Inferring time-varying recharge from inverse analysis of long-term water levels

Abstract

Water levels in alluvial aquifers of the Southwest typically vary in response to time varying rates of mountain-front recharge. This suggests the possibility of inferring time varying recharge rates on the basis of long-term water-level records. Presumably, these recharge rates depend on variations in precipitation rates due to known climate cycles such as the El Nino/Southern Oscillation index and the Pacific Decadal Oscillation. This investigation examined the inverse application of a one-dimensional analytical model for periodic flow [Townley, 1995] to estimate periodic recharge variations on the basis of water-level variations in long-tenn water-level records. Time-varying water level records at various locations along a flow path were obtained by simulation of ground-water flow in numerical models of idealized basins. Time-varying recharge was represented as sinusoidal recharge of a single period or as a composite of multiple periods of length similar to known climate cycles. Periodic water level components, reconstructed using singular spectrum analysis (SSA), were used to calibrate the analytical model to estimate each recharge component. The results demonstrate that periodic, time-varying recharge rates can be estimated from hydrologic time series such as long-term water level records. The results also demonstrate that periodic recharge estimates were most accurate in basins with nearly uniform transmissivity, and that the accuracy of the recharge estimates depended on the monitoring well locations.