Long Term Human-Flood Dynamics in the City of Dresden: Modeling and Inference

We postulate, based on literature, a "socio-hydrological" model that describes the interactions between floods, settlement density, people awareness and preparedness, and flood losses in a floodplain system. We then use Bayesian Inference as a method to infer the model parameters from empirical data on these variables. First, in order to evaluate the method, we simulate hypothetical time series using the model, we sample few data points for its variables and we check whether we can correctly estimate the model parameters with this information. Second, we calibrate the model using real data for the case study of Dresden, Germany, over a period of 200 years. Results show that the model parameters can be estimated with reasonable accuracy, even though large uncertainty is associated to the data. Combining socio-hydrological modeling and empirical data allows for acquiring additional knowledge about the socio-hydrological system, such as quantifying the "forgetfulness" of the society, i.e., how quickly awareness of flood risk decreases in time, which would not be easily identified by using uncalibrated socio-hydrological models or by standard statistical analysis of empirical data.