via the survey. To capture climate impact, eleven indicators were selected: drought, rainfall, temperature, fires, soil problems, wind, insect infestations, livestock diseases, floods, new plant species, and new animal species. These were identified from literature (Migrating Out of Poverty Program 2013; International Centre for Integrated Mountain Development 2014) and information collected in the villages prior to implementing the surveys. The survey explored perceptions of changes to these eleven indicators over the past ten years, from 2008 to 2018. Although considerably less than the usual thirty-year period for climatic measurement, this ten-year period was selected to balance degradation in recall accuracy over time (Berney and Blane 1997; De Nicola and Giné 2012) against the minimal period over which to infer trends in the climate over time, identified elsewhere (Australian Bureau of Meteorology 2007; Kolawole et al. 2016) as ten years.

The surveys were delivered in Khmer by an experienced team of two Cambodian researchers, using the electronic data collection tool Kobo Toolbox. On average one survey took forty-five minutes to conduct. The surveys were conducted in two phases: February to March 2018 and February to March 2019. They were analyzed using IBM SPSS Statistics 24.

In addition, eighty-five qualitative interviews were undertaken with survey informants and local officials to establish the background and context of the quantitative findings. From initial scoping interviews, via the supervision of the quantitative survey, through to qualitative follow-up interviews, the first author visited the surveyed villages on multiple occasions over two periods of two months each, adding insights into social practices and livelihood activities. Alongside informal conversations, the majority of interviews were undertaken in Khmer by a separate two-person team, one of whom is a native speaker and the other of whom is proficient in the language. The remainder of the interviews, either with native, fluent, or proficient English speakers, were conducted in English by the first author. All qualitative interviews were analyzed using the qualitative analysis tool NVivo 11.

Climate Change Data, Mobility, and Perception in Cambodia

In terms of both climate change impacts and migration prevalence, Cambodia is an exceptional site, ranked the world’s second most vulnerable country to climate change in 2014 (Kreft et al. 2014; Kreft et al. 2016). Climate change is now routinely described as “a major threat” to the economy and society of Cambodia (Khut 2017). Historical records indicate significant changes in rainfall patterns since the 1920s (Diepart 2015), and over the last two decades, increasingly sporadic, unpredictable, and intense rains have badly affected farmers’ livelihoods (Diepart 2015). In particular, the unpredictability of rainfall has transformed irrigation from agricultural luxury to basic necessity for many farmers, resulting in a growing need for capital investment (International Bank for Reconstruction and Development and World Bank 2015).

A large share of this capital is obtained via external and internal labor migration (Bylander 2015; Oudry, Pak, and Chea 2016). Over a third of Cambodians have lived in more than one province in their lifetime, with almost 10 percent of the population living abroad, predominantly for work (United Nations Educational, Scientific and Cultural Organisation 2018). Migration is also connected to the transition from the traditional transplanting method of farming—wherein seedlings are raised in a nursery plot before replanting—to a system of broadcasting, in which seeds are scattered directly onto larger plots (Liese et al. 2014; Shrestha et al. 2018). As the labor-intensive transplanting method has been replaced by the broadcasting system requiring only one fifteenth of the labor (Liese et al. 2014), a large proportion of rural people have been forced to look elsewhere for employment.

Despite widespread acceptance of the impacts of climate change and rural marketization on Cambodia’s population movement, detailed information on this relationship is sparse (Jacobson et al. 2019). This is in large part due to the low resolution and unreliability of climate data across the kingdom’s “instrument-poor” weather monitoring infrastructure (Chhinh and Millington 2015, 798). Sparsely positioned weather stations—long acknowledged by the World Bank and others to be lagging behind (United Nations Office for Disaster Risk Reduction 2010) others in the region—offer accurate data on only a small subsection of each province. Even in those areas, inaccuracies and gaps in data remain an issue (Chhinh and Millington 2015). Given the significant differences in rainfall patterns in evidence at both the provincial and subprovincial scales (Vance et al. 2004) and across the kingdom’s eight hydrological regions (Thoeun 2015), this constitutes a major barrier to effective analysis of the specific impacts of the changing climate. Moreover, the only properly maintained weather stations are usually positioned in close proximity to either the geographic or urban center of each province. Given that many of Cambodia’s most politically marginalized people—in particular those lacking citizenship and property rights—live far from these urban centers (Messerli et al. 2015), this leaves understanding of environmental shifts least accurate for those with the fewest resources and the least political agency.

