City-scale agricultural land use detection using soil moisture derived from GF-1 images

Agricultural land use is essential for prosperity and sustainability of a city. However, the performance of detecting agricultural land use based on vegetation parameters will be substantially reduced when the vegetation changed in type, quantity and condition. This study believed that soil moisture, a fundamental soil parameter, is able to detect agricultural land use. A method was proposed to evaluate city-scale agricultural land use detection capability based on three temporal statistics of soil moisture in both local and global perspectives. Using GF-1 images as a data source, an experiment in Wuhan city was performed to discuss the city-scale agricultural land use detection capability based on soil moisture. Soil moisture were inverted by the modified perpendicular drought index derived from GF-1 images. Under the visual and quantitative comparisons, it was found that the temporal mean of soil moisture performed much better than the temporal standard deviation of soil moisture and the temporal coefficient of variance of soil moisture. The overall accuracy and Kappa coefficient of the temporal mean of soil moisture reach 91.7% and 0.81, respectively, demonstrating its great capability of detecting city-scale agricultural land use. A global analysis is implemented by regression analyses with true zonal agricultural statistics, showing a high global accuracy in agricultural land use detection.