On quantifying agricultural and water management practices from low spatial resolution RS data using genetic algorithms: A numerical study for mixed-pixel environment

In this paper, we present a genetic algorithm-based methodology to quantify agricultural and water management practices from remote sensing (RS) data in a mixed-pixel environment. First, we formulated a linear mixture model for low spatial resolution RS data where we considered three agricultural land uses as dominant inside the pixel—rainfed, irrigated with two, and three croppings a year; the mixing parameters we considered were the sowing dates, area fractions of agricultural land uses in the pixel, and their corresponding water management practices. Then, we carried out numerical experiments to evaluate the feasibility of the proposed approach. In the process, the mixing parameters were parameterized by data assimilation using evapotranspiration and leaf area index as conditioning criteria. The soil–water–atmosphere–plant system model SWAP was used to simulate the dynamics of these two biophysical variables in the pixel. The results of our numerical experiments showed that it is possible to derive some sub-pixel information from low spatial resolution data e.g. the existing agricultural and water management practices in a region, which are relevant for regional agricultural monitoring programs.