Thus, there is increasing consensus surrounding the need for “a new paradigm that joins traditional climate research with research on climate adaptation” (Overpeck et al. 2011, 702) to bridge the gap between top-down and bottom-up approaches, both globally (Conway et al. 2019) and in Cambodia (Jacobson et al. 2019; Parsons and Chann 2019). The role of climate perceptions offers a possible bridge in this respect, offering the potential to relate the physical and social dimensions of climate response in a meaningful and productive manner—especially in relation to the role of marginality in driving climate migration (e.g., Marino 2012; Warner and Afifi 2014; Panda 2017; Conway et al. 2019).

Nevertheless, the subjectivity of climate change perceptions has historically been used as an argument against their usefulness (Weber 2010; Hansen, Sato, and Ruedy 2012). Individuals are assumed, with some basis, to lack the ability to make objective judgments on long-term climatic trends, due to their embeddedness in socioeconomic and cultural systems. In the United States, for example, Smith et al. (2014) identified gender, education, income, and political affiliation as key variables in the perception of climate change. Similarly, in Ethiopia, Deressa, Hassan, and Ringler (2011) summarized that “farmers’ perception of climate change was significantly related to the age of the head of the household, wealth, knowledge of climate change, social capital and agro-ecological settings” (23). Yet positive assessments of the accuracy of climate perceptions in some contexts, ranging from “qualitatively congruent” (Funatsu et al. 2019, 7) to “highly accurate” (Hein et al. 2019, 1) support the strength of the relationship between climate perception and measurement. Indeed, as outlined in Table 1, the data observed through this study appear to indicate that respondents observed climatic changes in all three sites.

Table 1. Perception of climate change indicators disaggregated by agro-ecological zone

Perceived climate changes	Drought-prone east (N = 175)	Flood-prone riverside (N = 308)	Tonle Sap and lakeside (N = 210)
Drought (%)	63.8	75.3	1.1
Rainfall (%)	81	74	52
Temperature (%)	100	84.1	98.9
Soil problems (%)	0	7.8	0
Wind (%)	4.8	19.8	55.4
Insect infestation (%)	9.5	44.8	10.3
Livestock diseases (%)	17.1	37.3	0
Floods (%)	0	2.9	1.1
New plant species (%)	0.5	5.2	0
New animal species (%)	3.3	45.1	22.3

Nevertheless, as with observed climate data, the perception data shown in Table 1 reveal significant differences across each of the regional contexts. The greatest consistency between sites is exhibited in the perception of temperature, in relation to which between 84.1 percent and 100 percent of respondents noted changes. The greatest range of responses was observed in relation to drought, with responses averaging between 1.1 percent and 75.3 percent according to site. Interestingly, variation was also observed within sites, indicating that the observed variance cannot be explained solely by agro-ecological differences but rather incorporates a socioeconomically determined subjective dimension (Figure 2). Indeed, as exemplified in Figure 2, multiscalar analysis reveals the limitations of a purely geographical interpretation of climate perception. Although community-scale differences might be apparently observed at the provincial scale, reducing the scale of analysis reveals the extent to which competing perceptions of climate phenomena are geographically interspersed.

Figure 2 Drought perception at the provincial, commune, and village scales. Each dot represents a respondent household. Green dots are those reporting changes in drought. Red dots are those reporting no change.

An analysis of the direction of environmental issues observed by respondents echoes this finding. Rather than establishing comparable trends regarding, for example, floods and droughts, the data indicate significantly different—often diametrically so—perceptions of environmental trends between households in close proximity of one another, severely undermining the potential of climate perception as a proxy for small-scale climatological data.

Indeed, as shown in Figure 3, respondents in some cases reported opposite experiences of changes to rainfall and drought over the past decade, despite their homes and agricultural land being in close (i.e., less than 1 km) proximity. Nevertheless, as this article aims to show, the lack of consistency in small-scale geographical reporting of climate perceptions does not render the data useless, nor should inconsistency be confused with randomness. Rather, if, as long noted in the literature on climate perception in the Global North, perceptions are predictably structured by social and economic characteristics, then the same might also be true in the Global South. What follows will explore this position in relation to climate perception and livelihoods specifically.

