Optimal design of surface networks for observation of soil moisture

By analyzing in situ soil moisture data, we show that soil moisture variability consists of two components, one of which is related to large-scale atmospheric forcing, and the other related to small-scale land surface variability and hydrologic processes. We use empirically estimated spatial autocorrelation functions for Illinois to estimate errors of spatial averaging of soil moisture observations, using the method of statistically optimal averaging of meteorological fields. The estimated dependence of the root-mean-square errors of averaging on the soil moisture station network density can be used to analyze existing observational networks and for designing new ones. For the application of providing information on a regular grid for numerical models of weather and climate, we show that the new, relatively high density networks of soil moisture observations in Oklahoma, may not provide estimates with very much more accuracy than the relatively low density currently operational network in Illinois. This prediction must be tested when we receive sufficiently long time series of observations from Oklahoma.