ABSTRACT

A pixel-based spectral matching method using enhanced vegetation index (EVI) time series from GaoFen-1 (GF-1) Satellite was developed based on spectral similarity value (SSV) and iterative thresholding to map irrigated areas for varying and scattered croplands at high resolution in order to improve understanding of dynamic irrigation pattern. The approach was tested in a typical irrigation district in Northwest China, resulting in 16-m irrigated area maps of the major crops (wheat and maize) for 2015–2018. The results showed an overall accuracy of 92.67% with a Kappa Coefficient of 0.89, producer’s accuracy of 97.42%, and user’s accuracy of 92.65% for the irrigated area. This method does not require a large number of ground-truth data as training samples, resulting in fast calculation speed, which is different from other supervised classification methods. After obtaining the standard spectrum of the study area, the irrigated areas can be continuously mapped for multi-years. Comparing with the existing global products, these high-resolution maps can depict small and fragmented irrigated croplands effectively with higher accuracy, which benefit local water management, food security analysis, and drought impact assessment, providing the prototype for the regions with fragmented croplands which requires high-resolution products.