Grantee Research Project Results
Final Report: Sensitivity Analysis of the Effect of Changes in Mean and Variability of Climate on Crop Production and Regional Economics in the Southeastern U.S.
EPA Grant Number: R824997Title: Sensitivity Analysis of the Effect of Changes in Mean and Variability of Climate on Crop Production and Regional Economics in the Southeastern U.S.
Investigators: Mearns, Linda , McCarl, B. , Katz, R. , Easterling, William Ewart , Carbone, Greg , Adams, Richard
Institution: National Center for Atmospheric Research , Texas A & M University , University of South Carolina at Columbia , Oregon State University , Pennsylvania State University
Current Institution: National Center for Atmospheric Research , Oregon State University , Pennsylvania State University , Texas A & M University , University of South Carolina at Columbia
EPA Project Officer: Chung, Serena
Project Period: November 1, 1996 through October 31, 1999 (Extended to September 30, 2001)
Project Amount: $1,200,901
RFA: Global Climate (1996) RFA Text | Recipients Lists
Research Category: Ecological Indicators/Assessment/Restoration , Climate Change
Objective:
This project investigated the effects of the spatial scale of climate change scenarios on crop production in the Southeastern United States. We have determined differences in crop responses to different scales of possible future climate, from a control and doubled CO2 experiment of a high resolution regional climate model, and from the coarse resolution General Circulation Model (GCM) that provided boundary conditions for the regional model. We applied the scenarios to crop models (i.e., CERES, CROPGRO, GOSSYM, and EPIC), which were run for conditions of: (1) climate change only; (2) climate change and elevated CO2 effects; and (3) climate change, elevated CO2 effects, and management adaptations. The baseline simulated crop responses were determined from a dense network of about 500 climate stations, organized in a GIS framework, as were other inputs to the crop models.Resulting changes in mean and variability of simulated yield from the different scenarios provided input to an agricultural sector economic model (ASM) for evaluation of economic sensitivity to the different sets of yield changes. The overarching goal was to establish the sensitivity of the regional economics to changes in crop yields resulting from a range of changes in climate at different spatial resolutions.
We also analyzed the effect of large-scale circulation indices on daily climatological characteristics of the region. Such phenomena as the El Niño Southern Oscillation (ENSO), the North Atlantic Oscillation (NAO), and the Bermuda High Index (BHI) were considered.
Summary/Accomplishments (Outputs/Outcomes):
I. High and Low Resolution Climate ScenariosWe analyzed the control runs and 2xCO2 projections of the CSIRO Mk 2 GCM and
the RegCM2 regional climate model (hereafter referred to as RegCM), which was
nested in the CSIRO GCM, over the Southeastern U.S., and developed the output
for use as input to crop models.
The biases of the model control runs are similar. In general, mean temperatures
are too cold, more so for the RegCM than the CSIRO. However, the apparent better
performance of the CSIRO results from compensating errors in the maximum and
minimum temperatures. The RegCM exhibits very small negative biases for minimum
temperature, while those of the CSIRO are larger and positive. Maximum temperature
biases are negative for both models but larger for the CSIRO for the domain
average in three of the four seasons. Precipitation biases are relatively small
in the winter, but both models produce large positive biases in summer, that
of the RegCM being the larger. Spatial pattern correlations of the model control
runs and an observed data set reveal that the RegCM better reproduces the pattern
of precipitation in all seasons compared to the CSIRO, whereas both models reproduce
well the spatial pattern of temperature in all seasons. However, precipitation
patterns are poorly reproduced by both models in spring and fall.
Under climate change conditions the most salient feature from the point of view of scenarios for agriculture is the very large decreases in precipitation (30 percent) in July, August, and September in the CSIRO, which become more extreme (50 percent) in the regional model. Precipitation increases (20 to 40 percent) generally are found in early spring, most uniformly for the CSIRO. In June, for the CSIRO, most of the domain experiences relatively small increases in precipitation (about 10 percent), whereas in the RegCM a more complex pattern emerges, with decreases in precipitation of 30 percent dominating the eastern portion of the domain, but some increases in the south central region. Precipitation decreases around 30 percent dominate in winter in the CSIRO, but a more complex pattern of increase and decrease is exhibited in the regional model. The spatial variability of the changes are of course much greater in the RegCM, which also experiences greater extremes of change.
