Research, part of a special feature on The Nature of Peace: Post-Conflict Peacebuilding and its Implications for Environment and Livelihoods

INTRODUCTION

Armed conflicts can harm ecosystems and wild species in many ways, from direct effects due to shelling and pollution from military activities to indirect effects from a breakdown in environmental governance (Butsic et al. 2015, Hanson 2018). While mitigating and repairing environmental damage that occurs as a result of armed conflict is important to protect biodiversity for its own sake (especially since conflict-affected areas are often located in biodiversity hotspots [Hanson et al. 2009]), it also helps protect the benefits that conflict-affected communities receive from functioning ecosystems, such as clean water, food, and cultural values (Brauman et al. 2019). In addition, there is evidence that environmental degradation can itself contribute to increased conflict (IUCN 2021); thus, restoring nature could contribute to wider peacebuilding efforts.

To devise effective strategies to mitigate and recover from environmental impacts of conflicts, it is necessary to know what and where impacts have occurred. However, assessing changes in environmental conditions and their consequences for people in a conflict setting can be difficult, for example, due to a lack of access to research sites or a breakdown in research infrastructure (Conca and Wallace 2013). This results in a lack of information on conflict-related impacts on nature, which could be a barrier to designing effective recovery strategies after the conflict has ended or once recovery of natural resources becomes feasible in a conflict context (Conca and Wallace 2013). For instance, information about the spatial distribution of conflict-related environmental damage can help prioritize resources for nature restoration and help identify effective restoration strategies.

Satellite remote sensing offers a unique opportunity to address this information gap because it provides information about environmental changes in areas where ground-based data collection is not feasible due to conflict; up-to-date information is provided without endangering researchers. Satellite data have been used extensively to document the environmental impact of armed conflict after a conflict has ceased (or during a peace interlude), especially on deforestation, across the world (e.g., in Colombia [Clerici et al. 2020] and the Democratic Republic of Congo [Butsic et al. 2015]). However, examples of using satellite data to monitor environmental degradation during an ongoing conflict, where the study site is still largely inaccessible to researchers, are, to our knowledge, rare in the scientific literature (though there are examples in the grey literature; e.g., CEOBS 2022).

To illustrate the usability of satellite data to monitor environmental impacts of ongoing conflicts, we investigated environmental change during the war in Tigray (Ethiopia), which started in November 2020, during its initial active phase. The main parties to the conflict are the Tigray regional government (represented by the Tigray People’s Liberation Front) on one side, and the Ethiopian federal government, along with allies, including the Eritrean government and ethnic militia and security forces from Ethiopia’s different regions, on the other side (Amnesty International 2022, International Commission of Human Rights Experts on Ethiopia 2023). On-the-ground impacts of this conflict were still ongoing during the time of data analysis (2021–22), although active fighting has largely ceased since a truce agreement was signed in November 2022 (see e.g., OCHA 2023, Human Rights Watch 2023). However, given that the situation in Tigray cannot be described as peaceful in the sense of positive peace (sensu Barash and Webel 2013), we refer to the conflict as ongoing.

Based on evidence from other contexts (Landholm et al. 2019), it is possible that the Tigray conflict caused deforestation. Increases in deforestation can occur during armed conflicts for multiple reasons, such as fuel needs by civilians or the sale of timber to fund military activities (e.g., Ordway 2015, Woods et al. 2021) (though note that deforestation is not an inevitable outcome of armed conflict [e.g., Stevens et al. 2011]). Woody vegetation cover is important for ecosystem functioning in Tigray; therefore, deforestation could potentially have harmful effects on agricultural yields. During the early 21st century, Tigray underwent nature restoration at a regional scale to improve the retention of water and soil (Gebremeskel et al. 2018). This included building small-scale infrastructure, such as terraces and dams, but also widespread reforestation and forest rehabilitation in grazing exclosures. Expansion of forest and other woody vegetation helps slow down water runoff, reduce erosion, and promote water infiltration into the soil, and thus helps retain water in the landscape (Descheemaeker et al. 2006). There is evidence that these efforts directly contributed to increased water availability (Nyssen et al. 2010), reduced erosion (Girmay et al. 2009), and increased woody vegetation cover (Nyssen et al. 2009, Belay et al. 2015) (Fig. 1), leading to improved yields (Gebremeskel et al. 2018) and thus contributing to decreasing food insecurity in the region.

Given the potential for deforestation during armed conflicts and the harmful effects this would have on both ecosystems and people in the region, we focus on two environmental questions arising from the conflict in Tigray: (1) to what extent has woody vegetation declined in Tigray since the start of the war, and (2) does the spatial pattern of these declines match known non-conflict drivers of woody vegetation loss? To answer these questions, we combined open-source satellite remote sensing data (collected during the war) with ground-based information (collected before the war) to (1) map the distribution of forests and other woody vegetation across Tigray before the war, (2) use a remotely sensed indicator of vegetation vigor to map areas potentially affected by woody vegetation decline, (3) explore to what extent non-conflict factors, namely changes in rainfall and large vegetation fires, spatially coincide with observed vegetation changes, and (4) consider the potential importance of conflict-related deforestation. We then used these results to speculate about the environmental impacts of the war in Tigray and consider how nature restoration and rehabilitation could help contribute to an eventual recovery process.

METHODS

Study area

Tigray is the northernmost state of Ethiopia, covering an area of approximately 50,000 km², and is characterized by an arid climate (Fekadu 2015). Tigray receives 590 mm of rainfall per year on average (International Food Policy Research Institute and Datawheel 2017), mostly during the kremti season from June to September, though this masks a high level of climatic variability across Tigray, with rainfall in the southwest being significantly higher than elsewhere (Fig. 2). Most people in Ethiopia depend on subsistence agriculture (Sibhatu and Qaim 2017), and agricultural output is highly dependent on rainfall during the wet season (Eze et al. 2022).