Figure 3 Spatially mapped drought and rainfall perception in an alluvial riverside village.

Climate Perception and Livelihoods

The relationship between climate change perception and livelihoods is both important and underresearched (Kolawole et al. 2016; Fadairo, Williams, and Nalwanga 2019). Although less readily available in cross-sectional data sets than socioeconomic and demographic indicators such as age, wealth, and gender, data on livelihoods constitute a potentially powerful predictor of climate perception, framing climatic factors according to particular economic sensitivities. Indeed, this relationship is supported by the data here. As highlighted in Table 2, a comparison of climate perception data among fishers and nonfishers across all three agro-ecological zones reveals statistically significant differences in climate perception in four out of ten climate variables.

Table 2. Climate perception of fishers and nonfishers

Perceived climate changes	Livelihood type
	Respondent’s household does not fish (n = 593)	Respondent’s household does fish (n = 99)
Drought (%)** (p = 0.000)	58.3	22.2
Rainfall (%)** (p = 0.000)	75.5	40.4
Temperature (%) (p = 0.172)	92.1	96
Soil problems (%) (p = 0.734)	3.4	4
Wind (%)** (p = 0.000)	20.9	42.4
Insect infestation (%) (p = 0.144)	24.5	31.3
Livestock diseases (%)** (p = 0.005)	23.6	11.1
Floods (%) (p = 620)	1.7	1
New plant species (%) (p = 0.088)	2.9	0
New animal species (%) (p = 0.726)	27	25.3

** indicates that the relationship is significant at the 0.01 level.

Although certain perceptions relate to the possession of assets—fishers are less likely to possess livestock, for example, and thus less likely to perceive changes in diseases affecting them—many of the perceptions relate specifically to the manner in which climatic variations influence the performance of livelihoods. For example, strong winds and high temperatures are viewed as a problem for fish stocks, heightening fishers’ sensitivity to changes in the wind and thus the likelihood of perceiving them. In addition, this point is further exemplified by an examination of the data on farmers’ perceptions of the climate. As shown in Table 3, farmers exhibit a notably different profile of climate perception, less sensitive than fishers to wind and temperature changes but more sensitive than both fishers and the data set average to all other indicators besides floods. This reflects, again, both the possession of assets—farmers are much more likely than fishers to possess livestock, for example—but also how perception of the climate is structured by farmers’ livelihoods. Simply put, droughts and changes in rainfall have a far greater impact on those who depend on readily available water supplies for their livelihood than those, such as fishers or modern sector workers, who do not.

Table 3. Climate perception of farmers and nonfarmers

Perceived climate changes	Livelihood type
	Respondent’s household does not farm (n = 540)	Respondent’s household does farm (n = 153)
Drought (%)** (p = 0.000)	46.3	77.1
Rainfall (%)** (p = 0.009)	68.1	79.1
Temperature (%) (p = 0.927)	92.6	92.8
Soil problems (%)** (p = 0.001)	2.2	7.8
Wind (%)** (p = 0.006)	26.5	15.7
Insect infestation (%)** (p = 0.000)	15.9	58.8
Livestock diseases (%)** (p = 0.000)	17.6	36.6
Floods (%) (p = 0.753)	1.7	1.3
New plant species (%)** (p = 0.000)	0.6	9.2
New animal species (%)** (p = 0.002)	23.9	36.6

** indicates that the relationship is significant at the 0.01 level.

Although differences in overall reporting of climatic changes are significantly related to livelihoods, reporting on the direction or type of change shows a far weaker relationship. As outlined earlier, respondents demonstrate a limited ability to perceive the same specific changes in their environment. What our respondents are perceiving is therefore likely related to differences in their livelihoods, which they associate with specific environmental needs, challenges, and features. This is highlighted by the lack of statistically significant variations between farmers’ and nonfarmers’ perceptions of the type of changes evident in rainfall, outlined in Table 4. Moreover, although the space available here is insufficient to present disaggregated data on each climatic feature, a similar lack of correlation was evident in the case of other indicators of climate perception, such as drought and new animal and plant species.

