72-The Use of Local Data in Estimating and Predicting Extreme Floods.

Abstract

Worldwide floods continue to cause the most damage of all natural hazards. Estimating floods can involve the use of estimates of extreme rainfall. In the UK, the Flood Estimation Handbook (FEH) has been shown to underestimate both rare rainfall depths and the Probable Maximum Precipitation (PMP) simply through a lack of use of local rainfall and historic storm information. Although the FEH encourages the use of historic flood data to improve estimates of extreme floods, it is dominated by desktop procedures. One consequence of this approach is that few practitioners now have the skills and experience needed to investigate floods in the field. This should be incorporated into national methodology for flood estimation in ungauged catchments. Its absence has resulted in upto an eight-fold difference in the estimate of the 100 year flood on a small chalk catchment. On clay catchments such as the Upper Brue, the 100 year flood has been underestimated by a factor of 2, and the Probable Maximum Flood (PMF) was underestimated by a similar margin. Local historic flood data can also be used to estimate flood volumes; this data is less sensitive to errors than for flood peaks. Historic floods can also be used to produce fl ood inundation maps or to test the reliability of existing maps and of dam-break floods. Predicting floods in real time requires accurate rainfall data, a reliable estimate of Soil Moisture Deficit (SMD), and a well calibrated runoff model that uses local soil hydraulic conductivity measurements. There can be serious underestimates of rainfall by the UKMO Nimrod radar based system. For one 5 km x 5 km grid square in East Somerset for the period 2004-2005 the error on daily rainfall is 40% at a depth of 10 mm, increasing to 64% at 28 mm. The errors are greatest at a time when floods may take place. Measurement of actual evaporation using weighing lysimeters also allows a reliable estimate of SMD to be made at low cost. In contrast to this both the UKMO MORECS (Meteorological Office Rainfall and Evaporation Calculation System), and MOSES (Meteorological Office Soil Exchange System) methods have errors upto 50 mm too high. This will prevent timely flood warnings. Locally measured soil hydraulic conductivity has produced higher estimates of percentage runoff than the FEH, while at low rainfall intensi;is there is an overestimate. Local soils data combined with local observed time to peak have been used to produce a real-time flood warning system for the upper Brue, operational now for three years.