Mechanistic modelling of hydro-ecological processes in a changing and uncertain world
Human activity is now changing the global earth system at unprecedented rates, including its water cycle and connected ecological processes. So how can we simulate connected hydro-ecological processes under climatic conditions that are different from those in our observational record? One possible way forward is the move from empirical models, which exploit correlation in historical datasets, to mechanistic models, which exploit process understanding – and are therefore by definition more suitable for modelling unobserved circumstances. In a changing world, we cannot extrapolate from historical observations into the future to understand change implications, but rather require models that can assess the behavior of a changed environment. These models should also recognize uncertainty originating from our data as well as from our gaps in understanding so that information for decision support is more robust.
I will present a multi-year study in which we brought together hydrologists, biologists, as well veterinary and computer scientists to understand environmental controls on infectious diseases. We integrated hydrological and epidemiological processes into a new mechanistic model, focusing on the widespread parasitic disease of fasciolosis as an example. The model simulates environmental suitability for disease transmission, explicitly linking the parasite life-cycle to key weather-water-environment conditions. Using epidemiological data, we show that the model can reproduce observed infection levels in time and space over two study regions in the UK. Second, to overcome limitations in the epidemiological dataset, we propose a calibration approach combining Monte Carlo sampling with expert opinion, which allows us to constrain the model in a process-based way. Finally, comparison with information from the literature and with a widely-used empirical risk index shows that the simulated disease dynamics agree with what has been traditionally observed, while the new model gives better insight into the time-space patterns of infection risk.
Host: Dörthe Tetzlaff