Monsoon Marauders and Summer Violence: Exploring the Spatial Relationship between Climate Change and Terrorist Activity in India

Abstract

Climate change is a global phenomenon that has been associated with a growing list of concerns in society today, often leaving more questions than answers. Thus, it is no surprise that questions are forming regarding the effects of climate change on global security, and more specifically, terrorism. India is the ideal case study for investigating the relationship between climate change and extremism, with average temperatures in the country reaching record highs as well as having 9,096 terrorist incidents occur during our 20-year study period between 1998 and 2017. Using daily temperature, precipitation, elevation, and distance to the equator data from the National Climatic Data Center and terrorist incidents from the Global Terrorism Data base (GTD), this study assesses the spatial relationship between these factors through geospatial analyses. Suitability analyses indicate that all the climatological variables tested—temperature, precipitation, and elevation—relate to shifting patterns of terrorist activity. We also found that beyond intensity, seasons result in a shifting of patterns in terrorist behavior to other locales. Implications for the global community and for India specifically are discussed.

Keywords:

• Terrorism
• seasons
• climate change
• India
• geospatial analysis

Extreme weather patterns are a global phenomenon that has been associated with a growing list of concerns in society today. Questions are forming regarding the effects of climate change on global security, and more specifically, terrorism and similar forms of political violence. Whereas past research mainly focused on the impact of regional temperature increases deviating from the national norms on environmental conditions, limited attention was given to the extent in which climate change and environmental factors impact levels of political and criminal extremism.

Increasing extreme weather variations have prompted renewed calls by nonprofit, transnational, and governmental entities to address what many describe as an ongoing “climate crisis.” For example, the European Union recently announced one trillion Euros will be designated to battle climate change; funds represent a fourth of the EU’s budget, with other appropriations coming from the private sector (The Hill, 2020). Beyond practitioners and governmental leaders, scholars have also given increased attention to this issue. Climate change and extreme weather events have been associated with increased vulnerability for humans. In their study focusing on Canada, Ho et al. (2018) found the risk for heat vulnerability is greatest in suburban areas. Similarly, in a study on coastal vulnerability in Maryland, United States, Jiang et al. (2015) found that extreme weather events increase the risk for disease in human populations. Mirza (2003) found that extreme climate events impose a disproportionate debt burden on developing countries compared to their developed counterparts. This is an important finding because developing countries also tend to be associated with higher levels of political violence.

Nevertheless, climate change can also facilitate changes in patterns of political violence via its in-direct impact on civilian centers and the operational needs of violent groups. While a growing number of interest groups and community leaders are interested in combating changing weather patterns, there remains doubt among some populations that climate change is occurring. Additionally, skepticism of climate change tends to be connected with personal experiences (Kaufmann et al., 2017). Some individuals acknowledge climate change is “real” but fail to recognize its effects surrounding them. A growing body of literature has argued that proximal distancing impacts psychological perceptions of climate change (Jones et al., 2017; Milfont et al., 2017). Specifically, scholars have found that individuals are more likely to feel that climate change is a salient issue if they are in areas exposed to greater extreme weather events, such as near coastlines or areas lacking precipitation (Brügger et al., 2015; Milfont et al., 2014; Spence et al., 2012).

Building on this body of literature, the current study seeks to use innovative geospatial methods to assess the suitability of climate change in relation to the study of terrorism using the Republic of India as a case study. Given anecdotal associations between climate change and violence, such as comparisons suggesting declines in terrorism in the north when India’s mountains are covered with a blanket of snow, we expect climatological factors to be related to terrorist hot spots. In the subsequent sections, we review the extant literature. We then provide an overview of India’s seasons (ritus), which is important for understanding the phenomenon of interest from a localized lens. Next, we document the complex methodological framework for the study, followed by the results and discussion sections.¹ We found a relationship between climatological factors and terrorist hot spots. While the intersection of all factors only yielded suitability analyses for smaller geographic areas, combining subsets of climatological factors produced more overlap with regions containing a higher density of terrorist activity.

Climate change, crime, and terrorism

A strong body of literature has documented the relationship between seasonality and crime more broadly. For example, Rees and Schnepel (2009) found that communities hosting college football games often see dramatic increases in a variety of crimes, including assaults, vandalism, and disorderly conduct. This connects with a plethora of anecdotal evidence documenting rioting and criminal activity following major sporting events (e.g., Kurland et al., 2014; Lewis, 2007; Roberts & Benjamin, 2000; Russell, 2004; Smith, 2018). Similarly, in their work, Caplan and colleagues (2020) identified shifting crime risk in Jersey City, N.J., across twelve discrete two-hour time intervals (see also: Drawve et al., 2020). While many factors can be associated to the individual crime observation instances, the occurrences follow regular patterns and seasonality. Much like a sports season, natural seasons follow cyclical patterns are likely to invoke similar effects of changing patterns of crime and violence over time; however, a breadth of literature has documented climate change (e.g., Dash et al., 2007; Thompson, 2010), showing that the periods themselves are changing as time progresses unlike the relatively stable length of the normal sports cycle. Despite past research demonstrating an association between weather patterns and violent crime in Chicago (Reeping & Hemenway, 2020), and both climatological events and terrorist activity increasing financial hardships in developing countries, there remains a lack of research investigating the intersection of these two constructs. Roy (2019) found weather variations in northern India to be increasing, particularly for the first six months of each year, while Hansen et al. (2010) found global surface temperature to also be rising.

