Threats to land and environmental defenders in nature’s last strongholds

Abstract

Land and environmental defenders are a major bulwark against environmental destruction and biodiversity loss resulting from unsustainable nature resource extraction. Resultant conflicts can lead to violence against and deaths of these defenders. Along with mounting environmental pressures, homicides of these defenders are increasing globally. Yet, this issue has only recently started to receive scientific attention. While existing studies indicate the importance of socio-economic processes in driving such murders, spatially explicit global analyses considering environmental components are largely missing. Here, we take a broad spatial approach to assess relative contributions of environmental factors to the killing of environmental defenders. We find higher rates of such homicides are typically found in areas where limited or underutilized resources (e.g., freshwater, land and forests) are more available. Our results point towards a prevalent global land scarcity that results in industries targeting the last remaining strongholds for biodiversity and the environmental defenders within.

Supplementary Information

The online version contains supplementary material available at 10.1007/s13280-021-01557-3.

Introduction

Globally, agribusiness, logging and mining have been implicated in massive habitat loss, with industrial activities linked to > 50% of the 2.3 million km² of forests lost over the last two decades (Hansen et al. 2013; Abood et al. 2015; Austin et al. 2017). Other activities, namely poaching and overfishing, have also been responsible for causing population declines of over 60% of species worldwide (Rands et al. 2010; Maxwell et al. 2016). On the front lines of conservation, local and regional civil society organizations and community groups—including indigenous peoples—are a major bulwark against such local environmental destruction and biodiversity loss (Gadgil et al. 1993; Ricketts et al. 2010). Their support for the protection of the natural environment has been credited for conservation of many habitats around the world—for instance the creation of extractive reserves in Brazil, and the preservation of Cockpit Country region in Jamaica (Lundy 1999; Brown and Rosendo 2000). These land and environmental defenders can be found throughout a wide socio-demographic gradient, from indigenous peoples, to park rangers, journalists, and workers from non-government organizations (Schwartzman and Zimmerman 2005; Rands et al. 2010; Ghazoul and Kleinschroth 2018; Glazebrook and Opoku 2018; Butt et al. 2019). They include men and women, from a range of age and ethnicity groups as well as educational levels (Glazebrook and Opoku 2018). Their endeavor to conserve and preserve natural resources and the people that depend on them, however, often places them at odds with industrial interests resulting in violent conflict and possible death (Bebbington et al. 2018; Glazebrook and Opoku 2018).

The killing of land and environmental defenders is a globally distributed phenomenon, reported in over 50 countries, and appears to be growing in frequency (Fig. 1). Such trends are only likely to continue with the apparent rise in frequency of indigenous people and leaders being murdered in areas such as the Amazon, especially since the systematic dismantling of environmental agencies and violation of environmental laws in Brazil in January 2019 (Rulli et al. 2013; Global Witness 2017; Glazebrook and Opoku 2018; Conselho Indigenista Missionario 2019; Prasniewski et al. 2020). Already, recent updates for 2019 report the greatest number of land and environmental defenders murdered in a single year (Global Witness 2020). With little governmental protection—as many extrajudicial killings linked to the paramilitary and police, as well as political instability and unsupportive administrations—the need for international attention and support is dire (Global Witness 2017; Ghazoul and Kleinschroth 2018; Glazebrook and Opoku 2018; Le Billon and Lujala 2020). This has led to promising developments on an international stage with the United Nations Environmental Programme and United Nations General Assembly recognizing and supporting the protection of land and environmental defenders in 2018 and 2019, respectively (United Nations Environment Programme 2018; United Nations General Assembly 2019).

Fig. 1.

Number of killings of environmental defenders per country across the world depicted a spatially; and b temporally. Significant positive correlations between the number of environmental defenders killed and c rates of homicides within a country; and d number of journalists killed. e Significant negative correlation between number of environmental defenders killed and change in wilderness extent within a country. Color of points in panels c–e follows the UNSD-defined regions depicted in b. Axes denoted with asterisks (*) are log-transformed

