Abstract:
Recently, there has been an increased interest in developing new methods to measure the impact of complex humanitarian interventions in hard-to-reach areas to help guide policy decisions. Quantifying agricultural interventions post-conflict remains a challenge. The advancement in Earth observations and remote sensing techniques can provide a timely and precise evaluation of agricultural activities and production in such settings. Little research has been done on the potential use of remote sensing for impact evaluation of agricultural interventions in humanitarian settings. Here, we evaluate a complex humanitarian intervention that aims at strengthening agricultural activity in conflict affected Syria. The overall objective of this study is to develop a framework for evaluating the effectiveness of agricultural interventions in a conflict setting using remote sensing and machine learning techniques. We use a combination of vegetation indices which were normalized by rainfall for three identified periods: pre-conflict, conflict, and post-intervention, and an unsupervised machine learning classifier. Examination of the multi-temporal time series of anomalies and irrigated agriculture revealed distinct patterns in active agricultural areas during the three defined periods of study. The results showed an overall improvement in vegetation and irrigated areas in intervention villages post-intervention. Remote-sensing analysis showed that rehabilitation of irrigation systems significantly increased irrigated areas in some villages like pre-conflict levels.
