Accuracy of spatio-temporal RARX model predictions of water table depths

摘要

Time series of water table depths (H_t) are predicted in space using a regionalized au-toregressive exogenous variable (RARX) model with precipitation surplus (P_t) as input variable. Using their physical basis, RARX model parameters are guessed from auxiliary information such as a digital elevation model (DEM), digital topographic maps and soil profile descriptions. In the 'direct' method (DM) P_t is transformed into predictions of H_t using the guessed RARX model parameters. In the 'indirect' method (IM) these predictions are corrected for observed systematic errors. In the Kalman filter methods (KF) the parameters of regionalization functions for the RARX model parameters are optimized conditional to observations on H_t. External drift kriging and simple kriging with varying means are applied as regionalization functions, using guessed RARX model parameters or DEM data as secondary variables. Both actual H_t and H_t which occurs in a given hydrological regime are predicted in a study area of 1375 hectares. The prediction performance is tested by cross-validation using observed values of H_t in 27 wells which are positioned following a stratified random sampling design. IM performs significantly better with respect to systematic errors than the alternative methods in predicting H_t for a given hydrological regime. KF methods perform better than both DM and IM in predicting the temporal variation of actual H_t, as is indicated by significantly lower random errors.