The impacts of warfare and armed conflict on land systems

ABSTRACT

Land-use decisions can change abruptly in response to shocks, and warfare and armed conflicts are among the most drastic and globally frequent shocks. Yet, our understanding of where armed conflict affects land systems, how land-use patterns are impacted, and how far-reaching and persistent these changes are, is partial. We used a spatially detailed dataset on armed conflict and a literature review to explore these questions. A number of key insights emerged from our study: (1) warfare and armed conflict affect land systems mainly in more densely populated areas, regardless of the dominating land use; (2) warfare and armed conflict can impact land systems in major ways, but these effects are diverse and not unidirectional; (3) warfare and armed conflict primarily affects land systems locally, but can forge telecouplings; and (4) although the impact of warfare and armed conflict is often immediate, it can instigate long-lasting land-use legacies.

KEYWORDS:

• Warfare
• armed conflict
• case studies
• land-use change
• land-cover change
• deforestation
• agricultural abandonment
• re-cultivation
• remote sensing

Introduction

Understanding what drives people’s land-use decisions is important to understand how land systems may evolve in the future and for identifying effective policies to steer land systems towards desired outcomes (Meyfroidt, 2015; Rounsevell et al., 2012; Turner et al., 2013). Much research in this context has focused on how gradually changing drivers affect transitions between broad land-use regimes (Foley et al., 2005; Rudel et al., 2005), including population change (Jepsen et al., 2015; Vanacker, Govers, Barros, Poesen, & Deckers, 2003; Vesterby & Heimlich, 1991), economic development (Bičı́k, Jeleček, & Štěpánek, 2001; Rounsevell et al., 2006), technology diffusion, or climate change (Hanewinkel, Cullmann, Schelhaas, Nabuurs, & Zimmermann, 2013; Pielke, 2005).

Many drivers of land-use change, however, can shift abruptly too, sometimes leading to drastic and widespread land-use change. For example, economic crises can trigger illegal logging (Sunderlin, Angelsen, Resosudarmo, Dermawan, & Rianto, 2001), waves of deforestation (Zak, Cabido, Cáceres, & Díaz, 2008), and shifts from cash crops to food crops (Sunderlin et al., 2000). Revolutions and the institutional overhaul that typically follows, for instance when economic or land reforms are implemented, can also result in marked episodes of land-use change, such as in case of the breakdown of the Soviet Union (Baumann et al., 2011; Schierhorn et al., 2013). Similarly, technological breakthroughs, such as the discovery of the Haber–Bosch process or the development of genetically modified (GMO) soybean variants can lead to rapid land-use change (Lambin & Geist, 2006; Rudel et al., 2009), as can technological disasters such as the reactor meltdown in Chernobyl (Hostert et al., 2011). Finally, policy interventions, such as the EU’s nitrogen directive in 1991 (Levers, Butsic, Verburg, Müller, & Kuemmerle, 2016), the accession of a country to an economic trade zone (Corbelle-Rico, Butsic, Enríquez-García, & Radeloff, 2015), or agricultural subsidies for biofuel productions (Suttles, Tyner, Shively, Sands, & Sohngen, 2014), may result in rapid land-use change. While these examples highlight the importance of shock events, our understanding of their effects on land-system change still remains limited (Ramankutty & Coomes, 2016). This is unfortunate given that such shocks are fairly frequent – especially at broad spatial scales and over longer time periods (Jepsen et al., 2015; Ramankutty & Coomes, 2016). This lack of understanding limits our ability to predict future land use (Müller et al., 2014). Moreover, understanding shock events is important because they have the potential to shift land systems to alternative states (Dearing, Braimoh, Reenberg, Turner, & van der Leeuw, 2010; Ramankutty & Coomes, 2016). Despite this, research on regime shifts in land systems, and the shock events that contribute to them, is lagging behind (Jepsen et al., 2015; Müller et al., 2014; Ramankutty & Coomes, 2016).

Warfare and armed conflicts (hereafter: armed conflict) are among the most drastic shocks that can impact societies and thus land systems. Casualties, both among those fighting and among civilians (Slim, 2007), the displacement of people fleeing from unsafe areas (Paasche, 2012), human-rights abuses, and the destruction of local livelihoods (Justino, 2011; Seddon & Hussein, 2002) are just some of the grave consequences of armed conflict, and a wide body of literature exists assessing the causes and outcomes of armed conflicts for societies. Armed conflicts can also have drastic environmental outcomes (Dudley, Ginsberg, Plumptre, Hart, & Campos, 2002; Machlis & Hanson, 2008), including biodiversity loss where protected areas become havens for insurgents or are left unguarded, where rebels plunder natural resources during times of conflict (Gleditsch, 1998; Hanson et al., 2009; Irland, 2008), or where fighting affects natural ecosystems directly (e.g. defoliation with Agent Orange during the Vietnam War (Dudley et al., 2002; Rustad, Rød, Larsen, & Gleditsch, 2008)). On the other hand, nature and wildlife may benefit from armed conflict, for example where human pressure decreases in contested areas or areas contaminated with landmines (e.g. North/South Korea Demilitarized Zone (Kim, 1997)).

