Future Land Use Change projection under SSP-RCP scenarios over Ethiopia

Abstract

Land use land cover (LULC) data are crucial for modeling a wide range of environmental conditions. So far, access to high-resolution LULC products at a global and regional scale for public use has been difficult, especially in developing countries/regions (Doelman et al., 2018). Land Use Land Cover (LULC) change simulation models are a powerful tool for analyzing the causes and effects of LULC dynamics under different scenarios. Scenario-based simulations of future land-use change can provide important information for evaluating the impacts of land strategies under different conditions. In this study, we project the future land use data at a 1-km resolution that comprises six land use types, adopting the newest integrated scenarios of the shared socioeconomic pathways and the representative concentration pathways (SSPs-RCPs) over Ethiopia. To generate this high-resolution land-use product, we use the FLUS model to simulate future land-use dynamics. The process of developing a future land dataset for Ethiopia can be divided into two parts. The first part is the estimation of the future area demands of different land use types under different SSP-RCP scenarios extracted from the LUH2 (Land-Use Harmonization 2) datasets which is available for free at http://luh.umd.edu/index.shtml. This dataset comprises a global projection of multiple land types for successive years from 2015 to 2100 under different SSP-RCP scenarios with a 0.25° resolution (approximately 25 km at the equator). The second part is conducting a 1-km spatial land simulation using the future land use simulation (FLUS) model under the macro constraints of the demands. In this sense, we select a series of relevant spatial driving factors, such as socioeconomic (GDP, population), distance factors (urban center, roads, and rivers), and natural factors (climate, topography, and soil quality). On this basis, a new set of land use projections, with a temporal resolution of 10 years and a spatial resolution of 1km, in eight SSP-RCP scenarios, comprising six land use types in Ethiopia is produced. This dataset shows good performance compared to remotely sensed ESA CCI-LC data. The results show that our land use simulation yields a satisfactory accuracy (Kappa = 0.8, OA = 0.9, and FoM = 0.1). Because of the advantages of the fine resolution, current scenarios, and multiple land types, our dataset provides powerful data support for environmental impact assessment and climate research, including but not limited to climate models.