IIASA Integration Assessment via Downscaling of Population, GDP, and Energy Use Urbanization, Development Pathways and Carbon Implications NIES, Tsukuba, Japan March 28-30, 2007 gruebler@iiasa.ac.at
Why Downscaling? Need for spatially explicit scenario drivers, e.g. for land-use change and forestry models Description of spatial heterogeneity (adds to scenario uncertainty, even if illustrative) Necessary input to impact and vulnerability assessments (e.g. people and cities at risk of sea level rise) Can help to identify additional constraints: spatial energy and pollution densities, infrastructure needs, Input to country-level policy analysis Core research question: Where are key drivers of change and of vulnerabilities?
Downscaling Philosophy Focus first on main drivers for land availability and economics of agriculture and forestry (population and GDP) Capture scenario uncertainty (3 IIASA-GGI scenarios: A2r, B2, B1) Avoid proportional scaling techniques if possible Occam s Razor: In absence of data/models apply simplest assumption/algorithm possible Calibrate with global data sets as they become available (G-ECON, GRUMP, ongoing activity) Complement top-down with bottom-up assessments (plausibility, missing scenario uncertainty, )
Downscaling Approach Interdisciplinary team incl. demographers, economists, geographers, land-use modelers, 2-step approach: Global/regional national grid-cell level reflecting distinctly different user needs Combination of constrained optimization and simulation techniques Reflects data/methods available 2004/5
Thanks to: Anne Brian Erik Keywan Peter Serguei Vadim
IIASA Integrated Assessment & Scenario Analysis Feedbacks Scenario Storyline Economic development Demographic change Technological change Policies Feedbacks Global and Regional Scenarios Population Economy Downscaling Tools Spatially explicit and national scenarios Spatially explicit socio-economic drivers National, regional & spatially explicit socio-economic drivers DIMA Forest Management Model Consistency of land-cover changes (spatially explicit maps of agricultural, urban, and forest land) AEZ-BLS Agricultural Modeling Framework CLIMATE and ACIDIFICATION IMPACT MODELS Carbon and biomass price Potential and costs of forest bioenergy and sinks MESSAGE-MACRO Systems Engineering / Macro-Economic Modeling Framework (all GHGs and all sectors) Endogenous Climate Model Agricultural bioenergy potentials and costs Drivers for land-use related non-co2 emissions NATIONAL POLICY MODELS (GAINS) Emissions Emissions & Abatement Costs
Scenario Taxonomy
Scenario Overview (World by 2100) 2000 A2r B2 B1 Population, 10 9 6 12 10 7 GDP, 10 12 $ 35 190 240 330 PE, EJ 440 1750 1300 1050 Efficiency, %/yr -0.6-1.2-1.7 Zero-C, % share 15 36 47 61 GtC energy 7 27 16 6 GtC forests 1 <1-2 -1 GtC-e all others 3 10 5 4 GtC-e total 11 38 19 9 ppmv (CO 2 -equiv) 370 1390 980 790 Stabil. Levels (ppm-equiv) 670-1090 520-670 480-670
Downscaling Flow Chart POPULATION GDP Regional 11 regions National 185 countries Sub- National Cells GRID 7.5 x 7.5 Cells GRID 7.5 x 7.5 Sub- National National 185 countries Regional 11 regions World NAM Per capita urban GDP NAM World WEU Austria Belgium Urban POP Spatial datasets Urban GDP Austria Belgium WEU national projections urban share UK Urban Rural gravity type models Rural POP PEOPLE per square km GDP at MER per ha Rural GDP Urban Rural GDP urban/rural UK Optimization SAS Per capita rural GDP SAS
Approach Population GDP existing methodology: global and world regional scenarios National population projections (constrained downscaling using UN) Estimation of future urban population (UN scenario extensions H/M/L) Depicting urbanized areas Distributing of rural/urban population (downscaling) Projections (based on gravity-type models) National GDP projections (constrained optimization) Urban and rural per capita GDP estimates for base year Projections of urban and rural per capita GDP disparities Distributing per capita GDP over rural/urban population
