ERROR ANALYSIS IN ESTIMATING SOIL WATER CONTENT FROM NEUTRON PROBE MEASUREMENTS

Abstract

We present a variance analysis for quantifying the different sources of errors induced on volumetric water content measurements and calculation of soil water storage with the use of a neutron moisture meter in one single access tube. For comparative purposes, we apply the analysis to field data obtained with two different probes. In each case the calibration curve is established by measuring simultaneously and independently neutron count rates and corresponding water contents. Two different approaches are considered, i.e. the unbiased treatment and the biased treatment. The unbiased treatment differs from the biased by its correction for measurement errors using the leastsquare technique. For the site under consideration, we show that the calibration component is the major contribution to the total variance associated with an individual water content estimation. The use of the unbiased statistical treatment notably decreases the total variance. In cases where the calibration curve is established very accurately, the instrument component can be reduced by increasing the number of count replications at each sampling point or the counting time or both. The loss of precision due to using neutron count rate ratios instead of simple count rates is negligible if several count replications are made in a standard medium or if the counting time is long enough. We show that the rule of integration of water content profiles in order to calculate water storage has a great importance: the use of Simpson's rule drastically decreases the associated variance as compared with the classical trapezoidal method.