4 - Aspects of uncertainty, reliability, and risk in flood forecasting systems incorporating weather radar

Summary

ABSTRACT

Uncertainty in flood forecasts is dominated by errors in the measurements and forecasts of rainfall used as input. Error magnitudes are influenced by raingauge network density (possibly used in combination with weather radar), by rainfall intensity, and by the method of rainfall forecasting employed. An empirical approach to quantifying uncertainty associated with rainfall measurements and forecasts is taken, supported by data from two dense raingauge networks, together with weather radars, in southern Britain. The impact of uncertainty in rainfall on flood forecasts is examined through a comprehensive case study within the Thames basin in the vicinity of London. This study also allows the relative effect of model and catchment on flood forecast uncertainty to be better appreciated. Reliability of flood forecasts is considered in the context of the complexity of region-wide flood forecasting systems and the need to ensure that forecasts are made under all situations, including the possible loss of significant telemetry data. The River Flow Forecasting System's Information Control Algorithm is outlined as a solution to providing reliable forecasts, coping with both complexity and data loss. Risk is considered here in the context of when to issue a flood warning given an uncertain flood forecast. The use of both informal and more formal methods of ensemble forecasting is introduced as a means of quantifying the likelihood of flooding implied by a flood forecast.

INTRODUCTION

Flood forecasting systems function in real-time to transform telemetered field measurements (principally relating to river level and rainfall) and external forecasts (especially of weather) to forecasts of river level and flow, possibly along with settings associated with river control structures and reservoirs.