ROBUST DOWNSCALING APPROACHES TO DISAGGREGATION OF DATA AND PROJECTIONS UNDER UNCERTAINTIES: CASE OF LAND COVER AND LAND USE CHANGE SYSTEMS

摘要

The interdependences among land use systems at national and global levels motivate the development of advanced systems analysis approaches for integration of land use models operating at different weights. The paper develops novel general approaches based on cross entropy principle for downscaling aggregate data and projections, which are robust with respect to feasible priors. Robust downscaling methods account for so-called non-Bayesian uncertainties, i.e., incomplete, unobservable, or erroneous information or data. In numerous case studies in China, African countries, Brazil, and Ukraine, the approaches allowed deriving local development projections of land use and land use change consistently with existing trends and expectations.