Temperature increases in the CSIRO GCM tend to remain in the range of 3 to 5ºC, the higher values dominating in winter and spring. However, high (6 to 7ºC) increases in minimum temperature are found in March and April. In the RegCM, temperature increases are much more spatially and temporally variable, ranging from around 1 to 7ºC across seasons. In summer, large increases (7ºC and above) in maximum temperatures are found in the northeastern part of the domain, where maximum drying occurs. In spring (May) small increases (1ºC) in temperature dominate in the Southcentral area, where precipitation substantially increases.
II. Crop Model Applications
CERES/CROPGRO models included: corn, rice, sorghum, soybean, and wheat. Two cotton models were used, COTTAM and GOSSYM, but only the results of GOSSYM were carried through all three climate change cases. EPIC models included corn, soybean, and wheat.
A. Validation of Crop of the CERES/CROPGRO Models
The CERES and CROPGRO models were validated against experimental yields from experimental farms throughout the Southeast. The most complete validation was performed with SOYGRO. It was found that the crop models, in general, reproduced relatively well mean yields at the given locations. SOYGRO also fairly reproduced the interannual variance of the yield; however, no model was adequate at reproducing the specific year-to-year pattern of variability.
B. Applications of Crop Models to Baseline and Climate Scenarios
All crop models were exercised with the baseline (current) climate and two spatial scales of climate change scenarios for three cases: (1) climate change only, (2) climate change plus elevated CO2 effects, and (3) the latter plus adaptations. The level of CO2 used in the crop models was 330 ppm for current conditions, the same for the climate change only case, and 540 ppm for the climate change plus elevated CO2 effect and the adaptation case. Adaptations used were change in sowing dates and change in variety for both the coarse and fine scale climate change scenarios. The elevated CO2 effect was assumed in the adaptation runs. Thirty-six years (1960-1995) of crop yields were produced at each 50 km grid point for each case/scenario.
1. Application of Cotton Models
Initially, the cotton model COTTAM was applied to the climate change scenarios. This model resulted in large yield increases with climate change. After consultation with other crop modelers and cotton experts, we determined that the response of COTTAM to the climate change scenarios likely was unrealistic. We decided to apply a different cotton model, GOSSYM, which we believe provides a more realistic response to the climate changes. This is a much more detailed model, which recently has been updated to be particularly suited for climate change research (R. Reddy, personal communication).
In comparing the results of COTTAM with those of GOSSYM, we note that the baseline yields are very similar for both models except for the far northern states, Tennessee and North Carolina, where the GOSSYM yields are lower. The GOSSYM model is rather sensitive to temperature thresholds, and tends to produce rather low yields in areas that may be marginal for cotton production from the point of view of temperature. For most states, the percentage change in yield regardless of scenario is lower for the GOSSYM model compared to the COTTAM model. Exceptions include Tennessee and North Carolina, where the baseline yields for GOSSYM are relatively low. Here, the improved temperature environment with climate change results in rather large increases in yield, and hence, high percentage increases. Further details on the GOSSYM results for climate change are presented in the next section.
2. CERES/CROPGRO and GOSSYM Models
We examined the response of the CERES/CROPGRO and GOSSYM crop models on three different spatial scales: at the level of the 50 km grid scale (the scale of the observations and regional model), the scale of the entire region (Southeast), and also the scale at which the yield results were aggregated for input to the ASM (state level). On the domain-wide level, for the climate change only case, all crops showed decreases in crop yields compared to the base yields, with decreases significantly larger for the RegCM (fine resolution scenario), except for wheat, for which there was no significant difference between the scenario yields (see Table 1). For the elevated CO2 case, simulated yields of course improve, but the change from base conditions remains negative for all crops/scenarios, except for cotton for the CSIRO scenario, where an increase of 8 percent is found (see Table 1). With adaptation domain, wide percentage changes in yield for corn, cotton, and rice become positive for both scenarios.
Table 1. Percentage change in dryland crop yields for the three climate change cases for all crops (CERES and GOSSYM models), Southeast domain-wide averages.
