A spatial frequency/spectral indicator-driven model for estimating cultivated land quality using the gradient boosting decision tree and genetic algorithm-back propagation neural network

Cultivated land quality (CLQ) is related to national food security. Rapid and high-precision monitoring of CLQ is crucial for the sustainable development of agriculture. However, current satellite image-based evaluation methods that only consider the crop's spatial spectrum characteristics in the key growth stages cannot accurately estimate CLQ. This study proposes a new method based on time-series spectral data of crop growth to improve the accuracy of CLQ estimation. This study was conducted in the Conghua District of Guangzhou, Guangdong Province, China. The results showed that seven spectral indicators were determined as the optimal indicators based on the gradient boosting decision tree (GBDT) and variance inflation factor (VIF). And the genetic algorithm-back propagation neural network (GA-BPNN) model provided more accurate CLQ estimates than the partial least squares regression (PLSR) model, indicating a nonlinear relationship between CLQ and the indicators. In addition, the GA-BPNN model with a normalized root mean square error (NRMSE) of 9.91% demonstrates the excellent potential for mapping CLQ over large areas. The model based on the seven optimal indicators of crop phenology provided higher accuracy than the GA-BPNN model based on the normalized difference vegetation index (NDVI) indicators in the spatial domain, significantly decreasing the NRMSE of the CLQ estimates by 3.17%. This further implied that the spectral indicators in the spatial frequency domain can improve the accuracy of estimating CLQ.