ArticlePDF Available

Climate has contrasting direct and indirect effects on armed conflicts


Abstract and Figures

There is an active debate regarding the influence that climate has on the risk of armed conflict, which stems from challenges in assembling unbiased datasets, competing hypotheses on the mechanisms of climate influence, and the difficulty of disentangling direct and indirect climate effects. We use gridded historical non-state conflict records, satellite data, and land surface models in a structural equation modeling approach to uncover the direct and indirect effects of climate on violent conflicts in Africa and the Middle East (ME). We show that climate–conflict linkages in these regions are more complex than previously suggested, with multiple mechanisms at work. Warm temperatures and low rainfall direct effects on conflict risk were stronger than indirect effects through food and water supplies. Warming increases the risk of violence in Africa but unexpectedly decreases this risk in the ME. Furthermore, at the country level, warming decreases the risk of violence in most West African countries. Overall, we find a non-linear response of conflict to warming across countries that depends on the local temperature conditions. We further show that magnitude and sign of the effects largely depend on the scale of analysis and geographical context. These results imply that extreme caution should be exerted when attempting to explain or project local climate-conflict relationships based on a single, generalized theory.
Content may be subject to copyright.
Environ. Res. Lett. 15 (2020) 104017
Environmental Research Letters
19 April 2020
8 July 2020
27 July 2020
21 September 2020
Original content from
this work may be used
under the terms of the
Creative Commons
Attribution 4.0 licence.
Any further distribution
of this work must
maintain attribution to
the author(s) and the title
of the work, journal
citation and DOI.
Climate has contrasting direct and indirect effects on armed
David Helman1,2,5, Benjamin F Zaitchik1and Chris Funk3,4
1Department of Earth and Planetary Sciences, Johns Hopkins University, Baltimore, Maryland, United States of America
2Currently at the Department of Soil & Water Sciences, The Robert H. Smith Faculty of Agriculture, Food & Environment, Rehovot,
and Advanced School for Environmental Studies, The Hebrew University of Jerusalem, Jerusalem, Israel
3U.S. Geological Survey, Earth Resources Observation and Science Center, Sioux Falls, South Dakota, United States of America
4Climate Hazards Center, University of California, Santa Barbara, California, United States of America
Keywords: climate, conflict, violence, warming, Africa, Middle East, structural equation model
Supplementary material for this article is available online
There is an active debate regarding the influence that climate has on the risk of armed conflict,
which stems from challenges in assembling unbiased datasets, competing hypotheses on the
mechanisms of climate influence, and the difficulty of disentangling direct and indirect climate
effects. We use gridded historical non-state conflict records, satellite data, and land surface models
in a structural equation modeling approach to uncover the direct and indirect effects of climate on
violent conflicts in Africa and the Middle East (ME). We show that climate–conflict linkages in
these regions are more complex than previously suggested, with multiple mechanisms at work.
Warm temperatures and low rainfall direct effects on conflict risk were stronger than indirect
effects through food and water supplies. Warming increases the risk of violence in Africa but
unexpectedly decreases this risk in the ME. Furthermore, at the country level, warming decreases
the risk of violence in most West African countries. Overall, we find a non-linear response of
conflict to warming across countries that depends on the local temperature conditions. We further
show that magnitude and sign of the effects largely depend on the scale of analysis and
geographical context. These results imply that extreme caution should be exerted when attempting
to explain or project local climate–conflict relationships based on a single, generalized theory.
1. Introduction
Although there is a suggested linkage between viol-
ent conflict and climate, the underlying mechanisms
of the link are still under debate [1,2]. One com-
monly suggested mechanism is of a climate–conflict
link through economic disruption [3,4]. Though
plausible, there is currently no robust evidence for
such a direct climate–economy–conflict nexus [5].
Instead, many studies suggest that climate-driven
depressions may lead to conflict through a combina-
tion of socioeconomic and political failures, particu-
larly in agricultural-dependent regions where people
depend directly on such resources [4]. That is, cli-
mate influences economy, which influences social and
political systems relevant to conflict.
5Author to whom any correspondence should be addressed.
It is also possible that the climate–conflict con-
nection is less direct, operating through the influence
that climate-induced changes in economy, food
security, or group interactions cascade to influ-
ence the probability of inter-group violent con-
flicts. This indirect influence is relevant to theor-
ies like the ‘engagement’ hypothesis, which claims
that when climate crisis reduces economic productiv-
ity people become more likely to engage in con-
flicts than in economic activities [6,7], or the
‘inequality’ hypothesis, which argues that conflict
may upsurge when climate crisis increases economic
inequality because of increasing efforts to redistrib-
ute assets [8], and the ‘state weakness’ hypothesis
that suggests a weakening of governmental insti-
tutions and their ability to suppress violence due
to decline in economic productivity following cli-
mate crisis [9]. All these, suggest that climate has
© 2020 The Author(s). Published by IOP Publishing Ltd
Environ. Res. Lett. 15 (2020) 104017 D Helman et al
an indirect rather than a direct effect on violent
conflicts [10].
While these hypotheses were first studied in the
context of civil wars and other state-engaged conflicts,
research in the past decade on communal, non-state
violence has also emphasized the mediated pathways
through which climate can influence conflict. This
includes the potential for harmful climate anomalies
like drought to drive conflicts in times of scarcity due
to resource competition, lowered opportunity cost, or
other mechanisms [11,12]. But it also includes the
potential for beneficial climate anomalies to increase
conflict due to rent seeking or available resources to
support violent activities during times of abundance
[13,14]. Studies have also found that climate variab-
ility in either direction can lead to increased conflict,
due to the presence of multiple mechanisms driving
conflict or to the presence of qualitatively different
categories of conflict [15,16].
The direct influence of climate on individual
tendency toward violence may also play a role.
Warming, for example, has been shown to enhance
violence through a direct psychological mechanism
[the General Aggression Model—GAM] by mak-
ing people uncomfortable and irritated [17]. Altern-
atively, warming may enhance violence in cooler
environments because warm, more favorable weather
conditions lead to increased activity and interac-
tion between people [Routine Activity Theory—
RAT], which may lead to more opportunities for
conflict [18].
To assess climate impacts on violence and uncover
whether the underlying mechanisms are direct, indir-
ect, or a combination of both, ‘non-climatic’ effects
must be isolated. Some studies do this by pooling
data across locations and applying statistical mod-
els that control for non-climatic factors explicitly.
The climate influence is then examined through its
partial effect on violence [19,20]. Other researchers
argue that controlling for non-climatic factors expli-
citly can absorb most of the climatic impact and,
therefore, may result in an underestimation of the cli-
mate effect [21]. For this reason, it is argued, pooling
analysis across sites is misleading, and climate effects
should be studied by comparing each place with itself
in time rather than with other places. Studies using
this site self-comparison approach have reached more
conclusive results regarding climate impacts on viol-
ence than cross-sectional studies using explicit con-
trols [21,22]. The problem with this self-comparison
approach, however, is that it cannot identify under-
lying ‘universal’ mechanisms because the analysis is
conducted location-by-location rather than across
locations [23].
To some extent, the contrasting results published
in the literature are a reflection of that disagree-
ment [24], with this inconsistency leading to cri-
ticism of climate–conflict research. Some research-
ers have claimed that the link between climate and
conflict is unsupported by the evidence [25]. Fur-
thermore, researchers have been accused of bias in
their approach to the problem [26,27]. Yet, most
experts do believe that climate has a significant effect
on human conflicts [28], though the generality of the
links and the underlying mechanisms are yet to be
Here we use a powerful assemblage of disaggreg-
ated data (table S1 (available online at stacks.iop.
org/ERL/15/104017/mmedia)), which includes the
Uppsala Conflict Data Program (UCDP) conflict
dataset [29] as well as climatic [temperature and rain-
fall anomalies] and non-climatic [anomalies in water
availability, infant mortality rates, agricultural yield,
and economic welfare] datasets derived from satel-
lites and land surface models to test generalizability of
climate–conflict relationships from national to con-
tinental scale. To leverage the strengths of the two
approaches—the site self-comparison and the use of
explicit controls in a cross-sectional analysis—and
explore general mechanisms, we make use of struc-
tural equation modeling [SEM] [30] in which non-
climatic factors are explicitly controlled while dir-
ect and indirect effects of climate—through the non-
climatic factors—are quantified in order to uncover
the underlying mechanisms.
We choose to focus on non-state conflicts rather
than civil wars because small-scale conflicts are likely
to be more sensitive to environmental and climatic
changes [19,28]. Also, we focus on Africa and the
Middle East [ME] because these two regions exper-
ienced a large number of armed conflicts in the last
three decades. Finally, we hypothesize that compar-
ing these two ethnically and culturally distinct, but
yet geographically close regions may reveal contrast-
ing mechanisms.
2. Data and methods
2.1. Armed conflict dataset
2.1.1. The UCDP geolocated violent conflict dataset
We used the most updated Georeferenced Event Data-
set [GED] Global version 18.1 (2017) of the UCDP
[29] for location-specific information on armed con-
flicts in Africa and the ME. The GED.v18.1 is UCDP’s
most disaggregated data set, covering individual
events of organized violence as phenomena of lethal
violence occurring at a given time and place. Events
are sufficiently fine-grained to be geo-coded down to
the level of individual villages, with temporal dur-
ations disaggregated to single, individual days [31].
Conflicts used here are ‘non-state’ conflicts, defined
by UCDP as ‘the use of armed force between two
organized armed groups, neither of which is the gov-
ernment of a state, which results in at least 25 battle-
related deaths in a year’ [31]. Information on specific
conflict is freely available at [], and
questions regarding the definitions used by UCDP
as well as the content of the dataset can be directed
Environ. Res. Lett. 15 (2020) 104017 D Helman et al
to that site. In the GED dataset, each conflict has a
unique identifier [conflict ID], while the start date is
recorded as precisely as possible with the level of pre-
cision for day, month and year indicated alongside
[‘Startprec’ variable in GED.v18.1].
