Evaluation of Reflectance Methods for Soil Organic Matter Sensing

Multivariate calibration methods, including stepwise multiple linear regression, principal components regression, and partial least squares regression, were applied to soil color and spectral reflectance data to predict the organic matter content of Illinois soils. Data were obtained with a visible/near infrared (NIR) spectro-photometer and a commercial colorimeter on a calibration set consisting of 30 Dlinois mineral soils at 0.033 MPa and 1.5 MPa moisture tension levels. Good correlation with soil organic matter (r2 = 0.91, standard error of prediction = 0.34% organic matter) was obtained when using as few as 12 NIR reflectance data points in the range from 1720 to 2380 nm. Color coordinate and visible reflectance data provided considerably poorer predictions of soil organic matter content.