SUPPORTING ONLINE MATERIAL
1 The GCAM Integrated Assessment Model
GCAM[1] (Kim et al., 2006, Clarke, et al., 2007, Edmonds and Reilly, 1985) is an integrated assessment (IA) model that links a global energy-economy-agricultural-land-use model with a climate model of intermediate complexity. [2] GCAM has 14 global regions defined on geopolitical boundaries: the United States, Canada, Western Europe, Japan, Australia & New Zealand, Former Soviet Union, Eastern Europe, Latin America, Africa, Middle East, China and the Asian Reforming Economies, India, South Korea, and Rest of South & East Asia. GCAM is a long-term model, typically operating in five-year time steps through the year 2095 (though the code is written to accommodate any arbitrary time-step or end year). As part of GCAM’s modeling of human activities and physical systems, GCAM tracks emissions and concentrations of the important greenhouse gases and short-lived species (including CO2, CH4, N2O, NOx, VOCs, CO, SO2, BC, OC, HFCs, PFCs, and SF6).
GCAM version 3.0 is marked by a new approach for modeling agriculture and land use at a finer level of spatial resolution, as well as the general ability to run the model at shorter time intervals. In the core version of GCAM 3.0, the modeling and data representing agriculture and land is specified at a resolution of 151 land use subregions around the globe. These land use subregions are based on a division of the extant agro-ecological zones (AEZs), which we derived from work performed for the GTAP project (Monfreda et al, 2009), within each of GCAM’s 14 global geo-political regions. These changes provide a substantial enhancement to GCAM’s ability to model crops and land use decisions and implications in much more physical, technological, and spatial detail while maintaining tight integration with the rest of the GCAM.
GCAM models the energy, agriculture, and land use in an economically and physically consistent global framework. In each model period, GCAM explicitly models markets and solves for equilibrium prices in energy, agriculture and other land uses, and emissions; that is, the set of prices that ensures that supplies are equal to demands in all markets. GCAM is a dynamic-recursive model, which means that it solves for each period’s market equilibrium sequentially.[3] GCAM models energy and agriculture technologies as discrete and linear rather than by abstracting them with economic production functions. However, economic choices and allocations among factors such as energy resources, technologies, and uses of land are based on nonlinear functions such as logit choice models. In contrast to linear optimization models, which are often characterized by winner-take-all solutions that must be limited by explicitly imposed constraints, the logit choice approach used in GCAM ensures some degree of heterogeneity in economic decisions.
The size and composition of the global population and the flow of GDP are the principle terms shaping the scale of energy, agriculture, and land-use systems. The modeling of the future regional economies and populations is highly aggregated. GDP is calculated as the product of labor force and average labor productivity modified by an energy-service cost feedback elasticity that captures the effect of mitigation on GDP. The labor force and labor productivity are both exogenous inputs to GCAM, developed offline from detailed demographic analyses.
1.1 The Energy System in GCAM
GCAM was originally designed to answer questions about the role of the energy system and energy technologies in mitigation, and energy remains an area of. The GCAM energy system models energy from its point of origin to its final end-use. This includes primary energy resources, production, energy transformation, and energy consumption. GCAM includes both depletable (coal, gas, oil, uranium) and renewable (wind, solar, geothermal) resources, which are represented through graded resource curves. As more energy is extracted and used, costs rise, though those cost increases can be ameliorated by technological improvement. The supply of bioenergy is determined by the agriculture and land-use submodel within GCAM, which is discussed below.
GCAM models the transformation of primary energy resources into final energy forms (electricity, hydrogen, refined liquids, refined gas, coal, and solid bioenergy) through a set of conversion sectors, each of which may include a range of conversion technologies. For example, GCAM includes multiple technologies for producing electricity from coal, natural gas, oil, bioenergy, wind energy, nuclear power, solar energy, hydropower, and geothermal energy. Final energy forms are consumed by three end-use sectors (buildings, industry, and transportation). GCAM includes detailed representations of each of these demand sectors in the U.S. It also includes a globally detailed representation of transportation demands. Consumption of energy in these sectors is determined by the demand for final energy services, as well as the characteristics of the technologies used to provide those services.
1.2 The Agriculture, Forest, and Land Use Systems in GCAM
As integrated assessment research has evolved, understanding agriculture and land use systems has become increasingly important. Land use interacts with mitigation both as a supplier of bioenergy and as a source or sink of terrestrial emissions. For this reason, GCAM includes a spatially-disaggregated land use model that models land cover, land use, and production of agricultural and forest products, as well as ecosystem types. Energy, agriculture, forestry, and land markets are integrated in GCAM, along with unmanaged ecosystems and the terrestrial carbon cycle. GCAM determines the demands for and production of products originating on the land and the carbon stocks and flows associated with land use. GCAM 3.0 divides the globe into 151 land use subregions based on a mapping of up to 18 AEZs (Monfreda et al, 2009) within each of GCAM’s 14 global geo-political regions, as shown in Figure 1. These AEZs are defined as zones with similar temperature and precipitation levels, and as such are a useful division of land for modeling agriculture and other land use. A complete description of GCAM 3.0 modeling of agriculture and land use is provided in Wise and Calvin (2011) and Kyle et al (2011).