Mapping areas with trees and other woody vegetation before the war

Pre-conflict land cover in Tigray was mapped using both radar and multispectral data from ESA’s Sentinel 1 and 2 satellite missions, respectively (Lopes et al. 2020). Satellite data were chosen to coincide with two different seasons: (1) January and February 2020 (at the end of the last pre-war dry season), and (2) September and October 2020 (just before the start of the war, and at the start of the new dry season). Using satellite remote sensing data from multiple seasons helps differentiate between land cover classes (see Table 1) because their reflective properties often change in characteristic ways over the year (Lopes et al. 2020).

Pre-processing was carried out in Google Earth Engine. A radiometric slope correction was applied to the Sentinel 1 (radar) data to correct for topographic distortions. For pixels with multiple observations in September/October 2020, the median backscatter value for each band (cross- and co-polarization) was calculated; the same process was carried out for the January/February data. For the Sentinel 2 (multispectral) data, the cloud probability mask was used to mask out any pixels with a likelihood of > 15% of being covered by a cloud. Though this necessarily resulted in some gaps in coverage, they were negligible across most of Tigray (Fig. A1.1, Appendix 1). Finally, for pixels with multiple observations, the median spectral value for each band (B1-8, B8A, B9, B11, B12) was calculated for January/February and September/October 2020, respectively. Further analysis was carried out in R and QGIS.

Given the large size of Tigray, the inability to access the region due to the security situation, and the urgent need to monitor environmental degradation, groundtruth data were obtained by drawing polygons around homogeneous patches of different land cover classes that were visible in very-high-resolution imagery available on Google Earth Pro (see Table 1). We followed an opportunistic sampling design, which was limited mainly by the availability of high-resolution imagery from the correct time period; however, in order to ensure relatively even coverage across Tigray within these constraints, we used a 30- x 30-km grid to guide groundtruth collection (Fig. A1.2, Appendix 1). Despite the opportunistic sampling approach, the different climatic zones in Tigray are well represented in the groundtruth dataset (Fig. A1.2, Appendix 1). We used spatially referenced land cover data collected on the ground before the war (Annys et al. 2017) as examples of land cover classes. The groundtruth data collected on Google Earth were checked by two researchers with extensive experience in mapping land cover across Tigray to ensure that they were correct.

The groundtruth data were then combined with the Sentinel 1 (mean cross-polarization [i.e., the VH band] and mean co-polarization [i.e., the VV band]) and 2 (using bands B1-8, B8A, B9, B11, B12) data to conduct a supervised land cover classification using the Random Forest algorithm (Breiman et al. 2001). Of the groundtruth data, 60% (n = 21,014 pixels) were used as training data, and the remainder (n = 14,009 pixels) were used for validation to ensure that the points in these two sets came from different polygons. This minimizes the risk of overfitting the classification model to the data. A Random Forest model was created based on the training data alone, growing 1000 trees and using five variables to split at each tree node.

To make the land cover mapping robust to the choice of training and validation data, nine further random training and validation sets were sampled, which resulted in a total of 10 training/validation set combinations. The land cover classification was repeated for each, so a total of 10 land cover maps were generated. In each land cover classification, the same variables emerged as the most important (i.e., those with the highest decrease in mean accuracy): Sentinel 2 band 1 from both seasons (which corresponds to light at the blue edge of the visible spectrum), Sentinel 2 band 9 from both seasons (a mid-range infrared), and Sentinel 1’s cross-polarization (VH) band from January/February.

A consensus map was created from these 10 maps by majority vote. The consensus map was then re-classified into two classes, namely land cover with woody vegetation (forest, other woody vegetation; e.g., scattered trees) and land cover without woody vegetation (all other classes), and the accuracy was assessed by comparing it against an independent groundtruth dataset (i.e., which had not contributed to any of the land cover maps). This independent groundtruth set consisted of 18,664 samples of woody vegetation, and 17,780 samples of other land cover classes.

Mapping areas where vegetation cover or condition has changed

The Normalized Difference Vegetation Index (NDVI) was used to track changes in vegetation cover and/or condition over time (Pettorelli 2013). The NDVI describes the difference between the amount of near-infrared and red light reflected by vegetation, and is commonly used to assess vegetation status, including in Ethiopia (Zewdie et al. 2017). Photosynthesizing vegetation tends to absorb red light and reflect near-infrared light, so areas with lush vegetation have NDVI values of close to 1, whereas areas with less vegetation, or where photosynthetic activity is reduced (e.g., due to a drought), have lower NDVI values.

We used the Break for Additive and Seasonal Trend Monitor algorithm (Bfast Monitor) to identify abrupt changes in NDVI dynamics (which are potentially caused by deforestation) while accounting for normal seasonal changes (Verbesselt et al. 2012). The Bfast Monitor algorithm was developed to detect significant change that occurs toward the end of a time series (the monitoring period) when compared to the rest of the time series (the historical period). It also allows the magnitude and timing of this break in the time series to be calculated.

In Google Earth Engine, we collected all Landsat 7 or 8 Collection 2 Tier 1 surface reflectance data (30-m spatial resolution) for Tigray between 1 January 2018 and 30 September 2022 (see Table A1.1 in Appendix 1 for the number of unique dates for which there were Landsat 7 or 8 observations). NDVI values were calculated using bands 4 and 5 for Landsat 8 products, and bands 3 and 4 for Landsat 7 products.

Landsat pixels that did not contain forest or woody vegetation in 2020 were ignored in the following analysis. This required resampling the woody vegetation map to match the coarser 30-m resolution of the Landsat imagery. This necessarily makes the map less accurate at very small spatial scales (i.e., tens of meters), but overall, the coarser map adequately reflected where trees and shrubs were found in 2020.

