USING HYPERSPECTRAL REMOTE SENSING TO ESTIMATE SOIL ORGANIC MATTER IN AGRICULTURAL FIELDS

Knowledge of soil properties is important to guide the application rate of fertilization. When the variation of soil properties is available, farm management can be improved so that higher productivity is obtained and the chance of environmental pollution is reduced. Remote sensing has proven to be a rapid and efficient tool to quantify soil properties. This study aimed to examine the capacity of hyperspectral reflectance for mapping soil organic matter content, and this is achieved through partial least squares (PLS) modeling of lab measured spectral reflectance. A total of 60 agricultural soil samples were collected in the Eagle Creek, Cicero Creek and Fall Creek Watersheds, Central Indiana. These soil samples were analyzed for organic matter content ranging from 2.5 to 13%, total phosphorus ranging from 200 to 1500 ppm, total nitrogen ranging from 0.041 to 0.489%, total carbon ranging from 0.77 to 6.37% and spectral reflectance of the samples was measured with an ASD spectroradiometer. The total samples were split into subsets for calibration (45 samples) and validation (15 samples). PLS modeling resulted in a coefficient of determination (r²) between measured and predicted organic matter 0.84, RMSECV 1.41 for calibration and r² 0.78, RMSEP 1.08 for validation. The result implies the potential of this method for remotely estimating organic matter content of agricultural soils.