While generally believed to have important effects on land use change, research on how armed conflicts affect land-use decisions and thus land-use patterns is scarce. This is unfortunate, considering that the few existing studies suggest that these effects can be drastic (Machlis & Hanson, 2008), long-lasting (Baumann, Radeloff, Avedian, & Kuemmerle, 2015), and far-reaching (Binns & Maconachie, 2005). Assessing how armed conflicts affect land systems should also be a priority because armed conflicts are, sadly, relatively frequent (Gleditsch, Wallensteen, Eriksson, Sollenberg, & Strand, 2002). For example, in 2014, there were 40 wars ongoing globally. This is the highest number since 1999 (Pettersson & Wallensteen, 2015), affecting regions on four continents in both in the temperate zone and the tropics. In total, 259 wars have taken place since World War II (Pettersson & Wallensteen, 2015). Moreover, some regions have been particularly war-ridden, such as the Congo Basin, the Middle East, or the Caucasus, suggesting the potential for long-lasting or repeated effects on land systems.

We carried out a systematic review of how armed conflict affects land-use decision making and land-use patterns, and thus land systems. Specifically, we sought to answer three interrelated research questions:

1. Which land systems are predominantly affected by armed conflict?

2. How does armed conflict affect land-use patterns?

3. What are the mechanisms through which armed conflict affects and restructures land systems?

To address these questions, we first compiled a database incorporating information on land systems, as well as spatially explicit information of armed conflict events. We then conducted a literature review on studies that established a relationship between armed conflict on the one hand and land-system change on the other. We synthesized the findings of these studies with regards to the types of land-use/cover change, its proximate drivers, and the causal mechanisms behind these land system changes. Based on these findings we assessed past research efforts into the effects of armed conflicts on land systems and identified key research issues.

Methods

We used the Georeferenced Event Dataset from the Uppsala Conflict Data Program (UCDP GED, version 4.0 (Croicu & Sundberg, 2015; Sundberg & Melander, 2013)) to explore which land systems are primarily affected by armed conflicts (research question 1). This database contains all individual conflict events defined as ‘an incident where armed force was used by an organized actor against another organized actor, or against civilians, resulting in at least 1 direct death at a specific location and a specific date’ (Sundberg & Melander, 2013). The UCDP GED distinguishes between three types of conflicts (state-based, non-state, and one-sided conflict¹), and in its current version (i.e. 4.0) contains conflict events for the period 1989–2014 for Asia, Africa, and the Middle East, and for 2005–2014 for the Americas and Europe (Sundberg & Melander, 2013). Identified conflict events are obtained from global newswire reporting, global monitoring of local news performed by the British Broadcasting Corporation (BBC), and secondary sources such as local media reporting or reports by non-governmental organizations (NGOs), and international governmental organizations (IGOs). All entries in the database underwent a rigorous quality assessment, and are consistently coded (Sundberg & Melander, 2013). The database also contains information on the accuracy of the conflict location, ranging from 1 (i.e. exact location known) to 7 (i.e. conflict coordinates estimated (Croicu & Sundberg, 2015; Sundberg & Melander, 2013)).

We downloaded the entire UCDP GED database and evaluated the location of conflict events in two ways. First, we counted the number of conflict events per country considering all events of precision category 1–6 (i.e. events that have the precision at the country-level). Second, we selected conflict events of precision categories 1 and 2 (i.e. the location of the event is estimated with at least 25-km accuracy). We overlaid these conflict events onto two maps depicting key aspects of land systems. First, we compared conflict in relation to anthropogenic biomes (hereafter: anthromes (Ellis & Ramankutty, 2008)), which characterize human transformation of terrestrial biomes based on global patterns of human population and land use/cover (Ellis, Klein Goldewijk, Siebert, Lightman, & Ramankutty, 2010). Second, we compared conflict locations to the global land-system map by van Asselen and Verburg (2012), which characterizes land systems based on land use/cover, livestock density, and agricultural intensity. Both maps were generated using a hierarchical, expert-based classification scheme, represent the year-2000 situation, and have a spatial resolution of 5 arc-minutes.