It has been well established that enough similarities exist between criminals and terrorists that it is appropriate to use criminological theory in explaining terrorist behavior (Griffiths et al., 2017; Marchment et al., 2018). Prior research has shown that there is a particular benefit to applying the journey to crime literature in explaining terrorist target selection (Gill et al., 2019; Gill & Horgan, 2013; Gruenewald et al., 2013; Hasisi et al., 2019; Marchment et al., 2018; Rossmo & Harries, 2011; Smith et al., 2008). Target selection, while obvious once the attack has occurred, is a strategic, complicated task that requires a multifaceted approach in its understanding.

This logic is not restricted to civilian populations. Some terrorist organizations, for example, tend to operate in arid, remote regions, such as mountains of Afghanistan or Pakistan’s Federally Administered Tribal Areas. As climatological factors continue to shift, such as through increasing temperatures and decreasing rainfall, these illicit actors may be forced to vacate their current base locations in search of geographic terrain that is more favorable for conditions of human life. The possible forced relocation of terrorist organizations, due to extreme weather patterns, from remote, isolated areas closer to societal centers, remains a void in the scholarly literature, despite grave policy implications of such a scenario.

Security studies scholars have sought to understand if conflicts are related to climate change. For example, a 2012 special issue of the Journal of Peace Research focused specifically on the causal mechanisms linking climate change to conflict. Hendrix and Salehyan (2012) found variability in rainfall to be significantly related to political conflict, even when controlling for more traditional conflict predictors, emphasizing the need for further research specifically linking climatological factors to terrorist hot spots. Related, Bergholt and Lujala (2012) found that climate-related natural disasters do not result in increased armed conflict risk. While climate-related phenomena may not directly impact broader conflicts, scholars have found some evidence to suggest that it could have an impact on lower intensity conflicts or more decentralized efforts, such as terrorist attacks.

Political violence in India

Terrorism is a social phenomenon, deeply intertwined within the fabric of society. Influences of terrorism range from airport security to the proliferation of television shows depicting terrorist and counterterrorist narratives. Perhaps most intriguing of this phenomenon is its reliance on a wide variety of variables and considerations. The social nature of terrorism makes it dependent on such factors as political and economic situations, support networks, even geography. Recent scholarship has even found weather and seasons affect terrorist levels (Berrebi & Ostwald, 2011; Bohannon, 2014; Braithwaite & Johnson, 2010; Weimann & Brosius, 1988). However, this research is restricted to Afghanistan (Bilby, 2012; Dobias & Sprague, 2010; Hultman, 2012; Lawrence, 2010; Rashid, 2001; Shortland & Bohannon, 2014). Replicating these studies in the context of other countries is desperately needed to generalize these observations to a wider observation.

India is an ideal case study, as the country has faced extreme levels of terrorism. The country is home to the thuggees, which were one of the longest surviving terrorist campaigns in the world, persisting for approximately 600 years with some figures crediting the movement with upward of two million deaths by strangulation (Dmello, 2018; Laqueur, 2001; Rapoport, 1983). Extremism in India has primarily orbited around two spheres: Indo-Pakistani tensions and ethnically motivated insurgencies. Relations between India and Pakistan have been fragile since the two nations gained independence in 1947. Several of the high-profile attacks in India, such as Lashkar-e-Taiba’s bombing campaign throughout Mumbai and its prior attack on the Indian Parliament in New Delhi in partnership with Jaish-e-Mohammed could be connected back to Pakistan and Islamabad’s Inter-Services Intelligence (ISI) agency (Dmello & Perliger, 2019; Javaid & Kamal, 2013). As Blank (2013) noted, the ISI wages a shadow war with India, funding initiatives to destabilize Kashmir while maintaining plausible deniability. Terrorism in the eastern region is largely attributed to separatism and ethno-nationalism. Hussain (2013) describes how this region’s geographic factors position it to become a hotbed for violence, being an area surrounded by international borders, connecting to the main part of India by a corridor home to the longest-running insurgency led by Naga tribal separatists. The author’s analysis highlights the intersection of various categories of political violence (i.e., insurgency and terrorism) operating in the area. He then goes on to describe in half of India's eight north-eastern Indian states (i.e., Assam, Manipur, Nagaland and Tripura), violence levels had increased so much that they could justifiably be categorised as Low Intensity Warfare (Hussain, 2013).