Figure 1. GCAM 3.0 Geopolitical and Land Use Region Map
Within each of these 151 subregions, GCAM categorizes land into approximately a dozen types based on cover and use. Some of the land types in the subregions, such as tundra and desert, are not considered arable. Among arable land types, further divisions are made for lands historically in non-commercial uses such as forests and grasslands as well as commercial forestlands and croplands. Within each subregion, arable land is allocated across a variety of uses based on expected profitability, which depends on the productivity of the land, the non-land costs of production (labor, capital, fertilizer, etc.), and the price of the product. Production of approximately fifteen crops and commercial forest product is currently modeled, with yields of each specific to each of the 151 subregions. The model is designed to allow specification of different options for future crop management for each crop in each subregion, and the model structure itself allows for other regional breakdowns besides the AEZs.
Because it is an IA model, land use modeling in GCAM is not limited to agriculture but is instead comprehensive in scope of modeling land use and land cover. Land in each of each subregions is divided into one of several land use and land cover categories, with the soil and vegetative terrestrial carbon in each category in each subregion modeled (Kyle et al, 2011). These land categories include lands for commercial uses such as cropland, commercial pasture, forest products, and bioenergy crops, as well as non-commercial but arable lands such as non-commercial forestlands, grasslands, and shrublands. For both commercial and unmanaged forestlands, the GCAM models the temporal accumulation of terrestrial carbon based on growth profiles specific to each subregion. Non-arable lands such as tundra, desert, and urban land are also tracked but considered fixed for this study. The amounts of land in each of the arable land categories, including the distinction between commercial and non-commercial land coverage, are not rigid in GCAM. As GCAM models future periods, the amount of land devoted to each of these categories and uses changes in response to socioeconomic, policy, and technology drivers, and the net terrestrial carbon emission or net carbon uptake from these land use changes are computed.
As with the GCAM energy system, the economic modeling approach for GCAM agriculture, forest, and land is that of an integrated economic equilibrium in the products, sectors, and factors that are modeled. Markets, for products such as corn, wheat, wood, or bioenergy crops must be cleared so that supplies are equal to demands in each model period. Depending on the product or on user specifications, markets can be cleared globally, regionally, or across groups of regions.
GCAM models the production of several types of bioenergy: traditional bioenergy (straw, dung, fuel wood, etc.), bioenergy from waste products (including crop residues, municipal solid waste, and black liquor from the pulp and paper industry), and purpose-grown bioenergy crops. Purpose grown bioenergy crops, including perennial grasses like switchgrass, and woody crops such as willow, are modeled as economically competing for land with all other agriculture, forestry, and other uses of land. Food crops, such as corn, soybeans, and sugar, can also be used as energy feedstocks to be supplied to GCAM’s energy transformation and use sectors.
1.3 The Climate System in GCAM
All integrated assessment models must include some meaningful representation of global bio-geophysical processes that govern the fate of greenhouse gas and other anthropogenic emissions. GCAM uses the MAGICC model (Wigley and Raper 2001) as its default biophysical component. MAGICC provides a representation of important physical Earth system elements: carbon cycle, atmospheric chemistry, ocean systems, and climate systems.
MAGICC operates by taking anthropogenic emissions from the other GCAM components, converting these to global average concentrations (for gaseous emissions), then determining anthropogenic radiative forcing relative to preindustrial conditions, and finally computing global mean temperature changes. The MAGICC climate system model is an energy-balance climate model that simulates the energy inputs and outputs of key components of the climate system (sun, atmosphere, land surface, and ocean) with parameterizations of dynamic processes such as ocean circulations.
The carbon cycle in MAGICC is modeled with both terrestrial and ocean components. The terrestrial component includes CO2 fertilization and temperature feedbacks; the ocean component is a modified version of the Maier-Reimer and Hasselmann (1987) model that also includes temperature effects on the terrestrial biosphere. Reactive gases and their interactions are modeled on a global-mean basis using equations derived from results of global atmospheric chemistry models (Wigley et al. 2002).
Global mean radiative forcings for CO2, CH4, and N2O are determined from GHG concentrations using analytic approximations. Radiative forcing for other GHGs is proportional to concentrations. Radiative forcing for aerosols (for sulfur dioxide and for black and organic carbon) is taken to be proportional to emissions. Indirect forcing effects, such as the effect of CH4 on stratospheric water vapor, are also included. Given radiative forcing, global mean temperature changes are determined by a multiple box model with an upwelling-diffusion ocean component. Climate sensitivity is specified as an exogenous parameter.
