Flood forecasting for everywhere-PUB in flood forecasting

In this study the Prediction in Ungauged Basins has been taken very literally in that we present a system that enables setting up a rainfall-runoff model, the Distance Distribution Dynamics (DDD) model for any catchment in Norway. The system can be used in operational flood forecasting since hydrological simulation results for an arbitrary catchment are obtained in a few minutes. A GIS map tool is used to calculate catchment boundaries, a hypsographic curve and other catchment characteristics such as vegetation and mean annual discharge needed to estimate DDD model parameters. Derived terrain information and catchment boundaries are furthermore used to extract meteorological information from gridded (1 x 1 km) maps for both historical and forecast periods. The historical period may be of such length (>30 years, daily resolution) that the mean annual flood can be reasonably estimated and compared to forecasted runoff values for hazard assessments. In this way a flood forecaster is no longer limited to only be looking at hydrological simulation results from calibrated models set up for a few gauged catchments. Rather, she can set up a model for ungauged catchments where the forecasted precipitation is the most intense or where vulnerable infrastructure is located. The relative comparison between simulated forecasted runoff and simulated mean annual flood is of value for hazard assessments. Regarding absolute values, the DDD model has been tested for prediction in ungauged basins for 25 gauged catchments and obtains an average Kling-Gupta efficiency (KGE) of 0.77. The mean annual flood is, however underestimated by 40 %. Better results are expected when improved gridded meteorology and estimates of mean annual discharge are available. Future developments include higher temporal and spatial resolutions so that flood forecasting and flood estimation can be carried out for smaller and faster responding ungauged catchments.