For our analysis we used conflicts indicated with a
‘Startprec’ level of at least five meaning that ‘Day and
month are assigned, year is precisely coded; day and
month are set as precisely as possible’. A violent event
was defined as a coded event, which is unique in terms
of starting and ends dates, and is not a continuation or
part of a previous event. All events were first binned at
a spatial resolution of 0.5×0.5for African and ME
regions by summing the total number of events per
grid per year. Events were assigned to a specific year
by indicated starting date. A layer of violent events by
0.5per year was produced alongside another layer
with the sum of events for the entire period of 1990–
2017 . Because we look for effects on the risk of viol-
ent conflict outbreak, each layer was converted into
a binary layer in which each grid was assign a value
of 1 for grids that experienced violence during this
year, or 0 for grids that did not experience violence.
Although we had information on violence for 1990–
2017, we used only layers for years 1992–2012 in the
SEM analysis because this was the period in which we
had a complete data set of climate and non-climate
variables (see below). We included Syria in our ana-
lysis, but excluded the years after 2010 because of the
poor information on violent events during the period
of the Syrian civil war [31,32].
2.2. Climate data
We used temperature and rainfall data sets, described
below, to seek direct and indirect effects of tem-
perature and rainfall anomalies on non-state con-
flicts. Direct effects of climate could be only at the
time of occurrence, so relationships were analyzed
for the same year (present-year violence). However,
indirect effects—through food, water, and economic
welfare—may occur at a certain time-lag. Because
linking climate anomalies indirectly to conflicts at too
long time-lag periods may be problematic (because of
the uncertainty that such climatic changes are really
related to conflicts many years after), we looked only
for links with a one-year time lag (next-year violence).
2.2.1. Temperature anomaly
We used monthly maximum temperatures from
the newly derived Climate Hazards center InfraRed
Temperature with Stations [CHIRTS] dataset [33].
CHIRTS provides monthly 2 m maximum air tem-
peratures at a high spatial resolution of 0.05and
a quasi-global coverage [60S–70N] from 1983 to
2016. Temperature estimates are derived using a com-
bination of thermal imagery from a constellation of
geostationary satellites, a high-resolution climatology
from the Climate Hazards Center’s Tmax climatology,
and in situ monthly 2 m Tmax air temperature obser-
vations obtained from the Berkeley Earth and Global
Telecommunication System [GTS]. We used the tem-
perature estimates from CHIRTS because these were
shown to be suitable for monitoring temperature
anomalies and extremes in data-sparse regions like
Africa and the ME [33]. The high spatial resolution
temperature estimates were averaged over 0.5×0.5
for the period of the analysis [1992–2012], and the
yearly anomaly was calculated per grid as z-score [the
long-term mean annual temperature was subtracted
from the specific year mean temperature and divided
by the standard deviation].
2.2.2. Rainfall anomaly
For rainfall anomaly, we used the Climate Hazards
center Infrared Precipitation with Stations [CHIRPS]
dataset, available at a high spatial resolution of 0.05
[34]. This product is a quasi-global precipitation
product with daily to seasonal time scales and a
1981 to near real-time period of record. CHIRPS
uses three main types of information: (1) global
0.05rainfall climatologies, (2) time-varying grids of
satellite-based rainfall estimates, and (3) in situ rain-
fall observations. CHIRPS is built on ‘smart’ inter-
polation techniques and high resolution, long period
of record estimates based on infrared cold cloud
duration [CCD] observations as well as on satellite
information, used to represent ungauged locations.
CHIRPS is quite reliable in regions like Africa and
the ME where most rainfall products fail to accurately
represent the high temporal and spatial variability in
rainfall [35] due to the sparse gauge network in this
region [36].
We used CHIRPS monthly rainfall sums [from
January to December] to assess the annual rainfall
anomaly for 1992–2012, calculated as z-scores [the
long-term mean annual rainfall subtracted from spe-
cific year rainfall sum, divided by the standard devi-
ation]. Each year a z-score map was produced while
pixels were aggregated to the spatial resolution of the
analysis [0.5×0.5]. Annual rainfall is not a com-
prehensive proxy for conflict-relevant rainfall variab-
ility, but it offers a practical, objective measure that
can be applied consistently across our diverse study
2.3. Non-climate data
2.3.1. Infant mortality rate
As a proxy of socioeconomic development, we used
information on infant mortality rate [IMR] from
the Global Subnational Infant Mortality Rates, Ver-
sion 1 [GSIMR.v1] [37]. The GSIMR.v1 dataset is
produced by the Columbia University Center for
International Earth Science Information Network
[CIESIN] at a high spatial resolution of 5 km and is
freely available for download as a raster data layer
from [].
The GSIMR.v1 consists of IMR estimates for the
Environ. Res. Lett. 15 (2020) 104017 D Helman et al
year 2000, which was collected from vital registration
data, surveys and models or estimated using repor-
ted live births and infant deaths data. Though our
analysis spans the period of 1992–2012, we assume
that the 2000 GSIMR.v1 is, in average, representat-
ive of the entire period following previous studies
[38]. The IMR is calculated as the number of deaths
of infants of less than one year old divided by the
number of live births and multiplied by 1000. We
preferred using the IMR as a proxy of poverty and
socioeconomic status instead of using other variables
because measures like gross domestic product [GDP]
or population living on less than one U.S. dollar per
day, are difficult to obtain at sub-national levels, par-
ticularly for the regions of this study. Moreover, using
IMR has several advantages over other socioeconomic
metrics. For example, IMR is a highly standardized
measure compared to other measures, which means
that it can be used to compare between countries
with different economic systems better than GDP, for
example [38]. Also, IMR is less likely to be influenced
by skewed wealth distribution. And, information on
IMR is available for�90% or more of the population
in medium- and low-income countries. The original
5 km IMR data layer was binned at the spatial resol-
ution of 0.5×0.5, which is the resolution of the
analysis and used as a static map layer.
2.3.2. Distance to border
Distance from/to political borders was assessed using
a geographical information system and a shapefile
layer of the political borders of African and the ME
countries. The minimal distance from each grid cell
to the nearest border was recorded and used in the
SEM analysis. Because this information is static [i.e.
it does not change during the period of analysis] the
same value was used in all years.
2.3.3. Agricultural dependence
To assess agricultural dependence as share of cropland
area in a 0.5grid cell, we used the Climate Change
Initiative [CCI] of the European Space Agency [ESA]
Land Cover product. The ESA CCI product is an
annual global land cover time series from 1992 to
2015 [now available also for 2016 to 2018], available
at an unprecedent high spatial resolution of 300 m
( =node/175).
This unique dataset was produced by reprocessing
and interpretation of daily surface reflectance of five
different satellite missions. It uses the full archive of
the MEdium-spectral Resolution Imaging Spectro-
meter (MERIS) [2003–2012], with 15 spectral bands
and 300 m spatial resolution and the 1 km time series
from AVHRR [1992–1999], SPOT-VGT [1999–2013]
and PROBA-V [2014 and 2015]. The baseline was
established through MERIS data and use of machine
learning and unsupervised algorithms [39].
The advantage of this product over other products
that are derived from several observation systems is
that it maintains a good consistency over time. This
is done by confirming changes observed in earlier
and later MERIS era satellites via back- and forward-
checking through the 10 year MERIS base-line Land
Cover (LC) maps. The ESA CCI LC product was
evaluated with a global independent validation data-
set according to international standards, testing the
accuracy of both LC classes and LC change in time
[39]. It was also found accurate through a compar-
ison using country-level information provided by the
Food and Agriculture Organization of the United
Nations [FAO-STAT] in several countries [40].
We used the 1992–2012 ESA CCI LC maps to clas-
sify pixels into agricultural versus non-agricultural
classes. More specifically, LC classes #10, 20, 30, and
40, which include also mosaics of croplands and nat-
ural vegetation, were designated as agricultural pixels
while others were assigned as non-agricultural pixels.
We then aggregated the 300 m pixels into the coarser
resolution of 0.5[resolution of analysis] and calcu-
lated the total share of agricultural area in each 0.5
grid cell [as the percentage of total area]. These estim-
ates were used to examine influence of agricultural
dependence [larger crop share of area equals higher
agricultural dependency [38]] on violence risk as well
as to derive yearly change in agricultural yield produc-
tion [see next sub-section].
2.3.4. Yield production
To quantify changes in agricultural yield produc-
tion, we used NASA’s VIPPHEN EVI2 satellite
product [41]. The VIPPHEN EVI2 data product
is provided globally at 0.05[�5600 m] spatial res-
olution and contains 26 Science Datasets [SDS],
including phenological metrics such as the start,
peak, and end of season as well as the maximum,
average, and background calculated EVI2 (https:// It is
currently the longest and most consistent satellite-
based global vegetation phenology product available.
VIPPHEN SDS are based on the daily VIP product
series and are calculated using a 3 year moving win-
dow average to eliminate noise.
The modified 2-band enhanced vegetation index
[EVI2] is highly correlated with the commonly-used
EVI [42], which was found to be useful for tracking
changes related to vegetation dynamics [43] as well
as gross primary productivity [44]. EVI2 differs from
the traditional EVI by its use of two bands, the red and
near infrared, instead of the use of three bands, which
also includes the blue band in the index calculation.