Table 4. Type of change in rainfall among farmers and nonfarmers

	Livelihood type
Reported changes in rainfall	Respondent’s household does not farm (n = 540)	Respondent’s household does farm (n = 153)
% Becoming heavier	0.2	0
% Becoming less frequent	20.7	19.6
% Becoming less predictable	11.7	19.6
% Becoming more frequent	1.3	1.3
% Becoming less heavy	33.9	38.6
% Cannot specify change	31.9	20.9

Climate Perception, Migration, and Marginality

The findings reported here have implications for the literature on climate change and migration (e.g., Jacobson 1988; Myers 1997, 2002; International Organisation for Migration 2019) because of their implications for the strength and directness of the relationship between climate change, the perception of climate change, and action in response to perceived climate change. In some cases, climate change has been viewed as driver of migration (McLeman and Hunter 2010; Black et al. 2011), whereas in others (Mortreux and Barnett 2009; Kelman et al. 2019) it has been viewed as a less significant factor. Nevertheless, the debate is characterized by the assumption that the perception and physical incidence of climate change are covariant, a premise that is rendered problematic by the intermediary role of livelihoods.

As shown in Table 5, migration rates vary significantly across the three agro-ecological zones under study. Whereas migration rates in the Tonle Sap and lakeside zone are well below the national average, the flood-prone riverside zone and drought-prone east zones demonstrate far higher levels of migrant households (as well as migrants per household). Informants’ residence across one of the three locations is a significant indicator of migration behavior, yet cross-cutting indicators of climate change perception reveal smaller scale relationships with migration behavior. Indeed, as shown in Table 6, significant correlations with household migration are in evidence across all but one indicator of climate perception. Of these, the perception of drought, soil problems, new animal species, and new plant species—all dimensions of climate whose perception is likely to be influenced by agricultural livelihoods—exhibit strong effects on the incidence of migration. Smaller but nevertheless significant relationships with temperature and rainfall also support the link between climate perception and migration incidence. This relationship is further supported by inferential analysis. Constructing an ordinal regression with the presence of migrants in the household as the independent variable and all nine significant climate perception indicators as dependent variables produces a model (χ² = 124.4, p = 0.00) indicating a significant relationship between climate perception and migration. Given that this significant relationship is evident both within each of the three agro-ecological zones and across the data set as a whole, the data suggest that climate perceptions might offer a geographically cross-cutting piece of data with which to examine migration behavior in relation to climate change.

Table 5. Migrants and migration rates disaggregated by location

Data set	Households containing migrants (%)	Migrants per migrant household
Drought-prone east (n = 210)	69.5	1.9
Flood-prone riverside (n = 308)	51.3	1.8
Tonle Sap lake and lakeside (n = 175)	6.3	1.2

Table 6. Climate perception and household migration incidence

Perception of climate change	Households with and without migrant members
	No migrant(s) (n = 378)	Migrants (n = 315)
Drought (%)** (p = 0.000)	39.2	69.8
Rainfall (%)** (p = 0.000)	61.1	81.9
Temperature (%)** (p = 0.000)	90.7	94.9
Soil problems (%)** (p = 0.000)	2.6	4.4
Wind (%)** (p = 0.000)	29.4	17.8
Insect infestation (%)** (p = 0.000)	24.3	26.7
Livestock diseases (%)** (p = 0.000)	14	31.1
Floods (%) (p = 1.000)	1.6	1.6
New plant species (%)** (p = 0.000)	1.9	3.2
New animal species (%)** (p = 0.000)	21.7	32.7

** indicates that the relationship is significant at the 0.01 level.

As shown in Figure 4, for example, one finding of this approach is to oppose the contention (see Richards and Bradshaw 2017; Conway et al. 2019) that marginal groups are more likely to migrate than others, as a result of their inability to adapt in situ. Taking the most marginal, hyperprecarious migrants from each community—in the drought-prone east, those who migrate to beg; in the alluvial riverside regions, those who migrate to debt bondage; and in the Tonle Sap and lakeside zone, ethnic Vietnamese people who lack citizenship and thus migrate without documentation—the data indicate that marginal households in each of the three agro-ecological zones are not only less likely to migrate but also less likely to perceive changes in the climate. Given the strength of the relationship between climate perception and livelihoods demonstrated earlier, this might be due to marginal groups being less likely to possess the productive, predominantly agricultural assets (e.g., farmland or livestock) associated with greater sensitivity to climate change.

