Peter Ahnert

A procedure for the objective reduction of a rain gauge network is developed and applied for the Susquehanna River Basin (SRB) in the United States. The procedure utilizes evaluation of the theoretical error variance associated with precipitation analysis using a variant of the National Weather Service’s (NWS) Multisensor Precipitation Estimator (MPE). The uncertainty analysis is carried out for 16 different combinations of the precipitation accumulation period, season, magnitude, and areal extent, and use or nonuse of the Flash Flood Potential Index (FFPI), which is used as a proxy for the spatially varying runoff ratio. To estimate the statistical parameters of the procedure, the historical archive of the MPE products operationally produced by the Middle Atlantic River Forecast Center (MARFC) was used. The marginal value of each rain gauge in the Susquehanna Flood Forecasting and Warning System (SFFWS) network to the parent network is assessed by calculating the increase in the theoretical error variance in radar-gauge precipitation analysis over the SRB following hypothetical removal of the gauge. The parent network is made of 73 gauges in the SFFWS network plus 120 high-quality hourly and subhourly rain gauges within and in the vicinity of the SRB. The results show that significant variability exists in the rankings of the marginal value of the SFFWS gauges among the 16 cases considered. The most significant source of this variability is the seasonal variation in the spatial structure of precipitation. The second most significant source is the use or nonuse of the FFPI. Given the significant sensitivity to these and possibly other factors, one may not expect a unique solution for optimal network reduction. For robust decision making, an ensemble of solutions should be considered that reflects the range of variability in such factors.