Addressing change of support problems in forcing data for hydrological models.

---

Thursday 2 july 2015

17:51 - 17:54h at Central America (level 0)

Themes: (T) Water resources and hydro informatics (WRHI), (ST) Catchment hydrology, Poster pitches

Parallel session: Poster pitches: 13H. WRHI - Catchment

---

Hydrological models assume major relevance in water resources management activities, providing sound decision basis at the operational and strategic levels alike. They are employed in a wide set of tasks, ranging from operational forecasting to long-term planning. Regardless of the chosen application, their value typically increases with the accuracy of the produced simulations. It is widely recognized that the response of most hydrological models is highly dependent on features of forcing data such as rainfall or temperature. Furthermore, it is known that at least some hydrological models become "specialized" in the forcing data used for calibration and validation. Because most of this data does not necessarily represent reality, but a mere approximation, it is hard to ascertain a priori the performance of these models when different forcings are adopted, in what can be denominated a change of support scenario. In fact, in some cases the forcing data used to calibrate and validate the hydrological models does not match the forcing data which will be used to run simulations. This can happen, for example, when rain gauge networks change in time, when satellite, radar and rain gauge data are combined in order to obtain simulations covering periods of several decades, or when outputs from global circulation models are used to assess the impact of climate change on water resources. Because models' responses are associated with the features of the forcing data, one should acknowledge that, when change of support occurs, simulations potentially depict not only trends contained in the employed forcing data, but also unwanted artifacts resulting solely from that change. In some cases these artifacts can be highly relevant, greatly masking the trends contained in the data. In this contribution, the challenges related to the change of support are illustrated for the case of rainfall by combining satellite and ground data. Namely, it is shown how interpolation techniques can affect daily hydrological simulations and how a particular rainfall interpolation technique – the pattern-oriented memory interpolation – can contribute to mitigate the change of support problem.