The integral over the growing season of EVI2 [EVIGSI;
figure S1] was used here as a proxy of agricultural
yield production. Growing season integrals of veget-
ation indices are usually well correlated with biomass
of green tissues, particularly in annual vegetation sys-
tems [4547], and as such may serve as a good proxy
of crop yield production [48]. EVIGSI was derived per
year for agricultural pixels with > 50% of agricultural
Environ. Res. Lett. 15 (2020) 104017 D Helman et al
area cover [estimated from the ESA CCI LC 300 m
product]. Pixels with < 50% of agricultural area cover
were discarded from the analysis in order to remove
influences of non-agricultural vegetation systems on
Because agricultural fields differ in crop type and
different crop types may have similar EVIGSI values,
we used the relative anomaly of EVIGSI as a proxy of
relative anomaly in local yield production instead of
the absolute EVIGSI value. In order to assess the valid-
ity of this approach, we compared yearly anomalies of
national yield production, derived from the food and
agriculture data provided by the Food and Agricul-
ture Organization of the United Nations [FAO-STAT
[49]], with country-level EVIGSI anomalies (z-scores)
for the period of analysis [1990–2012; see supplement-
ary material and figures S2 to S5]. Yield is provided
in FAO-STAT as hectograms per hectare [hg/ha] for
cereals, citrus fruit, coarse grain, fibre crops, oil-
crops, pulses, roots and tubers, treenuts, vegetables
and fruits ( The
total annual yield and the long-term mean annual
yield [1990–2012] from FAO-STAT were calculated
to derive the relative anomaly in percentages of the
long-term average yield [%]. The same procedure
was applied for the calculation of the EVIGSI annual
anomaly [as percentages of the mean EVIGSI].
2.3.5. Satellite night-time lights as a proxy of economic
We used night-time lights intensity from the Defense
Meteorological Satellite Program [DMSP [50]] to
estimate grid-based economic welfare status and
dynamics in Africa and the ME. This night-time light
product dates back to 1992 and is considered to be
well correlated with GDP, built-up area, energy con-
sumption, poverty, and other socioeconomic welfare
variables [5154]. We used the DMSP yearly average
stable night-time lights intensity product at a spa-
tial resolution of 30 arcsec [�1 km] for 1992–2012 to
calculate the percentage area of light per pixel [Lit-
Area]. Method was followed by that described in
[55]. In short, light intensity in DSMP is given as a
digital number [DN] from 0 to 100 for each pixel.
A DN threshold value is then used to assign each
pixel with a binary 1/0 for presence/absence of light.
The threshold of DN > 31 was used following [55].
The total LitArea per 0.5grid—i.e. the sum of the
squared kilometers of light in a 0.5grid cell—was
derived by aggregating pixels with values to the spatial
resolution of the analysis. The total number of square
kilometers was then converted into square meters and
divided by the population density in the same grid cell
to derive the relative LitArea [R-LitArea].
We divided the LitArea by the population dens-
ity because places with denser populations are expec-
ted to have higher LitArea, which will not neces-
sarily indicate a higher economic welfare status
but may just reflect a larger build-up area. By
dividing the LitArea by the population density,
we thus normalize for such an effect, remaining
with a relative measure of economic welfare. We
used the WorldPop dataset []
for grid-based information on population density.
This dataset uses an ensemble learning method for
classification, combining 30 m Landsat Enhanced
Thematic Mapper (ETM) satellite imagery for high-
resolution mapping of settlements and gazetteer
population numbers to produce gridded popula-
tion density maps at high spatial resolutions [56].
Yearly population maps for Africa and the ME
are available from 2000 to date [downloaded from: =3]
at the same resolution of the DMSP dataset [1 km ×
1 km]. We used simple linear interpolation to derive
population density for 1992–1999, and aggregated the
original resolution to the coarse spatial resolution of
the analysis [0.5×0.5]. R-LitArea was derived per
0.5grid cell as the ratio between LitArea and popula-
tion density. Finally, R-LitArea z-score was calculated
to get yearly economic welfare anomaly.
2.3.6. Grid-based water resources information from
land surface models
Gridded estimates of soil moisture and hydrolo-
gical fluxes, along with river network estimates of
streamflow, were generated using the NASA Land
Information System [LIS] [57] software frameworks.
In this implementation, LIS was implemented using
the Noah-MultiParameterization [Noah-MP] [58]
Land Surface Model and the Hydrological Modeling
and Analysis Platform [HyMAP] [59] river routers.
All simulations were performed using meteorolo-
gical forcing data drawn from the NASA Modern
Era Reanalysis for Research and Applications, v2
[MERRA-2] [60], with the exception of precipitation,
which came from the Climate Hazards InfraRed Pre-
cipitation with Stations, v2 [CHIRPSv2] [34] dataset.
Simulations were performed at 0.1horizontal res-
olution with a timestep of 30 min. A 30 year spin-
up was performed to equilibrate model soil moisture
states, and the simulation was then run from 1990–
2018. In this application, Noah-MP was used with
four soil moisture layers [thicknesses of 0.1, 0.3, 0.6
and 1.0 m, descending from the surface] and a simple
unconfined aquifer. Soil moisture and surface runoff
were aggregated to the spatial resolution of the ana-
lysis and the z-score of each 0.5grid cell was calcu-
lated to derive the inter-annual anomaly.
3. Assessing direct and indirect causal
The SEM approach was used because it allows one
to evaluate direct and indirect effects of climate and
non-climate factors on violence risk, as well as to
quantify relationships among factors. In that sense,
SEM has an advantage over univariate regression
Environ. Res. Lett. 15 (2020) 104017 D Helman et al
approaches, such as general additive models (GAM)
and general linear models (GLM), because it can be
used to evaluate direct effects while controlling for
joint effects. For example, it provides a way to eval-
uate the direct effect of yield on conflict risk while
controlling for the joint effects of climate variables
on yield and conflict. The ability of SEM to quantify
direct and indirect relationships makes it particularly
suited for confirming causal relationships based on a
priori hypotheses.
Our SEM was developed based on a conceptual
model designed to test a priori hypothesis that relates
climate to food and water security, economic welfare
and—directly and indirectly—to conflict risk [69].
It was then applied on a 0.5grid basis in a time-
for-space model design for 1992–2012 (see supple-
mentary material). The SEM model was applied for
Africa, the ME, and both regions together, as well
as for each country separately. To enable comparison
between datasets with different normal distributions,
we used the relative anomaly—quantified as a stand-
ard score [z-score]—instead of the absolute values of
the climate and non-climate factors. The control vari-
ables [IMR, agricultural dependence and distance to
border], on the other hand, were maintained with
their absolute values in order to quantify the abso-
lute influence of these factors on the climate-conflict
relationships. The conflict data were converted to a
binary dataset, with 0 for non-conflict and 1 for con-
flict years/grids. The results of the SEM are presented
as standardized effects indicating the magnitude and
sign of effect.
4. Results and discussion
Non-state armed conflicts in the last three decades
were not restricted to certain climatic conditions in
Africa and the ME but rather occupied the entire cli-
matic domain (figure 1(A) to (B)). Consistent with
previous studies, violent conflicts are mostly found in
agriculture-dependent areas [38], low socioeconomic
areas [61], and close to political borders [19] in both
regions (figures 1(C) to (E)).
Non-state conflict grid cells also have higher than
average rainfall (figure 1(F)) on account of the fact
that population and agricultural activities are lim-
ited in arid regions. However, the association between
violence and agricultural dependence was about four-
fold greater in the ME (table 1), in spite of the lar-
ger average agricultural area in Africa [14% compared
to 11% for the ME] (figure 1(C)), likely because of
lower mean annual rainfall and therefore greater agri-
cultural vulnerability to drought and water scarcity
(figure 1(F)).
4.1. Contrasting climate effects in Africa and the
Middle East
When applying the SEM to both regions together
[general model] (figure 2(B)), yield and economic
welfare had the strongest effect on present-year viol-
ence risk. Increases in yield and welfare reduced the
chance of violence in both present and following year,
while warming increased the risk and rain decreased
this risk.
While these results are in accordance to previ-
ously reported by others [19,21,38], unexpected
complex climate-conflict links were revealed when
SEMs were applied to each region separately (fig-
ures 2(C) and (D)). Warming increased the risk of
violence in Africa (figure 2(C))—similar to the gen-
eral model—but unexpectedly decreased this risk in
the ME (figure 2(D)). There was no effect of rain and
yield on conflict risk in Africa and no effect of wel-
fare in the ME. But there was a weak, though signi-
ficant (P< 0.05), indirect negative effect of rain on
the risk of conflicts in Africa (table 1), which was,
surprisingly, through the effect of water availability
on welfare and not through yield (figure 2(C)). This
may be in part because satellite-based estimates of
yield have limited skill in some conflict-prone African
regions (figures S3 and S4), but could also be due to
a more complex link between rainfall, yield, and viol-
ence than that drawn by our model. In all models, the
risk of violence was greater in places where conflict
already occurred in the previous year (figures 2(B)
to (D)), which likely indicates the roles of political
instability and historic background on such conflicts.
4.2. A non-linear response of conflict to warming
To examine the generality of the contrasting effects,
we further applied the SEMs per country. After
analyzing the countries with enough conflicts to pro-
duce a statistically significant model (table S4), we
found that not only ME countries, but also some
African countries—particularly West African—had a
negative, direct temperature effect on violence risk
(figure 3(A)). This negative direct temperature effect
was in spite of the fact that some of these coun-
tries are warmer than those showing a positive effect
[e.g. East African countries], which would theoretic-
ally make them more vulnerable to heat-induced viol-
ence [23]. In general, the country data show a non-
linear relationship between the warming effect—i.e.
the standardized direct effect of temperature anomaly
on conflict risk—and the mean temperature condi-
tions across countries, with a peak response at around
32C (figure 3(C)). Countries with lower and higher
mean annual temperatures (MAT) tend to exhibit
lower effects of temperature anomalies on the risk of
violence, with even negative effects in some cases.