For each conflict event, we extracted the anthrome and land system it fell into and counted the number of fatalities resulting from the conflict events. We then summarized the total number of conflict events and fatalities per anthrome and land system, and calculated the share of conflicts per anthrome and land-system type. We divided this share by the global share of an anthrome and land system. The resulting values indicate whether conflicts in a specific anthrome or land system occur more often (values >1) or less often (<1) than expected by the global prevalence of this anthrome or land system. For all analyses, we used aggregated classes for both maps (Table SI 1 in the supplementary material) and assumed the broad land systems and anthrome classes to be constant (i.e. unchanged compared to the time period covered by a particular case study). We provide results for the number of conflict events here, and results for the number of fatalities in the Supporting Information.

To better understand how armed conflicts affect land-use/cover change (research question 2) and which underlying mechanisms may have caused these changes (research question 3), we conducted a literature review based on Thompson Reuters Web of Science database with combinations of search terms as outlined in Table 1. For this, we followed the guidelines of the PRISMA framework [(Moher, Liberati, Tetzlaff, Altman, & Grp, 2009) for the PRISMA statement see Supporting Information].

Table 1. Search term used in the Thomson Reuters Web of Science database.

Search conducted on	Search term
23 February 2016	TI = [(land cover*) or (land use*) or (land system*) or (land conversion*) or (land transformation*) or (land degradation*)] or [(cropland) or (pasture) or (forest*) or (agricultur*)] or [(deforestation) or (afforestation) or (reforestation) or (urbanization) or (logging) or (irrigation) or (grazing) or (abandonment)] and TS = [(warfare) or (war) or (conflict) or (armed conflict) or (violence)]

Table 2. Summary of categories under which the selected case studies were evaluated.

Variable group	Categories	Description/example
General information	Year of publication	Year when study was published
	Study location	Country in which the analysis took place
	Length of study	Time-period covered by the study
	Name of war/conflict	(e.g. Iran–Iraq war)
Data sources	Land-use/cover change data source	(e.g. remote sensing, statistical data)
	Conflict data source	(e.g. conflict database, surveys)
	NA
Land-use/cover changes	Agricultural change	Changes in croplands (rainfed + irrigated) and grazing lands; subdivided into agricultural expansion, agricultural abandonment, and re-cultivation
	Forest change	Summarizing different terminology; subdivided into forest gain (including forest regrowth, forest expansion, etc.) and forest loss (including deforestation, logging, etc.)
	Other	Changes in settlement/urban structures
Descriptive/quantitative	Quantitative	A model or another statistical method (e.g. matching) was used to assess the relationship of armed conflict and land-use/cover change
	Descriptive	Qualitative description of the relationship between armed conflict and land-use/cover change
	NA	Armed conflict only served as argumentative help to explain detected land-use/cover changes
Study design	Conflict – local	Land-use/cover changes at the conflict location during the war
	Post-conflict – local	Land-use/cover changes at the conflict location after the end of the war
	Conflict – distant	Land-use/cover changes afar from the conflict location during the war
	Post-conflict – distant	Land-use/cover changes afar from the conflict location after the end of the war

The search was conducted on 23 February 2016 and yielded an initial number of 1319 references. We evaluated the title and the abstract of each of these papers for suitability and relevance in the context of our study. Specifically, a study had to fulfil two requirements to be considered. First, a study had to analyze land-use/cover changes in relation to a period with an armed conflict, enabling us to relate the two. Thus, studies examining land-use/cover change after an armed conflict without reference to the pre-conflict situation were not considered. Second, a study had to use or provide spatial information on land-use/cover changes. Such information included, for example, maps derived from satellite imagery or statistical data such as changes in the extent of irrigated areas. In other words, studies that provided only qualitative information of how land use/cover changed during or after conflicts were not considered. These two requirements reduced the sample to 38 studies, for which we derived detailed information according to five groups of criteria. (1) General information included the year of publication, study location, length of the study period, and additional information about the conflict where available (e.g. name and duration of conflict event). (2) Data sources included information on the sources of the land-use/cover change data (i.e. remote sensing or statistical data) and the conflict data (e.g. conflict databases such as UCDP or ACLED). (3) Land-use/cover changes included the type of land-use/cover change examined in a study. To remain consistent across studies, we used a common terminology throughout. Changes in forest cover contained various types of forest loss (including deforestation, i.e. conversion to another land use/cover, but also logging or forest loss due to the fighting itself) and forest gain (including forest expansion, post-logging regrowth). Changes in agricultural land included changes in grazing lands as well as croplands (both rain-fed and irrigated), and consisted of agricultural abandonment, agricultural expansion, and agricultural re-cultivation. Other land-use/cover changes included changes in urban structures (e.g. settlement destruction or urban expansion) or oil spills.