However, research has found that political violence extends far beyond just dynamics of religion and ethno-nationalism. For example, Dmello et al. (2020) found that violence perpetrated by organizations ideologically aligned with the ruling party would increase following an electoral victory, demonstrating a case of majoritarian empowerment in India. Similarly, Wilkinson (2006) found that power-sharing, specifically the ruling party’s reliance on minority votes, will play a large role in determining if the government seeks to intervene in anti-minority aggression. Varshney (2003) has investigated the mechanisms for extremism in India more broadly, noting several drivers of aggression and communal violence.

India’s six Ritus

When analyzing violence in the context of climate change, seasons likely impact the relationship, as climate change could manifest differently throughout the year. India has six distinct seasons, referred to as “Ritus,” following the lunar Hindu calendar, rather than aligning with Western, solar-based models of seasons. This refined definitional standard of “seasons” allows for a more accurate prediction of the effect of weather patterns on terrorist effectiveness in the Indian context. India’s seasons are tied to the specific climatological patterns of the country.

The Ritus occur during the same month each year, though specific dates of transition may slightly deviate. Like with Western seasons, each Ritu is characterized by its own weather patterns. The first Ritu is Śiśira, the Hindu Winter, lasting January to March. This is the coldest part of the year. Second, is Vasanta, Spring, lasting March to May. The Vernal Equinox occurs at the midpoint of this Ritu, where temperatures start to warm, and flowers can blossom. Third, is Grishma, Summer, starting in May and ending with the summer solstice, followed by Varshā, India’s Monsoon season, lasting July to September, which sees heavy monsoon rains, providing the moisture necessary for the country’s sprawling agricultural economy. Śarad, India’s Post-Monsoon season (also referred to as India’s “Autumn”), lasts September to November. Last, is Hemanta, the Pre-Winter Ritu, lasting November to December, ending with the winter solstice.

The current study

Our study setting is India, a highly populated country with a rich history of both terrorist incidents and climate extremes. This exploratory study focuses on macro-level patterns of extremist activity, seeking to evaluate the relationship between climatological factors and incidents of extremism in India as a whole, not based on individual group-level displacement across time. To give some context regarding size, India is one of the largest countries in the world both in geography at 308,928 km² and population of approximately 1.3 billion in 2019 according to The World Bank (2020). As reviewed in our literature, the country has experienced a multitude of conflict and terrorist activities throughout the years. During our study period from 1998 to 2017, the country experienced 9,096 terrorist attacks. This high level of violence, coupled with the fact that India is well known for its extreme weather throughout its six seasons, makes it a prime candidate to explore the relationship between terrorist conflict and climate change. Thus, the current study is attempting to assess the suitability of climate change in relation to the study of terrorism and what climatological variables, if any, should be considered when contemplating these issues. A clear relationship would show an overlap (suitability) between one or more climatological factors and terrorist hot spots in a given region.

Methodology

Data & variables

The data for this project come from two open sources. The first dataset used was the Global Terrorism Database (GTD), which is an open-source dataset that includes comprehensively and systematically coded data on incident-level violent activity from around the world (Institute for Economics & Peace, 2016), including geographic coordinates for attacks. The GTD codebook defines a terrorist act as, “the threatened or actual use of illegal force and violence by a non-state actor to attain a political, economic, religious, or social goal through fear, coercion, or intimidation” (START, 2021, p. 10). For the current study, the authors adopt the GTD’s definition of “terrorism.” We included all terrorist incidents in a twenty-year span from 1998 to 2017 (N = 9,096), which were then stratified by month for analysis purposes (Table 1). Our study period was largely dictated by the availability of climatological factors data and complete terrorism data.

Table 1. Number of terrorist attacks by month in India 1998–2017.

Month	Number of terrorist incidents (n)
January	757
February	597
March	640
April	899
May	872
June	764
July	810
August	898
September	641
October	793
November	788
December	637
TOTAL	9,096

It should be noted that the GTD’s inclusion of conflict acknowledges the possibility that incidents not exclusively terrorism, such as those that may have some characteristics of other forms of violence, such as insurgency, may also be captured, if they meet the aforementioned criteria of “terrorism” (START, 2021). For our purposes, we include incidents as they meet the operationalized definition of “terrorism,” we also acknowledge that activity by insurgent and terrorist actors does not always overlap, and decision-making often varies based on differences in motivations and objectives.² For example, Ubilava et al. (2023) found that violence perpetrated by militias during harvest periods in Africa increased by 10%, while observing no such change for state or rebel groups, suggesting that insurgent movements located primarily in rural or agricultural areas may have different violence patterns from extremist groups that are operating closer to urban centers.

