On April 7, I gave 2 presentations in the “hydrological forecasting and predictive uncertainty” session at EGU2011.
My first presentation summarised research into the value – expressed in flood risk – of producing hydrological forecasts in terms of probability distributions. It shows that there is added value compared to producing ‘single value’ forecasts. This research will shortly be submitted for publication.
The second presentation considered two different techniques for estimating predictive hydrological uncertainty: the Hydrological Uncertainty Processor (by Krzysztofowicz et al) and Quantile Regression (by Roger Koenker). Both techniques were applied in Ovens basin in Australia. Preliminary results show that the resulting distributions are independent of the technique used, but quite sensitive to the configuration, i.e. to the choice of parameters the distributions are conditioned on. This research will be continued and should lead to a submission to a peer-reviewed journal later this summer.