An Investigation of Soil Moisture Dynamics Using FLUXNET Data

Soil moisture data, obtained from four FLUXNET sites in the US, were examined using an ecohydrological framework. Sites were selected for the analysis to provide a range of plant functional type, climate, and soil grain size distribution. Data at a selected site included at least two years of measurements of volumetric soil water content, air temperature, precipitation, atmospheric pressure, net radiation, and latent heat flux. The Rosetta database program, based on pedo-transfer functions, was used to generate water retention curves from site soil grain size distributions. Using these curves and plant parameters found in the literature, ranges for each critical soil moisture point were determined. For all sites, the hydroscopic point (Sh) and wilting point (Sw) had the smallest range, while more uncertainty was associated with the stress point (S*) and field content (Sfc). Soil moisture trends revealed the importance of measuring water content at several depths throughout the rooting zone; soil moisture at the surface (above 10 cm) was around 20 to 30 percent less than that at 50 to 60 cm. Frequently, the surface soil moisture would fall below Sw while remaining between S* and Sw at deeper intervals. While daily variability of soil moisture was high due to the timing of precipitation events, yearly variability was lower than anticipated. However, a broader range of years should be examined to confirm this finding. A steady state soil moisture dynamics model was used to generate soil moisture probability density functions (pdfs) at each site. The model was altered to accommodate the year-round growing seasons at two of the sites, a compromise between a fully transient model and the typical steady state model. The modeled pdfs were then compared to histograms generated from the measured data. Model accuracy depended heavily on proper parameter selection. Most parameters could be found using available FLUXNET data for the site, however, S* and Sfc were not known with sufficient certainty. A simple inversion technique was used to find these parameters and calibrate the model. The inversion results demonstrated that the commonly used soil matric potential values for finding S* and Sfc may not be appropriate.