A version of the Bfast Monitor algorithm implemented in Google Earth Engine (Hamunyela 2014) was applied to NDVI observations in each pixel. The start of the monitoring period was set to 1 January 2021; the end to 30 September 2022 (the historical period was 1 January 2018 to 31 December 2020). To be able to assess the effect of parameter settings on the NDVI change detection, we ran six variations of the Bfast Monitor algorithm, setting the width of the moving window to either 25% or 50% of the entire time series, and varying the harmonics term (which describes the shape of the seasonal variation component) between 1, 2, and 3. We used alpha = 0.05 to determine whether the difference in NDVI between the historical and monitoring period was statistically different.

To estimate the agreement between the varying parameter combinations, the number of statistically significant decreases and increases were counted for each pixel; these could vary between 0 (a statistically significant decrease/increase in NDVI was never detected) and 6 (all parameter combinations detected a significant decrease/increase). We restricted all further analysis to pixels where NDVI change had been detected with high certainty; i.e., all six parameter combinations agreed that a significant decline or increase had taken place.

Many changes seemed to be detected very close to the start of the monitoring period (i.e., in the first few days of 2021). This could be an artifact of earlier disturbances that took place during the historical period. We thus restricted all further analysis to changes that were detected on or after 1 February 2021. This most likely resulted in omissions of some significant NDVI increases and decreases, but it will have reduced the commission error (erroneous detection of NDVI changes where none have occurred), and so represents a conservative estimate of vegetation change.

To visualize the results of the analysis, we aggregated the data into 100- x 100-pixel blocks, which lead to 3-km “megapixels.” For each megapixel, we calculated the total area of forest and other wooded vegetation that had undergone significant decreases and increases in NDVI.

These steps were then repeated for three other years to gauge what effect, if any, the conflict had on changes in NDVI dynamics. To show the potential impact of interannual climate variability on NDVI changes, an exceptionally dry year (2015) and a comparatively wet year (2019) were analyzed. The year immediately before the war (2020) was also analyzed to compare post-war to pre-war changes in trees and other woody vegetation. The monitoring period was set to 1 January to 31 December for each year; the historical period was set to three full years before the start of the focal year to match the length of the historical periods used for NDVI change detection in 2021/2022. Only NDVI changes detected after 1 February were included in the analysis because, again, many NDVI changes were detected in the first few days of each year.

Non-conflict-related drivers of vegetation changes

We also investigated whether any NDVI declines that we observed in 2021/22 could be attributed to droughts or declines in rainfall more broadly. We calculated the sum of rainfall residuals from November 2020 (the start of the first dry season during the war) to September 2022, based on monthly rainfall data from CHIRPS v2 (Funk et al. 2015). Residuals refer to the difference between the median rainfall in a given month between 2002 and 2022 and the rainfall that was received, which accounts for the strong seasonal signal in rainfall in this region. For each CHIRPS pixel (approximately 5.5 x 5.5 km), we then extracted the median change magnitude of each smaller (Landsat) pixel that had undergone significant changes in NDVI to compare how lower- or higher-than-expected rainfall was correlated with NDVI changes.

Increases in fire activity could also have contributed to vegetation declines. We mapped the extent of burned areas using satellite data to gauge whether vegetation declines may have been the result of large vegetation fires. We downloaded all available VIIRS S-NPP standard Active Fire and Thermal Anomalies products (nominal resolution: 375 m [NASA FIRMS 2022]) generated after 3 November 2020 (the latest product at the time of writing was from 30 June 2022). To approximate the spatial cover of the fire corresponding to each point, a 460-m buffer was generated (the median pixel size as recorded in the VIRRS product). To compare the spatial distribution of fire activity during the war to that before the war, we also processed and visualized VIIRS-detected fires in the 3 years before the war (1 January 2018 to 2 November 2020). It is important to note that this approach misses small fires (< 250 m²), especially when they are relatively cold and/or occur during the day (Schroeder et al. 2014). Such small fires may have proliferated during the war as a result of burning houses or other infrastructure, and while they are harmful to people and the environment, we assume that they are unlikely to have a significant effect on woody vegetation at a large scale because they will be concentrated in villages and towns.

The potential effect of cloud gaps on the vegetation change analysis was assessed by plotting the amount of missing data per megapixel against the proportion of woody vegetation, NDVI increases, and NDVI declines.

RESULTS

Woody vegetation loss

The classification accuracy of the binary woodland/no woodland map (the proportion of validation points for which the classification algorithm identified land cover class correctly) was 88%. The producer accuracy (the proportion of validation points identified as woody vegetation that were correctly identified [Foody 2002]) and the user accuracy (the proportion of validation points corresponding to woody vegetation correctly identified as such [Foody 2002]) were 86% and 90%, respectively (Fig. 2). Forest and other woody vegetation (such as scattered shrubs) covered approximately 47% of Tigray’s total area before the war.

Between February 2021 and September 2022, approximately 4% of the forest and woody vegetation (approximately 930 km²) experienced significant declines in NDVI, which can indicate removal of vegetation or a decline in the health of vegetation (Table 2, Fig. 3). These NDVI declines appear to be concentrated mostly in distinct areas (decline “hotspots”), many along major roads (Fig. 4). During the same time, 11.1% of the areas with woody vegetation experienced a significant increase in NDVI, indicating growth or expansion of that vegetation. To be able to compare the rate of NDVI changes between pre-war years (2015, 2019, and 2020) and the war period, we also present NDVI changes that occurred between 1 February 2021 and 31 December 2021. In this way, we account for the fact that NDVI change accumulates over time, meaning a direct comparison between the entire war period (i.e., until September 2022) and pre-war years would lead to an overestimation of the relative frequency of NDVI changes during the war.