An example of the latter is Sierra Leone and
Liberia, which had the strongest effects, with a stand-
ardized negative effect of 0.43 and 0.39, respect-
ively (figure 3(A)). These two countries are charac-
terized by extremely warm and humid conditions,
with high temperatures and large amounts of rain-
fall year-round (�3000 mm y1). Contrary to the
Environ. Res. Lett. 15 (2020) 104017 D Helman et al
Figure 1. Armed conflicts in Africa and the Middle East, and associated factors. (A) Log number of armed conflicts by 0.5
grids for 1990–2017. (B) Mean annual rainfall and temperature binned by log number of conflicts. Gray line in (B) marks the
region’s climatic domain (95th quantile of all grids). Boxplots show median, 1st and 3rd quantiles of (C) relative agricultural area,
(D) infant mortality rate (IMR), (E) distance to border, and (F) mean annual rainfall for violent (red), non-violent (blue) and all
(violent +non-violent) grids. Violent grids are significantly different from non-violent grids in (C) to (F) at P< 0.001.
Table 1. Direct, indirect and total standardized effects of rain, temperature and yield anomalies on risk of violence, with and without
explicit controls (marked in italic). High infant mortality rate (IMR) means low socioeconomic status. Positive (negative) relationships
are shown in regular (bold) font.
Without controls With controls
Predictor Direct Indirect Total Direct Indirect Total
General model Rain n.s. 0.003
Temperature 0.011
Yield 0.009
Agricultural area – – 0.110
– –
IMR – – 0.017
– –
Distance to border – – 0.033
– –
Africa Rain n.s. n.s. n.s. n.s. 0.001
Temperature 0.020
∗∗ ∗
n.s. 0.020
∗∗ ∗
n.s. 0.019
Yield n.s. 0.002
∗∗ ∗
n.s. n.s. 0.002
Agricultural area – – 0.073
– –
IMR – – 0.023
– –
Distance to border – – 0.030
– –
ME Rain 0.015
∗∗ ∗
∗∗ ∗
Temperature 0.026
∗∗ ∗
∗∗ ∗
∗∗ ∗
Yield 0.048
∗∗ ∗
n.s. 0.047
∗∗ ∗
n.s. 0.058
Agricultural area – – 0.264
– –
IMR – – 0.083
– –
Distance to border – – n.s. – –
n.s. not significant; P< 0.05; ∗∗ P< 0.01; ∗∗∗ P< 0.001
GAM, which suggests that uncomfortable environ-
mental conditions increase violent perceptions [3],
in this case uncomfortable extreme weather con-
ditions [extra warming in already warm, humid
countries] seemed to decrease the risk of viol-
Some experimental studies suggested that phys-
ical aggression may have a rather complex, curvilinear
Environ. Res. Lett. 15 (2020) 104017 D Helman et al
Figure 2. Structural equation models showing causal effects on conflict risk. (A) The conceptual model. Models were applied to
(B) Africa and Middle East together (general model), and to (C) Africa and (D) Middle East separately. Factors not affecting
present-year violence are colored gray. Numbers alongside arrows indicate the standardized direct effects, with the color of the
arrow indicating its sign (black for positive; red for negative) and width indicating its importance in the model. Constructs in our
SEM are indicated by ovals while indicators are shown as rectangles. Only significant effects at P< 0.05 are shown.
response to heat [6264]. Aggression was shown
to increase with temperature rise, but decrease
at excessive heat in several experimental settings,
particularly when other negative-affect-producing
factors are present [64]. The explanation given
for this is that it is the urgent ‘need’ to escape
or minimize discomfort that overcomes tenden-
cies to aggressive behavior [65]. Taking this to
Sierra Leone and Liberia, a further increase in tem-
perature resulting in extremely unpleasant condi-
tions might have increased discomfort and reduced
the level of engaging in violence through such
‘escape’ mechanism. In this context, the RAT—
suggesting that people interact more under pleas-
ant conditions, which lead to more opportunities
for violence—may be another possible explanatory
mechanism [18].
The positive and negative temperature effects in
Yemen and Turkey (figure 3(B)) suggest that GAM
might be the primary mechanism in the ME rather
than the RAT. The contrasting sign effect may be
explained by a relaxation mechanism in which a
decrease in unpleasant conditions—being warming
in a cold area [Turkey] or cooling in a warm area
[Yemen]—reduces the chance of violence [66]. This
does not necessarily contradict the abovementioned
‘escape’ theory because warm and humid conditions
in both Turkey and Yemen are more tolerable than in
Sierra Leone and Liberia (figures 3(A), (B)).
4.3. Climate effects in the context of geography and
To further show how complex this climate–conflict
link may be, we focus on two cases—Algeria and
Mali. Algeria is the largest country in Africa, with
an economy relying heavily on energy exports.
Though Algeria’s government has promoted agricul-
tural development, yield is highly unstable due to cli-
mate variability [67]. This instability is likely to pro-
mote violence, particularly in agriculture-dependent
areas as shown from our results (figure 3(D)). The
contrasting indirect [negative] and direct [positive]
effects of temperature in Algeria are likely due to
a positive temperature effect on yield and a direct
adverse influence of heat, which may be explained by
Environ. Res. Lett. 15 (2020) 104017 D Helman et al
Figure 3. Contrasting effects of temperatures on conflict risk. Direct (close symbols) and indirect (open symbols) standardized
effects of temperature on the risk of conflict from SEMs in (A) African and (B) Middle Eastern countries. Mean annual
temperature and rainfall over 1992–2012 are shown. Effects with bars not crossing the vertical line in (A) and (B) are considered
significant at P< 0.05. (C) The standardized direct effect of temperature vs. local temperature conditions expressed as the mean
annual temperature [MAT]. Each symbol in (C) is a single country (Sierra Leone and Liberia are excluded for clarity), with the
size indicating the average temperature change [�T] for the period of analysis and the line in depicting the nonparametric
regression with corresponding confidence interval. (D) The standardized direct (black) and indirect (white) effects of
temperature, rain, and yield on conflict risk in Algeria and Mali. Inserts in (D) show the effects of agricultural dependence and
infant mortality rate (IMR) in Algeria and Mali. Asterisks denote significant effects at P< 0.05.
the GAM. In contrast, the influence of yield on viol-
ence risk was positive and significantly smaller in Mali
[50% smaller than in Algeria]. This is in spite of the
fact that Mali’s economy is more centered on agricul-
ture than Algeria [68]. Moreover, the positive yield
effect was limited to the northern part of Mali, which
is less agricultural than its southern [and central] part
(insert in figure 3(D)).
Putting this in context, we know that most con-
flicts in Mali during the period of analysis were
intra-state conflicts between the government and the
Tuareg nomadic inhabitants living in the northern
part of the country. Because our analysis is limited
to small-scale conflicts, the non-state, inter-group
aspects of the Tuareg conflict, which occur primar-
ily in the northern, less agricultural part of the coun-
try, is well noted (figure 1(A)) [68]. The Tuaregs are
primarily pastoral and as such continuously compete
for scarce resources between pastoral groups and with
the few crop farmers and settled villagers in the north
[68]. Tuareg conflict is believed to be an example of
a resource conflict driven by climatic changes [69]
and the positive effect of yield on violence risk in our
SEM is likely a reflection of this struggle, with periods
of increased yield in the northern region being a
potential driver of ethnic tension and inter-group
These contrasting complex links in Algeria and
Mali imply that the climate–violence linkage should
be investigated in the context of historical, geograph-
ical and ethnical backgrounds of each location rather
than as a general cross-sectional analysis. Such an
approach can shed light on contrasting effects of
5. Concluding remarks
Our findings reveal previously unreported effects of
climate on risk of conflict outbreak. More specifically,
contrasting effects of temperature were detected at a
regional scale and in numerous countries in Africa
and the ME. Importantly, temperature and rainfall
direct effects on conflict risk seem to be stronger than
any indirect effect through resources such as water
availability and agricultural production (figure 3and
table 1). This could mean one of three things: that
climate affects violence mostly through psychological
and/or interactive mechanisms [e.g. GAM and RAT
Environ. Res. Lett. 15 (2020) 104017 D Helman et al
[17,18]]; that indirect effects depend on aspects of
climate variability that we have not considered [70];
or that underlying mechanisms in which resource
scarcity or conflict-relevant abundance patterns affect
violence are more complex than those modeled by our
As in previous studies [21], the use of explicit
controls affected the strength of the climate effect in
our SEMs [i.e. the difference between total and direct
effects in table 1], in our case by up to 46%. This was
important enough to expose indirect rain effects in
Africa (table 1). It is important to note that the SEMs,
although statistically significant (table S4), confirm-
ing the validity of the a priori hypothesis, had very
little predictive power [with 1st and 3rd quantiles
being 1.6% and 11%, respectively, across countries]
(table S5). This means that although our SEMs did
confirm impacts of climate on armed conflicts by
effectively quantifying its direct and indirect effects,
these effects were relatively small compared to unob-
served factors like political, ethnic and likely other
unaccounted socioeconomic factors. These were only
partly considered in our analysis, due to the difficulty
to account for such factors at a grid-cell level, in the
form of next-year violence risk, which was shown to
be greatly affected by present year violence (explain-
ing between 10% and 16% of the variance, at the con-
tinental level; figures 2(B) to (D)).
