Multi-temporal mapping of flood damage to crops using sentinel-1 imagery: a case study of the Sesia River (October 2020)

Abstract

ABSTRACT Monitoring large-scale flood damage can be complicated and costly. Damages caused by floods affect also the agricultural sector. Permanence, height and quantity of stagnant water can significantly influence crop yield. Many studies exploit satellite data to map flooded areas, but only a few are focused on the timing of water persistence. This work refers to the river Sesia flooding event which occurred on 3 October 2020 in Northwest Italy with the aim of detecting damages to local crops. The analysis was based on Sentinel-1 data processed by Google Earth Engine platform. In particular, the Otsu’s method was applied to test the difference between pre- and post-event images. Areas that were mapped as flooded were successively analysed to estimate local water persistence: specifically, 1-2-6 days after the event. According to the available Corine Land Cover 2018 dataset, it was found that flood mainly affected agricultural areas (about 3288 ha). Since damage also relies on water persistence, a focus area was selected to test the effectiveness of S1 multi-temporal in mapping its distribution. Results show that only 3.5% of the agricultural fields in the focus area remained underwater for at least 6 days and 69% for only 1 day.