HR: 13:55h
AN: H12F-02    [PDF]
TI: Estimation of Drainage and Evapotranspiration from Time Series of Soil Moisture, Potential Evaporation, and Precipitation
AU: * Salvucci, G D
EM: gdsalvuc@bu.edu
AF: Boston University, 675 Commonwealth Ave., Boston, MA 02215 United States
AU: Gioioso, M
EM: mgioioso@bu.edu
AF: Boston University, 675 Commonwealth Ave., Boston, MA 02215 United States
AB: A previous study demonstrated that the dependence of soil water outflow on soil moisture can be estimated by averaging precipitation conditioned on soil moisture. The methodology is non parametric and relies only on the assumed stationarity of the soil moisture time series. Here we present a method for partitioning out the evapotranspiration component of total outflow. One goal is to structure the model with as few assumptions about model form as possible. for example we set evapotranspiration efficiency to increases monotonically with moisture and to be concave down, while the net drainage (capillary rise to or percolation from the root zone) is made to depend on moisture in a concave upward fashion. The functions used to represent these behavior are piecewise continuous polynomials or line segments. After generating a set of feasible partitions using a linear programming technique, we evaluate the relative likelihood of each by estimating the entropy of the time series of soil water storage that results from integrating the fluxes. We show that the entropy of the series is proportional to the likelihood that the increments that make it up come from a stationary process, and use this as a basis for model selection. We also estimate the growth of variance of the time series, and decompose this into an equilibrium process (that saturates with time due to a negative correlation among increments) and an error process which (for white noise model, measurement and sampling errors) leads to a random walk term. A unique feature of the method is that it does not fit model predictions to soil moisture, but instead evaluates the stationarity of the running series of soil water storage values implied by the partitioning. Because of this feature the method can be driven with indices of soil moisture (like brightness temperatures) rather than site-specific water contents.
DE: 1836 Hydrologic budget (1655)
SC: Hydrology [H]
MN: 2003 Fall Meeting