Abstract
A soil sensor, which consisted of a spectrophotometer using band pass filters, working resistance detectors using load cells, and a roughness meter using a laser distance meter, was assembled to estimate soil moisture content with the PLS method. The characteristics of the output of each sensor in the frequency region were examined. The data of all sensors from the field experiment were collected at the same place on different days. Accuracy of estimation was improved by combining the data with light reflection spectra, working resistances and cut-surface roughness. The root mean square error of predicting soil moisture content estimated by using all data was 2.0% w.b., an improvement of 0.5% w.b. from that estimated by each data, and 4.8 times larger than the standard deviation measured by the oven dry method.