% Change from Base Yield
|
|||||||
CROP |
Simulated Observed (T/ha)
|
CSIRO
330CO2 |
RegCM
330CO2 |
CSIRO
540CO2 |
RegCM
540CO2 |
CSIRO
540+A |
RegCM
540+A |
Corn |
8.1
|
-13
|
-16
|
0
|
-2*
|
+7
|
+6*
|
Cotton |
1.2
|
-4
|
-17
|
+8
|
-3
|
+29
|
+18
|
Rice° |
9.6
|
-16
|
-19
|
-3
|
-5
|
+2
|
+6
|
Sorghum |
6.0
|
-36
|
-51
|
-26
|
-42
|
-17
|
-28
|
Soybean |
2.4
|
-49
|
-69
|
-26
|
-54
|
-8
|
-46
|
Wheat |
4.5
|
-36
|
-32*
|
-26
|
-21*
|
-25
|
-21*
|
°Irrigated (paddy) | |||||||
*CSIRO and RegCM yields are NOT significantly
different ( = 0.05) CSIRO 330CO2: coarse-scale climate change only; RegCM 330CO2: fine-scale climate change only; CSIRO 540CO2: coarse-scale elevated CO2; RegCM 540CO2: fine-scale elevated CO2; CSIRO 540+A: coarse-scale adaptation; RegCM 540+A: fine-scale adaptation. |
Wheat, which showed the least change in yield with adaptation, remained negative, as did sorghum and soybean. Larger losses to yield remain with the fine scale scenario.
When we considered the changes in yield aggregated at the state level (the level of aggregation needed for the economic model), contrasts were more complex. For example, corn yields, which had been similar for both scenarios on the domain-wide scale, exhibit greater scenario contrasts on the state level (see Table 2). For the climate change only case, the CSIRO scenario now produces significantly larger decreases for most states, compared to the RegCM scenario, particularly in the South central and Delta states. With elevated CO2, five states show increased yields with the fine scale scenario, but only Alabama and Arkansas for the coarse scale. With adaptation, corn yields increased for most southern states for both scenarios, but still decreased for North and South Carolina and Tennessee for the fine scale scenario. Also, contrasts in yield changes between the scenarios diminished under the adaptation case, indicating that adaptations mitigate the unique spatial scale effects of the climate change scenarios.
3. EPIC Models
Domain wide results for the climate change only case show decreases for both scenarios for corn and soybean, with larger decreases with the fine scale scenario. For wheat, the change is not different from zero for either scenario. The yield decreases under climate change only conditions were, in general, not as great as for the CERES models. Contrasts between the scenarios are significant on the state level for corn and soybeans, but not for wheat. With elevated CO2 , domain-wide percentage changes range between about 10 and 25, the largest values being for wheat. On a state level, percentage change in yield is positive for all crops in all states. For soybean, the CSIRO scenario produced larger increases for all states, except Arkansas and Florida, compared to the RegCM scenario. As with CERES, little contrast in
Table 2. Percentage change in dryland CERES corn yields for states in the Southeast, for all three climate change cases
% Change from Base Yield
|
|||||||
State | BaseYield (T/ha) | CSIRO 330CO2 |
RegCM 330CO2 |
CSIRO 540CO2 |
RegCM 540CO2 |
CSIRO 540+A |
RegCM 540+A |
Alabama | 7.6 | -12 | -5 | +3 | +9 | +13 | +20 |
Arkansas | 7.4 | -12 | -5 | +2 | +9 | +22 | +25* |
Florida | 6.3 | -21 | -7 | -5 | +10 | +3 | +11 |
Georgia | 7.5 | -19* | -16 | -5 | -1* | +1 | +6* |
Louisiana | 7.9 | -14 | -9 | -2 | +3 | +6 | +4* |
Mississippi | 8.4 | -14 | -8 | -3 | +3 | +9 | +12* |
North Carolina | 9.1 | -15 | -34 | -1 | -18 | +1 | -9 |
South Carolina | 8.3 | -19 | -30 | -5 | -14 | +0 | -6* |
Tennessee | 9.8 | -14 | -25 | -3 | -12 | +1 | -5 |
Domain Mean | 8.1 | -13 | -16 | 0 | -2 | +7 | +6 |
wheat yields between the scenarios was found in all three cases. For both climate change cases and all three crops, results for EPIC are more positive than for CERES, but the effect of the spatial scale of the scenarios is still seen. A different approach was taken for EPIC adaptations. The effect of transient climate change was mimicked by dividing up the climate change into discrete periods. Agronomic adaptations were lagged behind evolving climate changes to simulate farmer dependence on historical climate information in making current management decisions. Results clearly show that adaptation is less effective when the assumption of farmer "clairvoyance" is relaxed.
II. Economic Modeling ASM
A. Crop Yield Changes for the Rest of the U.S.
To usefully model the response of the ASM to the different sets of changes in yield, changes in yield in the rest of the United States were needed, as well as changes in yields of crops necessary for the economic model, but not explicitly modeled in our study. We developed a coarser resolution observed climate data set for the rest of the United States, and also we developed two different scales of climate change scenarios based on the CSIRO results over the rest of the U.S. and on earlier runs of the RegCM for the western two thirds of the United States (Giorgi, et al., 1998) and the Great Lakes region (G. Bates, personal comunication), which also used the boundary conditions from the CSIRO/GCM.