National POP Scenarios Input: 3 SRES scenarios incl. one substantial revision (A2r, developed at IIASA) Based on UN long-range (300yr) scenarios Regional population scenario downscaled to national level using UN scenario with closest match in demographic characteristics Improved over previous efforts CIESIN, MEA Remaining problem: some discontinuities after 2050 (halt of migration in UN scenarios)
Comparison of population downscaling for China and Afghanistan
Comparison of 2 Downscaling Methods for a Low Population Scenario (B1)
National GDP Scenarios 186 National GDP scenarios downscaled from 11 world regional level for 3 scenarios Optimization algorithm with constraints: sum of national GDPs = regional GDP GDP growth = f(gdp/capita) different pathways for clusters of countries within region upper and lower bounds of income disparities (B1 only)
1: Topological Relationship Between GDP Growth ad GDP/Capita Levels (scenario dependent) 3 Western Europe A2: the rich slow down f(x) = a * log(x) + b x GDP/CAP GDP growth (percent) 2 1 Region: Western Europe (a2) Model approximation 0 10000 20000 30000 40000 50000 60000 Per capita income (US$) 14 South Asia B1: the poor catch up f(x) = a * x / (x 2 +b) + c x GDP/CAP a 2 * xmax * ymax b xmax 2 ymax max growth rate xmax GDP/CAP@ ymax GDP growth (percent) 12 10 8 6 4 2 0 Region: South Asia (B1) Model approximation 0 10000 20000 30000 40000 50000 Per capita income (US$)
2: Model Application for all Countries in Region, Constrained by Regional Total GDP scenario GDP Growth - LAM - A2 GDP Growth - LAM - B1 12 12 10 10 8 8 Growth 6 Growth 6 4 4 2 2 0 0 10000 20000 30000 40000 50000 60000 10 GDP/CAP GDP Growth - FSU - A2 0 0 10000 20000 30000 40000 50000 60000 GDP/CAP 10 GDP Growth - FSU - B1 8 8 6 6 Growth 4 2 Growth 4 2 0 0-2 -2-4 0 10000 20000 30000 40000 50000 60000 GDP/CAP -4 0 10000 20000 30000 40000 50000 60000 GDP/CAP
Result GDP/CAP GDP per Capita - B1 OECD90 versus ALM GDP per Capita - A2 OECD90 versus ALM 100000 100000 10000 10000 1000 1000 OECD90 OECD90 100 ALM 1990 2010 2030 2050 2070 2090 100 ALM 1990 2010 2030 2050 2070 2090
Disparities in Projected Country GDPs Lorenz Curves based on 185 Countries Fraction of GDP 1 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 0.133 0.528 0.741 0 0.2 0.4 0.6 0.8 1 Fraction of Population 1990 2100 (A2) 2100 (B1) Equality
Urbanization Scenarios Combination of country level projection (to 2030) and 3 scenarios (to 2100) Based on UN urbanization projections (2003) Extension of UN Projection by 3 scenarios: High (A2r), Medium (B2), and Low (B1) urbanization
Urbanization Trends IIASA scenarios: High/Medium/Low UN data and projection
Sub-National Scenarios 1 (POP) Estimation of base-year sub-national rural/urban population/area allocation (constrained by UN urbanization statistics) Spatially explicit allocation for 3 scenarios: -urban: based on gravity model (with density saturation) w. limited range -rural: proportional scaling (weak)
Population Density, A2 and B1
Sub-National Scenarios 2 (GDP) Estimation of base year sub-national rural/urban GDP per capita 3 scenarios of rural/urban income convergence: High (B1), Medium (B2), Low (A2r) Constrained by national total GDP scenarios Spatial allocation: based on population density and rural/urban income differential scenarios (weak)
Base Year GDP comparison (1): National Statistics Sub-National Shares of GDP (USA) 40 Sub-National Shares of GDP (Brazil) 15 y = 1.0261x - 0.0968 R 2 = 0.9647 y = 1.0734x - 0.1439 R 2 = 0.9693 30 Model, %, 1990 10 Model, %, 1990 20 5 10 0 0 5 10 15 Sub-National Statistics, Shares of %, 1995 GDP (India) 15 y = 0.9917x + 0.5064 R 2 = 0.6982 0 0 10 20 30 40 Sub-National Shares Statistics, of %, GDP 1998 (China) 10 y = 0.724x + 1.0006 R 2 = 0.5802 8 Model, %, 1990 10 Model, %, 1990 6 4 5 2 0 0 0 5 10 15 0 2 4 6 8 10 Statistics, %, 1994 Statistics, %, 1994