A fourth category of attributes, (4) methodological approaches, describe how the influence of the armed conflict on land-use change was measured. Studies that simply highlighted land-use/cover changes without further evaluation (such as ‘fire from oil spills’), where labeled as NA. Finally, we assessed the (5) study design with which land-use/cover changes were analyzed in the case studies. We assessed if land-use/cover changes were analyzed for the conflict period itself or the period after the conflict (i.e. post-conflict), as well as if the land-use changes associated with the conflicts were analyzed for the same area or if the conflict-induced land-use/cover change occurred in distant locations. This resulted in four possible combinations of period of analysis and local/distal land-use change: (1) conflict – local, (2) post-conflict – local, (3) conflict – distant, (4) post-conflict – distant. Not all studies fell into one of these categories (Table 2).

Results

Global distribution of armed conflicts in land systems

The UCDP GED (version 4.0) contained a total of 109,148 conflict events since 1989, of which 62,352 were geo-located with an accuracy of at least 25-km (precision category 1 or 2, Figure 1).

Figure 1. Distribution of conflict events globally. The conflict events stem from the UCDP georeferenced dataset (Sundberg & Melander, 2013), and were summarized at the country level (top), and overlaid over a map of global land systems for the year 2000 by van Asselen and Verburg (2012) (middle) and the map of anthropogenic biomes of the world by Ellis et al. (2010) (bottom).

These conflict events were not distributed equally in space, but were geographically highly concentrated. Regions that were particularly affected by armed conflicts were Central Africa, the Middle East, and Central America (Figure 1). Other regions with frequent conflict events were Bangladesh and eastern India. The countries with the most recorded conflict events were Afghanistan (18,051) and India (13,452), followed by Nepal (5651), Pakistan (5340), Iraq (5103), Turkey (4144), and Algeria (3879). Overall, 26 countries had at least 1000 conflict events recorded. In contrast, some regions had no or only very few conflict events, including most parts of Europe, North and South America, and Asia. In total, 153 countries had one or less conflict event.

Comparing conflict events across anthromes suggests that conflicts occurred concentrated in specific anthromes. For example, 65% of all conflict events (40,125) were recorded within dense settlements (18,063) or villages (22,062). These anthromes comprised less than 10% of the global surface in 2000, but were almost 25 times (dense settlements) and 7 times (villages) more likely to suffer from armed conflicts than could be expected based on the global area share of these anthromes. In cropland anthromes, ~15% (9799) of all conflict events were recorded, roughly corresponding to the global share of this anthrome (14.7%). In contrast, rangelands (6821 events) and semi-natural lands (4883) combined had less conflict events than the dense settlement anthrome. Finally, although covering more than 50% of the Earth’s surface, wildlands had only 0.002% of all conflicts (119 events, Figure 2).

Figure 2. Distribution of conflict events across anthromes (top row), and land-systems (bottom row). The left graphs show the proportion of conflicts within each anthrome and land system (upper bar) and the global area share of each anthrome and land system (lower bar). The right graphs show the share of the conflicts per anthrome and land system share over the proportion of each anthrome and land system. For more details on how the ratios were calculated and how the classes of anthromes and land systems were aggregated, please refer to the ‘Methods’ section of this article and the supplementary material.

Similarly, most armed conflict events fell into only a small number of land systems, and some land systems were particularly conflict-prone. For example, settlement systems contained most armed conflict events (24% of all events) although comprising only 0.02% of the Earth’s surface. Conflict events in extensive cropland systems were about 3.5-times more likely than expected based on the share of extensive cropland systems (11% of all conflicts in extensive cropland systems). In contrast, the proportional amount of conflict events corresponded to the global area share in medium cropland systems (~8% of all conflicts (5155)), grassland systems with livestock (~7.8%, 4911), and intensive agricultural systems (~3.5%, 2180 events). In relation to their global share, bare lands (15.5% of all conflict events (9605) compared to 25% of this land system’s global share), forest systems (7.4% of all conflict events (4626) compared to 20.6% of this land system’s global share), and grassland systems without livestock (16.8% of all conflict events (10,501) compared to 26.8% of the system’s global share) showed less conflicts than would be expected by chance (Figure 2).

Systematic literature review

Our systematic review of case studies assessing land-use/cover change in areas affected by armed conflict showed that despite armed conflict being frequent, scholars have only recently focused on studying the land-system implications of such shocks. Of the 38 studies we analyzed, only 5 were published before 2000 and more than 50% of all studies were published after 2010 (Figure 3). The existing body of literature also focused on a limited number of world regions and armed conflicts, with the vast majority of studies assessing armed conflict impacts on land use/cover in Central and Western Africa (12 studies) and the Middle East (13). Only four studies assess armed-conflict-effects on land use/cover in Central and South America and three assessed armed-conflict-effects in Southeast Asia (Figure 3). Studies were particularly scarce for East India, Myanmar and the Philippines despite numerous armed conflicts in these countries (Figure 3).

