Information content of prior hydrological knowledge in EGU visitors for streamflow prediction

Citizens living in flood-prone areas sometimes outperform hydrological models in terms of flood prediction. An interesting question is whether this is due to more access to information, more prior knowledge or better pattern recognition capabilities. In this interactive pico presentation, we plan to quantify the information content of the prior hydrological knowledge present in the brains of members of the audience, by making them participate in a probabilistic prediction game and comparing their performance to hydrologically ignorant computer-based time series prediction models. Results will be analysed in an information-theoretical perspective, where we will try to reverse engineer the best performing forecasters (without physical contact; the software, not the hardware).