Sensitivity analysis of stream and river runoff to spatial rainfall variability using dense rain gauge observations

Abstract:

Precipitation input is crucial in hydrological modeling, especially for heavy precipitation events. Therefore, we studied the impact of station density and interpolation scheme to runoff simulations of short and long duration heavy precipitation events. The WegenerNet, a highly dense meteorological station network (about 1 station per 2 km²) in the southeastern alpine foreland of Austria that partially covers the Styrian Raab catchment (986 km²), gives us the opportunity to this event-based study. We run our simulations with the calibrated and validated process-oriented hydrological model WaSiM for the Raab catchment and selected small sub-catchments (areas of 10 km² to 50 km²) covered by the WegenerNet observations. A set of six station densities is used, starting with the ordinary precipitation stations, 5-stations and 8-stations case (mean station distance around 10 km), plus adding stations from the WegnerNet up to 16, 32, 64 and 158 stations (mean station distance 1.4 km). The 158 stations case is our reference, which was also used for model calibration. Additional, to include the influence of the interpolation scheme, we studied the precipitation interpolation with Inverse Distance Weighting and a weighting power of 2 and 3 (IDW2 and IDW3), as well as the Thiessen polygon interpolation (TP).

Our results from analyzing the variation of the simulated runoff peaks suggest that on the one hand a strong catchment and event dependency exists, which is especially pronounced for short duration events and small sub-catchments. On the other hand, the mean over all catchments of the long duration heavy precipitation events shows a “threshold” with 16 stations (mean station distance around 6 km) after which no big improvement exists. A lower number of stations shows a stronger effect to the interpolation scheme, but the station density is clearly more relevant than the interpolation scheme.

Overall, the study suggests to be aware of the uncertainty of station density when modelling runoff, especially for short duration heavy precipitation events. Long duration heavy precipitation events and bigger catchments are less vulnerable to the precipitation input. To be able to include these effects for precipitation uncertainty studies, we suggest to include the analysis of precipitation datasets in combination with rainfall-runoff modelling.