2 Estimation of Climate Impacts on Crop Yields
The data available from Easterling et al. (2007) is disaggregated by crop type but aggregated over large geospatial regions, dividing the world into two regions (mid-to-high and low latitude). Crop yield impacts depend on many factors including soil and terrain characteristics as well as the regional impacts of changes associated with global climate means. Crop yields can be expected to vary significantly within latitudinal bands and across crop models. In addition, estimates of climate change can be expected to vary across climate models. For this reason, we have opted to use the edges of the distribution.
The effect of climate change is quite different along each of these impact pathways. Figure 1 summarizes wheat yield studies of climate change impacts. The envelope of the high damage (labeled “Most Decrease” in Figure 1) studies shows a monotonic decline in crop yields beginning in the present. In contrast, the envelope of the lowest damage (labeled “Most Increase” in Figure 1) studies show an increase in crop yields rising until climate change reaches almost two degrees Centigrade beyond which further increases in global mean surface temperature has little effect until global mean surface temperature change reaches 3.5 degrees Centigrade. When temperature rises above this level, yields decline from their peaks, but do not return to present crop yield levels until temperature rise more than 4.5 degrees Centigrade.
Similar relationships were extracted for each of the three crops and both latitudinal bands from (Easterling et al, 2007, p286).
3 Estimation of Climate Impacts on Agricultural Systems
Climate impacts on agricultural systems depend on the effects of climate change on crop yields, as discussed above, as well as the response of human agricultural and economic systems to those changes in crop yields through management and policy actions. We estimate the consequences of climate change for agricultural and land-use systems for two cases—Reference and 550 ppm stabilization.
The concentrations of greenhouse gases in the atmosphere for each scenario were estimated using MAGICC, which in turn computed radiative forcing and the transient global mean surface temperature. The associated change in crop yields for the two different latitude classes were then calculated from the above data and these values were used to adjust crop yields relative to the case without climate change impacts.
Because GCAM is a model of human and natural Earth systems crop yield changes trigger changes in the choices that farmer and consumers make. As crop yields decline, farmers produce less per unit land area. The reduced production results in higher crop prices, which in turn increases the value of crop land relative to land in other uses. The land area used to produce crops therefore expands at the expense of other activities. Thus, GCAM incorporates endogenous adaptation to climate impacts driven by the market equilibrium process. That is, it includes shifts in both where crops are grown and the composition of the crops that are grown, as well as shifts in the total demand for crops. GCAM does not, however, include other adaptive responses. For example, it does not consider the development of new crop strains. It does not consider shifts in management practices beyond those included in the underlying source literature assembled by Easterling, et al. Importantly, GCAM does not include the effects of changing water supplies. Thus, a major response to climate change could be to change the extent to which crops are irrigated. However, since water is not explicitly incorporated in GCAM, this line of accommodation is not currently available. Similarly, the effects of reduced water availability would not be incorporated.
4 References
Clarke, L., J. Edmonds, H. Jacoby, H. Pitcher, J. Reilly, R. Richels, 2007. Scenarios of Greenhouse Gas Emissions and Atmospheric Concentrations. Sub-report 2.1A of Synthesis and Assessment Product 2.1 by the U.S. Climate Change Science Program and the Subcommittee on Global Change Research. Department of Energy, Office of Biological & Environmental Research, Washington, D.C., USA, 154 pp.
Edmonds, J. and J. Reilly. 1985. Global Energy: Assessing the Future, Oxford University Press, New York. 1985.
Kim, SH, JA Edmonds, SJ Smith, M Wise, J Lurz. 2006. The Object-oriented Energy Climate Technology Systems (ObjECTS) Framework and Hybrid Modeling of Transportation The Energy Journal Special Issue 2.
Kyle, G. Page, Patrick Luckow, Katherine Calvin, William Emanuel, Mayda Nathan, and Yuyu Zhou. 2011. GCAM 3.0 Agriculture and Land Use: Data Sources and Methods. Pacific Northwest National Laboratory. PNNL-21025. http://wiki.umd.edu/gcam/images/2/25/GCAM_AgLU_Data_Documentation.pdf
Maier-Reimer, E., and Hasselmann, K. 1987. "Transport and Storage of CO2 in the Ocean: An Inorganic Ocean-Circulation Carbon Cycle Model," Climate Dynamics, 2:63-90.
Monfreda, C., N. Ramankutty and T. W. Hertel (2009). “Global Agricultural Land Use Data for Climate Change Analysis.” in Economic Analysis of Land Use in Global Climate Change Policy. T. W. Hertel, S. Rose and R. S. J. Tol, Routledge.
Wigley, T.M.L. and Raper, S.C.B., 2001: Interpretation of high projections for global-mean warming. Science 293, 451-454.
Wigley, T.M.L., Steven J. Smith, and M.J. Prather (2002) Radiative Forcing due to Reactive Gas Emissions. Journal of Climate 15(18), pp. 2690–2696.