Figure 4 Climate perception and migration among marginal and nonmarginal households.

Discussion

Climate migration tends to be conceived, a priori, as an undesirable and unusual state of affairs. Both reflecting and shaping trends in governance and policy, public discourse, inclusive of the media, has characterized climate migration as a significant global problem in recent years. Described in the media as a “crisis” (Wernick 2019, 1) that will “unleash millions of global warming refugees” (Rowse 2018, 1), modeled estimates of up to 2 billion climate migrants by 2100 (Geisler and Currens 2017) reflect both widespread concern and a sense of the exceptionalism of the phenomenon. Indeed, the latter is exhibited in relation to the type as well as the extent of climate migration. It is frequently asserted in climate migration scholarship that those who migrate in response to the changing climate are likely to be the most vulnerable and marginal members of the community due to their limited ability to adapt in situ (Marino 2012; Stapleton et al. 2017; Conway et al. 2019). Moreover, this is an assertion that exists across multiple scales of climate migration analysis. Climate change, marginality, and migration are assumed to be linked at the national (Grecequet et al. 2017), community (Raleigh and Dowd 2018), household (Warner and Afifi 2014), and even intrahousehold (Jungehülsing 2010) scales. Policy efforts to manage, mitigate, or reduce the risks of climate migration are therefore often directed at stemming the untrammeled flow of those displaced by disaster: the “climate barbarians at the gate” (Bettini 2013, 63) whose very vulnerability casts them as a national security threat (K. A. Thomas and Warner 2019).

Policy perspectives of this sort derive in large part from top-down scholarship on the climate change–migration nexus (e.g., Myers 2002; Geisler and Currens 2017; World Bank 2018). Such top-down analyses continue to be dominated by large-scale Ricardian modeling of climate data against existing socioeconomic data sets, resulting in analysis showing relatively direct relationships between meteorological phenomena and human behavior (Falco, Donzelli, and Olper 2018). As the body of scholarship exploring the climate change–migration nexus grows, however, empirical evidence increasingly reveals “that environmental factors shape human migration patterns in combination with numerous other micro-, meso-, and macro-level contextual factors” (Hunter, Luna, and Norton 2015, 382). Subjective positionality and marginality shape the relationship between climate and migration (Clare et al. 2017; Vulturius et al. 2018), entangling environmental factors in a variety of other socioeconomic characteristics and processes, from age to gender to education. A tension has therefore arisen between bottom-up and top-down analysis of climate migration (Conway et al. 2019). Physical climate data indicate large-scale trends, whereas local behavior and impact analyses suggest far smaller and more complex effects rooted in subjective and contextually determined perceptions of the climate.

By presenting data on a range of climate change perception indicators, as reported by people engaged in a variety of livelihoods, this article has aimed to elucidate this complexity and add the aspect of livelihoods to analysis of the climate change–perception–migration nexus. As shown here, livelihoods structure not only the extent of climate perception but also how and which climatic changes are perceived. Those engaged in livelihoods that place a particular emphasis on certain aspects of the climate are more likely than others to notice changes in those aspects—even if they live in close proximity to each other. By extension, those engaged in certain livelihoods are more likely also to react to those changes. Livelihoods therefore play an active role in determining whether or not somebody perceives changes that may constrain or facilitate migration.

From this standpoint, the results question the common contention that climate change will hit the most marginal communities hardest, arguing that this statement requires greater nuance. Rather than enhancing perception and, by extension, experiences of the changing climate, marginality instead shapes these in specific ways. Indeed, marginality might overall diminish marginal populations’ sensitivity to the changing climate. This was highlighted by, first, the relationship between livelihoods and the type of changes to the climate perceived by respondents; second, marginal groups’ relatively low engagement in climatically sensitive primary livelihoods, due to their lack of assets and resources; and, third, the lower levels of migration observed among marginal than nonmarginal groups. Thus, the common explanation among migration scholars, that the most vulnerable lack the social and material resources to migrate, requires further analytical work. Rather than being highly vulnerable to the shifting environment but unable to adapt to it, socially marginalized groups appear in the results presented here to be less likely to perceive changes in their environment but—as shown in Figure 4—similarly likely to have migrants in their households if they do so. This suggests that it is not only a lack of physical resources to migrate but also the social embeddedness of livelihoods shaping perception of the changing climate itself that might contribute to marginalized communities’ low rates of overall migration.