Potential drivers of woody vegetation loss during the conflict in Tigray

Comparing NDVI changes between an exceptionally dry year (2015) and an exceptionally wet year (2019) gave an indication of the responses of woody vegetation to either end of the climatic variability spectrum (Fig. 3). During the dry year (2015), NDVI declines were common and increases were rare, whereas the reverse was true for the unusually wet year (2019), as expected (Table 2). In comparison, the last pre-war year (2020) showed overall subdued NDVI changes, with unremarkable NDVI declines and NDVI increases of a lower magnitude than in 2019 (Table 2, Fig. 3). Specifically, the ratio of NDVI decreases to NDVI increases in the last pre-war year (2020) was more similar to that in the wet year (2019) than that during the war. NDVI increases continued during the war, across roughly the same area as in the last pre-war year (Table 2). This suggests that NDVI increases were possible in the climatic conditions during the war (i.e., there was no drought), but NDVI declines seemed to be much more widespread than expected, which suggests that a non-climatic factor was driving the observed changes.

Supporting this conclusion is the observation that there was no correlation between rainfall residuals during the war and median NDVI change magnitude (Fig. 5). Large negative residual sums indicate that an area experienced many months with less rainfall than expected (and vice versa for large positive residual sums); i.e., areas that received less rainfall did not systematically show larger declines in NDVI (and vice versa for areas that received a lot of rain). Overall, this means that rainfall is unlikely to drive the observed NDVI changes.

Approximately 13,000 fires were detected in Tigray between November 2020 and June 2022 (Fig. 6A), though due to the spatial resolution of the underlying data, only relatively large fires (> 250 m²) can be detected. This means that smaller fires, such as those affecting settlements or individual fields, were not considered in this analysis. Fires were concentrated mostly in the west, with little apparent systematic overlap with areas of vegetation decline. The distribution of relatively large fires during the war was similar to that before the war (Fig. 6B), and the incidence of these types of fires does not appear to have increased during the war (Fig. 6C).

There was no obvious relationship between the extent of missing satellite data due to clouds and the resulting land cover classification (see Fig. A1.1B, Appendix 1). While the proportion of both NDVI increases and decreases appeared to increase with the amount of cloud-free satellite data (Fig. A1.1C and A1.1D, Appendix 1), these relationships did not appear to differ between increases and decreases, meaning that while some significant NDVI changes may have been missed, the results are unlikely to be biased in either direction.

DISCUSSION

Drivers of woody vegetation loss during the conflict in Tigray, and implications for future restoration efforts

Using satellite remote sensing data, we have provided an initial quantitative assessment of one of the potential environmental impacts of the ongoing conflict in Tigray, namely woody vegetation loss. Specifically, we have shown that vegetation declines, potentially linked to deforestation, intensified since the start of the war in some areas, though vegetation recovery appears to continue in others. In addition, there is evidence that the potential woody vegetation loss observed is driven by the conflict because other candidate factors (fire, rainfall) cannot explain the spatial distribution of vegetation declines. It is important to highlight, however, that our analysis does not allow for the establishment of a causal link between the conflict and woody vegetation loss; this could be achieved by, for example, combining spatially explicit information about conflict occurrence and intensity and ground-validated observations of deforestation (data that were not available at the time this study was conducted). While these results will need to be validated via ground-based observations of deforestation, they provide important insights into the spatial distribution of potential deforestation, which can guide future environmental assessments, as well as nature recovery strategies. On-the-ground assessments will also be necessary to map ecological impacts, such as declines in wild animal populations (which can, for example, result from overhunting during periods of conflict [Gaynor et al. 2016]), small-scale erosion, or declines in water quality that cannot (or not as reliably) be detected via satellite data alone.

Since the start of the war, landlocked Tigray has been under a blockade, so almost no imports have reached the region (The Guardian 2022). Food, fuel, and other essential materials have been in short supply, which has led to a humanitarian crisis, including widespread famine (Jerving 2022, World Food Programme 2022). Given the absence of fuel imports, and extensive electricity blackouts (Associated Press 2022), local wood resources have likely been used as alternative fuel or alternative sources of income. Such conflict-enabled deforestation has been observed in other conflict contexts; e.g., in the Democratic Republic of the Congo (Butsic et al. 2015). Given the importance of trees in regulating local water supply (Asfaha et al. 2015), this could have negative impacts on the yield and livelihood gains achieved during a successful, decades-long landscape restoration program in Tigray (Gebremeskel et al. 2018). The blockade has been making ground-based assessments of the extent of this environmental impact exceptionally difficult for researchers inside and outside of Tigray, especially given that telecommunication services are widely unavailable (Associated Press 2022). Under these circumstances, remotely sensed information such as satellite imagery is the only source of information that can provide an initial, yet thorough, assessment of potential environmental impacts of the war.

Our observations are consistent with anecdotal observations received from contacts within Tigray who report that trees are being cut down locally for fuel from charcoaling as the region has remained disconnected from the national energy supply grid throughout the period of the war. Charcoal hearths are visible in high-resolution satellite imagery from Google Earth that was taken during the war (Fig. 4). We did not find evidence that changes in rainfall or fire activity were responsible for the observed vegetation declines. Rainfall deficits did not systematically overlap with areas of NDVI decline (Fig. 5), and while NDVI declines were observed during the war, so were increases, which suggests that vegetation growth was not fundamentally limited during this time. While fires have resulted from warfare since November 2020 (e.g., Reuters 2021), there is no evidence that the incidence or spatial distribution of relatively large fires (> 250 m²) has changed significantly during the war; therefore, forest fires are unlikely to be responsible for the observed potential woody vegetation loss (Fig. 6). In particular, fires during the war were concentrated in the west, as they were before the war (van Breugel et al. 2016). Finally, the potential impact of locusts, which can decimate vegetation (Kimathi et al. 2020), is more difficult to rule out because locust observations have not been possible in Tigray during the war (OCHA 2021). However, based on observations of locusts outside of Tigray’s border, widespread locust impacts on woody vegetation seem unlikely, except potentially in northeastern Tigray (FAO 2022).

