Search

International audience; The present study assessed the visible and short wave infrared (VNIR-SWIR) laboratory spectroscopy coupled random forest regression (RF) technique for predicting soil properties in the southern Karnataka Plateau, India. The spectral data acquired for about 228 profile samples were used to predict key soil properties. The RF model fits well for the spectral prediction of clay (R-2 = 0.65), sand (R-2 = 0.60), cation exchange capacity (R-2 = 0.74), field capacity (R-2 = 0.65) and permanent wilting point (R-2 = 0.72). Wherein soil organic carbon was poorly predicted with an R-2 of 0.22 and RPD of 1.2 due to its lower content and narrow range (0.8 to 20 g kg(-1)). The spectral assessment by PCA showed that the first (50%) and third (34%) components had high spectral variation and significantly correlated with soil properties such as pH, CEC, clay, FC, and PWP related to wavelengths indicating clay minerals and iron oxides. However, the second component had less spectral variation (13%) that is related to wavelengths indicating various organic components and correlated well with SOC. Thus, the VNIR-SWIR spectroscopy could be a suitable supplementary method for rapidly predicting soil properties related to clay minerals and iron oxides.

The strategy for increasing crop production is either based on bringing new lands under cultivation or through intensive cultivation with improved management practices. Nevertheless, both the strategies require constant monitoring of soil fertility status to avoid nutrient deficiency that is one of the major constraints for crop production and it is a decisive tool for farmers and policy planners in soil nutrient management. The integrated watershed management programme called SUJALA by the watershed department, Karnataka, facilitated to diagnose the soil nutrient status in rainfed areas of Karnataka by collecting soil samples at grid interval from the farmer’s field. A total of 34,011 georeferenced composite samples (0–15 cm) were collected at a 325 × 325 m grid interval to analyse the fertility status of soils. The samples were aggregated under different agro-climatic zones of Karnataka to assess the soil fertility status under each zone. The results showed that per cent sample of low soil organic carbon (