Research Interests - Florian Pappenberger

From numerical weather and climate predictions to flood inundation outlines

Quantifying and understanding the cascade of uncertainties from numerical weather and climate predictions to flood inundation outlines in real time forecasting and climate change scenarios. The use of numerical weather and climate predictions to derive flood inundation outlines in real time and for climate change scenarios is an exciting challenge. Such demands can be met by a system consisting of three different model components (weather/climate forecast, rainfall-runoff forecast, flood inundation forecast) in addition to several post and pre-processing routines and computations which convert hazard into risk maps. All of those components have considerable uncertainties in the structure, boundary conditions model parameters and many other factors. I am working on understanding the uncertainty in different parts of this cascade such as uncertainty analysis of the analogue approach for the downscaling of climate change scenarios, post-processing of numerical weather predictions as well as real-time updating of river discharges and flood inundation maps. Moreover, I have led the development of a novel methodology to cascade uncertainties from rainfall forecasts to flood inundation predictions.

I am now developing this methodology further in the context of real time global flood predictions by including a global routing module in the ECMWF atmospheric models. I am also part of a Natural Environment Research Council (NERC program: Flood Risk from Extreme Events) project which investigates the importance of uncertainty in individual model components under climate change conditions using the new ensemble forecasts of climate change by the UK MetOffice.

Functional Classification and evaluation of Hydrographs

Visual graphical evaluation based on the simple plotting of two curves is the most intuitive and favoured approach by many modellers. I have adopted this approach for a novel performance measure (the Multicomponent Mapping method) which pioneers the functional classification of model factors.

I am currently working on including this methodology in the evaluation of global discharge predictions matching spatial and temporal discrepancies between point measurement and large scale grid predictions.

Advanced Sensitivity Analysis (SA) of models

SA tries to understand how the variations in model outcome are based on variations of factors. Understanding the sensitivity of models can help to reduce the uncertainty in for example flood inundation or flood hydrograph modelling by identifying the larges sources of uncertainty in the modelling process. I have developed a unique methodology which determines global sensitivity based on local properties.

Recent publication show my research on comparing different types of sensitivity measures and deriving information from a multi-objective analysis.

Socio-economic calibration of flood models

Models are traditionally calibrated on physically based observations of the system such as flood inundation outline, hydrographs or precipitation fields. In particular the calibration of fields will in most case result in a trade-off of areas which are better represented and those which are not regardless of the socio-economic geography of the area. The traditional approach is hazard based and I have argue that model calibration has to shift to a vulnerability weighted approach. For example, it might be necessary to optimize your uncertainty and model calibration routines towards location of hospitals than for an empty pasturs. This can lead to vastly different flood hazard maps.

I am now developing this concept of vulnerability based calibration further for climate change scenarios and investigate methods to apply a vulnerability based calibration of flood inundation to the parameter estimation of rainfall-runoff and climate change scenarios.

Communication of uncertainty by meteorological/hydrological/hydraulic modellers to the public and between scientists

Communicating uncertainty to the public requires the commitment of scientists to accept that uncertainty analysis is necessary and can be communicated. In this context it is imperative to explore the communication between the hydrological experts constructing and receiving flood risk alerts, and to learn how to best represent information in these flood alerts. In order to achieve this aim I would like to find out how different members of the expert community perceive and interpret a variety of flood risk based information.