It is important to note that vegetation recovery appeared to proceed in some areas, overall outweighing declines (Table 2), especially in the northeast of Tigray, as well as in a cluster around 30 km southeast of Mekelle (Figs. 2, 3). This means that loss of trees and other woody vegetation is not inevitable in a conflict context but is more likely to be shaped by an interaction between different local and regional factors (Burgess et al. 2015, Butsic et al. 2015), such as woody vegetation cover, accessibility, and demand, all of which can be affected by conflict. For instance, proximity to roads likely facilitates access to timber and transport of charcoaling products (Barber et al. 2014); thus, areas close to roads may be more vulnerable to conflict-driven deforestation than are other areas (Fig. 4). In addition, differences in the socioeconomic conditions across Tigray during the war (such as differences in fuel demand from local and displaced inhabitants) could have contributed to the regional differences in vegetation change. Areas where vegetation continues to thrive will be less affected by the loss of soil and water retention capacity (Descheemaeker et al. 2006, Gebremeskel et al. 2018), which may contribute to the ability of local communities to recover after the war. Identifying the local factors that make areas more (or less) vulnerable to negative environmental impacts during conflicts, whether due to differences in exposure to conflict-related drivers or regional differences in vulnerability to their effects, is necessary to be able to mitigate and respond to conflict-driven environmental impacts (e.g., Mobaied and Rudant 2019).

The ongoing famine (or, at the very least, food security crisis [FEWS NET 2023]) in Tigray has multiple roots in the conflict. It results from the blockade, which prevents the import of fertilizer and seeds (Nyssen et al. 2022); from looting and destruction of farms and other economic infrastructure; and from insecurity making it difficult for farmers to tend to their fields (Ghebreyohannes et al. 2022). In the long-term, food insecurity will likely be compounded by the loss of woody vegetation and subsequent loss of water and soil retention capacity of the land, especially in areas where our analysis shows a concentration of woody vegetation loss (Fig. 3). Reversing the conflict-driven landscape degradation would contribute to increasing food and livelihood security, and is also likely to play a key role in building the region’s resilience against the impact of climate change, as Tigray is expected to experience more intense droughts in the future, and agricultural yields are expected to decline as a result (Murken et al. 2020, Ranasinghe et al. 2021). Minimizing soil erosion and boosting water availability, and thereby safeguarding agricultural productivity, is thus an important part of ensuring the sustainability of rural livelihoods, as part of an eventual peacebuilding process.

We considered changes in NDVI only in areas that were covered in forest or other woody vegetation before the war. It is possible, given the abandonment of farms and the difficulty of carrying out agricultural activities during the war (Ghebreyohannes et al. 2022), that woody vegetation encroached onto cropland (though our analysis cannot quantify whether this has occurred). Where this affects productive agricultural land, clearing of such vegetation will be an appropriate strategy to improve agricultural productivity. However, in some areas (e.g., where agricultural output is severely limited), allowing woody vegetation to encroach could benefit regional soil and water retention, which, overall, would contribute to landscape recovery. On-the-ground assessments that take into account local social-ecological context will be necessary to decide where this will be appropriate.

The war has resulted in the deaths of hundreds of thousands of civilians and immense human suffering (Vanden Bempt et al. 2021), and the displacement of millions of people within Tigray (UNHCR 2022). Aside from these primary impacts of the large-scale humanitarian catastrophe the conflict has caused, there are likely to be secondary impacts: specifically, the loss of human life and displacement of survivors has likely eroded the capacity of social-ecological systems in Tigray to contribute to recovery and environmental peacebuilding processes by harming the social, economic, and political structures that facilitated ecological restoration efforts in the past. For instance, the labor force available for the large-scale public works that underpinned Tigray’s decades-long restoration program has very likely been reduced in many places; under such conditions, local expertise and ecological knowledge, which may be key to supporting successful restoration, can also be lost or reduced (Boger et al. 2019). Given the importance of environmental health for the well-being of people in Tigray, rebuilding the institutions and structures that resulted in its previous nature restoration success have to be at the heart of any future recovery strategies.

Conclusions

We have demonstrated how satellite remote sensing data can assist in the assessment of environmental impacts of an ongoing conflict. Where deforestation in Tigray has taken place, it is likely to ultimately reduce the ability of the landscape to support societal recovery from the conflict, which will be compounded by the dismantling of the social-ecological structures underpinning agricultural productivity during the conflict. As a result, rehabilitating the landscape after the war and rebuilding the social and physical structures that previously enabled large-scale nature recovery in Tigray should be an important part of any future peacebuilding process. The pre-war landscape restoration, as well as the continued vegetation recovery observed across large parts of Tigray, shows that this is possible, even when the landscape has been severely degraded, but it takes considerable input of resources, especially labor, and occurs on the timescale of years and decades. More broadly, the loss of trees and other woody vegetation during the ongoing conflict in some areas of Tigray shows how armed conflicts can harm the results of nature restoration efforts and similar nature-based solutions to societal problems. Given that nature restoration and protection are increasingly being recognized as key tools in combating the global climate and biodiversity crises, it is crucial that we understand in which circumstances these efforts are threatened by armed conflicts and how they can be made resilient to conflict. This is particularly important in fragile or conflict-prone contexts where healthy ecosystems can help increase the resilience of communities, and so reduce the humanitarian and socioeconomic consequences of armed conflicts.