Perhaps in part because of the connections to human rights violations, violence and conservation, attention towards the murders of land and environmental defenders has been largely limited to policy. Only recently has the scientific community begun to be engaged. Supported by the Global Witness annual reports and data collection efforts, recent studies have identified key socio-economic and political drivers associated with environmentally-motivated murders based on country-level statistical analyses (Butt et al. 2019; Middeldorp and Le Billon 2019; Global Witness 2020). For instance, killings were found to be highly associated with rule of law and corruption as well as general homicide rates, democracy levels and per capita income (Jeffords and Thompson 2016; Butt et al. 2019; Middeldorp and Le Billon 2019; Le Billon and Lujala 2020). These findings support the general movement away from neo-Malthusian assumptions in the field of environmental conflicts, and supports the importance underlying socio-economic factors that determine the valuation of nature as a resource (Harvey et al. 1996; Le Billon 2015; Scheidel et al. 2020). It also reinforces the idea that environmental resources are a relative social construct, and the inequity and uneven distribution of these resources in relation to human populations are sources of conflicts (Kallis 2019; Scheidel et al. 2020).

Yet, environmental conflicts differ from social conflicts due to the importance of their environmental components. Negative impacts to the environment, such as waste disposal and contamination and exploitation of natural resources, are known to increase marginalization and burdens on poorer actors (Demaria and D’Alisa 2013; Le Billon and Lujala 2020). In the same way, despite the observed close link between environmental and non-environmental homicide (Middeldorp and Le Billon 2019), it is important not to discount the ecological dimensions (Collier and Hoeffler 2004; Szalavitz 2018). Environmental factors such as resource scarcity and abundance are implicitly tied to land and environmental defenders, and a nuanced understanding of its role is key to forming targeted strategies to reduce environmental conflicts and killings (Ghazoul and Kleinschroth 2018; Neimark 2019).

In order to assess elucidate the nuanced relationship between environmental killings and socio-ecological factors, we draw upon fundamental discussions on greed/grievance drivers of conflict and political ecology of violence (Collier and Hoeffler 2004; Bohle and Fünfgeld 2007). These studies highlight the importance of both resource availability and scarcity, as well as social versus environmental reason for historical conflicts and violence (Collier and Hoeffler 2004; Bohle and Fünfgeld 2007). Incorporating these perspectives into a socio-ecological framework (Ostrom 2009), we form a theoretical framework (Fig. 2) that serves as the basis of our analyses, with environmental defender homicides as a measure of a specific outcome of environmental conflicts (Collier and Hoeffler 2004; Bohle and Fünfgeld 2007; Virapongse et al. 2016). Our framework further builds on the conceptual framework by Scheidel et al. (2018). In our framework, social and ecological systems interact (“integration”) through resource extractive activities (e.g. mining, dams, agriculture, hunting, logging) (Fig. 1). These extractive activities and their impacts on the environment and local actors depend on the socio-metabolic configuration of each socio-ecological system (Scheidel et al. 2018). This interaction may involve varied actors with diverse views and values of nature that may conflict (“actors”). Conflict emerges as a result of the extractive activities leading to ecological distributional inequalities that generate environmental justice movements (Scheidel et al. 2018). These justice movements can, in turn, lead to the opposition of extractive activities potentially turning land and environmental defenders from victims to protectors (Conde and Le Billon 2017; Bebbington et al. 2018; Scheidel et al. 2020).

Fig. 2.

Theoretical framework describing the influence of social-ecological factors on conflict and specifically environmental conflict, mediated through a globalized assemblage—a series of actors from a local to subnational and national to international levels. Social-ecological system is adapted from (Virapongse et al. 2016)

The study of environmental conflict has been richly addressed through local case studies, with global syntheses studies only recently emerging (e.g. the Atlas of Environmental Justice) (Conde and Le Billon 2017; Bebbington et al. 2018; Scheidel et al. 2020). Despite these large advances, empirical global spatially explicit analyses of land and environmental defenders’ homicides are lacking. These global analyses would represent an empirical contribution that would allow understanding land and environmental defenders’ homicides from a perspective of global land scarcity (Lambin and Meyfroidt 2011), telecoupled land-use pressures and rapidly increasing human pressures rapidly reducing the remaining areas of intact habitat (Liu et al. 2014; Watson et al. 2018).

