Development and validation of 1 km resolution PCR-GLOBWB for Africa

Abstract:

PCR-GLOBWB (van Beek et al., 2011; Sutanudjaja et al., 2018) is a global hydrology model that has been developed over the past two decades at the Department of Physical Geography, Utrecht University. The latest version of the model has a spatial resolution of 5 arc minutes (~ 10 km) and runs on a daily resolution. Different initiatives have expressed the need for hyper-resolution global hydrological modelling (see e.g. Bierkens et al., 2015), and in compliance we aim to further refine the resolution to 30 arc seconds (~1 km).

We have been developing a version of PCR-GLOBWB that has a spatial resolution of 1 km for the African continent. In doing so, two major challenges need to be overcome. First, computation time constitutes a hurdle, in particular to model river water routing/flow using a time-explicit scheme. To reduce run time, we apply a river basin partitioning strategy decomposing the model domain into several groups of river basins such that each river basin, connecting from their most upstream to downstream cells, could be run as an independent process. We aim to further advance this with a domain decomposition method based on the Pfafstetter stream levels and capitalizing on the hierarchical structure of a drainage network in combination with massive parallel computing to make this possible. Second, the parameterization of the model at 1 km poses a major challenge as this resolution approaches that of available global datasets. We have derived the model parameters at 1 km for Africa using global datasets, combining our past experiences (see e.g. Sutanudjaja et al., 2018) and recent high resolution datasets such as the SoilGrids and MERIT DEM.

We have tested this version of the high resolution model with the meteorological forcing from CHIRPS and ERA5-Land for the period 1981-2019. Results are promising and validated to various independent datasets such as GRDC discharge observation data, GLEAM evaporation, and GRACE (including GRACE-FO) total water storage thickness, as well as Sentinel soil moisture products.

References:

Bierkens, M.F.P. et al. (2015), https://doi.org/10.1002/hyp.10391
van Beek et al. (2011), https://doi.org/10.1029/2010WR009792
Sutanudjaja, E.H et al. (2018), https://doi.org/10.5194/gmd-11-2429-2018