Our results demonstrate that no single pro-
posed climate–conflict mechanism can alone explain
the empirical patterns that underlie the climate–
conflict linkage across contrasting regions or coun-
tries, and that this linkage is more complex than
some analyses have previously suggested [21]. We
conclude that extreme caution should be exercised
when attempting to explain or project local climate–
violence relationships on the basis of a single, gen-
eralized theory. Large-scale cross-sectional studies
can be useful for identifying general associations
and trends, but an appropriately scaled and struc-
tured analysis is required to explain and, potentially,
address climate–violence risk factors in geographic
Authors thank C Helman for helping to organize
the data for the analyses, A Mussery for helping
to organize the SEMs results in the SM, and two
anonymous reviewers for insightful comments. D
Helman was a USA-Israel Fulbright Post-Doctoral
Fellow at Johns Hopkins University 2018/19. C Funk
is supported by the U.S. Geological Survey Drivers of
Drought program and the U.S. Agency for Interna-
tional Development’s Famine Early Warning Systems
Data availability statement
The data that support the findings of this study are
available upon reasonable request from the authors.
David Helman
[1] Scheffran J, Brzoska M, Kominek J, Link P M and Schilling J
2012 Climate change and violent conflict Science
336 869–71
[2] Hsiang S M, Meng K C and Cane M A 2011 Civil conflicts
are associated with the global climate Nature 476 438–41
[3] Bernauer T, Böhmelt T and Koubi V 2012 Environmental
changes and violent conflict Environ. Res. Lett. 715601
[4] Koubi V 2019 Climate change and conflict Annu. Rev. Polit.
Sci. 22 343–60
[5] Koubi V 2017 Climate change the economy, and conflict
Curr. Clim. Chang. Rep. 3200–9
[6] Hodler R and Raschky P A 2014 Economic shocks and civil
conflict at the regional level Econ. Lett. 124 530–3
[7] Dube O and Vargas J F 2013 Commodity price shocks and
civil conflict: evidence from Colombia Rev. Econ. Stud.
80 1384–421
[8] Harris G and Vermaak C 2015 Economic inequality as a
source of interpersonal violence: evidence from Sub-Saharan
Africa and South Africa South African J. Econ Manage. Sci.
18 45–57
[9] Zhang D D et al 2011 The causality analysis of climate
change and large-scale human crisis Proc. Natl. Acad. Sci.
108 17296 LP–17301
[10] Carleton T A and Hsiang S M 2016 Social and economic
impacts of climate Science 353 6304, 9837
[11] Fjelde H and von Uexkull N 2012 Climate triggers: rainfall
anomalies vulnerability and communal conflict in
sub-Saharan Africa Polit. Geogr. 31 444–53
[12] Wischnath G and Buhaug H 2014 Rice or riots: on food
production and conflict severity across India Polit. Geogr.
43 6–15
[13] Salehyan I and Hendrix C S 2014 Climate shocks and
political violence Glob Environ Change 28 239–50
[14] Witsenburg K M and Adano W R 2009 Of rain and raids:
violent livestock raiding in Northern Kenya Civil Wars
11 514–38
[15] Raleigh C and Kniveton D 2012 Come rain or shine: an
analysis of conflict and climate variability in East Africa J.
Peace Res. 49 51–64
[16] Nordkvelle J, Rustad S A and Salmivalli M 2017 Identifying
the effect of climate variability on communal conflict
through randomization Clim. Change 141 627–39
[17] Dewall C N, Anderson C A and Bushman B J 2011 The
general aggression model: theoretical extensions to violence
Psychol. Violence 1245–58
[18] Cohen L E and Felson M 1979 Social change and crime rate
trends : A Routine Activity Approach Am. Sociol. Rev.
44 588–608
[19] O’Loughlin J, Linke A M and Witmer F D W 2014 Effects of
temperature and precipitation variability on the risk of
violence in sub-Saharan Africa, 1980-2012 Proc. Natl. Acad.
Sci. USA 111 16712–7
[20] O’Loughlin J et al 2012 Climate variability and conflict risk
in East Africa, 1990–2009 Proc. Natl. Acad. Sci. 109 18344–9
[21] Hsiang S M, Burke M and Miguel E 2013 Quantifying the
influence of climate on human conflict Science
341 1235367
Environ. Res. Lett. 15 (2020) 104017 D Helman et al
[22] Hsiang S M and Burke M 2014 Climate, conflict, and social
stability: what does the evidence say? Clim. Change
123 39–55
[23] Helman D and Zaitchik B F 2020 Glob. Environ. Change 63
[24] Ide T 2017 Research methods for exploring the links between
climate change and conflict Wiley Interdiscip. Rev. Clim.
Change 8
[25] Buhaug H 2010 Climate not to blame for African civil wars
Proc. Natl. Acad. Sci. USA 107 16477–82
[26] Adams C, Ide T, Barnett J and Detges A 2018 Sampling
bias in climate-conflict research Nat. Clim. Change
[27] Buhaug H et al 2014 One effect to rule them all? A comment
on climate and conflict Clim. Change 127 391–7
[28] Mach K J et al 2019 Climate as a risk factor for armed
conflict Nature 571 193–7
[29] Sundberg R and Melander E 2013 Introducing the UCDP
georeferenced event dataset J. Peace Res. 50 523–32
[30] Chou C-P and Bentler P M 1995 Estimates and tests in
structural equation modeling Structural Equation Modeling:
Concepts, Issues, and Applications (Newbury Park, CA: Sage
Publications, Inc) pp 37–55
[31] Sundberg R and Croicu M 2017 UCDP Non-State Conflict
Codebook Version 18.1
[32] Carpenter T G 2013 Tangled web: the Syrian civil war and its
implications Mediterr. Q. 24 1–11
[33] Funk C et al 2019 A high-resolution 1983–2016 Tmax
climate data record based on infrared temperatures
and stations by the climate hazard center J. Clim.
32 5639–58
[34] Funk C et al 2015 The climate hazards infrared precipitation
with stations—a new environmental record for monitoring
extremes Sci. Data 2150066
[35] Dinku T et al 2018 Validation of the CHIRPS satellite rainfall
estimates over eastern Africa Q. J. R. Meteorol. Soc.
144 292–312
[36] Maidment R I, Allan R P and Black E 2015 Recent observed
and simulated changes in precipitation over Africa Geophys.
Res. Lett. 42 8155–64
[37] Center for International Earth Science Information
Network.—CIESIN—Columbia University 1999 Poverty
Mapping Project: global Subnational Infant Mortality Rates
(Palisades, NY: NASA Socioeconomic Data and Applications
Center (SEDAC)) (Accessed: 17 February 2019)
[38] Von Uexkull N, Croicu M, Fjelde H and Buhaug H 2016
Civil conflict sensitivity to growing-season drought Proc.
Natl. Acad. Sci. USA 113 12391–6
[39] ESA 2019 Land Cover CCI Product User Guide Version 2.0
(Accessed: 17 June 2019) https://www.esa-landcover-\protect$\relax=$webfm_send/84
[40] Liu X et al 2018 Comparison of country-level cropland areas
between ESA-CCI land cover maps and FAOSTAT data Int. J.
Remote Sens. 39 6631–45
[41] Didan K et al 2015 Multi-Sensor Vegetation Index and
Phenology Earth Science Data Records Algorithm Theoretical
Basis Document and User Guide Version 4.0
[42] Jiang Z, Huete A R, Didan K and Miura T 2008 Development
of a two-band enhanced vegetation index without a blue
band Remote Sens. Environ. 112 3833–45
[43] Huete A et al 2002 Overview of the radiometric and
biophysical performance of the MODIS vegetation indices
Remote Sens. Environ. 83 195–213
[44] Huang X, Xiao J and Ma M 2019 Evaluating the performance
of satellite-derived vegetation indices for estimating gross
primary productivity using FLUXNET observations across
the globe Remote Sens. 11
[45] Helman D, Mussery A, Lensky I M and Leu S 2014 Detecting
changes in biomass productivity in a different land
management regimes in drylands using satellite-derived
vegetation index Soil Use Manag. 30 32–39
[46] Helman D, Lensky I M, Mussery A and Leu S 2014
Rehabilitating degraded drylands by creating woodland
islets: assessing long-term effects on aboveground
productivity and soil fertility Agric. For. Meteorol. 195–6
[47] Helman D 2018 Land surface phenology: what do we really
‘see’ from space? Sci. Total Environ. 618 665–73
[48] Gitelson A A et al 2006 Relationship between gross primary
production and chlorophyll content in crops: implications
for the synoptic monitoring of vegetation productivity J.
Geophys. Res. Atmos. 111
[49] FAO 2016 Production database. crops dataset Latest update:
November 2016 (Accessed 18 June 2019)
[50] NOAA 2020 Defense Meteorological Satellite Program
(DMSP)—Data Archive, Research, and Products [Internet]
Earth Observation Group, Boulder
[51] Levin N, Ali S and Crandall D 2018 Utilizing remote sensing
and big data to quantify conflict intensity: the Arab Spring as
a case study Appl. Geogr. 94 1–17
[52] Ivan K, Holobˆ
a I-H, Benedek J and Török I 2020 Potential
of night-time lights to measure regional inequality Remote
Sens. 12
[53] Bagan H, Borjigin H and Yamagata Y 2018 Assessing
nighttime lights for mapping the urban areas of 50 cities
across the globe 46 1097–114
[54] Li S, Zhang T, Yang Z, Li X and Xu H 2017 Night time light
satellite data for evaluating the socioeconomics in Central
Asia ISPRS–Int. Arch. Photogramm. Remote Sens. Spat. Inf.
Sci. 42W7 1237–43
[55] Proville J, Zavala-Araiza D and Wagner G 2017 Night-time
lights: A global, long term look at links to socio-economic
trends PLoS One 12 1–12
[56] Stevens F R, Gaughan A E, Linard C and Tatem A J 2015
Disaggregating census data for population mapping using
Random forests with remotely-sensed and ancillary data
PLoS One 10 1–22
[57] Kumar S V et al 2006 Land information system: an
interoperable framework for high resolution land surface
modeling Environ. Model. Softw. 21 1402–15
[58] Niu G-Y et al 2011 The community Noah land surface model
with multiparameterization options (Noah-MP): 1. Model
description and evaluation with local-scale measurements J.