Despite problematizing dominant analytical approaches that seek to draw direct causal linkages between climate and population data, the aim here is not to undermine the importance of the role of climate change on migration patterns. Rather, it is a call to recognize the embeddedness of climate impacts within economic and social systems and the importance of the perceptions–livelihood nexus in shaping people’s experiences of climate change. Although the climate, as a statistical phenomenon, operates (in part) at a global scale, climate perceptions—and by extension climate impacts—are mediated locally by a broad milieu of social, economic and cultural factors. Simply put, climate change in the physical sense is a top-down phenomenon, whereas climate perceptions emerge from the bottom up. Given the issues related to the use of climate data as a predictor of migration, this is of key importance, offering the potential for locally gathered climate perception data to be used to produce more effective models of migration in a changing climate. Yet it is with further work that the greatest benefits of this position might be gleaned, with the compilation of larger scale data sets on climate perception having the potential to deepen our understanding of the extent to which correlations can be identified, and inferences drawn, between physical climate data and human behavior. Viewed thus, climate perception data have the potential not only to enhance understanding of climate mobility but, in doing so, to provide an epistemological bridge between competing scientific and social scientific ontologies of climate in the analysis of climate change impacts.

Conclusion

This article has sought to further understanding of the determinants and outcomes of climate change perception by exploring three aspects of the perception of the climate: first, the relationship of climate perceptions to geography and space; second the influence of livelihoods on climate change perception; and, third, the relationship between climate change perception and migration.

In relation to the first point, the results show that climate perception is not significantly influenced by spatial distance, with respondents living in close proximity reporting divergent or even opposite perceptions of various aspects of the climate. Crucially, the issue of scale is key here. Although larger scale analyses might indicate coherent differences in climate perception between agro-ecological zones, in many cases this coherence breaks down at smaller scales. This suggests that climate perceptions are influenced by factors other than climate, the influence of which is generally observed at a larger scale.

The second contribution of the article is to emphasize the role of livelihoods in shaping climate perceptions. The data here indicate that certain climate perceptions are significantly influenced by engagement in certain primary livelihoods. Farmers are more likely to observe changes in rainfall, for example, whereas fishers are more likely to observe changes in the wind. In contrast to Ricardian approaches, which assess the likelihood of responding to changes in the climate, therefore, this approach emphasizes instead the subjective nature of climate change perception.

Nevertheless, although the subjectivity of climate perceptions has generally been viewed as undermining their utility, this article demonstrates that they retain a potentially useful role in understanding behavior in response to the climate. Specifically, the article demonstrates the utility of climate perception data for understanding the relationship of climate perception to migration, outlining a statistically significant relationship between the presence of migrants in the household and nine out of ten climate change indicators. In addition, the data here consider the role of marginality on climate perceptions, concluding that marginality has a significant impact on the likelihood of perceiving climate change but not on the likelihood of having migrants in the household.

In broad terms, the article therefore offers two contributions to the literature on climate change and migration. First, it demonstrates that climate perception data constitute a useful means through which to understand the motivations and conditions that shape climate migration. Second, by demonstrating the subjective entanglement of climate perceptions—and thus responses—it problematizes efforts to draw direct correlations between physical climactic phenomena and human behavior. Further work should therefore be directed toward understanding the geographical and social distribution of climate perception, as a complementary lens through which to interpret the climate’s impacts.

Additional information

Funding

Funding for this research was provided by the British Academy Postdoctoral Fellowship scheme (Ref: pf170152) and the ESRC-DFID Development Frontiers Research Fund (Grant No. ES/R00238X/1).

Notes on contributors

Laurie Parsons

LAURIE PARSONS is British Academy Postdoctoral Research Fellow in the Department of Geography at Royal Holloway, University of London. E-mail: [email protected]. His work explores the contested politics of climate change through its impact on socioeconomic inequalities, patterns of work, and mobilities.

Jonas Østergaard Nielsen

JONAS ØSTERGAARD NIELSEN is Professor in Integrative Geography in the Department of Geography and the Integrative Research Institute on Transformations of Human–Environment Systems (IRI THESys), Humboldt University, Berlin, Germany. E-mail: [email protected]. His research covers human adaptation to climate change and the globalization of land use change.