LITERATURE CITED

Amnesty International. 2022. “We will erase you from this land”: crimes against humanity in Western Tigray Zone. https://www.amnesty.org/en/latest/news/2022/04/ethiopia-crimes-against-humanity-in-western-tigray-zone/

Annys, S., B. Demissie, A. Z. Abraha, M. Jacob, and J. Nyssen. 2017. Land cover changes as impacted by spatio-temporal rainfall variability along the Ethiopian Rift Valley escarpment. Regional Environmental Change 17:451-463. https://doi.org/10.1007/s10113-016-1031-2

Asfaha, T. G., A. Frankl, M. Haile, A. Zenebe, and J. Nyssen. 2015. Determinants of peak discharge in steep mountain catchments – case of the Rift Valley escarpment of Northern Ethiopia. Journal of Hydrology 529:1725-1739. https://doi.org/10.1016/j.jhydrol.2015.08.013

Associated Press. 2022. Electricity, telecoms return to parts of Tigray following cease-fire with Ethiopia. https://www.pbs.org/newshour/world/electricity-telecoms-return-to-parts-of-tigray-following-cease-fire-with-ethiopia

Barash, D. P., and C. P. Webel. 2013. Peace and conflict studies. Sage Publications, Thousand Oaks, California, USA.

Barber, C. P., M. A. Cochrane, C. M. Souza, Jr., and W. F. Laurance. 2014. Roads, deforestation, and the mitigating effect of protected areas in the Amazon. Biological Conservation 177:203-209. https://doi.org/10.1016/j.biocon.2014.07.004

Belay, K. T., A. Van Rompaey, J. Poesen, S. Van Bruyssel, J. Deckers, and K. Amare. 2015. Spatial analysis of land cover changes in Eastern Tigray (Ethiopia) from 1965 to 2007: are there signs of a forest transition? Land Degradation & Development 26(7):680-689. https://doi.org/10.1002/ldr.2275

Boger, R., S. Perdikaris, and I. Rivero-Collazo. 2019. Cultural heritage and local ecological knowledge under threat: two Caribbean examples from Barbuda and Puerto Rico. Journal of Anthropology and Archaeology 7(2):1-14.

Brauman, K. A., L. A. Garibaldi, S. Polasky, C. Zayas, Y. Aumeeruddy-Thomas, P. Brancalion, F. DeClerck, M. Mastrangelo, N. Nkongolo, H. Palang, L. Shannon, U. B. Shrestha, and M. Verma. 2019. Status and trends – nature’s contributions to people (NCP). Chapter 2.3 in E. S. Brondízio, J. Settele, S. Díaz, and H. T. Ngo, editors. Global assessment report of the Intergovernmental Science-Policy Platform on Biodiversity and Ecosystem Services. IPBES Secretariat, Bonn, Germany.

Breiman, L. 2001. Random forests. Machine Learning 45(1):5-32. https://doi.org/10.1023/A:1010933404324

Burgess, R., E. Miguel, and C. Stanton. 2015. War and deforestation in Sierra Leone. Environmental Research Letters 10(9):095014. https://doi.org/10.1088/1748-9326/10/9/095014

Butsic, V., M. Baumann, A. Shortland, S. Walker, and T. Kuemmerle. 2015. Conservation and conflict in the Democratic Republic of Congo: the impacts of warfare, mining, and protected areas on deforestation. Biological Conservation 191:266-273. https://doi.org/10.1016/j.biocon.2015.06.037

Clerici, N., D. Armenteras, P. Kareiva, R. Botero, J. P. Ramírez-Delgado, G. Forero-Medina, J. Ochoa, C. Pedraza, L. Schneider, C. Lora, and C. Gómez. 2020. Deforestation in Colombian protected areas increased during post-conflict periods. Scientific Reports 10(1):4971. https://doi.org/10.1038/s41598-020-61861-y

Conca, K., and J. Wallace. 2013. Environment and peacebuilding in war-torn societies: lessons from the UN Environment Programme’s experience with post-conflict assessment. In D. Jensen, and S. Lonergan, editors. Assessing and restoring natural resources in post-conflict peacebuilding. Routledge, London, UK.

Conflict and Environment Observatory (CEOBS). 2022. Report: the war in Tigray is undermining its environmental recovery. https://ceobs.org/the-war-in-tigray-is-undermining-its-environmental-recovery/

Descheemaeker, K., J. Nyssen, J. Poesen, M. Haile, B. Muys, D. Raes, J. Moeyersons, and J. Deckers. 2006. Soil and water conservation through forest restoration in exclosures of the Tigray highlands. Journal of the Drylands 1(2):118-133.

Eze, E., A. Girma, A. Zenebe, C. C. Okolo, J. M. Kourouma, and E. Negash. 2022. Predictors of drought-induced crop yield/losses in two agroecologies of southern Tigray, Northern Ethiopia. Scientific Reports 12:6284. https://doi.org/10.1038/s41598-022-09862-x

Famine Early Warning Systems Network (FEWS NET). 2023. Ethiopia key message update, March 2023: Despite favorable start to gu/genna, Emergency! (IPC Phase 4!) persists. https://fews.net/east-africa/ethiopia/key-message-update/march-2023

Fekadu, K. 2015. Ethiopian seasonal rainfall variability and prediction using canonical correlation analysis (CCA). Earth Sciences 4(3):112-119. https://doi.org/10.11648/j.earth.20150403.14

Food and Agriculture Organization (FAO). 2022. Desert locus upsurge (2019-2021). https://www.fao.org/ag/locusts/en/info/2094/index.html

Foody, G. M. 2002. Status of land cover classification accuracy assessment. Remote Sensing of Environment 80(1):185-201. https://doi.org/10.1016/S0034-4257(01)00295-4