Geophys. Res. Atmos. 116
[59] Getirana A C V et al 2012 The Hydrological Modeling and
Analysis Platform (HyMAP): evaluation in the Amazon
Basin J. Hydrometeorol. 13 1641–65
[60] Gelaro R et al 2017 The Modern-Era Retrospective Analysis
for Research and Applications, Version 2 (MERRA-2) J.
Clim. 30 5419–54
[61] Hegre H and Sambanis N 2006 Sensitivity analysis of
empirical results on civil war onset J. Conflict Resolut.
50 508–35
[62] Baron R A and Ransberger V M 1978 Ambient temperature
and the occurrence of collective violence: the ‘long, hot
summer’ revisited J. Pers. Soc. Psychol.
36 351–60
[63] Bell P A and Baron R A 1976 Aggression and Heat: the
Mediating Role of Negative Affect1 J. Appl. Soc. Psychol.
[64] Baron R A 1972 Aggression as a function of ambient
temperature and prior anger arousal J. Pers. Soc. Psychol.
21 183–9
[65] Baron R A and Bell P A 1976 Aggression and heat: the
influence of ambient temperature, negative affect, and a
cooling drink on physical aggression J. Pers. Soc. Psychol.
33 245–55
[66] Anderson C A 1989 Temperature and Aggression: ubiquitous
Effects of Heat on Occurrence of Human Violence Psychol.
Bull. 106 74–96
[67] Amine B M and Fatima B 2016 Determinants of on-farm
diversification among rural households: empirical evidence
from Northern Algeria Int. J. Food Agric. Econ. (IJFAEC)
04 87–99
Environ. Res. Lett. 15 (2020) 104017 D Helman et al
[68] Keita K 1998 Conflict and conflict resolution in the Sahel:
the Tuareg insurgency in Mali Small Wars Insur.
[69] B¨
achler G 1998 Violence through Environmental
Discrimination: Causes, Rwanda Arena, and
Conflict Model vol 2 (Alphen aan den Rijn: Kluwer
[70] Buhaug H 2015 Climate–conflict research: some reflections
on the way forward Wiley Interdiscip. Rev. Clim. Change
... Together with other drivers, climate change and variability threaten human life in many ways including increasing the occurrence of natural disasters, undermining livelihoods security and peace. Concerning human security and peace, an increasing stream of research over the past decades has addressed the climate-conflict nexus (Burke et al., 2009;Fjelde, 2015;Froese and Schilling, 2019;Helman et al., 2020). An ongoing debate within this stream of research revolves around the arguments of causality and the mechanism or the contextual pathways through which climate may affect human security, peace and stability (Busby, 2018;Martin-Shields and Stojetz, 2019). ...
... On the one hand, studies that hold the view of a direct relationship between climate variability and conflict are framed from the General Aggression Model which state that higher temperatures trigger human aggression (DeWall et al., 2011), and Routine Activity Theory which holds that higher temperatures force people to spend more time outdoors increasing chances of that may undermine peace (Groff, 2008). On the other hand, those that take the indirect effect stance argue that climate variability affects conflict trough some intervening factors such as food insecurity (Koren and Bagozzi, 2017;Anderson et al., 2021), crop production (Wischnath and Buhaug, 2014;Caruso et al., 2016;Jun, 2017), and poverty and inequality (Harris and Vermaak, 2015;Helman et al., 2020) and country's economic growth (Bergholt and Lujala, 2012). ...
... In this study we investigate the empirical associations between climate variability (as measured by the temperature and precipitation anomalies) maize production, household food insecurity and conflict. Given the complexity of the associations, we employ the structural equation modeling (SEM) approach which has previously used to unravel complex relationships such as the association between climate and conflict through different pathways (Helman et al., 2020;Yue and Lee, 2020). The SEM continues to gain popularity for modeling and estimating pathspecific associations within a complex set of relationships. ...
Full-text available
Climate continues to pose significant challenges to human existence. Notably, in the past decade, the focus on the role of climate on conflict and social unrest has gained traction in academic, development, and policy communities. This article examines the link between climate variability and conflict in Mali. It advances the argument that climate is a threat multiplier, in other words, climate indirectly affects conflict occurrence through numerous pathways. We take the view that maize production and household food security status sequentially mediate the relationship between climate variability and the different conflict types. First, we provide a brief review of the climate conflict pathways in Mali. Second, we employ the path analysis within the structural equation modeling technique to test the hypothesized pathways and answer the research questions. We use the Living Standards Measurement Study-Integrated Surveys on Agriculture (LSMS-ISA), a nationally representative data from Mali merged with time and location-specific climate and the Armed Conflict Location and Event Data (ACLED) data. Results show that an increase in positive temperature anomalies when sequentially mediated by maize production and household food security status, increase the occurrence of the different conflict types. The results are robust to the use of negative precipitation anomalies (tendency toward less precipitation compared to the historical norm). Our findings highlight two key messages, first, the crucial role of climate change adaptation and mitigation strategies and interventions on influencing household food security status and thus reducing conflict occurrence. Second, that efforts to build peace and security should account for the role of climate in exacerbating the root causes of conflict.
... As sudden changes in temperature and precipitation are expected to become more frequent in some areas due to climate change, some researchers recognized climate change as a threat multiplier and linked it to armed conflict [12][13][14]. For example, using a panel regression that links climate change and conflict events, Burke et al. [15] found a strong historical positive correlation between local temperatures and armed conflict in sub-Saharan Africa. ...
... Previous literature has also shown that armed conflicts tend to occur in areas dependent on agriculture [30,34]. Vegetation index and land cover data are often used to measure agricultural yield production [14]. In this study, to capture the extent of agricultural dependence, we obtain the normalized difference vegetation index (NDVI) from the Terra Moderate Resolution Imaging Spectroradiometer (MODIS) Vegetation Indices Monthly (MOD13C2) Version 6.1 product [35] and land cover from the European Space Agency (ESA) GlobCover [36]. ...
Full-text available
Sub-Saharan Africa has suffered frequent outbreaks of armed conflict since the end of the Cold War. Although several efforts have been made to understand the underlying causes of armed conflict and establish an early warning mechanism, there is still a lack of a comprehensive assessment approach to model the incidence risk of armed conflict well. Based on a large database of armed conflict events and related spatial datasets covering the period 2000–2019, this study uses a boosted regression tree (BRT) approach to model the spatiotemporal distribution of armed conflict risk in sub-Saharan Africa. Evaluation of accuracy indicates that the simulated models obtain high performance with an area under the receiver operator characteristic curve (ROC-AUC) mean value of 0.937 and an area under the precision recall curves (PR-AUC) mean value of 0.891. The result of the relative contribution indicates that the background context factors (i.e., social welfare and the political system) are the main driving factors of armed conflict risk, with a mean relative contribution of 92.599%. By comparison, the climate change-related variables have relatively little effect on armed conflict risk, accounting for only 7.401% of the total. These results provide novel insight into modelling the incidence risk of armed conflict, which may help implement interventions to prevent and minimize the harm of armed conflict.
... However, given the complexity of the effect of COVID-19 on conflict, descriptive and limited statistical analysis using only conflict and COVID-19 data might be inadequate to yield persuasive results as they can omit critical information regarding, for example, the role of climate [22,23]. Moreover, since the effect of COVID-19 on conflict risk might vary between regions and types of conflict, it is doubtful that the research focusing on a single type of conflict or specific countries could comprehensively capture the combined effects of COVID-19 on multiple types of conflict worldwide. ...
... Third, we identified the environmental background variables (control variables) that may determine regional differences that lead to the inconsistency of conflict incidence across regions. Some studies suggested that water stress and food shortage may increase the chance of conflict [22].To measure water stress, we obtained monthly soil moisture data from the National Aeronautics and Space Administration (NASA) Goddard Earth Sciences Data Information Services Center (GES DISC). To quantify changes in crop yield, we used the monthly Normalized Difference Vegetation Index (NDVI) from the Aqua Moderate Resolution Imaging Spectroradiometer (MODIS). ...
Objectives: Understand whether and how the COVID-19 pandemic affects the risk of different types of conflict worldwide in the context of climate change. Methodology: Based on the database of armed conflict, COVID-19, detailed climate, and non-climate data covering the period 2020–2021, we applied Structural Equation Modeling specifically to reorganize the links between climate, COVID-19, and conflict risk. Moreover, we used the Boosted Regression Tree method to simulate conflict risk under the influence of multiple factors. Findings: The transmission risk of COVID-19 seems to decrease as the temperature rises. Additionally, COVID-19 has a substantial worldwide impact on conflict risk, albeit regional and conflict risk variations exist. Moreover, when testing a one-month lagged effect, we find consistency across regions, indicating a positive influence of COVID-19 on demonstrations (protests and riots) and a negative relationship with non-state and violent conflict risk. Conclusion: COVID-19 has a complex effect on conflict risk worldwide under climate change. Implications: Laying the theoretical foundation of how COVID-19 affects conflict risk and providing some inspiration for the implementation of relevant policies.
... The vegetation condition used in the study was obtained from a daily vegetation index and phenological dataset (VIP), which was created using Advanced Very High-Resolution Radiometer (AVHRR) N07, N09, N11, and N14 datasets from 1981 to 1999 and Moderate Resolution Imaging Spectroradiometer (MODIS)/Terra MOD09 surface reflectance data from 2000 to 2014. It provides a long time series (34 years) of vegetation index and landscape phenological information and has been widely used in previous studies [23][24][25][26][27][28][29]. A two-step filtering approach for the input AVHRR and MODIS daily data was adopted to retain high-quality cloud-free data. ...
