Estimation of Soil Nutrient Content Using Hyperspectral Data

Soil nutrients play a vital role in plant growth and thus the rapid acquisition of soil nutrient content is great significance for agricultural sustainable development. Hyperspectral remote-sensing techniques allow quick monitoring nutrients. However, at present, obtaining accurate estimates proves to be difficult due weak spectral features low accuracy estimation models. This study proposed new method improve estimation. Firstly, characteristic variables, we employed partial least squares regression (PLSR) fit degree select an optimal screening algorithm from three algorithms (Pearson correlation coefficient, PCC; absolute shrinkage selection operator, LASSO; gradient boosting decision tree, GBDT). Secondly, linear (multi-linear regression, MLR; ridge RR) nonlinear (support vector machine, SVM; back propagation neural network with genetic optimization, GABP) 10-fold cross-validation were implemented determine most model estimating total nitrogen (TN), phosphorus (TP), potassium (TK) contents. Finally, was used map TK regional scale using component variables retrieved by fully constrained (FCLS) based on image HuanJing-1A Imager (HJ-1A HSI) Conghua District Guangzhou, China. The results identified GBDT-GABP observed as TN ( 0.69, root mean square error (RMSECV) 0.35 g kg−1 ratio performance interquartile range (RPIQ) 2.03) TP 0.73, RMSECV 0.30 RPIQ = 2.10), LASSO-GABP proved estimations 0.82, 3.39 3.57). Additionally, highly LASSO-GABP-estimated (R2 0.79) reveals feasibility retrieve scale.