Ensemble flood forecasting: A review

Summary

Operational medium range flood forecasting systems are increasingly moving towards the adoption of ensembles of numerical weather predictions (NWP), known as ensemble prediction systems (EPS), to drive their predictions. We review the scientific drivers of this shift towards such ‘ensemble flood forecasting’ and discuss several of the questions surrounding best practice in using EPS in flood forecasting systems. We also review the literature evidence of the ‘added value’ of flood forecasts based on EPS and point to remaining key challenges in using EPS successfully.

Introduction

Flood protection and awareness have continued to rise on the political agenda over the last decade accompanied by a drive to ‘improve’ flood forecasts (Demeritt et al., 2007, DKKV, 2004, Parker and Fordham, 1996, Pitt, 2007, van Berkom et al., 2007). Operational flood forecasting systems form a key part of ‘preparedness’ strategies for disastrous flood events by providing early warnings several days ahead (de Roo et al., 2003, Patrick, 2002, Werner, 2005), giving flood forecasting services, civil protection authorities and the public adequate preparation time and thus reducing the impacts of the flooding (Penning-Rowsell et al., 2000). Many flood forecasting systems rely on precipitation inputs, which come initially from observation networks (rain gauges) and radar. However, for medium term forecasts (∼2–15 days ahead), numerical weather prediction (NWP) models must be used, especially when upstream river discharge data is not available (Hopson and Webster, in press) or when the equipment or transmission of data fails as is often the case in extreme floods. In general NWPs are essential to establish longer leadtimes than the catchment concentration time, but even for shorter leadtimes NWPs provide added value in that predictions can be made for parts of the catchment near to the outlet/point of reference.

Operational and research flood forecasting systems around the world are increasingly moving towards using ensembles of NWPs, known as ensemble prediction systems (EPS), rather than single deterministic forecasts, to drive their flood forecasting systems. This usually involves using EPS as input to a hydrological and/or hydraulic model to produce river discharge predictions (Fig. 1), often supported by some kind of Decision Support System (Fig. 2).

Several different hydrological and flood forecasting centres now use EPS operationally or semi-operationally (Table 1; note that not all ensemble forecasts are publicly available), and many other centres may be considering the adoption of such an approach (Bürgi, 2006, Rousset Regimbeau et al., 2006, Sene et al., 2007). The move towards ensemble prediction systems (EPS) in flood forecasting represents the state of the art in forecasting science, following on the success of the use of ensembles for weather forecasting (Buizza et al., 2005) and paralleling the move towards ensemble forecasting in other related disciplines such as climate change predictions (Collins and Knight, 2007). For hydrological prediction in general, the Hydrologic Ensemble Prediction Experiment (HEPEX) initiative has been setup to investigate how best to produce, communicate and use hydrologic ensemble forecasts (Schaake, 2006, Schaake et al., 2006, Schaake et al., 2005, Schaake et al., 2007), which are now often referred to as Ensemble Streamflow predictions (ESP) (note that historically the term ESP was used in a slightly different context, usually for longer term predictions (seasonal to yearly) of total volume or peak flows, and with ensembles created from an analysis of historical observations rather than ensemble weather forecasts (Twedt et al., 1977, Day, 1985) – for more discussion see (Seo et al., 2006, Wood and Schaake, 2008)).

In addition, other international bodies are demonstrating their interest in ensemble predictions for hydrological prediction, for example, the International Commission for the Hydrology of the Rhine Basin (CHR) and the World Meteorological Organization (WMO)’s Expert Consultation in March 2006 on ‘ensemble predictions and uncertainties in flood forecasting’, and the International Commission for the Protection of the Danube River’s (ICPDR) recent move to adopt the ensemble forecasts of the European Flood Alert System (EFAS) in their flood action plan.

However, there is currently no rigorous critique of the scientific drivers of the move towards the use of EPS in medium range flood forecasting, and in addition there remain many assumptions in the practice of this, for example, the over-reliance on a disjointed set of case studies for evaluation (see later discussion). In this paper we address these issues and outline some of the challenges ahead. First we review the reasons why ensembles of NWPs are so attractive for flood forecasting systems. Second we discuss how uncertainty is represented in, and cascaded through, these systems and some of the assumptions behind these methodologies. Third we discuss the methods used to calculate flood forecasts probabilistically. Fourth, we review the case studies in this field which mostly find that ensemble prediction gives useful information (‘added value’) for flood early warning and highlight some weaknesses in current practice. Finally we identify the key challenges of using EPS for flood forecasting.