Full-text available
The rapid intensification of drought, commonly known as flash drought, has recently drawn widespread attention from researchers. However, how the characteristics and drivers, as well as the ecological impacts of rapidly intensified droughts, differ from those of slowly intensified ones still remains unclear over the globe. To this end, we defined three types of droughts based on the root zone soil moisture (RZSM) decline rates, flash droughts, general droughts, and creep droughts, and then implemented a comparative analysis between them across the globe and the 26 Intergovernmental Panel on Climate Change Special Report on Extremes (IPCC-SREX) regions. The ensemble of RZSM from multiple reanalysis datasets was used to reduce the uncertainties. According to the frequency analysis, our findings suggest that flash droughts contributed to the majority of drought events during 1980–2019, indicating the prevalence of rapid transition from an energy-limited to a water-limited condition in most of the regions. The comparative results of vegetation responses show that flash droughts are more likely to happen in the growing season, leading to faster but relatively minor vegetation deterioration compared to the slowly intensified ones. By analyzing the precipitation and temperature anomalies in the month of drought onset, we found the role of temperature (precipitation) on drought intensification can be generalized as the warmer (drier) the climate is or the faster the drought intensifies, but the main driving forces vary by region. Unlike temperature dominating in midwestern Eurasia and northern high latitudes, precipitation plays a prominent role in the monsoon regions. However, the temperature is expected to be the decisive driver in the warming future, given its monotonically increased contribution over the past four decades.
... Other major producers, by contrast, face both pre-existing political instability and a very high vulnerability to climate change, in particular Angola, Iraq, Libya and Nigeria (Phillis et al. 2018;Fund for Peace 2022). This is in line with recent studies that have found extreme climatic events increase the risk of violent conflict among oil-producing states in Western Africa and the Middle East (Helman, Zaitchik, and Funk 2020;Ide et al. 2021). However, there is still exists some scepticism about such a link (Daoust and Selby 2022). ...
Full-text available
Climate change can undermine human, national and planetary security in various ways. While scholars harve explored the human security implications of climate change and climate security discourses in Australia, systematic scientific assessments of climate change and national security are scarce. I address this knowledge gap by analysing whether climate change impacts the national security of Australia before 2050, focussing particularly on climate-related threats within Australia and on countries of high strategic importance for Australia. The results indicate that climate change will very likely undermine Australia’s national security by disrupting critical infrastructure, by challenging the capacity of the defence force, by increasing the risk of domestic political instability in Australia’s immediate region, by reducing the capabilities of partner countries in the Asia-Pacific region, and by interrupting important supply chains. These impacts will matter most if several large-scale disasters co-occur or if Australia becomes involved in a major international conflict. By contrast, international wars, large-scale migration, and adverse impacts on key international partners are only minor climate-related risks.
Full-text available
Previous research has demonstrated that climate change can escalate the risks for violent conflict through various pathways. Existing evidence suggests that contextual factors, such as migration and livelihood options, governance arrangements, and existing conflict dynamics, can influence the pathways through which climate change leads to conflict. This important insight leads to an inquiry to identify sets of conditions and processes that make climate-related violent conflict more likely. In this analytic essay, we conduct a systematic review of scholarly literature published during the period 1989–2022 and explore the climate-conflict pathways in the Middle East and North Africa (MENA) region. Through the systematic review of forty-one peer-reviewed publications in English, we identify that society’s ability to cope with the changing climate and extreme weather events is influenced by a range of factors, including preceding government policies that led to the mismanagement of land and water and existing conflict dynamics in the MENA region. Empirical research to unpack the complex and diverse relationship between the climate shocks and violent conflict in the MENA region needs advancing. Several avenues for future research are highlighted such as more studies on North Africa and the Gulf region, with focus on the implications of floods and heatwaves, and exploring climate implications on non-agriculture sectors including the critical oil sector.
Full-text available
International, regional, and national organizations and policymakers are increasingly acknowledging the implications of climate on peace and security, but robust research approaches that embrace the complexity of this nexus are lacking. In this paper, we present the Integrated Climate Security Framework (ICSF), a mixed-methods framework to understand the mechanisms of climate–conflict linkages at different scales. The framework uses conventional and non-conventional methods and data to provide state-of-the-art policy-relevant evidence that addresses four main questions: how, where and for whom climate and conflict risks occur, and what can be done to mitigate this vicious circle. The framework provides a comprehensive assessment of the complex social-ecological dynamics, adopting systems approaches that rely on a combination of epistemological stances, thereby leveraging diverse qualitative, quantitative, locally relevant, and multifaceted data sources; and on a diversity of actors involved in the co-production of knowledge. Using a case study from Kenya, we show that the climate security nexus is highly complex and that there exists strong, theoretical, and statistical evidence that access to natural resources, livelihoods and food security are important pathways whereby climate can increase the risk of conflict, and that conflict undermines resilience objectives. We also find that communities in climate security hotspots are aware and highly knowledgeable about the risk that the climate crisis poses on existing drivers of conflict and yet, online issue mapping and policy coherence analysis indicate that policymakers have not been acknowledging the nexus appropriately. The policy-relevant evidence that is collected through the ICSF and collated in the CGIAR Climate Security Observatory aims to fill this gap and to help transform climate adaptation into an “instrument for peace”.
This article examines the relationships between conflict, climate change, and disaster in forced displacement contexts. We present these nexus dynamics through the case of Rohingya refugees in Bangladesh who are exposed to climatic hazards and other vulnerabilities that threaten their lives and livelihoods. Having fled persecution by the Myanmar military, Rohingya refugees face a range of conflict- and climate-related risks, both in the overcrowded and disaster-prone camps in Cox's Bazar and on the island of Bhasan Char where Bangladeshi authorities have relocated tens of thousands of people. The protracted refugee crisis has exacerbated social tensions between the Rohingya and host communities; limited access to resources and exposure to significant hazards that exacerbate conflict-induced displacement challenges. This paper contributes to the nascent literature on the region's conflict, climate change, and disaster displacement nexus by examining how cascading risks and state fragility contribute to increased instability. The article demonstrates the need for a more nuanced understanding of how conflict-induced displacement leads to new threats and vulnerabilities in hazard-prone environments.
Full-text available
Scientists have shown a relationship between climate conditions and social unrest, specifically related to temperature and precipitation trends. Though deviations from historically normative patterns are becoming more pronounced, it remains unclear to what extent they affect protests, linearly and otherwise. Similarly, we know little about the extent to which climate anomalies create contemporaneous versus lagged effects on protests. To address these questions, we examine determinants of protest levels between 1995 and 2013 in India, Pakistan, and Bangladesh. Our innovative methodology involves modeling geocoded, media-reported protests (derived from the Integrated Crisis Early Warning System) at 0.5x0.5 degree geospatial resolution, using unobtrusive, satellite-derived data on temperature and precipitation patterns and historical deviations. We also integrate comparable satellite data on nightlight pollution. Properties of our outcome variable require negative binomial regression analysis. We find that, net of controls gauging local demographic and country-level dynamics, temperatures and precipitation directly impact protests in unexpected ways: In any given year and location, protests are strongly, positively associated with heavy precipitation and with higher temperature levels. Protests are also strongly, positively associated with historical deviations in precipitation patterns. This applies to anomalies related to both wetness and dryness. Finally, historically aberrant precipitation has pronounced curvilinear, lagged effectson protest levels. We situate these complex findings in light of classic, structurally based arguments grounded in social science theory.
Full-text available
The aim of the paper is to develop a model for the real-time estimation of local level income data by combining machine learning, Earth Observation, and Geographic Information System. More exactly, we estimated the income per capita by help of a machine learning model for 46 cities with more than 50,000 inhabitants, based on the National Polar-orbiting Partnership–Visible Infrared Imaging Radiometer Suite (NPP-VIIRS) nighttime satellite images from 2012–2018. For the automation of calculation, a new ModelBuilder type tool was developed within the ArcGIS software called EO-Incity (Earth Observation–Income city). The sum of light (SOL) data extracted by means of the EO-Incity tool and the observed income data were integrated in an algorithm within the MATLAB software in order to calculate a transfer equation and the average error. The results achieved were subsequently reintegrated in EO-Incity and used for the estimation of the income value at local level. The regression analyses highlighted a stable and strong relationship between SOL and income for the analyzed cities. The EO-Incity tool and the machine learning model proved to be efficient in the real-time estimation of the income at local level. When integrated in the information systems specific for smart cities, they can serve as a support for decision-making in order to fight poverty and reduce social inequalities.
Full-text available
Several studies have linked high temperatures to increases in violent conflicts. The findings are controversial, however, as there has been no systematic cross-sectional analysis performed to demonstrate the generality of the proposed relationship. Moreover, the timescale of temperature/violence relationships have not been fully investigated; it is unclear how short versus long-term, or seasonal and inter-annual temperature variability contribute to the likelihood or frequency of violent events. We here perform systematic regional and grid-based longitudinal analyses in Africa and the Middle East for the period 1990-2017, using geolocated information on armed conflicts and a recently released satellite-based gridded temperature data set. We find seasonal synchrony between temperature and number of armed conflicts at the regional scale (climatic region), as well as a positive relationship in temperature and conflict anomalies on inter-annual timescales at the grid cell level (for the entire African and ME region). After controlling for location effects, we do not find that long-term warming has affected armed conflicts for the last three decades. However, the effects of temperature anomalies are stronger in warmer places (~5% increase per 10°C, P<0.05), suggesting that populations living in warmer places are more sensitive to temperature deviations. Taken together, these findings imply that projected warming and increasing temperature variability may enhance violence in these regions, though the mechanisms of the relationships still need to be exposed.
