Robust Rescaling Methods for Integrated Water, Food, and Energy Security Management under Systemic Risks and Uncertainty

Yermoliev, Y., Ermolieva, T., Havlik, P. ORCID: https://orcid.org/0000-0001-5551-5085, Mosnier, A., Obersteiner, M. ORCID: https://orcid.org/0000-0001-6981-2769, Leclère, D., Fritz, S. ORCID: https://orcid.org/0000-0003-0420-8549, Kyryzyuk, S., et al. (2015). Robust Rescaling Methods for Integrated Water, Food, and Energy Security Management under Systemic Risks and Uncertainty. In: Systems Analysis 2015 - A Conference in Celebration of Howard Raiffa, 11 -13 November, 2015, Laxenburg, Austria.

Preview

Text Robust Rescaling Methods for Integrated Water, Food, and Energy Security Management under Systemic Risks and Uncertainty.pdf - Published Version Available under License Creative Commons Attribution Non-commercial Share Alike. Download (654kB) | Preview

Abstract

The aim of this presentation is to discuss robust, non-Bayesian, probabilistic, cross-entropy-based disaggregation (downscaling) techniques. Systems analysis of global change (including climate) processes requires new approaches to integrating and rescaling of models, data, and decision-making procedures between various scales. For example, in the analysis of water security issues, the hydrological models require inputs that are much finer than the resolution of, say, the economic or climatic models generating those inputs. In relation to food security, aggregate national or regional land-use projections derived with global economic land-use planning models give no insights into potentially critical heterogeneities of local trends. Many practical studies analyzing regional developments use cross-entropy minimization as an underlying principle for estimation of local processes. However, the traditional cross-entropy approach relies on a single prior distribution. In reality, we can identify a set of feasible priors. This is relevant, in particular, for land-cover data. Existing global land cover maps (GLC2000, MODIS2000, GLOBCOVER2000) differ in terms of spatially resolved estimates of land use, (e.g., crop, forest, and grass lands). We present novel general approach to achieving downscaling results that are robust with respect to a set of potential prior distributions reflecting non-Bayesian uncertainties, that is, data that are incomplete or not directly observable. The robust downscaling problem is formulated as a probabilistic inverse problem (from aggregate to local data) generally in the form of a non-convex, cross-entropy minimization model. The approach will be illustrated by sequential downscaling aggregate model projections of land-use changes using the Global Biosphere Management Model, with case studies from Africa, Brazil, China, and Ukraine. The approach is being used to harmonize alternative land-cover maps and to develop hybrid maps.

Item Type:	Conference or Workshop Item (Poster)
Research Programs:	Ecosystems Services and Management (ESM)
Depositing User:	Michaela Rossini
Date Deposited:	19 Jan 2016 15:18
Last Modified:	14 Jun 2023 13:23
URI:	https://pure.iiasa.ac.at/11804

Actions (login required)

View Item