The other open-sourced data consists of the climatological variables included in the study. These variables include temperature, precipitation, distance from equator, and elevation. This data was obtained from and produced by National Oceanic and Atmospheric Administration’s (NOAAs) National Center for Environmental Information (NCEI). The files included data from between 83 and 97 unique weather stations, dependent on the month and year, over the past twenty years and in total yielded 7,751 unique climate measurements (Table 2). Temperature was given as a monthly average that was computed by adding the unrounded monthly means of the maximums (average of daily maximum temperatures) and minimum (average of the daily minimum temperatures) temperatures and dividing by two; values were set to missing if either the maximum or minimum was missing. Precipitation was based on daily or multi-day (if daily is missing) rainfall levels in millimeters. Elevation was given in meters. Finally, the distance from equator was measured by its latitude. It should be noted that there was no data for June 2011 or April 2015 during our study’s period.

Table 2. Monthly data points, stations, & kriging results.

Month	Number of points	Number of unique stations	Temperature RMSE	Precipitation RMSE	Elevation RMSE	Distance from equator RMSE
January	668	89	1.06	3.02
February	805	93	2.03	2.01
March	765	92	1.91	2.96
April	602	85	2.52	14.65
May	543	87	2.45	81.00
June	393	83	2.45	160.72
July	508	86	2.35	260.81
August	585	88	2.13	224.06
September	636	93	2.07	91.09
October	726	94	2.15	22.16
November	774	93	2.056	21.534
December	746	93	2.037	2.304
Total	7751	97			362.53	0.99

Analytic strategy

Analyses are focused on the monthly level. This decision was made for three reasons. First, the exact date of the seasonal transition shifts slightly from year to year, with some months consistently split between seasons. For example, May is split between Vasanta (Spring) and Grishma (Summer). Second, an analysis at the monthly level provides a more granular level of detail, enabling not only a discussion of suitability across seasons, but also within. However, the Indian ritus must also be taken into consideration to ensure the representation of localized context. The ritus are inherently oriented around the climatological factors; thus, it is expected that variance between suitability would occur across seasons. Finally, in the interest of producing generalizable knowledge, orienting analyses around months instead of years could also be used to compare findings from India to other geographic locations in subsequent research initiatives.

The analytic strategy for this paper involved several methodologies. We first wanted to empirically document climate change was occurring in India. So as not to detract from the purpose of the current study, which is to explore the relationship between climate change and terrorism, we limit this validation to extreme temperature variation, as other scholars have documented the existence of climate change (e.g., Dash et al., 2007; Intergovernmental Panel on Climate Change (IPCC), 2023). To do this we used a one-sample t-test to compare mean temperatures from the 30 years prior to our study, 1967–1996, to our study period 1998–2017. This was done only on temperature for a couple reasons. One was that temperature rising over several years is often associated with climate change. This is consistent with methods employed as past research has indicated that a region’s climate is determined by analysis of variables over a period of at least 30 years (Arguez & Vose, 2011; Daron & Stainforth, 2013; Richardson et al., 2005). Further, temperature was the only variable that was consistently reported in the 30 years prior to our study period. Precipitation was often left out of the data files prior to 1980 or was only recorded at a few of the weather stations. Finally, elevation and distance to the equator were not good candidates for this as they remain constant over the years and should be used as companions or control variables to other climatological variables in connection with climate change.

Second, we utilized hot spot mapping to understand the clustering of terrorism in India. Specifically, we used Kernel Density Estimation (KDE), a method that has been found to have greater predictive accuracy than other methods when analyzing crime (Chainey et al., 2008; Hart & Zandbergen, 2014). KDE is an interpolation method wherein individual points, more specifically, clusters of these points represent “hot” areas for our variable of interest (i.e., terrorism) (Bailey & Gatrell, 1995; Hart & Zandbergen, 2014).

ArcGIS 10.6 was used to run a KDE for terrorist incidents across the entirety of India. Generally, kernel density maps are done on a smaller geographical surface. In contrast, India takes up quite a large surface area at 308,928 km². Prior research has suggested that cell size should be the measurement of the shortest side of the minimum bounding rectangle (MBR) divided by 150 (Chainey, 2011). Though later findings suggested that the cell size has little to no impact on the ability of KDE to predict hotspots and is more representative of boosting the visual appeal of the map, with higher resolutions avoiding poor pixel quality, the MBR is still a useful starting point for analysis (Chainey, 2013). Using the “minimum bounding geometry” tool in ArcGIS it was determined that the shortest side was 3,300,000 meters. Keeping the research in mind, the cell size selected was 22,000 meters. The bandwidth was also determined based on prior literature wherein it was found that a good baseline should be five times the cell size, (Chainey, 2011). Unlike cell size, bandwidth can affect the abilities of KDE to predict spatial patterns (Chainey, 2013). Various search radii were tested; however, the radius was finally set to 110,000 meters—five times our cell size—after other fixed intervals showed no significant variation. Finally, the KDE was classified using the “standard deviation” method to symbolize how area counts varied from the mean. The resulting maps indicated areas that were prime for terrorist incidents to occur. Whether these clusters changed over the course of a day informed the researchers of the necessity to take a closer look at the climatological factors within these hot spots that could be influencing terrorism rates.

