STOCHASTIC ESTIMATION OF PLANT-AVAILABLE SOIL-WATER UNDER FLUCTUATINGWATER-TABLE DEPTHS

Preservation of native valley-floor phreatophytes while pumping ground water for export from Owens Valley, California, requires reliable pred ictions of plant water use. These predictions are compared with stored soil water within well field regions and serve as a basis for managin g groundwater resources. Soil water measurement errors, variable recha rge, unpredictable climatic conditions affecting plant water use, and modeling errors make soil water predictions uncertain and error-prone. We developed and tested a scheme based on soil water balance coupled with implementation of Kalman filtering (KF) for (1) providing physica lly based soil water storage predictions with prediction errors projec ted from the statistics of the various inputs, and (2) reducing the ov erall uncertainty in both estimates and predictions. The proposed KF-b ased scheme was tested using experimental data collected at a location on the Owens Valley floor where the water table was artificially lowe red by groundwater pumping and later allowed to recover. Vegetation co mposition and per cent cover, climatic data, and soil water informatio n were collected and used for developing a soil water balance. Predict ions and updates of soil water storage under different types of vegeta tion were obtained for a period of 5 years. The main results show that : (1) the proposed predictive model provides reliable and resilient so il water estimates under a wide range of external conditions; (2) the predicted soil water storage and the error bounds provided by the mode l offer a realistic and rational basis for decisions such as when to c urtail well field operation to ensure plant survival. The predictive m odel offers a practical means for accommodating simple aspects of spat ial variability by considering the additional source of uncertainty as part of modeling or measurement uncertainty.