Improving hydrological modelling and prediction at the European and Global scale

Hydrological models are crucial for evaluating the water cycle, offering decision makers vital insights into floods, droughts, and water resource management, while enabling scenario analysis under different natural and anthropogenic conditions. One example is the open-source hydrological model OS-LISFLOOD that is used to generate flood forecasts and drought indicators for the European and Global Flood Awareness Systems (EFAS & GloFAS) as well as the European and Global Drought Observatories (EDO & GDO) of the Copernicus Emergency Management Service (CEMS).

OS-LISFLOOD is a distributed, physically based rainfall-runoff model. Being used in an operational setting, the hydrological model and its European and global model domain set-up benefit from regular upgrades. In its current operational version, the global model set-up (GloFAS v4.x) uses a spatial resolution of 3 arcminutes (~5.4 km) and a daily time step, whereas the European model set-up (EFAS v5.x) uses a spatial resolution of 1 arcminute (~1.8 km) and a 6-hourly time step. Both set-ups are used to provide a hydrological reanalysis as well as hydrological predictions.

A specific feature of the European and global model set-up of OS LISFLOOD is that not only the model and associated tools for pre-/post-processing, calibration, etc. are open-source, but also the required input and calibrated parameter maps are freely accessible. This allows users to benefit from the latest developments and, more importantly, it enables a wider community in contributing to further extending and improving the model and its set-up.

In this presentation we describe the next major evolution of OS LISFLOOD and its set-up for the European (EFAS v6.x) and global domain (GloFAS v5.x). The main foreseen changes can be grouped into three categories: 1.) model input; 2.) model improvements; and 3.) calibration and regionalization.

The main changes in the model input concern the meteorological forcings. For the European domain, the meteorological forcings benefit from an increased number of meteorological observations, improved quality control, and a modified interpolation method. In the global model domain, enhancements include a correction of spurious rainfall and a modified downscaling of ERA-5 meteorological variables. Furthermore, changes in the surface fields related to soil properties, lakes and reservoirs as well as water demand for anthropogenic use integrating the latest available datasets have been included. Hydrological model advancements focus on river routing, in particular for mild sloping rivers, and a modified reservoir routine. Furthermore, the model state initialization has been enhanced and a new modelling routine called transmission loss which accounts for streamflow leakage has been added. For model calibration and regionalization, it is foreseen to increase the number of calibration stations, improve the overall performance of the objective function along the whole flow duration curve, add more hydrological performance statistics, and to utilize the power of deep learning during the regionalization of model parameters.

The improved model, its new set-up as well as the hydrological model reanalysis and predictions will be freely available. Its release as part of the operational EFAS, GloFAS, EDO, and GDO of CEMS is foreseen during 2025.