Figure 3. Location of the case studies and publication year as well as number of publications per year between 1990 and 2016.

In terms of methodological approaches, 76% of the studies we assessed used remote sensing to quantify land-use changes (29), whereas 16% of all studies (6) relied on statistical information such as forest inventories or agricultural production statistics to assess land-use/cover change (Figure 4). Regarding data on the conflict itself, only a small proportion of all case studies used spatial data, most commonly conflict databases (7 studies, 18%) or targeted field surveys (1 study), to link armed conflict to land use. The remaining 30 studies (79%) conducted their land-use/cover change study in the context of the armed conflict event that had happened, but did not spatially link conflict events and land-system change. Eleven studies (29%) used a statistical model whereas 13 studies (34%) used descriptive analysis to associate land-use/cover changes with armed conflict (Figure 4). The study periods covered by the case studies ranged widely from 35 years to a single year, with an average 15.8 years (standard deviation of 8.3).

Figure 4. Quantitative evaluation of the literature review following different categories used in this study. (a) Type of land-use/cover change; (b) sources of the conflict data; (c) sources of the land-use/cover change data; (d) overview on how armed conflict as a driver of land-use/cover change was assessed; (e) number of studies examining the different complexities of the relationship between armed conflict and land-use/cover change.

The majority of the selected studies (20, equaling 52%) exclusively focused on land-use/cover changes at the location of the conflict site or nearby and only during the conflict period itself. About 34% of the studies (13) analyzed land-use/cover changes during the conflict and/or post-conflict period only at the conflict location, whereas only 8 studies assessed changes both at the location of the armed conflict and in at least one distant location. Out of these, five studies focused on the conflict-period and three assessed both the conflict and the post-conflict periods (Figure 4).

Of the 38 studies we assessed, 42% (16 studies) evaluated how armed conflict influenced forest change, 55% (21) evaluated agricultural change, and 13% evaluated other types of land-use/cover change (e.g. settlement destruction, fires, 5 studies).² Of the 16 studies analyzing forest change in relation to armed conflict, 12 found forest loss (75%), while 4 found both forest gain and forest loss. The main driver of forest loss at conflict location was, in our sample of case studies, forest loss caused through the fighting itself (e.g. through the use of defoliating agents or fires caused by the fighting or bombing, etc. (Sánchez-Cuervo & Aide, 2013; Van et al., 2015; Williamson, 1990)), timber harvesting (Basnet & Vodacek, 2015; Qamer et al., 2012), and mining (Butsic, Baumann, Shortland, Walker, & Kuemmerle, 2015; Hecht, Kandel, Gomes, Cuellar, & Rosa, 2006)). Another driver of forest loss was political decisions associated with the armed conflict (e.g. removal of forest to eliminate the trafficking of drugs (Sánchez-Cuervo & Aide, 2013)).

Of the 21 studies that assessed agricultural changes, 15 found agricultural land to be abandoned during and after times of armed conflict, 2 study found both agricultural abandonment and agricultural expansion, and 3 studies found agricultural abandonment and subsequent re-cultivation. Only one study identified all these agricultural change processes (Baumann et al., 2015). The studies we assessed identified three key drivers of agricultural abandonment. First, agricultural abandonment was driven by a diminishing agricultural workforce, because farmers were killed (Eklund, Persson, & Pilesjö, 2016), people fled or were evicted from conflict areas (Baumann et al., 2015), or because farmers engaged in the fighting (Suthakar & Bui, 2008). Second, the presence of the fighting actions themselves, sometimes including the placement of land mines, results in unsafe conditions and thus a cessation of agriculture (Dinar & Keck, 1997). Finally, an outflow of capital due to insecurity of investments resulted in a decrease in irrigation agriculture, due to dilapidated infrastructure after maintenance investment discontinued (Jaafar, Zurayk, King, Ahmad, & Al-Outa, 2015).

Re-cultivation of previously abandoned agricultural areas often was attributed to the return of previously evicted people (Baumann et al., 2015), and simultaneous agricultural abandonment at the conflict zone and agricultural expansion afar from the conflict zone was often a consequence of migration patterns of internally displaced people (IDPs, (Alix-Garcia, Bartlett, & Saah, 2013)).

