PRECISION AGRICULTURE APPLICATIONS OF AN ON-THE-GO SOIL REFLECTANCE SENSOR

This work demonstrates the utility of an on-the-go optical reflectance sensor in mapping soil attributes in central Kansas (USA). The sensor measures the reflectance of the soil at wavelengths ranging from 950nm to 1650nm and at a depth of approximately 70 mm below the soil surface. The sensor was used to map eight fields on approximately 20 meter transects. Once each field was mapped, the reflectance data was compressed using principal components analysis (PCA) and then clustered using a fuzzy c-means algorithm. A fuzzy logic algorithm was used to determine representative sample locations within each cluster and samples were acquired for laboratory analysis. Once this process was completed for all eight fields, calibrations for various soil attributes were created using partial least squares regression (PLS). Validation techniques indicated that the calibrations provided reliable prediction of organic matter, pH buffering capacity, and Mehlich 1 phosphorus. Less accurate, but potentially usable calibrations were obtained for pH and Mechlich 3 phosphorus. These calibrations were then applied to the complete set of field spectra in order to create soil attribute maps. In turn, these data layers were used to create maps to control the rate of applied lime and fertilizer.