As such, here we aim to provide a global empirical contribution to the literature while building on previous theoretical and conceptual contributions (Ostrom 2009; Scheidel et al. 2018) and take a broad spatial analytic approach to clarify the relative contribution of social and environmental correlates to global patterns of extrajudicial killings of land and environmental defenders. In particular, we explore these relationships by focusing on extrajudicial killings of land and environmental defenders as a consequence of environmental conflicts, and adding to existing literature examining resistance to extractive projects (Conde and Le Billon 2017; Bebbington et al. 2018; Scheidel et al. 2020). By assessing proxies of resource availability and scarcity alongside socio-economic factors, such as inequality, development and affluence, we clarify existing patterns of environmental murders, and also enable the development of predictions to better monitor and hold those responsible accountable (Ghazoul and Kleinschroth 2018; Neimark 2019).

Materials and Methods

Land and environmental defender homicides dataset

Global occurrences of land and environmental defender homicides was gathered from a dataset collected by Global Witness (Global Witness 2017). This dataset utilized information from national and international reports and news articles, of which a randomly selected subset was inspected and crosschecked by the authors. The dataset documented a total of 1069 homicides, including the names, date, location (of homicide), gender, and industries the victims were campaigning against (Global Witness 2017). The dataset was collected between 2002 and 2017. Missing information surrounding exact locations and associated industries were secondarily collected from a variety of sources (e.g., local newspaper websites or reports) and georeferenced. Excluding reports that could not be georeferenced resulted in a full dataset of 964 cases to an accuracy of 10 km². While these 964 cases likely represent an underreporting of global land and environmental defender homicide numbers, and a further underestimate of violence against land and environmental defenders, it remains one of the most comprehensive collation of such data.

Correlates of land and environmental defender homicides

To evaluate the relative importance of socio-economic and environmental factors surrounding these killings, a series of geospatial modeling techniques was then applied. As the study was implicitly driven by the difference between environmental and social conflicts, an important component of our analysis involved accounting and controlling for the potential influence of general homicide rates. This allowed us to disentangle and isolate correlates associated with the murder of land and environmental defenders rather than the general population. We also controlled for potential underreporting in some areas such as Africa and Oceania (see Fig. 1) (Le Billon and Lujala 2020). To achieve this, we relied on the sampling of points of comparison (or pseudo-absences) from a spatial specification layer, a technique previously shown to successfully control for such issues in other spatial analyses (Barbet-Massin et al. 2012). This specification layer was formed based on proxies of reporting likelihoods such as corruption perception index and general accessibility of locations, as well as a proxies of suppression of reporting, specifically rates of murders of journalists per country. A more detailed description of the variables and method is in Table S2. Ultimately, 5000 points of comparison was selected from a 2-degree radius around reports of land and environmental defender homicide.

We then used land and environmental defender homicide locations and these points of comparison to extract values associated with the conditions associated with each location. Utilizing the theoretical framework developed (Fig. 2), we identified proxies of socio-economic components that reflected levels political systems, inequality, development and affluence—population numbers, gross domestic product (GDP), Gini index (income inequality measure), human development index and social demographic index. We also identified proxies of environmental components that reflected both scarcity and abundance—average forest loss, amount of cropland, amount of freshwater available, human pressure, total nature rent (% of GDP due to the sum of oil rents, natural gas rents, hard and soft coal rents, mineral rents, and forest rents), amount of land already grabbed by foreign nations, and whether or not an area was protected. Rents here refer to the difference between commodity prices and production costs. A detailed description of the definitions and motivating hypothetical links to murders of land and environmental defenders (and environmental conflicts) are listed in Table S3. While a majority of the factors considered were spatially explicit, five variables (Gini, human development and social demographic indices, and natural resource rents and land grabbed) did not possess spatially explicit data. Instead, we used country-level data as the best approximation of spatial conditions. These were represented as scaled global raster layers (0.05° resolution) and possessed variable inflation factors (VIF) below 4.0.

