Estimation of soil and crop residue parameters using AVIRIS-NG hyperspectral data

Detailed ground cover information and efficient modelling approaches are needed for estimating soil properties from hyperspectral remote sensing (HRS) data. With the objective to estimate both soil and crop residue (CR) parameters using HRS data from the airborne visible-infrared imaging spectrometer-next generation (AVIRIS-NG) sensor, soil and CR samples were collected from 101 locations in the Western Catchment of Chilika lagoon, India. Nonlinear unmixing and two chemometric models were examined for estimating basic soil properties, nutrient contents, soil and CR wetness, and masses of fresh crop residue (FCR) and dry crop residue (DCR). Both soil and CR parameters showed wide variation with FCR and DCR masses varying from 0 to 13.33 Mg ha(-1) and 0 to 10.16 Mg ha(-1), respectively. Soils were relatively dry with an average gravimetric water content (?(g)) of 14% whereas CR moisture content (?(c)) ranged from 2.4 to 93%. Estimated coefficient of determination (R-2) values in the validation datasets varied from 0.51 for soil base saturation to 0.91 for exchangeable Mg+2 using soil spectra obtained through linear polynomial unmixing of AVIRIS-NG spectra. The R-2 values were more than 0.70 for clay content, soil organic carbon (SOC), cation exchange capacity (CEC), and several exchangeable cations although the CR parameters showed lower R-2 values than soil parameters. With the chemometric models calibrated, high spatial resolution maps for soil and CR parameters were generated, which offer a continuous mapping capability for such parameters. Extensive soil property data also allowed us to evaluate SOC sequestration capability of different land use systems. A critical value of 1/25 for the SOC/Clay content ratio was observed for most of our agricultural land uses and degraded landscapes showed lower SOC/Clay ratios. Thus, high spatial resolution soil and crop residue parameters may be accurately assessed for large areas with multiple land use and soil cover conditions.

Automated summary

Detailed ground cover information and efficient modelling approaches are needed for estimating soil properties from hyperspectral remote sensing (HRS) data. In this study, soil and crop residue samples were collected from 101 locations in the Western Catchment of Chilika lagoon, India, and nonlinear unmixing and two chemometric models were examined. The results showed that high spatial resolution soil and crop residue parameters can be accurately assessed for large areas with multiple land use and soil cover conditions using these models.