Applications of Wavelets to the Prediction of Soil Moisture Profiles.

Soil moisture is an important environmental variable that is measured at various temporal and spatial scales. Despite progress in sensor technology and availability, real-time measurements of soil water content at depths below 1.5 meter are rare. Such information is needed to predict transport processes influenced by factors other than conditions at the soil surface. The objective of this work was to predict profiles of soil moisture from information on soil moisture measured at 0.1 m together with vadose zone properties. In 2003, five CS616 probes per site were installed at the Rutgers Agricultural Research & Extension Center (site 1) and Fruit & Ornamental Research Extension Center (site 2) at depths that varied from 0.1 m to 3.0 m and from 0.1 m and 2.7 m, respectively. Detailed particle size distribution and water retention properties were measured on samples from each of the instrumented horizons. Wavelet analysis was used with daily averages of soil moisture to identify and isolate the frequency components that make up the water content signals. Cross-correlation was used to identify frequency-specific lags and to develop transfer functions to predict water content at depth from surface measurements. Wavelet analysis was completed via the cwtft and icwtft functions in Matlab Release 11a. These functions use an FFT algorithm to compute the 1-D continuous wavelet transform and the inverse continuous wavelet transform. The procedure yielded reasonable predictions of water contents at the deepest depths from surface measurements with coefficients of linear regression of 0.65 (site 1) and 0.56 (site 2). The influence of vadose zone composition in the performance of the wavelet approach will be discussed.