Third, we utilized spatial interpolation of our climatological variables—temperature, precipitation, elevation, distance from equator—for them to be operationalized properly for the final suitability analyses. Our climatological variables were given as point data at specific weather stations across India. In its current form this data, then, would only provide temperature, for example, at each given point, not for the national landscape. Climatological factors are dynamic across geographies, meaning that temperature at point A can be 50 degrees whereas temperature at Point B can be 0 degrees. This indicates a presence of temperature at any given location in that space. If we think about India as having an infinite number of dots covering its entire geography so that it appears blacked out by these dots, each dot will have its own temperature. Given the consistency of climatological factors in space and time, it is appropriate to take the values given to us at each unique weather station and spread them over the entirety of India to create complete layers of temperature, precipitation, distance from equator, and elevation, respectively. This would yield a complete spread of temperature, for example, as opposed to the temperature at only those weather station locations.

Ultimately, the interpolation method selected for this study was kriging. Kriging is a type of spatial interpolation that forms weights from surrounding measured values to predict values at unmeasured locations. With this method the closest measured values hold the most influence, which comes from a semi-variogram (Janssen et al., 2004). The equation for this looks like this: $${\mathrm{\text{Semi}}‐\mathrm{\text{variogram}}\left(\text{distance h}\right)=0.5\text{* average}\left[\right(\text{value at location}i–\text{value at location}j)}^2]$$

The use of this interpolation method in conjunction with climatological variables is supported by research (Boer et al., 2001; Kilibarda et al., 2014). Further supporting our use of kriging over other interpolation methods was the root mean squared error (RMSE); a low RMSE in comparison to other interpolation techniques, is considered an indicator of good model fit. This is because the RMSE indicates how close the observed data points are to the model’s predicted values. The result is a smooth surface with values assigned to every cell.

We interpolated the surfaces for temperature, precipitation, elevation, and distance from the equator. This was done monthly for both temperature and precipitation as these values varied during these times, resulting in 12 surfaces per variable (24 total). Elevation and distance from the equator were done only once because they remained static in time, resulting in two interpolated surfaces total. For the most part we saw low RMSE’s (Table 2), indicating good model fit. There was some higher variability in precipitation likely because of the wetter climate during the monsoon season.

The final step of the interpolation process was to classify the raster surfaces so that they were prepped for the final analyses. To do this we utilized a couple different strategies dependent on the variable. For temperature, precipitation, and elevation the standard deviation method of classification was used to break the surfaces into three categories: extreme high, normal, and extreme low. For example, temperatures that were above one standard deviation above the mean were deemed as the “extreme high” category, those that were one standard deviation below the mean were “extreme low,” and anything within one standard deviation of the mean was “normal.” The one surface for elevation and all surfaces for the temperature variable yielded three categories, whereas precipitation had five months/surfaces wherein only two categories could be established because of lack of a lower extreme. It did not make sense to use extremes to classify the distance from the equator, therefore, equal distance was used to separate the country into six equal parallels across the country.

The next step of the analytic strategy is the suitability analysis. Generally, a suitability analysis identifies geographic areas that meet the criteria set forth by a researcher to deem it appropriate or not for something or phenomenon. For example, this has been used for determining agricultural land suitability (Akıncı et al., 2013), wildlife habitat and corridor analysis (Ahmad et al., 2018; Walker & Craighead, 1997), and siting oil and gas stations (Peprah et al., 2018), just to name a few. Essentially, our study is seeking to determine suitability for terrorist activity via climate. This is done in a few steps. This begins by overlaying the KDE results on top of our interpolated climate data. When overlaying these results, we are looking to see where our terrorism hotspots and the extreme categories or distance from the equator overlap. We can identify a spatial relationship in areas where these variables overlap.

The next step is to test the combined power of our climatological variables by creating monthly Boolean rasters for temperature and precipitation, and one single Boolean raster for elevation. A Boolean raster assigns one of two values, a zero or a one, to the current values in a raster. In the case of our three variables, a one was assigned to the climatological factors we deemed most suitable for terrorist attacks to occur in and a zero for those unsuitable. In all three variables the extremes were associated with suitability while the average or normal ranges were considered unsuitable for terroristic violence to occur. Once the Boolean rasters were created the raster calculator was used to multiply the surfaces together. The resulting surface would show only those areas where all three of the variables overlapped.

The final step is to overlay our respective surfaces of terrorism hotspots and climatological interpolations. We overlay the KDE for the respective month on top of each climatological variable individually and Boolean raster (combined climatological factors effect) to determine climate effects, if any exist. These maps act to determine the plausibility of a statistical spatial relationship existing between terrorism and climate while visualizing their results.