We applied these scenarios to the appropriate CERES/CROPGRO and GOSSYM crop models for the different units of the United States (usually states) used in the economic model. The changes in yields for crops we did not explicitly consider (e.g., citrus, tomatoes, potatoes) were determined via a proxy method based on other crop simulations for similar studies of agricultural crop yield change over the United States.
B. The ASM and Final Crop Yield Input for ASM
The ASM is a spatially disaggregate model that simulates the economic equilibrium that arises in the U.S. agricultural sector. It considers the regional impacts of yield changes with endogenous price adjustments. Percentage changes in crop yields for all 63 primary spatial units of the ASM were produced, based on the two different climate changes scenario resolutions, for the elevated CO2 and adaptation cases, and served as input to the economic model.
C. ASM Results
1. Effects of Mean Yield Changes
On a country-wide basis, for the elevated CO2 case, the coarse scale scenario results in a large increased total surplus (or economic welfare) for the agricultural sector, but the fine resolution scenario produced a negligible increase. Change in total surplus depends primarily on the relative changes in consumer surplus and producer surplus.
Regional index numbers for the total value of production, which is a measure of economic activity within the regions, show interesting contrasts across the regions, based on the scenarios (see Table 3). The Southeast and Appalachia show the largest decreases in activity for both scenarios, but the decrease with the fine scale scenario is much larger. In the Cornbelt, the coarse scale scenario results in a gain in activity, while the fine scale scenario shows a clear decrease. Also note that considering all regions, there is greater variability of regional response with the fine scale scenario compared to the coarse scale.
With adaptation, the total surplus increases further for both scale scenarios, and the contrast between the scenarios narrows, but is still discernible. Regarding regional economic activity, conditions improve for most regions, but adaptations do not overcome the negative impacts of the climate change in all regions. For the coarse scale scenario, three regions (Appalachia, Southeast, and Delta States) still experience decreased economic activity, while four regions still experience this for the fine resolution scenario (see Table 3). The scenario/cases are CSIRO: coarse scale elevated CO2; and RegCM: fine scale elevated CO2; CSIROA: coarse scale adaptation; RegCMA: fine scale adaptation.
Table 3. Regional Economic Index Numbers From the ASM
Scenario/ Case | North East |
Lake States |
Corn Belt |
North Plains |
Appalachia | South East |
Delta States |
South Plains |
Mountains | Pacific |
CSIRO | 118 | 145 | 107 | 110 | 82 | 78 | 84 | 135 | 138 | 121 |
RegCM | 99 | 137 | 79 | 122 | 72 | 67 | 80 | 147 | 129 | 127 |
CSIROA | 122 | 158 | 116 | 123 | 85 | 83 | 86 | 147 | 144 | |
RegCMA | 106 | 146 | 83 | 141 | 79 | 80 | 86 | 163 | 138 | 138 |
The scenario/cases are: CSIRO: CSIRO (coarse) climate change + direct CO2 effect; RegCM: RegCM (fine) climate change + direct CO2 effect; CSIROA: CSIRO (coarse) + adaptation; RegCMA: RegCM (fine) + adaptation. Regions listed are the mega-regions used in the ASM. The base case has a value of 100.
Our analysis indicates that the spatial scale of climate change scenario substantially affects the simulation of changes in crop yields on various levels of spatial aggregation. Moreover, we have demonstrated that these contrasts in changes in yield are substantial enough to affect the results of an agricultural economic model, both on national and regional levels.
2. Additional Effect of Changes in Variability of Yield
The effect of yield variability was incorporated into the ASM. This version of the ASM is referred to as the Stochastic Programming with Recourse ASM or SPARM.
The inclusion of risk (variability) into the analysis changes the basic direction of total welfare effects. Specifically, the non-stochastic results discussed above (elevated CO2 effect case), (small positive in economic welfare for RegCM, and large positive change in economic welfare for CSIRO) are substantially altered. In the CSIRO scenario, economic welfare increases further (but only slightly), while for the RegCM scenario, welfare becomes substantially negative. This clearly indicates that including the effect of changes in yield variability can significantly affect economic results, and it presents the spatial scale effect of the climate scenarios in a more striking light.
