Impact of basin scale and initial condition on ensemble streamflow forecast uncertainty

The propagation of the uncertainty from quantitative precipitation forecasts (QPFs) to ensemble streamflow forecasts (ESFs) in a hydrometeorological prediction system is affected by two important factors: (i) the physical processes governing the rainfall-runoff transformation in a watershed, and (ii) the basin condition at the beginning of a forecast. In this study, we conduct a set of numerical experiments in the Baron Fork basin and its sub-basins (from ~0.8 to ~800 sq. km) to study the effects of these factors, quantified through the basin area (A) and the antecedent rainfall (AR), on the ESF dispersion, a metric of flood forecast skill. We first analyze the propagation of uncertainty associated with three types of ensemble QPFs –one consistent (observation and ensemble members are equally likely) and the other two affected by different kinds of deficiencies- into ESFs using a verification framework based on rank histogram and continuous ranked probability score. Next, we focus on consistent ensemble QPFs and corresponding ESFs, and we show that (i) for small basins (A< ~180 sq. km) the ESF dispersion is not affected by A and its value depends on the basin features controlling runoff generation (land cover and soil texture) and travel time (morphometric features); (ii) for larger areas, the ESF dispersion decreases with A according to a log-linear relation due to the decreasing variability of the QPFs and, possibly, to the response time in the channel network. In addition, analysis reveal that, regardless the basin area, the ESF dispersion is lower for larger AR, due to the increasing control of a deterministic component associated with the previous storm, and that the influence of AR tends to be larger for the basins with fast response times.