Discussion

Institutional, political, socioeconomic, or environmental shock events can affect coupled human–natural systems in major ways, sometimes leading to a lasting reorganization of these systems. Armed conflict is among the most drastic shocks and is, unfortunately, globally frequent. Using a spatially detailed dataset on armed conflict, as well as a structured literature review, we assessed (1) which land systems are most frequently impacted by armed conflict, (2) how armed conflict affects land-use/cover patterns, and (3) which mechanisms link land system change and armed conflict. Although the knowledge base in terms of empirical studies is surprisingly limited, a number of key insights about the relationship of armed conflict and land system change emerged from our study, which we detail in the following sections. We conclude by highlighting a number of research needs to further our understanding of the role of armed conflict for land-system change.

Armed conflict has widespread effects on land systems

Armed conflict affected land systems around the globe, with some regions being particularly conflict-ridden. Our analysis highlighted that specific land systems are particularly strongly affected by armed conflict, such as urban and densely-populated systems as well as agricultural systems (Figure 1). This seems noteworthy given rapid ongoing urbanization, with projections that more than 90% of the global human population will live in densely populated places by the end of the century (United Nations, 2010). Our findings also highlight that conflicts occur predominantly in areas most valuable for humans, possibly prompted by conflicts over access to these lands or its resources.

Our literature review also emphasizes that the effect of armed conflict on land systems can be pronounced. Or instance, several studies highlight widespread agricultural abandonment (e.g. 30% in Iraq (Eklund et al., 2016) or over 60% in the Caucasus (Baumann et al., 2015)) following armed conflict, and higher deforestation rates during times of armed conflict (Butsic et al., 2015). Thus, our review emphasizes the important role of armed conflict as a driver of rapid land-use/cover change, comparable to other shock events such as institutional shocks (Alcantara et al., 2013; Baumann et al., 2012; Prishchepov, Radeloff, Baumann, Kuemmerle, & Müller, 2012), technological disasters (Hostert et al., 2011), or technological innovation such as the introduction of GMO soybean variants (Leguizamón, 2014; Reenberg & Fenger, 2011).

The effects of armed conflict on land system change are diverse and not unidirectional

A key outcome from our review of case studies was that armed conflict effects can lead to both more intensive and less intensive land use. For example, armed conflict was associated with increased forest loss in some cases (Butsic et al., 2015; Hecht et al., 2006; Nackoney et al., 2014; Ordway, 2015), yet decreasing logging and forest recovery in others (Burgess, Miguel, & Stanton, 2015; Gorsevski, Kasischke, Dempewolf, Loboda, & Grossmann, 2012; Stevens, Campbell, Urquhart, Kramer, & Qi, 2011). In some cases, armed conflict triggered agricultural abandonment (Eklund et al., 2016; Landsberg, Vanhuysse, & Wolff, 2006; Suthakar & Bui, 2008; Wilson & Wilson, 2013; Witmer & O’Loughlin, 2009) and higher agricultural intensity in other situations, although agricultural expansions were mostly detected far from the conflict areas (Alix-Garcia et al., 2013; Basnet & Vodacek, 2015). Similarly, armed conflict led to outmigration and shrinking settlements in some cases (Witmer & O’Loughlin, 2011), and to urbanization and new settlements in others (Alix-Garcia et al., 2013; Baumann et al., 2015).

However, despite this diversity of land-use outcomes documented in the literature, these outcomes are neither arbitrary nor unpredictable, with a number of causal mechanisms linking armed conflict and its outcomes for land systems. Periods of violent conflict, particularly in the case of internal conflicts, are often characterized by weak law enforcement, rising corruption, growing black markets, eroding institutions, and, sometimes, failing states (Irland, 2008). Such conditions foster the excessive illegal use of natural resources such as firewood, valuable timber, or minerals and metals (Shortland, Baumann, Kuemmerle, & Fielding, in review), leading to deforestation, including in areas that would be inaccessible during times of peace, such as protected areas (Butsic et al., 2015). Moreover, insurgents may use natural resources to finance armed conflicts, leading to increased resource extraction (Harrison, 2015). Armed conflict can lead to increased land-use intensity where people flee from cities to the countryside (Eklund et al., 2016), or where IDP camps result in an increase of small-scale farming (Hagenlocher, Lang, & Tiede, 2012).