Results

Our study showed that when comparing temperatures in our study period, 1998–2017, to the 30 years prior that there is some variation throughout the months (Figure 1). A comparison of the means using a one-sample t-test to discern the difference between these two time periods revealed support for climate change in India (Table 3). There is a significant difference in four of the twelve months and another month approaching significance. In January, June, and October the temperatures are significantly lower than the 30 years prior, whereas in July the temperature is significantly higher. Regardless of the direction of the differences in temperature, it is clear we can establish an argument for climate change.

Figure 1. Average temperatures in India comparing 1968–1997 to 1998–2017.

Table 3. One-sample T-test comparing average temperatures from 1968–1997 to 1998–2017.

Month	Statistic (t)
January	−2.332**
February	−0.721
March	−0.343
April	−1.789*
May	1.252
June	4.634***
July	5.402***
August	−0.097
September	0.224
October	−3.09**
November	0.819
December	−0.492

Significance *0.1, **0.05, ***0.001.

Next, KDE’s were run for each month separately and one including all terrorist incidents during our time frame. Results did not differ much from month to month. In fact, we saw the same regions repeatedly affected by terroristic violence (Figure 2). These regions are primarily in the North and Eastern areas of the country. The States where we see terrorism hotspots occur completely within or spread into included: Jammu Kashmir,³ Jharkhand, Bihar, West Bengal, Assam, Meghalaya, Manipur, Tripura, Chhattisgarh, Orissa, Maharashtra, Tamil Nadu, Karnataka, Telangana, Andhra Pradesh, Haryana, Punjab, Delhi, and Kerala. These KDE’s were later used to overlay on top of climatological variables to establish any possibility of a spatial relationship between the variables. If a spatial relationship does, in fact, exist we would expect to see the hotspots occurring on top of the “extreme” areas of each, or those areas deemed outside of one standard deviation from the mean.

Figure 2. Kernel density of all terrorist incidents 1998–2017.

Next, we analyzed the climatological variables. More specifically, how these variables were spread out across India. Distance from equator was relative in the sense that India’s proximity to the equator is close (Figure 3). However, because the whole country is so close to the equator it would be difficult to ascertain this variable’s effect on terrorism. We can say that when terrorist attacks are overlayed on top of our distance from equator variable, we see that most attacks are occurring in the fourth, fifth and sixth layers. This would indicate that the farther away you get from the equator the more terrorist attacks we see. It should be noted that other factors, such as proximity to border conflict zones, could impact this finding; however, as shown in Figure 2 above, not all terrorist clusters are located along the border regions. This is relative only to India as the country itself is still much closer to the equator than many other countries in the world. Distance to the equator may be more relevant in a global study that compares several countries relationships with terrorism and climate change. Therefore, we cannot say with certainty that this variable has any relationship with terrorism or climate change.

Figure 3. India’s distance to the equator.

Analyzing extreme temperatures, extreme precipitations, and extreme elevations hinted at a much stronger spatial association. It was consistently found across the 12 months that the extremities in the climate variables aligned with hotspots for terrorist activity (Figure 4) (See online supplemental Appendix for all 12 months). To highlight these differences, in Figure 4, we compare May and July to show some variation in the months. Extreme temperatures in May were visualized in the north, north-west, east, and central regions of the country. This overlaps with terrorist incident hotspots in the north and east. This differs from extreme temperatures in July wherein we see these extremities mainly in the north, north-west, south, and a small area in the east. Again, we see the terrorist hotspot overlap in the north, which is the Jammu Kashmir region, and we see some overlap in the eastern part of Chhattisgarh. For precipitation, we see some months where there were little extremities. For example, one extreme zone in the east that overlaps with a terrorist incident hotspot. However, when we move into the monsoon season, we see many more extreme zones, as is evidenced in July, and significant more overlap with terrorist hotspots. Finally, elevation remained constant throughout all the months and because we see little variation in the terrorist incident hotspots, we can see that overlap occurring mainly in the north and east. From these analyses we can conclude that there is some level of spatial relationship with terrorist incidents in India. What we are unable to determine with these analyses is how strong that association is.

Figure 4. Comparison of May and July climate variables.

Finally, once a spatial relationship between each of the climatological variables could be entertained and found plausible it was necessary to determine the extent to which, if any, relationship exists between their combined effects and terrorist activity in India. These analyses offered little insight on their own as not much support could be seen for a spatial relationship and terrorist activities here (Figure 5). Keeping with our theme above, Boolean suitability rasters for both May and July are presented in Figure 5. In both cases the extreme values cover little geographic space and appear to have little to no spatial relationship with terrorist violence. Because individual support for each climatological variable’s relation to terrorist activity was verified with the above interpolation and KDE overlays, this would suggest that the extent to which each individual variable affects this will vary. Thus, the risk for terrorist attacks will differ based on climate variable and, presumably, by how extreme their measures are.

