Towards operational joint river flow and precipitation ensemble verification: considerations and strategies given limited ensemble records

Abstract

A framework for joint verification of river flow and precipitation ensembles is developed and demonstrated over Britain for eventual use in an operational flood forecasting setting. The river flow ensembles are obtained from a distributed hydrological model, the G2G model, using an ensemble of 15 min precipitation accumulations as input on a 1 km grid. The precipitation ensemble consists of operational Numerical Weather Prediction (NWP) forecasts from the Met Office Unified Model. Both hourly and daily precipitation accumulations are verified, and the relevance of different accumulation periods discussed in the context of timing errors and hydrological response. The implications of precipitation observation error are investigated by comparing verification results from raingauge- and radar-derived precipitation estimates. Challenges of verification using only a limited record of precipitation ensembles, from a system only relatively recently made operational, are addressed. Methods of obtaining more robust verification statistics, given the available ensembles, are presented and demonstrated for an example period in December 2015. For precipitation, percentile thresholds are used to ensure a given number of threshold crossing events for analysis using a contingency table and derived skill scores. For river flow, percentiles thresholds are of less relevance to operational flood guidance. Instead, exceedance of a flow threshold of given rarity (return-period) is used as a surrogate measure of flood severity. At the regional scale, both river flow and precipitation verification analyses are found to be dependent on the locations considered. This is linked to variations in precipitation amount. For river flows, catchment properties – and in particular catchment size – are found to be a key influence on verification. It is demonstrated how such behaviour can be used to obtain more-robust river flow verification statistics at sub-regional scales.