Soil organic matter sensing with an on-the-go optical sensor

Abstract

This research was conducted to develop an inexpensive on-the-go optical sensor for soil organic matter (OM) sensing. Diffuse reflectance for 86 soil samples from Kansas and Illinois was measured by a spectrometer in a laboratory. Stepwise multiple linear regression (SMLR) and B -matrix in partial least squares (PLS) were used to determine important wavelengths for soil OM measurement. The wavelengths of 660 and 940 nm, identified by both SMLR and PLS, were used for an optical sensor. The developed optical sensor with dual wavelength was evaluated with dry and wet soils in the lab and the relationship between reflectance and OM showed a coefficient of determination ( R 2 ) as high as 0.91. Gaps between soil and the sensor window reduced the ability to estimate soil organic matter, thus the sensor window should press firmly against soil. In field tests, all fields gave good results, with RPD (ratio of prediction to deviation = standard deviation/root mean square error) of 2 or greater for OM estimation. In comparison with a NIR spectrophotometer shank unit, the optical sensor showed similar results for OM mapping pattern with coefficient of determination as 0.86. The level of agreement between the two maps was 0.56 for overall accuracy and 0.34 for kappa coefficient. Further field tests need to be implemented to evaluate the soil organic matter estimation with the sensor over different types of soils in a wider set of locations.