To weigh the relative importance of these socio-economic and environmental components, an information theoretic framework was utilized to evaluate a total of 57 competing, hypothesized models (Table S4). Specifically, we utilized generalized linear mixed-effects models (GLMMs) to control for both country- and industry-specific conditions that are typically more associated with a socio-environmental conflicts over specific resource extraction and production projects, by including these factors as random-effects (Conde and Le Billon 2017; Bebbington et al. 2018; Scheidel et al. 2020). This modeling technique allowed us to maintain the broad focus on environmental conflicts and test hypotheses involving potentially important variables (fixed-effects) while controlling for variation among underlying units or groupings (random-effects) (Bolker et al. 2009). In this instance, underlying factors incorporated as country- and industry-specific random-effects capture the influence of country policies, or legal frameworks on specific industries, which cannot be directly quantified for typical fixed-effects analyses. Within-country spatial heterogeneities in these variables, for which we are not aware of more spatially-resolved maps, are certainly a limitation of our analyses. As future spatially-resolved datasets appear, these variables could be moved from the random effects into fixed effects, allowing for further insights to be generated. Model performance was evaluated with Akaike information criterion corrected for small sample sizes (AICc), as well as marginal and conditional R² to indicate model fit associated with fixed-effects and fixed- and random-effects, respectively. The top performing model (∆AICc = 0) was then bootstrapped 20 times, with average effect size (estimator means and confidence intervals) illustrated to indicate the relative effects of predictor variables on the occurrence of land and environmental defender homicides.

Hotspots of extrajudicial killings of land and environmental defenders

These models (top performing model, bootstrapped 20 times) were also used to generate averaged spatial projections of land and environmental defender homicides. Projections were limited to analyzed areas, excluding areas outside gathered datasets to reduce uncertainty surrounding reporting rates (Zeng et al. 2016). Projection outputs range from 0 to 1, with 1 indicating high risk of land and environmental defender homicides, and 0 indicating low risk.

Accounting for modelling uncertainty

To account for potential variability arising from modeling techniques (Barbet-Massin et al. 2012), we analyzed a total of 1368 individual models across 24 combinations of points of comparison (pseudo-absence) selection criteria. We used 1000, 5000, 10 000 and 50 000 points of comparison and from 0.5°, 1.0°, 2.0°, 5.0°, 8.0° and 10° radii. Ultimately, all combinations produced similar results, with similar social-economic and environmental factors represented in all models of ∆AICc < 2.0. We utilized 5000 points from a 2.0° radius as it best reflected the most parsimonious model evaluations and had good predictive ability (Table S5).

All analyses were performed within R 3.6.0 (R Core Team 2019). R packages “raster” was primarily used for raster analyses and illustrations (Hijmans and Van Etten 2016). GLMMs were modelled and evaluated using “MuMIn”(Barton 2009).

Results

Scope and scale of land and environmental defender homicides

Globally, the highest number of reported land and environmental defender homicides between 2002 and 2017 were found in Philippines (186 cases), Brazil (183 cases), Colombia (175 cases), Peru (76 cases), and Mexico (67 cases) (Fig. 1). The highest rates of killings were reported in the Americas (with a peak of 123 cases in 2016), followed by Asia (with a peak of 61 cases in 2016) and Africa (16 cases in 2016). Europe and Oceania reported the lowest rates of environmental killings (with two and zero cases respectively in 2017) (Fig. 1; Table S1). These results corroborates the existing trends of high rates of killings of indigenous people in Brazil in protected areas (Conselho Indigenista Missionario 2019), as well as the growing conflict surrounding anti-mining activists in Philippines (Wayland 2019).

Correlates of land and environmental defender homicides

Among the models analyzed, the top five models indicated correlative relationships between environmental homicides and socio-economic and environmental factors (Table 1). However, the model that possessed only environmental factors outperformed all models that contained socio-economic factors, with a ∆AICc < 2.0 (Table 1). The model (with random effects included) possessed conditional R² of 0.62 (Table 1). This model indicated that higher rates of land and environmental defender homicides were positively associated measures of natural resource usage or availability—human footprint (1.01 ± 0.09), freshwater available (0.27 ± 0.23), area protection status (0.14 ± 0.08), and forest loss (0.13 ± 0.05); but also negatively associated with the cropland (− 0.18 ± 0.13) and total nature rent (− 0.41 ± 0.33) (Fig. 3). The amount of land grabbed appears to have little influence on the overall land and environmental defender homicide (0.12 ± 0.29) (Fig. 3).

Table 1.