Figure 5. Boolean suitability rasters of combined climatological effects on terrorism in India.

These two months fall into different seasons. May is partitioned between Vasanta (Spring) in the earlier part of the month and Grishma (Summer) toward the latter, while July falls in Varsha (monsoon season). Although both months, and the respective seasons, are characterized by hot temperatures, other factors (most notably, precipitation) vary dramatically. Consistent with monsoon season, India has far more areas of extreme precipitation in July than in May, as shown in Figure 4 above. Because of season differences, it may be more beneficial to adapt which climatological factors are focused on when predicting terrorist targeting trends across years based on specific season. For example, as shown in the examples listed here, when comparing summer and monsoon seasons, temperatures are relatively similar in terms of extremity, but the deviation in precipitation levels could be a more salient correlate of terrorist hot spot suitability over evaluations of temperature. This is in line with past research, such as Hendrix and Salehyan (2012) analysis in the African context. Deviations across seasons are not restricted to just this example; to view extremism and suitability for each variable divided by each calendar month, see the online appendix.

Discussion

India has had no shortage of terrorist activity plaguing the nation. The notoriety of the terror plot in Mumbai perpetrated by Lashkar-e-Taiba has caused many to focus their attention on the western side of the country (see Tankel, 2013), in part, because of the large impact of the siege on economic and human security. However, political violence in the country has long been clustered in northern and eastern India, as shown in Figure 2. For example, the northeast region has a continued and sustained insurgency movements, connected to ethnic power relations in the region (Cline, 2006; Piazza, 2010). Similarly, Prakash (2008) documents violence to the north, particularly in the states of Punjab and Jammu and Kashmir. Whereas prior region centers on ethnic conflicts, the latter connects with hostile relations between India and Pakistan rooted in territorial disputes between the countries (Noor, 2007).

T-tests revealed significant temperature differences between the study period and the 30 preceding years for January, June, July, and October, empirical evidence that extreme temperature deviations are occurring outside of just the summer and monsoon seasons. Given that we see significant differences in temperature during periods with stratified precipitation (such as India’ monsoon and summer seasons) demonstrates the interaction of these two factors, reinforcing the necessity for research in this area to encompass a variety of climatological factors. The use of increasing temperatures as a proxy for climate change is not sufficient, as it only tells a fragment of the story. Research has shown that climate change has impacted India’s monsoon season, causing shifts to the timing and intensity of the season (Annamalai et al., 2013; Taraz, 2017). This is of importance, because it has far-reaching implications. For example, alterations to the monsoon season impact grain production (Auffhammer et al., 2012), which in turn affects food security in local populations and economic security from lost product sales. Thus, to holistically evaluate the impact of climate change on terrorism, multiple dimensions of “climate change” should be explored, as was included in this research effort. Turning back to our purposes, analyses reveal that extreme weather patterns also likely impact extremist operations and targeting practices.

Of the many climatological factors circulated in the scholarly literature, the suitability analyses indicate that all the individual variables—temperature, precipitation, and elevation—included are shown to be connected to terrorism. For each month of the year, the factors are present in some form. This suggests that the relationship between climatological factors and terrorist incidents may be stronger than previously anticipated. This argument is not novel; scholars have long argued a theoretical connection between these constructs (e.g., Chaturvedi & Doyle, 2015; Parenti, 2011). However, a demonstration of the empirical connection has been understudied to date in the extant literature. India’s weather extremes are dispersed throughout the country, which suggests that the impact of climatological factors on terrorist targeting associated with terrorist incident clustering could also be applicable in explaining attacks occurring elsewhere in the country.

Overall, we see some evidence in support of the hypothesis that season impacts terrorism, noting the additional layer of extreme variations in climatological factors. For example, during July, Monsoon season (Varshā), KDEs reveal large numbers of terrorist hotspots were pervasive throughout India, ranging from the eastern region (Assam, Nagaland, Manipur, Tripura, Meghalaya, Sikkim, West Bengal, Orissa, and Jharkhand), in the middle of the country (Chhatisgarh), north (Jammu and Kashmir and Punjab), south (Karnataka), and west (Gujarat and Maharashtra). Similarly, when overlaying each of the climatological factors (elevation, temperature, and precipitation), we see similar trends. In contrast, the analyses occurring in February, the peak of India’s winter (Śiśira), reveal more variation in other parts of the country; for example, Tamil Nadu and Kerala in the south become hotspots, as does Delhi in the north. Variations across seasons is in line with past research identifying the terrorist fighting season. However, the current analyses demonstrate that beyond intensity, seasons also result in a shifting of patterns in terrorist behavior to other locales, as there is a changing calculus in motivation and opportunities for groups engaging in violence. When controlling for climatological factors, we see similar intersections between the hotspots and areas of extreme variation.