III. Conditional Stochastic Models for Daily Weather in Southeastern United States
Teleconnections between patterns in large-scale atmosphere-ocean circulation
(e.g., El Niño-Southern Oscillation, North Atlantic Oscillation), and
climate in the Southeastern United States are well documented. Generally, such
teleconnections have been detected for
monthly or seasonal mean temperature and total precipitation averaged spatially
over subregions of the Southeast. Having stronger and more consistent correlations
than ENSO or NAO, a Bermuda High Index was considered. We attempted to convert
the connections with BHI into the corresponding effects on local daily statistics
of temperature and precipitation. Stochastic models for times series of daily
maximum and minimum temperature and precipitation amount were fitted conditional
on a two-state BHI (i.e., above or below normal). A shift in the mean, but not
necessarily the variance or autocorrelation, of daily minimum and maximum temperature
was found depending on BHI. Because of the relatively weak signal for seasonal
total precipitation, a statistical technique known as "borrowing strength"
was applied, constraining the relationship between daily precipitation and BHI
spatially. Under this constraint, a shift in the probability of occurrence of
precipitation, but not necessarily the mean or variance of the amount of precipitation
on wet days, is found depending on BHI.
Journal Articles on this Report : 8 Displayed | Download in RIS Format
Other project views: | All 38 publications | 10 publications in selected types | All 8 journal articles |
---|
Type | Citation | ||
---|---|---|---|
|
Adams RM, McCarl BA, Mearns LO. The effects of spatial scale of climate scenarios on economic assessments: an example from U.S. agriculture. Climatic Change 2003;60(1-2):131-148. |
R824997 (Final) |
Exit |
|
Carbone GJ, Mearns LO, Mavromatis T, Sadler EJ, Stooksbury D. Evaluating CROPGRO-soybean performance for use in climate impact studies. Agronomy Journal 2003;95(3):537-544. |
R824997 (Final) |
Exit Exit Exit |
|
Doherty RM, Mearns LO, Reddy KR, Downton MW, McDaniel L. Spatial scale effects of climate scenarios on simulated cotton production in the Southeastern U.S.A. Climatic Change 2003;60(1-2):99-129. |
R824997 (Final) |
Exit |
|
Easterling WE, Chhetri N, Niu X. Improving the realism of modeling agronomic adaptation to climate change: simulating technological substitution. Climatic Change 2003;60(1-2):149-173. |
R824997 (Final) |
Exit |
|
Katz RW, Parlange MB, Tebaldi C. Stochastic modeling of the effects of large-scale circulation on daily weather in the Southeastern U.S. Climatic Change 2003;60(1-2):189-216. |
R824997 (Final) |
Exit |
|
Mearns LO, Giorgi F, McDaniel L, Shields C. Climate scenarios for the Southeastern U.S. based on GCM and regional model simulations. Climatic Change 2003;60(1-2):7-35. |
R824997 (Final) |
Exit |
|
Mearns LO, Carbone G, Doherty RM, Tsvetsinskaya E, McCarl BA, Adams RM, McDaniel L. The uncertainty due to spatial scale of climate scenarios in integrated assessments: an example from U.S. agriculture. Integrated Assessment 2004;4(4):225-235. |
R824997 (Final) |
Exit Exit |
|
Tsvetsinskaya EA, Mearns LO, Mavromatis T, Gao W, McDaniel L, Downton MW. The effect of spatial scale of climatic change scenarios on simulated maize, winter wheat, and rice production in the Southeastern United States. Climatic Change 2003;60(1-2):37-72. |
R824997 (Final) |
Exit |
Supplemental Keywords:
ecosystem, regionalization, integrated assessment, scaling, downscaling., RFA, Scientific Discipline, Air, Geographic Area, climate change, Economics, Southeast, Atmospheric Sciences, Ecological Risk Assessment, Agronomy, environmental monitoring, sensitivity analysis, economic models, green house gas concentrations, regional ecosystems, socioeconomic indicators, carbon dioxide, circulation model, climate models, CO2 concentrations, GENESIS climate model, environmental stressors, agriculture, ecosystem sustainability, climate variability, crop productionRelevant Websites:
http://www.esig.ucar.edu/soeast Exit
Progress and Final Reports:
Original AbstractThe perspectives, information and conclusions conveyed in research project abstracts, progress reports, final reports, journal abstracts and journal publications convey the viewpoints of the principal investigator and may not represent the views and policies of ORD and EPA. Conclusions drawn by the principal investigators have not been reviewed by the Agency.