A uniform versus an aggregated water balance of a semi-arid watershed

Hydrologists have long struggled with the problem of how to account for the effects of spatial variability in precipitation, vegetation and soils. This problem is particularly acute in snow-fed, semi-arid watersheds, which typically have considerable variability in snow distribution and vegetation communities on scales much smaller than that addressed by most hydrological modelling. In this study, two approaches were used to compute a water balance using two years of data collected at the Upper Sheep Creek Watershed, a 26-ha semi-arid mountainous sub-basin within the Reynolds Creek Experimental Watershed in south-west Idaho, USA. The first water balance approach (uniform approach) assumed that the entire watershed was homogeneous; the second approach computed a partial water balance for each of three landscape units and then computed an aggregated water balance for the watershed. Runoff and change in groundwater storage were not distinguishable between landscape units; thus, the only difference between the two approaches was in the estimation of the two major components, precipitation and evapotranspiration (ET). Precipitation, which occurs predominantly as snow, was measured within each landscape unit directly and adjusted for drifting. ET was estimated using the simultaneous heat and water model (SHAW) and validated with measurements from Bowen ratio instruments. Precipitation input for the two years was approximately 480 and 700 mm, respectively; ET was approximately 450 and 410 mm, respectively. The water balance for the aggregated approach had a discrepancy of -17 and 55 mm, respectively for the two years, while the uniform approach was within 42 and 86 mm, respectively. (Negative values indicate more estimated outflow than inflow.) The differences in precipitation estimates for the two approaches were greatest for the second year owing to more variability across the watershed, which the uniform approach did not adequately address. The largest difference between the aggregated and uniform approach for both years was the estimated ET. This was attributed to the inability of the uniform approach to associate areas of the watershed having more vegetation and leaf area with areas having soil water available for transpiration. Differences in ET estimates for the two approaches were least during the second year when water was less limiting and potential ET was less. This suggests that ET can be aggregated more easily when water is not limiting using the average leaf area index independently of the spatial variability in leaf area. (C) 1998 John Wiley & Sons, Ltd.