Conversely, conditions of instability can also foster less intensive land use. For example, commercial timber harvesting can become too risky (Butsic et al., 2015; Gorsevski, Geores, & Kasischke, 2013), or mining concessions may lay idle because mining becomes unsafe (Shortland et al., in review). Similarly, timber harvesting or cash-crop production may decline because export markets diminish during times of war (e.g. due to sanctions or difficulties in transporting goods (Le Billon, 2000)), and because international investments typically come to a halt once war breaks out (Collier, 2000). Agricultural intensity can decline and forests can expand on former farmland because farmers are recruited to engage in the fighting as rural people are displaced (Chamarbagwala & Morán, 2011; Czaika & Kis-Katos, 2009). Similarly, armed conflict can result in decreasing investments in irrigation infrastructure (Dinar & Keck, 1997). Thus, the exact outcomes of armed conflict on land systems are complex and context-dependent, but clear cause–effect chains and typical constellations of actors and land-use/cover outcomes exist. In addition, our results suggest that armed conflict itself is rarely a direct driver of land-use/cover change, but rather affects other drivers that underlay land-use decisions, explaining the diverse land-use outcomes of armed conflict.

Armed conflict may create telecouplings and lead to land-system change afar from the conflict site

Our review highlighted that armed conflict has the potential to create distal linkages between the conflict zone and one or more areas afar, possibly forging telecouplings (Friis et al., 2016; Liu et al., 2013) and affecting land systems on either end of the coupling. Studies on telecouplings in the context of land use and armed conflict are still scarce, given that the telecouplings concept itself is still under development, but existing case studies suggest that people who are displaced by the conflict form the key mechanism in armed-conflict-related distal links between land systems. For example, the Nagorno–Karabakh conflict in the Caucasus resulted in a large number of refugees migrating to Azerbaijan where they likely benefited from Azerbaijan’s land reform in the 1990s, thus enabling them to engage in agriculture, and subsequently leading to agricultural expansion outside the conflict zone. Meanwhile, agricultural land use contracted substantially in the conflict areas following the conflict (Baumann et al., 2015). Similarly, refugees moving into new locations bring with them their resource needs, which can cause intensifying fuel wood harvesting, expansion of subsistence farming, or construction of new settlements (Alix-Garcia et al., 2013; Binns & Maconachie, 2005; Burgess et al., 2015; Wilson & Wilson, 2013). Refugees typically remain in contact with their home communities, for example via sending remittances, which in turn can be an important income source and feed back on land-use practices at the place of origin (Lambin & Meyfroidt, 2011; Seto et al., 2012). Distal links impacting land systems on either end also emerge where trade barriers or sanctions are established. For example, the conflict in Cambodia resulted in diminishing export markets for timber, leading to less forest harvesting, which reversed after the conflict ceased (Le Billon, 2000). The conflict between Russia and Ukraine led to trade barriers and strongly declining food imports of Russia, resulting in higher incentives for Russian agriculture to expand production, especially dairy farming, but also diminishing exports and a restructuring of agricultural trade-patterns of many European countries (Iwanski, 2014). All these examples highlight that armed conflict can couple and affect land systems across larger distances.

Armed conflicts affect land systems immediately, but these effects can be long-lasting

Many of the land-system outcomes of armed conflict are typically temporary and reverse once peace has been reestablished (Le Billon, 2000). However, the legacy from armed conflict can also last long into the future. For example, much of the areas in which conflicts played out in the Southern Caucasus (Baumann et al., 2015), Bosnia (Andersson, Dasousa, & Paredes, 1995), or Croatia (Landsberg et al., 2006) still suffer from substantial landmine contamination or are still disputed, thereby continuing to making farming unsafe across wide areas.

The effects of armed conflicts can also unfold with a time lag, sometimes even decades after the end of the conflict. For example, the Chaco War between Paraguay and Bolivia, ending with Paraguay seizing power over much of the dry Chaco, resulted in land being privatized (i.e. establishing a latifundia land tenure system (Caldas, Goodin, Sherwood, Campos Krauer, & Wisely, 2015)). The private land ownership over much of the Paraguayan Chaco became a key factor leading to rapid deforestation. In another example, Romania compensated its soldiers fighting during WW I with land, and this privatization of forest land was an important driver of deforestation since then (Olofsson et al., 2011). In addition, farm size and field size patterns in Poland to date differ substantially among those areas that were part of the German Reich until 1945 and those that were Polish before WW I, as collectivization during socialism, and thus land re-privatization after 1989 mainly took place in those areas that Poland acquired after 1945 (Dannenberg & Kuemmerle, 2010; Kuemmerle, Hostert, St-Louis, & Radeloff, 2009). Thus, although armed conflicts episodes are often short, their legacy on land system dynamics can be substantial and long-lasting.

