Fourier-Transform Infrared Spectral Inversion of Soil Available Potassium Content Based on Different Dimensionality Reduction Algorithms

Estimating the available potassium (AK) in soil can help improve field management and crop production. Fourier-transform infrared (FTIR) spectroscopy is one of most promising techniques for fast real-time analysis AK content. However, successful estimation content by FTIR depends on proper selection appropriate spectral dimensionality reduction techniques. To magnify subtle signals concerning understanding characteristic wavelengths content, a total 145 samples were collected an agricultural site located southwest part Sichuan, China, three typical methods—the successive projections algorithm (SPA), simulated annealing (SA) competitive adaptive reweighted sampling (CARS)—were adopted to select variable. Then, partial least squares regression (PLSR) was utilized establish inversion models incorporating optimal set variables extracted different algorithms. The accuracy each model tested based coefficient determination (R2), root mean square error (RMSE) absolute value (MAE), contribution explored. results show that: (1) application useful technique isolating specific components multicomponent spectra, as such powerful tool expand predicted potential Compared with SA CARS algorithms, SPA more suitable inversion. (2) showed that mainly around 777 nm, 1315 1375 1635 1730 nm 3568–3990 nm. (3) Comparing performances models, SPA–PLSR (R2= 0.49, RMSE = 22.80, MAE 16.82) superior SA–PLSR CARS–PLSR which has certain guiding significance rapid detection