Top five generalized linear mixed models and their associated evaluation metrics. All models had a marginal R² of 0.19 and conditional R² of 0.62

Formula	AICc	∆AICc
Homicide ~ NatureRent + LandGrabbed + HumanFootprint + ForestLoss + ProtectedAreas + Cropland + FWavailability	2576.6	0.0
Homicide ~ NatureRent + LandGrabbed + HumanFootprint + ForestLoss + ProtectedAreas + Cropland + FWavailability + SDI + GINI	2579.4	2.8
Homicide ~ NatureRent + LandGrabbed + HumanFootprint + ForestLoss + ProtectedAreas + Cropland + FWavailability + HDI + GINI	2579.4	2.8
Homicide ~ NatureRent + LandGrabbed + HumanFootprint + ForestLoss + ProtectedAreas + Cropland + FWavailability + SDI + HDI + GINI	2581.0	4.4
Homicide ~ NatureRent + LandGrabbed + HumanFootprint + ForestLoss + ProtectedAreas + Cropland + FWavailability + SDI + HDI + GINI + GDP + HumanPopulation	2582.9	6.3

Fig. 3.

a Projected spatially explicit risk map of the rates of environmental defender homicides globally. b Effect sizes of generalized linear mixed-effects model formed based on human footprint, freshwater availability, protected areas, forest loss, land grabbed, cropland and natural resource rents as fixed effects. c Projected rates of environmental defender homicides within Central America. d Projected rates of environmental defender homicides within Southeast Asia. Values within spatially explicit maps range from 0 to 1 with 1 indicating areas of high probability of environmental killing

Hotspots of extrajudicial killings of land and environmental defenders

Final model projections show that the predicted spatially explicit occurrence of land and environmental defender homicides largely match up with country-level prevalence (Figs. 1 and 3). Countries with large areas of hotspots include Philippines, Columbia, Guatemala and Honduras, but the prediction captures greater spatial resolution (Fig. 3). For instance, areas such as the northwestern area of Colombia possess higher rates of land and environmental defender homicides than the southeastern areas (Fig. 3); and the island of Mindanao possess the greatest rate of killings compared to other islands of the Philippines (Fig. 3). Predictions also identify potential areas of underreporting relative to overall reporting rates. Areas such as the southeastern coasts of Brazil and Java, Indonesia, could possess higher rates of land and environmental defender homicides than currently reported (Fig. 3).

Discussion

Overall, our results point to land and environmental defender homicides having stronger links to environmental rather than socio-economic factors (Table 1). While we find that socio-economic conditions can correlate with rates of environmental homicides, matching previous studies on socio-environmental conflicts associated with specific extraction and production projects (Conde and Le Billon 2017; Butt et al. 2019; Scheidel et al. 2020), our results show that models with only environmental factors have a stronger statistical link to land and environmental defender murders (Table 1). Specifically, we find that land and environmental defender homicides are more common in areas where resources (e.g. water and land) are abundant and relatively unexploited, but limited elsewhere in the country or region. This is indicated by the positive correlation to freshwater availability and protected areas, and negatively associated with cropland and natural resource rents—suggesting that the availability of untapped resources is a major motivation of environmental conflicts (Fig. 3).

This supports previous studies on ecological distribution conflicts where the uneven distribution of environmental resources leads to conflict (Martinez-Alier 2004; Scheidel et al. 2020). Necessary for many industries (e.g., dam creation, fishing, agribusiness and mining activities), the prevalence of usable freshwater is a major source of disputes (Aitken et al. 2018). Similarly, protected areas represent locations with natural resources that would be desirable for land grabbing, poaching and logging but are currently underutilized (Gaveau et al. 2007; Steinmetz et al. 2014; Busscher et al. 2018). These resources, which are also necessary for many indigenous people and local communities, would therefore result in higher levels of discord, and potentially killings, as other actors within the country or region attempt to gain access or control (Robbins 2011). One potential source of conflict can actually occur through the enforcement of strict protected areas leading to the eviction of local natural resource users, which may in turn result in tensions with protected area implementation and park rangers (Dowie 2011; Finegan 2018; Soliku and Schraml 2018). The duality of park rangers as victims and enforcers would require further research as larger datasets become available that would allow fitting separate models for rangers from other types of land and environmental defenders.

