A damage-based crop insurance system for flash flooding: a satellite remote sensing and econometric approach

This research attempted to develop a damage-based crop insurance model using satellite remote sensing imagery and field surveys of flash flood regions in Bangladesh. A normalized difference water index (NDWI) was prepared from Landsat-8 Operational Land Imager (OLI) data, and inundated areas were delineated using a very high-resolution land use/land cover (LULC) map of a Survey of Bangladesh (SoB) to generate an accurate damage map for boro rice. The boro rice area damaged by flash floods in 2017 was estimated for the three most vulnerable flash flood-prone areas and sorted into high, moderate and marginal damage categories. The results of the accuracy assessment showed excellent classification performance for all classes of LULC. Analysis of the classification of the damaged area showed that the Gowainghat (44.60%) and Kulaura (69.80%) subdistricts were in the highest marginal damage category and that Tahirpur (52.92%) was in the moderate damage category. The future value of expected losses was calculated to be $536.25, $442.00, and $416.00 per hectare (ha) for high, moderate and marginal damage areas, respectively. Moreover, findings concerning the damaged-based crop insurance premium rate suggested that the higher the coverage levels were, the lower the insurance premium, and the lower the damaged class was, the lower the insurance premium rate. The lowest insurance premium rate was observed for areas with high coverage and moderate damage and was $23.82/ha, and the highest insurance premium rate was observed for areas with marginal coverage and high damage and was $39.49/ha. Evidence from the classification of damaged areas and from regression findings suggests that farmers' socioeconomic features and environmental awareness (occupation, educational level, total income, bank account ownership and awareness of climate change) are relevant to the decision to adopt a damaged-based crop insurance system. The overall results show an empirical model for flash flood occurrence over both the temporal and spatial domains, which offers an effective measure for adopting a damaged-based crop insurance model.