Downscaling Global Land-Use Scenario Data to the National Level: A Case Study for Belgium

Abstract

Land use change scenarios, and their projected impacts on biodiversity, are highly relevant at local scales but not adequately captured by the coarse spatial resolutions of global land use models. In this study, we used the land use allocation tool of the GLOBIO 4 model to downscale the Land Use Harmonization v2 (LUH2) data from their original spatial resolution (0.25°) to 100 m and 10 m resolutions, using the country of Belgium as an example. Inputs to the tool included: (1) a reference present-day land cover map at the high spatial resolution, (2) regional land demand projections for three future scenarios, Sustainability (SSP1xRCP2.6), Regional Rivalry (SSP3xRCP6.0), and Fossil-fuelled Development (SSP5xRCP8.5), and (3) raster layers representing the suitability of the grid cells for different land use types. We further investigated the impact of using different reference land cover maps (CORINE at 100 m resolution and ESA WorldCover at 100 m and 10 m resolutions) on the downscaling outcomes. Comparison of downscaled current and future land use maps with the original LUH2 dataset showed that the use of ESA WorldCover as a reference map provides better agreement (RSR: 0.11–0.24, overall accuracy: 0.94–0.98, Kappa: 0.91–0.97) than CORINE (RSR: 0.28–0.33, overall accuracy: 0.90–0.93, Kappa: 0.90–0.91). Additionally, the validation of the present-day downscaled maps showed a good agreement with the independent Copernicus Global Land Service dataset. Our findings suggest that the choice of reference land cover map influences the degree of agreement between the downscaled and the original coarse-grain land-use maps. Moreover, the land use maps produced using our downscaling approach can provide valuable insights into the potential impacts of land use change on biodiversity and can guide local decision-making processes for sustainable land management and conservation efforts.