Crop area change in the context of civil war in Tigray, Ethiopia

Abstract

Abstract Smallholder, rain-fed agriculture has been practiced in Tigray, Ethiopia for thousands of years, so farmers have experience with natural disturbances. However, civil war began in November 2020 and disrupted the agricultural system through theft and destruction of farming implements, livestock, and crops, and threats towards human activity, impacting farmers’ ability to work their land. To investigate effects on agricultural activity we used remote sensing data and machine learning methods to map cropped area across Tigray from 2017 to 2022. Specifically, spectral mixture analysis was applied to Sentinel-2 data to produce green vegetation, non-photosynthetic vegetation, soil, and shade endmember fractions. Monthly medians of these fractions, along with reference data generated by manual interpretation of very high spatial resolution data, were used to drive random forests-based classifications of crop/no-crop for each year. Initially we used a greenness threshold to distinguish between active and abandoned fields for a given year, but when rainfall is adequate, fields abandoned due to conflict can green up with weeds rather than crops, leading to false positive crop detections. In the spring, abandoned fields have a bright soil crust due to a lack of plowing, so these fields were removed from the crop reference data if the March soil endmember fraction anomaly was greater than 0.15. Overall accuracies of the crop/no-crop maps ranged from 80% to 90% for the different districts. Producer’s/user’s accuracies for the crop class ranged from 55%–80%/69%–90%. In 2021, crop area declined by 29% and 20% in West and Northwest Tigray, respectively, corresponding with reports of intense conflict there. The rest of Tigray showed a mix of smaller increases and decreases, indicating more resilience to the regional conflict. Finally, in 2022 we found increases in cropped area relative to 2021, for all districts except West Tigray, indicating recovery except for the areas where conflict was most severe.