We explore sensitivity to spatially heterogeneous future impacts by incorporating a subset of the 35 yield change scenarios from the AgMIP global gridded crop model intercomparison study into GCAM’s exogenous agricultural assumptions. The scenarios selected have been found to span a range of global impacts in previous GCAM versions. Each GGCM assesses the effects on yields of RCP 8.5 changes in CO2 , temperature, and precipitation from five bias-corrected global circulation models for rainfed and perfectly irrigated versions of crops. Water for irrigation is constrained in the version of GCAM used for analysis. We apply crop impacts in one of two ways: only in the U.S. and everywhere . This work is consistent with earlier work using different models to consider similar scenarios regarding the importance of international impacts, adding robustness to the findings through the use of a different multi-sector model. We find that international impacts could be as important as domestic impacts for the financial value of U.S. agricultural crop production in across spatially and temporally varying agricultural impacts scenarios. Crucially, there are scenarios in which examining only domestic impacts would lead to a fundamentally different analysis of the future of U.S. agricultural crop production than if one considered the combined effects of domestic and international impacts. Therefore, while there is uncertainty about climate impacts on future crop yields, evidence suggests that the importance of international effects on the financial value of U.S. crop production is robust across this uncertainty.GCAM couples human and physical Earth systems to explore the impacts of economic and environmental policies. GCAM is calibrated to historical data through 2010 and then simulates forward from 2010 to 2100 in five year time steps by incorporating changes in quantities such as population, GDP, technology, and policy to produce outputs that include land use, emissions, and commodity prices. Specifically, GCAM can assimilate high spatial resolution information on the global distribution of crop yields and analyze its effects on the coupled system of global agriculture markets. This and previous versions of GCAM track long term trend behavior rather than inter annual variability . All scenarios in this study follow Shared Socioeconomic Pathway 2,vertical farm tower the “middle-of-the-road” socio-economic scenario.
The associated GCAM scenario provides the reference against which we measure impacts in this study. GCAM represents the energy system in 32 economic regions, and it represents global production in 384 agricultural and land-use regions. Each of the 384 land units in GCAM represents a water basin-economic region combination. Twenty-three of these lie within the U.S. With each land unit, GCAM allocates land across more than a dozen types based on cover and use. Allocation is based on a logit formulation to optimize profitability, with details provided in Wise et al . Important to this study, GCAM models production of a range of agricultural commodities , each with four different management types . S1 Fig illustrates the land competition nests used by GCAM v5.2 in each GCAM land unit . S1 Table in the S1 File provides the mappings between crops in the FAOSTAT database and GCAM commodities. Some GCAM commodities are straightforward , but some GCAM commodities are economic aggregates, such as Misc Crop and OilCrop . The span of GCAM commodities allows a relatively full modeling of the agricultural sector of economies. On the supply side of agricultural crop production in GCAM, the “no impacts” baseline yield change assumptions are read in exogenously. These yield changes are used by GCAM to calculate the profitability of a GCAM crop-irrigation-fertilizer combination in each GCAM land unit at each time step. This profitability determines land allocated to each land type . The combination of exogenous yields and endogenous land allocation gives production of each crop-irrigation-management combination in each land unit. Because land shares allocated to rainfed versus irrigated, high versus low fertilizer versions of each crop may change, the aggregate yield for each crop output by GCAM will differ from the input yields. In other words, there is endogenous yield intensification in GCAM. AgMIP yield impacts are incorporated as multipliers on the exogenous yield assumptions used by GCAM .The GCAM food demand system creates a slightly elastic portion of demand for each crop type, based on the exogenous population and GDP assumptions GCAM takes as inputs. Therefore, the minimum quantity defined by the food demand system must be met globally by GCAM agricultural crop production.
Other demand sectors are more elastic. For some GCAM agricultural commodities, such as Corn and Oil Crop, this leads to an overall more elastic total demand function because GCAM explicitly models the energy sector and the accompanying price-sensitive demand for use of these crops as bio-fuels. Crops such as Wheat and Rice that are primarily used to feed humans have nearly perfectly inelastic demand. This lends an extra layer of difficulty to analysis of the dynamics in any particular agricultural impacts scenario. Since both supply and demand schedules may shift, there is never a single mechanism that may be identified for any particular price change. GCAM includes mechanisms on both the supply and demand side that allow for adaptation behavior. Specifically, because there is a price elasticity of demand for GCAM agricultural crop commodities , reduction of quantity demanded is one available mechanism in the model. This includes changing demand for animal feed and bio-energy in response to changing prices. On the supply side, economic agents can endogenously adjust land allocation in response to changes in profit rates between rainfed or irrigated versions of crops . Additionally, the option to shift to higher nitrogen fertilizer per unit of land is included, which leads to an increase in both the yield and the cost per unit of land . Pesticide use is currently not explicitly modeled, and therefore changes to pesticide use are not included as an adaptation option in GCAM. Finally, GCAM includes trade of agricultural crop commodities and the ability for producers to shift land allocation among commodities as profit rates change with yield and price changes. A significant difference between the version of GCAM used in this paper and the version documented in detail, is that GCAM 5.2 features different agricultural crop trade behavior. Previously, GCAM modeled completely flexible trade with nearly all agricultural commodities traded freely on a global market with no explicit distinction between domestically-produced and imported products and a common global market price. The version of GCAM used for this study employs a system that is based on an Armington distinction between domestically-produced and imported products . In this new approach in GCAM, we specify region-specific agricultural markets at the 32-region level. Regional demand is an explicitly modeled choice between domestic production and imports from the global market via a nested logit structure, similar to our modeling of land use allocation in GCAM .