In comparison, areas with high amounts of cropland (spatially explicit) and natural resource rents (country-level) represents land that have already been exploited, typically at industrial levels and managed by powerful actors (Austin et al. 2017; Middeldorp and Le Billon 2019). The high rate of murders in areas with greater levels of untapped resources and lower levels of exploited resources highlights both the importance of the resource availability and scarcity in driving conflict (Collier and Hoeffler 2004; Szalavitz 2018), as well as the social implications of its uneven distribution (Martinez-Alier 2004; Scheidel et al. 2020). These patterns are also reflected in the high rate of murders among indigenous peoples or individuals involved in indigenous rights issues, as well as people defending against agribusiness, logging, mining, poaching and water/dams (Butt et al. 2019; Le Billon and Lujala 2020).

Our results also point towards a prevalent global land scarcity that makes extractive industries set their eyes on the last remaining strongholds for biodiversity (Lambin and Meyfroidt 2011). This is signaled by our finding that protected areas are a positive predictor of land and environmental defender homicides, as opposed to areas that have been converted to cropland, experienced land grabbing, and high natural resource rents (Fig. 3). This implies that industrial interests that drives land and environmental defender homicides appear to be more prevalent in areas with greater natural habitats such as protected areas (Butt et al. 2019).

Naturally, as our study maintains a broad scope of environmental conflicts, we did not consider more site-specific drivers of land and environmental defenders homicides—such as the interactions between local and state actors of each conflict (Conde and Le Billon 2017; Butt et al. 2019; Scheidel et al. 2020). This could have contributed to potential inaccuracies in predictions such as specific regions in Europe and underreporting in Ethiopia, Myanmar, Nigeria and Venezuela (Le Billon and Lujala 2020). We nevertheless find that higher rates of environmental killings are associated with greater human impact on land-use, specifically those with high forest loss and changes in human pressure (Fig. 3). This supports the idea that the last strongholds of biodiversity are particularly at risk (Lambin and Meyfroidt 2011). These key correlates directly indicate the levels of land conversion and degradation, as well as the level of industrial interests and activities (Hansen et al. 2013; Venter et al. 2016). It also supports the negative correlation between environmental homicides and amount of wilderness lost due to industrial processes (Austin et al. 2017) (Fig. 1).

Given the important role land and environmental defenders play in the preservation of habitats and its biodiversity, recognition of their efforts is not just a human rights issue, but also a conservation one (Gadgil et al. 1993; Ricketts et al. 2010; Larsen et al. 2020). As such, one of the aims of this study was to develop spatial projections to aid the call for stricter laws or special prosecutors within governments (Ghazoul and Kleinschroth 2018; Glazebrook and Opoku 2018; Neimark 2019). These spatial projections do not serve to predict future environmental killings under changing conditions, but instead provide a potentially useful tool to enable greater monitoring efforts, and thus efficient enforcement, as well as a means to identify areas of high risk of such conflicts. Hotspots within the map match country-levels, while providing greater resolution and identifying relatively underreported areas that could benefit from increased monitoring patterns (Figs. 1 and 3). This has the potential to aid existing efforts to hold governments and companies accountable, ensuring both upward and downward accountability and supporting international policy (Ribot 2003; Glazebrook and Opoku 2018; United Nations Environment Programme 2018; United Nations General Assembly 2019).

Conclusion

By identifying the nuanced interaction between resource scarcity and availability, we gain much needed insights into the impacts of protector struggles on global environmental issues such as habitat loss and resource depletion. These insights add to the existing understanding of important socio-economic factors that drive environmental conflict by providing key environmental perspectives. We also identify potential underreported hotspots that could benefit from increased monitoring. Through such efforts, this study represents an important step towards addressing global human rights, conservation and sustainability goals.

Supplementary Information

Below is the link to the electronic supplementary material.

Biographies

Yiwen Zeng

is a Research Fellow at the National University of Singapore. His research interests lie in the intersection between biodiversity conservation and human well-being.

Fangqi Twang

was a Final Year Student at the National University of Singapore in the Carrasco Bioecon Lab in 2019. Her research focused on extrajudicial killing of environmental defenders.

L. Roman Carrasco

is an Associate Professor at the National University of Singapore. His research focuses on identifying strategies for the reconciliation of biodiversity conservation, food security and economic development in the tropics.

Data availability

All data generated or analyzed during this study are included in this published article (and its supplementary information files) without geographical coordinates for confidentiality issues. Detailed data in our study was requested from Global Witness.