The effects of armed conflict on land systems remain under-researched

A number of research priorities emerge from our review. First, we found surprisingly few studies that empirically linked armed conflict and land-system change. Furthermore, these studies were spatially clustered, with few or no study in many areas in which armed conflict events have been frequent (e.g. India and Pakistan). This scarcity of studies makes it difficult to draw broader conclusions on how armed conflict affects land-systems across regions of different cultural background, land-use history, or political systems. Similarly, the small body of case studies does not allow us to quantify effect sizes of armed conflict impacts, or synthesize how different types of conflicts (e.g. interstate wares vs. civil wars) affect land systems. More empirical, quantitative case studies are therefore needed for under research regions. Methodologically our literature review highlights the power of using multiple data sources (e.g. satellite imagery and inventory data), and quasi-experimental statistical procedures and causal analyses to isolate the effect armed conflict has on land systems (Butsic et al., 2015; Wilson & Wilson, 2013), but such studies are still very scarce.

Second, systematic, global-scale analyses seem to be lacking at this time, despite an increasing availability of global-scale, high-resolution datasets on land system distribution and land-use/cover change (Hansen et al., 2013; Kuemmerle et al., 2013; van Asselen & Verburg, 2012), as well as increasing quality of armed conflict databases (Pettersson & Wallensteen, 2015; Sundberg & Melander, 2013), analyses making full use of these datasets to quantify the impact of armed conflict on land systems are missing. This is unfortunate as such consistent, broad-scale analyses can resolve issue related to the limited representativeness or number of local-scale case studies, as exemplified by the research on climate change effects on armed conflict (Buhaug, Gleditsch, Theisen, Mearns, & Norton, 2010; Burke, Miguel, Satyanath, Dykema, & Lobell, 2009; Fraser, 2011; Hsiang, Meng, & Cane, 2011; O’Loughlin et al., 2012).

Third, our literature review suggests that armed conflict itself often does not trigger land-system change, but amplifies or dampened other drivers. For example, the conflict in Cambodia resulted in sanctions to exporting Cambodian timber, and thus less timber harvesting (Le Billon, 2000), or the conflict in the Congo which resulted in a decline in foreign investment in mining (Shortland et al., in review). Further research is thus needed to better understand how armed conflicts affect and interact with other drivers of land system change.

Fourth, the manner in which armed conflict leads to the creation of distal linkages and how such telecouplings affect land systems on either end of the coupling remains largely unclear. Only a handful of studies have highlighted the role of distal links in the context of armed conflict and land system change, largely in relation to refugee movements. As the telecouplings concept is consolidated and operationalized, more studies assessing land systems experiencing armed conflict through the lenses of telecouplings, especially regarding the different flows that link systems (e.g. capital, refugees, information), the agents involved in establishing the flows between sending and receiving systems, and the possible spill-over effects to neighboring system would be beneficial (Liu et al., 2007, 2013, 2014).

Finally, whether or not armed-conflict-related land-use/cover changes are permanent and whether they lead to land system regime shifts (Ramankutty & Coomes, 2016) remains weakly understood. On reason may be that most studies so far we examined armed conflict effects on land system for fairly short time period, focusing mainly on the last decades. Assessments over longer time periods would be beneficial, especially for regions where good historic land use/cover information exists and where armed conflicts were prevalent, such as in Europe (Munteanu et al., 2015). Similarly, linking land system outcomes to peace-building success would be interesting, as regime shifts in land systems may depend on whether a conflict is resolved after fighting ends or whether the root cause of the conflict prevails (such as in the case of many disputed areas). Similarly, whether regime shifts on the time scale of human generations occurred may depend on the on contamination by landmines and ammunition, which are rarely removed until a conflict is fully resolved.

The outlined research issues emphasize that understanding the relationship between armed conflicts and land-change will require interdisciplinary research efforts, bringing together geography, environmental sciences, political sciences, and anthropology. It is unlikely that any research discipline alone will be able to uncover and delineate the causal linkages between armed conflict and land system change. Working together should thus be beneficial to improving our understanding of the impacts of shock events on land systems and thus the ability to understand how land systems may evolve in the future.

Acknowledgements

We are grateful to V. Butsic, L. Eklund, C. Hamilton, A. Shortland and three anonymous reviewers, whose comments greatly improved the manuscript. We are also thankful to D. Müller and D. Munroe for initiating this special issue. This research contributes to the Future Earth Global Land Project.

Disclosure statement

No potential conflict of interest was reported by the authors.

Supplemental Material

Supplemental data for this article can be accessed here.

Notes

1. ‘State-based’ conflicts describe an incompatibility between two parties, at least one of which is the government of a state; ‘non-state’ conflicts describe conflicts between two groups neither of which is the state; ‘one-sided’ conflicts describe direct and deliberate killings of civilians by any organized group, such as governments or rebel groups.

2. Some studies assessed more than one land-use/cover change (e.g. forest change and agricultural change), and the numbers therefore do not add up to 100%.