International trade is not modeled as explicit bilateral trade but instead as a single market for each commodity that contains all regional gross imports and gross exports. Additional details may be found in the S1 File.We examine the implications of agricultural impacts to U.S. producers through the use of four varying crop-climate model combinations in the AgMIP/ISIMIP global gridded crop model inter comparison study CO2 fertilization effects, driven by RCP 8.5 earth system changes. RCP 8.5 is a climate scenario that features large changes in temperature, CO2, and precipitation and therefore relatively larger local changes in crop yields produced by crop models than under the other RCPs. It was selected with the idea that the larger yield change signal in RCP 8.5 would aid in identifying the emergent mechanisms that dictate the future economic changes resulting from yield changes. The other RCPs are deserving of future examination, but one would expect that the policies applied to reduce emissions would then interact with the impacts of correspondingly smaller yield changes. The GCAM reference scenario to which these crop yield impacts are applied has no climate policies in place. We use AgMIP global gridded yields from each of the EPIC and LPJmL crop models driven by five GCMs under RCP 8.5. Each scenario models yields for different collections of crops. Further, these collections of crops differ from the commodities modeled in GCAM. Yield data is available for both perfectly irrigated and locally rainfed versions of each crop. These gridded yields are aggregated to the GCAM land units using MIRCA2000 harvested area data for weighting. At the basin scale, time series of yields are converted to multipliers by dividing by the historical baseline average yield for each crop-irrigation-basin combination. Finally, these crop-irrigation-basin specific multipliers are converted from the crop model specific crop types to multipliers for each GCAM commodity irrigation-basin combination using GTAP harvested area weights from the GCAM data system for aggregation. This method for incorporating climate driven yield changes as multipliers on GCAM’s exogenous yield assumptions follows methods used in the broader literature. GCAM’s biomass crop commodity receives the median of impacts to all other commodities,vertical plant tower for each irrigation-management practice combination, in each basin. Fig 1 is a schematic summarizing this processing pipeline. S2 Table in S1 File lists the mapping between AgMIP model crops and GCAM commodities used to estimate multipliers. Corresponding water supply constraints for irrigated crops are used for each scenario. Impacts are not applied to grassland or forest to isolate the role of impacts on crop yields.Results are presented for a variety of physical and economic output variables from GCAM. Due to both the variability and the relatively small sample size of spatially heterogeneous climate-crop impact scenarios available for this paper, summary statistics across scenarios are not presented. Rather, the results presented here focus on relationships between the Domestic and the Full scenarios that emerge in each of the different climate-crop impact combinations considered.
To this end, Figs 2 and 3, and S3 Fig illustrate the percent change relative to the no impacts reference GCAM scenario for several economic and physical variables in 2050 at the aggregate U.S. level for different GCAM commodities, under for all scenarios in the work. Fig 2 focuses results on Corn and OilCrop, globally important commodities for which the U.S. is a major producer. Fig 3 presents results for Rice and Wheat, important commodities with more spatially distributed production across the globe. S3 Fig presents results for the remaining GCAM commodities. The variables plotted include Area allocated to each commodity , the U.S. commodity Price, the production Prod of each commodity , and the aggregate endogenous yield change EndYld . Results are also included for changes in revenue Revto examine an aggregate direction of change between Price and Prod, as both variables are important to U.S. producers.Price changes are a primary economic mechanism through which international yield changes are transmitted to U.S. agricultural producers. Because agricultural commodities are traded across the globe, the prices of these commodities are affected by events that occur both in the U.S. and internationally. An event that affects U.S. yields will have consequences for U.S. production but also for prices in the rest of the world and will therefore affect production throughout the world agricultural system. The reverse also holds. Because the U.S. is a significant but not majority contributor to the global agricultural market for most crops, the magnitude of U.S. price changes in the Domestic is consistently smaller than in the Full scenario, across all climate-crop impact combinations for all GCAM commodities, illustrated in Figs 2 and 3 and S3 Fig. This is regardless of whether prices have increased or decreased relative to the baseline scenario, consistent with many of the results reported. Further, applying impacts in the Full case versus in the Domestic results in differing price changes from reference, even for commodities such as Corn that display very similar values in the Full and Domestic cases for physical variables such as area, production, and endogenous yield. It is possible that the relationship between the Full impacts price change from reference and the Domestic impacts price change from reference may break down as the structure of the system being modeled fundamentally changes. For example, in a more restrictive trade scenario, U.S. producers would be more restricted in their options to respond to future climate and prices would be increasingly dictated solely by the direct impacts on U.S. productivity. For commodities such as Corn and Oil Crop, for which the U.S. is a significant producer and exporter, a major shock only to U.S. production is closer in magnitude to a shock across the entire world than it would be for other commodities.