Funk, C., P. Peterson, M. Landsfeld, D. Pedreros, J. Verdin, S. Shukla, G. Husak, J. Rowland, L. Harrison, A. Hoell, and J. Michaelsen. 2015. The climate hazards infrared precipitation with stations—a new environmental record for monitoring extremes. Scientific Data 2:150066. https://doi.org/10.1038/sdata.2015.66

Gaynor, K. M., K. J. Fiorella, G. H. Gregory, D. J. Kurz, K. L. Seto, L. S. Withey, and J. S. Brashares. 2016. War and wildlife: linking armed conflict to conservation. Frontiers in Ecology and the Environment 14(10):533-542. https://doi.org/10.1002/fee.1433

Gebremeskel, G., T. G. Gebremicael, and A. Girmay. 2018. Economic and environmental rehabilitation through soil and water conservation, the case of Tigray in northern Ethiopia. Journal of Arid Environments 151:113-124. https://doi.org/10.1016/j.jaridenv.2017.12.002

Ghebreyohannes, T., J. Nyssen, E. Negash, H. Meaza, Z. Tesfamariam, A. Frankl, B. Demissie, B. Van Schaeybroeck, A. Redda, S. Annys, and F. Abay. 2022. Challenges and resilience of an indigenous farming system during wartime (Tigray, North Ethiopia). Agronomy for Sustainable Development 42:116. https://doi.org/10.1007/s13593-022-00812-5

Girmay, G., B. R. Singh, J. Nyssen, and T. Borrosen. 2009. Runoff and sediment-associated nutrient losses under different land uses in Tigray, Northern Ethiopia. Journal of Hydrology 376(1-2):70-80. https://doi.org/10.1016/j.jhydrol.2009.07.066

Hamunyela, E. 2014. Bfast monitoring algorithm implemented for Google Earth Engine. https://github.com/bfast2/geeBfastMonitor

Hanson, T. 2018. Biodiversity conservation and armed conflict: a warfare ecology perspective. Annals of the New York Academy of Sciences 1429(1):50-65. https://doi.org/10.1111/nyas.13689

Hanson, T., T. M. Brooks, G. A. Da Fonseca, M. Hoffmann, J. F. Lamoreux, G. Machlis, C. G. Mittermeier, R. A. Mittermeier, and J. D. Pilgrim. 2009. Warfare in biodiversity hotspots. Conservation Biology 23(3):578-587. https://doi.org/10.1111/j.1523-1739.2009.01166.x

Humanitarian OpenStreetMap Team. 2022. https://data.humdata.org/dataset/hotosm_eth_roads

Human Rights Watch. 2023. Ethiopia: ethnic cleansing persists under Tigray truce. https://www.hrw.org/news/2023/06/01/ethiopia-ethnic-cleansing-persists-under-tigray-truce

International Commission of Human Rights Experts on Ethiopia. 2023. Comprehensive investigative findings and legal determinations. https://www.ohchr.org/sites/default/files/documents/hrbodies/hrcouncil/chreetiopia/a-hrc-54-crp-3.pdf

International Food Policy Research Institute and Datawheel. 2017. https://dataafrica.io/profile/tigray-eth/RainfallBars

International Union for Conservation of Nature (IUCN). 2021. Conflict and conservation. Nature in a Globalised World Report No. 1. Gland, Switzerland. https://doi.org/10.2305/IUCN.CH.2021.NGW.1.en

Jerving, S. 2022. Fuel shortage leaves Ethiopia’s Tigray ‘running on fumes’. https://www.devex.com/news/fuel-shortage-leaves-ethiopia-s-tigray-running-on-fumes-102702

Kimathi, E., H. E. Z. Tonnang, S. Subramanian, K. Cressman, E. M. Abdel-Rahman, M. Tesfayohannes, S. Niassy, B. Torto, T. Dubois, C. M. Tanga, M. Kassie, S. Ekesi, D. Mwangi, and S. Kelemu. 2020. Prediction of breeding regions for the desert locust Schistocerca gregaria in East Africa. Scientific Reports 10:11937. https://doi.org/10.1038/s41598-020-68895-2

Landholm, D. M., P. Pradhan, and J. P. Kropp. 2019. Diverging forest land use dynamics induced by armed conflict across the tropics. Global Environmental Change 56:86-94. https://doi.org/10.1016/j.gloenvcha.2019.03.006

Lopes, M., P.-L. Frison, S. M. Durant, H. Schulte to Bühne, A. Ipavec, V. Lapeyre, and N. Pettorelli. 2020. Combining optical and radar satellite image time series to map natural vegetation: savannas as an example. Remote Sensing in Ecology and Conservation 6(3):316-326. https://doi.org/10.1002/rse2.139

Mobaied, S., and J.-P. Rudant. 2019. New method for environmental monitoring in armed conflict zones: a case study of Syria. Environmental Monitoring and Assessment 191:643. https://doi.org/10.1007/s10661-019-7805-5

Murken, L., M. Cartsburg, A. Chemura, I. Didovets, S. Gleixner, H. Koch, J. Lehmann, S. Liersch, S. Lüttringhaus, M. R. Rivas López, S. Noleppa, F. Roehrig, B. Schauberger, R. Shukla, J. Tomalka, A. Yalew, and C. Gornott. 2020. Climate risk analysis for identifying and weighing adaptation strategies in Ethiopia’s agricultural sector. A report prepared by the Potsdam Institute for Climate Impact Research for the Deutsche Gesellschaft für Internationale Zusammenarbeit GmbH on behalf of the German Federal Ministry for Economic Cooperation and Development.

