Quantifying the water balance in a planar hillslope plot: Effects of measurement errors on flow prediction

We examine comprehensively the water balance of an experimental plot in Seattle, WA, USA to determine the magnitude of hydrologic measurement errors and their effects on flow prediction. The plot represents a fully contained planar hillslope with known size and boundary conditions. Precipitation input to the plot is measured with a series of rain gauges installed above the ground surface and buried with rims at the ground surface level. Output from the plot via subsurface flow is measured by continuously recording tipping buckets. Evapotranspiration (ET) output is characterized using weather station and energy budget measurements. Analyses of water balance measurements show that the best precipitation measurements come from buried rain gauges, particularly simple funnel collectors, whereas surface rain gauges under-report rain by an average of 8%. ET is the most uncertain component of the water balance, with land surface energy budget, Penman–Monteith equation, and a calibrated empirical equation predicting seasonal ET rates that differ from one another by up to 18%. Standard un-calibrated empirical ET equations prove unsuitable for the plot site. To evaluate how differences in measured precipitation and evapotranspiration affect flow prediction, we run continuous flow simulations with the physically-based variably saturated flow model, HYDRUS-2D, for four scenarios, each with a different combination of forcing data (precipitation and reference ET time series). Results show that biases in the atmospheric forcing data propagate into flow predictions, causing biases in discharge predictions of up to 22%. Some scenarios show an apparent mass balance that results from compensating errors in precipitation and ET. Simulations demonstrate that errors in forcing data are not easily discernable from model performance metrics and highlight the importance of analyzing the water balance at multiple time scales to identify sources of bias in water balance predictions.