Conclusion

This paper sought to investigate the spatial relationships between climatological and terrorist incidents using the Republic of India as a case study. Although the scientific community is in concurrence that the Earth’s climate is changing, many in the policy sphere debate the veracity of these vast bodies of work. Thus, we first showed that climate change is occurring in India as a baseline to ensure applicability of the paper to both scholarly and practitioner audiences across the ideological spectrum. This was done through comparing the monthly means of the 30 years prior to our study period with the monthly means of our study period, empirically showing climate change in India across time.

Next, KDE hotspot analyses showed that there was little monthly variation in where terrorist activity occurred, with the core of the activity in the north and east regions of the country. We next conducted spatial interpolation on our four climatological variables—temperature, precipitation, elevation, and distance to the equator—using a kriging method. It was quickly realized that distance to the equator would not be a great indicator of a relationship between climate change and terrorism. Temperature, precipitation, and elevation, however, offered useful insights and plausibility to the argument that these variables are spatially related to terrorist activity. This was evidenced by overlaying the KDE’s of terrorist attacks in India by month on top of these interpolated surfaces. The areas that were most commonly found to have a spatial relationship with terrorism, included Assam, West Bengal, Andhra Pradesh, Jharkhand, Manipur, Tripura, and Jammu and Kashmir. Finally, we conducted Boolean suitability analyses to determine the overlap of terrorist hotspots and the combined effect of three of our climatological factors—temperature, precipitation, and elevation. While little support for the combined effect could be found, it was determined that because of the spatial associations visualized at the individual level that these variables likely have differing influence on the risk for terrorist activity.

While this paper works toward filling several gaps in the scholarly literature, it is not without limitations. First, as this is an exploratory study, we are documenting the relationships between constructs; however, we make no assertion to causality in this study. Prior to establishing a causal mechanism, it was essential to first provide a framework to empirically establish a relationship between factors, a methodologically rigorous undertaking documented here. Second, while India is a large and diverse country in terms of both climate and security of its regions, our analysis is still limited to one nation. While we cannot make claims about the generalizability of these findings to other geographic contexts, we note that climatological factors can and do transcend the artificially constructed national borders established by colonial powers. Thus, an argument can be made that similar trends would be seen if this analytical framework were to be replicated in another context. Additionally, as previously mentioned, the GTD includes incidents that may not be exclusively terrorism—future research could benefit from a more detailed parsing to evaluate deviations from groups engaging in pure terrorism versus other forms of extremism, such as insurgency or militias.

This paper establishes the relationship between climatological factors and terrorist incidents in India using geospatial methods. Exploratory in nature, we acknowledge there is much work left to be done in this space, as there remains much unknown about the nexus of climate change and terrorism studies. Specifically, future research can investigate the causal mechanisms driving the relationship between these constructs and how other factors, such as socio-demographic, political characteristics, and environmental features, impact the risk of terrorist activity in a given area. Additionally, Shako (2015) identified six climatological parameters (temperature, wind speed, relative humidity, rainfall, atmospheric pressure, and sunshine hours), though other scholars have included more; future research could benefit from further incorporating additional factors to more holistically analyze the extent to which climate change impacts terrorism. However, we acknowledge the technical difficulty associated with such a recommendation. Nonetheless, undertaking these scholarly efforts will benefit the field, given the complexity of the relationship between physical and social characteristics pertaining to climatological impacts on violence. The current study establishes a framework to begin these complex conversations, but further work is needed.

Finally, this research can also inform future policy. Upon his election in the November 2020 election, President Elect Joe Biden made climate change one of his administration’s top priorities, declaring his intent to tackle the issue from the onset of his presidency (Baragona, 2020). As the pendulum within the governing sphere again prepares to pivot, we can expect to see an increase in federal initiatives to tackle climate change. However, it is important, even essential, to understand the broader impact of climate change to ensure an effective national and global strategy is crafted. Brigadier General (Ret.) Stephen A. Cheney said, “Take it from the military: Climate security is national security” (Cheney, 2018, as cited in Motta et al., 2020). Thus, the only way to truly operationalize an effective national security strategy is to ensure we address nontraditional approaches to security, such as food, energy, water, human, and climate security.

Correction Statement

This article has been corrected with minor changes. These changes do not impact the academic content of the article.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Notes

1 We highlight a selection of the suitability analyses in the manuscript for brevity and page count limitations. For a detailed review of all suitability analyses, please refer to the online supplemental appendix.

2 Future research could benefit from analysis that explicitly separates terrorist activity from that of insurgent actors to evaluate if spatial deviations occur between those groups; however, such efforts would be reliant on the availability of reliable and reproducible data.

3 In October 2019, the Indian government dissolved the State of Jammu and Kashmir, splitting the region into two Union Territories: (1) Jammu and Kashmir and (2) Ladakh (Bhatt, 2019; Hussain & Sharma, 2019). However, because the current study includes data from 1998 to 2017, we refer to J&K as a State to reflect its status during that time.