Base Year GDP comparison (2): With G-ECON Data Set (W. Nordhaus) Dem. Rep. of Congo. 1995 GCP comparison 8 y = 1.1292x - 1.0044 R 2 = 0.9182 USA GCP comparison 10 TNT (LOG) 6 4 TNT, US$90, 10^ 8 6 y = 1.2427x - 2.0374 R 2 = 0.6151 2 0 sample: 1027 (out of 1156) cells 5729.24 (5753.25) billion US$1990 (Nordhaus) 5657.62 (5657.62) billion US$1990 (TNT) 0 2 4 6 8 Nordhaus (LOG) 4 4 6 8 10 Nordhaus, US$95, 10^
Urban/Rural per capita GDP in A2 and B1 (Pacific Asia) 100000 PAS, B1 100000 PAS, A2 GDP MER per capita, US$1990 10000 1000 GDP MER per capita, US$1990 10000 1000 100 1980 2020 2060 2100 100 1980 2020 2060 2100 PAS total PAS rural PAS urban PAS total PAS rural PAS urban
GDP Density with urban/rural residence and income differences
Spatial Resolution Base year (1990): 2.5 x 2.5 arc seconds Scenarios (2000-2100): 7.5 x 7.5 arc sec. Public Data Base (web access): 0.5 x 0.5 degrees http://www.iiiasa.ac.at/research/ggi/db
Use of Downscaled Scenarios Land price scenarios for determining biomass and forest C-sequestration potentials, and deployment in stabilization scenarios (iterated results, consistent C- prices) Impact and vulnerability assessments (people and GDP at risk) Energy access and energy density
Biomass Potentials Dynamic GDP maps (to 2100) Dynamic population density (to 2100) Top-down Downscaling Development of bioenergy potentials & use bottom-up assessment Consistency of land-price, urban areas, net primary productivity, biomass potentials/use (spatially explicit)
Biomass Potentials and Use: Significant reduction (compared to SRES/TAR) due to intersectorial linkages and consistent land and C-prices 500 400 EJ 300 200 pot_old pot_new use_old use_new 100 0 B1 B2 A2r
Downscaling Does it Matter? Yes for biomass and land-based forest C-sequestration (esp. in B1 low POP high income world) Main determinant: GDP distribution and to lesser extent rural population allocation (urbanization exerts indirect influence only) Wrong research question: bioenergy and sinks in C- controlled world less constrained by land availability, but rather how agricultural production and forest ecosystem and amenity services will be affected by energy and C prices (much larger economic leverage of biomass/bioenergy and sinks) Main influence of urbanization - Energy Densities: Transport infrastructure needs and costs underestimated (esp. for BECCS), urban energy demand determines fuel mix and quality (electricity and liquids rather than biomass)
Tokyo: Electricity Demand and Supply Densities kwh 100000 10000 vs. Solar Energy Supply Electricity demand 1000 100 Solar radiation Solar radiation converted to electricity 10 1 0 1000 2000 3000 km 2 Source: TEPCO & NIES, 2002
Europe: Power Density of Demand (W/m 2 ): Grey areas indicate where biomass or wind can satisfy local energy demand (< 0.5 W/m 2 ) England: Energy demand footprint larger than country area
Ongoing & Future Work Improved base-year calibration Experimental scenarios of spatially heterogeneous rural growth Mapping energy access and spatially explicit scenarios of final energy use Extensions to GHG and air pollutant (aerosols) emissions
Population Density
Population Density vs Final Energy per Capita
Data Available Online Full scenario data for 11 world regions, 3 scenarios to 2100 Population and GDP data plus urban/rural split for 185 countries for 3 scenarios Dynamic population and GDP maps, 3 scenarios http://www.iiasa.ac.at/webapps/ggi/ggidb/dsd?action=htmlpage&page=series Documentation: Special Issue Technological Forecasting & Social Change 74(8 9), October November 2007. Electronically already available via ScienceDirect