Rice Yield Loss Area Assessment from Satellite-derived NDVI after Extreme Climatic Events Using a Fuzzy Approach

Abstract

Extreme weather events pose high risks for agricultural production and cause yield losses in South and Southeast Asia. Tropical cyclones frequently cause significant yield losses in Bangladesh. In this regard, a fuzzy approach for satellite-derived normalized difference vegetation index was used to classify rice yield losses in a coastal region of Bangladesh adjacent to the Bay of Bengal. We used different fuzzy membership functions and overlaid a fuzzification gamma to calculate the expected crisp set to develop the classifiers for yield losses. There were five classes of yield loss: marginal, slight, moderate, very, and extreme. The natural breaks (Jenks) method was used to classify losses as marginal on 461 ha (1.5% of total), 2661 ha (8.6%), moderate 11811 ha (38.2%), very 5814 ha (18.8%), and extreme 10160 ha (32.9%). Field validation identified 29.5% of the reference yield information points in the moderate yield loss class, and 45.2% in extreme. These similarities indicate that the method can be used to estimate yield losses in South and Southeast Asian areas affected by cyclones.