Full-text available
Night-time lights satellite images provide a new opportunity to measure regional inequality in real-time by developing the Night Light Development Index (NLDI). The NLDI was extracted using the Gini coefficient approach based on population and night light spatial distribution in Romania. Night-time light data were calculated using a grid with a 0.15 km2 area, based on Defense Meteorological Satellite Program (DMSP) /Operational Linescan System (OLS satellite imagery for the 1992–2013 period and based on the National Polar-orbiting Partnership–Visible Infrared Imaging Radiometer Suite (NPP-VIIRS) satellite imagery for the 2014–2018 period. Two population density grids were created at the level of equal cells (0.15 km2) using ArcGIS and PostgreSQL software, and census data from 1992 and 2011. Subsequently, based on this data and using the Gini index approach, the Night Light Development Index (NLDI) was calculated within the MATLAB software. The NLDI was obtained for 42 administrative counties (nomenclature of territorial units for statistics level 3 (NUTS-3 units)) for the 1992–2018 period. The statistical relationship between the NLDI and the socio-economic, demographic, and geographic variables highlighted a strong indirect relationship with local tax income and gross domestic product (GDP) per capita. The polynomial model proved to be better in estimating income based on the NLDI and R2 coefficients showed a significant improvement in total variation explained compared to the linear regression model. The NLDI calculated on the basis of night-time lights satellite images proved to be a good proxy for measuring regional inequalities. Therefore, it can play a crucial role in monitoring the progress made in the implementation of Sustainable Development Goal 10 (reduced inequalities).
Full-text available
Satellite-derived vegetation indices (VIs) have been widely used to approximate or estimate gross primary productivity (GPP). However, it remains unclear how the VI-GPP relationship varies with indices, biomes, timescales, and the bidirectional reflectance distribution function (BRDF) effect. We examined the relationship between VIs and GPP for 121 FLUXNET sites across the globe and assessed how the VI-GPP relationship varied among a variety of biomes at both monthly and annual timescales. We used three widely-used VIs: normalized difference vegetation index (NDVI), enhanced vegetation index (EVI), and 2-band EVI (EVI2) as well as a new VI - NIRV and used surface reflectance both with and without BRDF correction from the moderate resolution imaging spectroradiometer (MODIS) to calculate these indices. The resulting traditional (NDVI, EVI, EVI2, and NIRV) and BRDF-corrected (NDVIBRDF, EVIBRDF, EVI2BRDF, and NIRV, BRDF) VIs were used to examine the VI-GPP relationship. At the monthly scale, all VIs were moderate or strong predictors of GPP, and the BRDF correction improved their performance. EVI2BRDF and NIRV, BRDF had similar performance in capturing the variations in tower GPP as did the MODIS GPP product. The VIs explained lower variance in tower GPP at the annual scale than at the monthly scale. The BRDF-correction of surface reflectance did not improve the VI-GPP relationship at the annual scale. The VIs had similar capability in capturing the interannual variability in tower GPP as MODIS GPP. VIs were influenced by temperature and water stresses and were more sensitive to temperature stress than to water stress. VIs in combination with environmental factors could improve the prediction of GPP than VIs alone. Our findings can help us better understand how the VI-GPP relationship varies among indices, biomes, and timescales and how the BRDF effect influences the VI-GPP relationship.
Full-text available
Research findings on the relationship between climate and conflict are diverse and contested. Here we assess the current understanding of the relationship between climate and conflict, based on the structured judgments of experts from diverse disciplines. These experts agree that climate has affected organized armed conflict within countries. However, other drivers, such as low socioeconomic development and low capabilities of the state, are judged to be substantially more influential, and the mechanisms of climate–conflict linkages remain a key uncertainty. Intensifying climate change is estimated to increase future risks of conflict.
Full-text available
Long-term time series of spatially explicit cropland maps are essential for global crop modelling and climate change studies. The spatial resolution and temporal continuity of global cropland maps have been improving and several global data sets are released recently. Here, we calculated country-level cropland areas from the annual land-cover (LC) maps produced by the European Space Agency Climate Change Initiative (ESA-CCI) project and from the Food and Agricultural Organization of the United Nations statistical data (FAOSTAT) from 1992 to 2014. Because these two data sets used different approaches for generating the cropland data, we further quantified the consistency/difference in cropland areas and temporal changes between both products. Using log-transformed the time-averaged country-level cropland areas, a good linear relationship was found between these two products across different countries. However, only 8% of countries (mostly Organization for Economic Co-operation and Development countries) showed cropland area difference smaller than 1% between ESA-CCI and FAOSTAT. The cropland areas (without mosaic cropland types) from ESA-CCI are lower than the areas from FAOSTAT in 26% of countries but higher in 66% of countries. The magnitude of the latter difference (i.e. higher estimates of ESA-CCI than FAOSTAT) would be further amplified if crop intensity was taken into account. In addition, opposite temporal trends of cropland areas were found between these two data sets in 41% of countries. Although there are uncertainties in ESA-CCI LC maps, resulting from remote-sensing techniques such as mixed pixels, spectral similar objects, and same subject with different spectrum, the long time series and relatively high resolution of this product help us to understand the differences between satellite-based and inventory-based data sets and thus identify the possible strategies to improve the accuracy of satellite-based LC products.
Full-text available
Long and temporally consistent rainfall time series are essential in climate analyses and applications. Rainfall data from station observations are inadequate over many parts of the world due to sparse or non‐existent observation networks, or limited reporting of gauge observations. As a result, satellite rainfall estimates have been used as an alternative or as a supplement to station observations. However, many satellite‐based rainfall products with long time series suffer from coarse spatial and temporal resolutions and inhomogeneities caused by variations in satellite inputs. There are some satellite rainfall products with reasonably consistent time series, but they are often limited to specific geographic areas. The Climate Hazards Group Infrared Precipitation (CHIRP) and CHIRP combined with station observations (CHIRPS) are recently produced satellite‐based rainfall products with relatively high spatial and temporal resolutions and quasi‐global coverage. In this study, CHIRP and CHIRPS were evaluated over East Africa at daily, dekadal (10‐day) and monthly time scales. The evaluation was done by comparing the satellite products with rain gauge data from about 1200 stations. The CHIRP and CHIRPS products were also compared with two similar operation satellite rainfall products: the African Rainfall Climatology version 2 (ARC2) and the Tropical Applications of Meteorology using Satellite data (TAMSAT). The results show that both CHIRP and CHIRPS products are significantly better than ARC2 with higher skill and low or no bias. These products were also found to be slightly better than the latest version of the TAMSAT product at dekadal and monthly time scales, while TAMSAT performed better at daily time scale. The performance of the different satellite products exhibits high spatial variability with weak performances over coastal and mountainous regions.
Full-text available
Critics have argued that the evidence of an association between climate change and conflict is flawed because the research relies on a dependent variable sampling strategy1–4. Similarly, it has been hypothesized that convenience of access biases the sample of cases studied (the ‘streetlight effect’⁵). This also gives rise to claims that the climate–conflict literature stigmatizes some places as being more ‘naturally’ violent6–8. Yet there has been no proof of such sampling patterns. Here we test whether climate–conflict research is based on such a biased sample through a systematic review of the literature. We demonstrate that research on climate change and violent conflict suffers from a streetlight effect. Further, studies which focus on a small number of cases in particular are strongly informed by cases where there has been conflict, do not sample on the independent variables (climate impact or risk), and hence tend to find some association between these two variables. These biases mean that research on climate change and conflict primarily focuses on a few accessible regions, overstates the links between both phenomena and cannot explain peaceful outcomes from climate change. This could result in maladaptive responses in those places that are stigmatized as being inherently more prone to climate-induced violence.
Understanding the dynamics and physics of climate extremes will be a critical challenge for twenty-first-century climate science. Increasing temperatures and saturation vapor pressures may exacerbate heat waves, droughts, and precipitation extremes. Yet our ability to monitor temperature variations is limited and declining. Between 1983 and 2016, the number of observations in the University of East Anglia Climatic Research Unit (CRU) T max product declined precipitously (5900 → 1000); 1000 poorly distributed measurements are insufficient to resolve regional T max variations. Here, we show that combining long (1983 to the near present), high-resolution (0.05°), cloud-screened archives of geostationary satellite thermal infrared (TIR) observations with a dense set of ~15 000 station observations explains 23%, 40%, 30%, 41%, and 1% more variance than the CRU globally and for South America, Africa, India, and areas north of 50°N, respectively; even greater levels of improvement are shown for the 2011–16 period (28%, 45%, 39%, 52%, and 28%, respectively). Described here for the first time, the TIR T max algorithm uses subdaily TIR distributions to screen out cloud-contaminated observations, providing accurate (correlation ≈0.8) gridded emission T max estimates. Blending these gridded fields with ~15 000 station observations provides a seamless, high-resolution source of accurate T max estimates that performs well in areas lacking dense in situ observations and even better where in situ observations are available. Cross-validation results indicate that the satellite-only, station-only, and combined products all perform accurately ( R ≈ 0.8–0.9, mean absolute errors ≈ 0.8–1.0). Hence, the Climate Hazards Center Infrared Temperature with Stations (CHIRTS max ) dataset should provide a valuable resource for climate change studies, climate extreme analyses, and early warning applications.
Tracking global and regional conflict zones requires spatially explicit information in near real-time. Here, we examined the potential of remote sensing time-series data (night lights) and big data (data mining of news events and Flickr photos) for monitoring and understanding crisis development and refugee flows. We used the recent Arab Spring as a case study, and examined temporal trends in monthly time series of variables which we hypothesized to indicate conflict intensity, covering all Arab countries. Both Flickr photos and night-time lights proved as sensitive indicators for loss of economic and human capital, and news items from the Global Data on Events, Location and Tone (GDELT) project on fight events were positively correlated with actual deaths from conflicts. We propose that big data and remote sensing datasets have potential to provide disaggregated and timely data on conflicts where official statistics are lacking, offering an effective approach for monitoring geopolitical and environmental changes on Earth.