NASA FIRMS. 2022. NRT VIIRS 375 m active fire product VNP14IMGT distributed from NASA FIRMS. https://earthdata.nasa.gov/firms

Nyssen, J., W. Clymans, K. Descheemaeker, J. Poesen, I. Vandecasteele, M. Vanmaercke, A. Zenebe, M. Van Camp, M. Haile, N. Haregeweyn, and J. Moeyersons. 2010. Impact of soil and water conservation measures on catchment hydrological response—a case in north Ethiopia. Hydrological Processes 24(13):1880-1895. https://doi.org/10.1002/hyp.7628

Nyssen, J., M. Haile, J. Naudts, N. Munro, J. Poesen, J. Moeyersons, A. Frankl, J. Deckers, and R. Pankhurst. 2009. Desertification? Northern Ethiopia re-photographed after 140 years. Science of the Total Environment 407(8):2749-2755. https://doi.org/10.1016/j.scitotenv.2008.12.016

Nyssen, J., E. Negash, and S. Annys. 2022. How Ethiopia’s conflict has affected farming in Tigray. https://theconversation.com/how-ethiopias-conflict-has-affected-farming-in-tigray-166229

Ordway, E. M. 2015. Political shifts and changing forests: effects of armed conflict on forest conservation in Rwanda. Global Ecology and Conservation 3:448-460. https://doi.org/10.1016/j.gecco.2015.01.013

Pettorelli, N. 2013. The normalized difference vegetation index. Oxford University Press, Oxford, UK. https://doi.org/10.1093/acprof:osobl/9780199693160.001.0001

Ranasinghe, R., A. C. Ruane, R. Vautard, N. Arnell, E. Coppola, F. A. Cruz, S. Dessai, A. S. Islam, M. Rahimi, D. Ruiz Carrascal, J. Sillmann, M. B. Sylla, C. Tebaldi, W. Wang, and R. Zaaboul. 2021. Climate change information for regional impact and for risk assessment. Pages 1767-1926 in V. Masson-Delmotte, P. Zhai, A. Pirani, S. L. Connors, C. Péan, S. Berger, N. Caud, Y. Chen, L. Goldfarb, M. I. Gomis, M. Huang, K. Leitzell, E. Lonnoy, J. B. R. Matthews, T. K. Maycock, T. Waterfield, O. Yelekçi, R. Yu, and B. Zhou, editors. Climate change 2021: the physical science basis. Contribution of Working Group I to the Sixth Assessment Report of the Intergovernmental Panel on Climate Change. Cambridge University Press, Cambridge, United Kingdom and New York, NY, USA.

Reuters. 2021. Hundreds of buildings burned around Tigray town, research group says. https://www.reuters.com/article/us-ethiopia-conflict-fires-idUSKBN2AP196

Schroeder, W., P. Oliva, L. Giglio, and I. A. Csiszar. 2014. The New VIIRS 375 m active fire detection data product: algorithm description and initial assessment. Remote Sensing of the Environment 143:85-96. https://doi.org/10.1016/j.rse.2013.12.008

Sibhatu, K. T., and M. Qaim. 2017. Rural food security, subsistence agriculture, and seasonality. PLoS ONE 12(10):e0186406. https://doi.org/10.1371/journal.pone.0186406

Stevens, K., L. Campbell, G. Urquhart, D. Kramer, and J. Qi. 2011. Examining complexities of forest cover change during armed conflict on Nicaragua’s Atlantic Coast. Biodiversity and Conservation 20:2597-2613. https://doi.org/10.1007/s10531-011-0093-1

The Guardian. 2022. Tigray still without aid eight days after deal to end Ethiopia’s blockade. https://www.theguardian.com/world/2022/nov/10/tigray-without-aid-eight-days-after-deal-end-ethiopia-blockade

United Nations High Commissioner for Refugees (UNHCR). 2022. Ethiopia emergency situation, regional update #27. https://reporting.unhcr.org/document/1304

United Nations Office for the Coordination of Humanitarian Affairs (OCHA). 2021. Desert locus situation update 23 September 2021. https://reliefweb.int/report/ethiopia/desert-locust-situation-update-23-september-2021

United Nations Office for the Coordination of Humanitarian Affairs (OCHA). 2023. Ethiopia situation report, updated 2nd June 2023. https://reports.unocha.org/en/country/ethiopia/

van Breugel, P., I. Friis, S. Demissew, J.-P. B. Lillesø, and R. Kindt. 2016. Current and future fire regimes and their influence on natural vegetation in Ethiopia. Ecosystems 19(2): 369-386. https://doi.org/10.1007/s10021-015-9938-x

Vanden Bempt, T., S. Annys, E. Negash, R. Ghekiere, and J. Nyssen. 2021. Tigray: one year of conflict—casualties of the armed conflict, 2020–2021—Tigray (Ethiopia). Ghent University, Department of Geography & London (U.K.): every casualty counts. https://www.researchgate.net/publication/356635731_Tigray_one_year_of_conflict_-_Casualties_of_the_armed_conflict_2020-2021_-_Tigray_Ethiopia

Verbesselt, J., A. Zeileis, and M. Herold. 2012. Near real-time disturbance detection using satellite image time series. Remote Sensing of Environment 123:98-108. https://doi.org/10.1016/j.rse.2012.02.022

Woods, K. M., P. Wang, J. O. Sexton, P. Leimgruber, J. Wong, and Q. Huang. 2021. Integrating pixels, people, and political economy to understand the role of armed conflict and geopolitics in driving deforestation: the case of Myanmar. Remote Sensing 13(22):4589. https://doi.org/10.3390/rs13224589

World Food Programme (WFP). 2022. Emergency food security assessment Tigray region, Ethiopia. https://www.wfp.org/publications/tigray-emergency-food-security-assessment

Zewdie, W., E. Csaplovics, and L. Inostroza. 2017. Monitoring ecosystem dynamics in northwestern Ethiopia using NDVI and climate variables to assess long term trends in dryland vegetation variability. Applied Geography 79:167-178. https://doi.org/10.1016/j.apgeog.2016.12.019