Concept of dealing with uncertainty in radar-based data for hydrological purpose

Abstract

Precipitation radar-based data constitute essential input to Numerical Weather Prediction (NWP) and rainfall-runoff models, however the data introduce a number of errors. Thus their uncertainty should be determined to provide end-users with more reliable information about forecasts. The common idea is to use Quality Index (<I>QI</I>) scheme for some number of quality parameters on the assumption that: (1) relationship between the parameters and relevant quality indexes is linear; (2) averaged <I>QI</I> is a weighted average of all particular indexes. The uncertainty parameters can be topography-dependent, resulting from spatial and temporal distribution of data, etc. Uncertainty in radar-based data is described by gamma PDF of precipitation, and it is proposed to determine the probability density function (PDF) parameters basing on <I>QI</I> values. Practically, precipitation is presented as ensemble of quantiles of the PDF and such an ensemble can constitute input to rainfall-runoff modelling. Since the ensemble is a precipitation input, the hydrological model needs to be activated according to a number of input members.