Temperatures rise modestly (1-2°C) by 2030 and precipitation increases by 25 to 125 mm y^-1 over most of the corn-growing region. By 2095, temperatures increase by 2.0 to 3.5°C and precipitation increased by more than 175 mm y^-1 over the entire region. Yield in the EPIC model used for this analysis is directly proportional to biomass production which is favored by a reduction in cold stress and a lengthening of the growing season in the Lake region, Cornbelt and Northeast (Fig. 3.1). Table 3.1 shows that yields increase at current CO₂ and improve still more at the higher concentration. Yields are slightly depressed in the Delta, Appalachian and Southeastern region where higher temperatures shorten the growing season (Fig. 3.1). With no CO₂-fertilization, regional yields are reduced in both 2030 and 2095 in the Delta and Southeast but only in 2030 in the Appalachian region. Climate-related losses are more than offset by CO₂-fertilization in all cases.

Baseline winter wheat yields and deviations due to the climate and CO₂ scenarios are shown for the Northern and Southern Plains, Mountain (Great Plains portions of Montana, Wyoming and Colorado) and the Western regions in Figure 3.2 and summarized by USDA production region in Table 3.2. Temperatures in these regions increase by 1-2°C by 2030 scenario but are considerably higher by 2095. By 2030, precipitation increased by 25-50 mm y^-1 over much of the Plains and Mountain growing regions and in Washington and Idaho, but was lower in California. By 2095 precipitation increases still more in the Plains and Mountain regions, increases in California and is variable in the northwestern states. CO₂-fertilization alone increases yields in all regions. The C-3 crops, of which wheat is one, experience increased photosynthetic and decreased transpiration rates under elevated CO₂. The reduction in transpiration is particularly important for wheat, which is generally grown in semi-arid regions. Aggregate regional production increases under all scenarios in the Pacific region, and most scenarios in the Mountain and Plains regions. Decreases in aggregate production were predicted for the Mountain and Plains regions when wheat growth was simulated without CO₂-fertilization effect in 2030 and for the Southern Plains in 2095 also without the CO₂-fertilization effect. Higher temperatures reduce the frequency of cold stress and increase the length of the growing season by shortening the winter dormancy period. In the more northerly regions the crop matures before the extreme heat of summer.

3.3 National level yield changes

Translating crop simulation results into data that can be used in a national economic model raises two methodological issues: (1) How to treat crops for which crop simulations were not conducted? and (2) How to extrapolate from sites to regional scale impacts? In this section we discuss these issues and present the aggregate national yield changes derived from our extrapolation assumptions.

With regard to omitted crops, the basic issue is that production and resource effects in the economic model depend on relative changes in yield and water use among crops. As a result, the production of crops omitted from the simulation studies are affected in the economic model even if no direct climate effects are assumed for them. This could create regional and resource use shifts that reflected the relative importance of omitted crops rather than the estimated climate effects. Left unaffected by climate change, the omission of impacts on some crops could lead to a bias in the estimate of the overall economic impact of climate change. Generally improving conditions would be underestimated if no yield increase were included with the converse true if conditions were generally worsening, leading to an underestimate of the impact, either positive or negative. Yield changes of omitted crops could potentially be opposed to the general direction of other crops, their omission leading to an overestimate of impact. These considerations lead to the conclusion that, for assessment purposes, it is useful to make a best guess for these omitted crops.

We assumed that, for each omitted crop, one of the crops for which yields were simulated in the crop studies could serve as a proxy, an assumption that has been commonly used. The exception was cotton, an economically important crop for which there was no obvious proxy crop. For cotton, we adapted the results of NCAR/Southeastern US study. The specific approach for each omitted crop is discussed below.

Proxy Crops. A direct proxy crop approach was used for the crops as shown below. For example, silage sensitivity was assumed to be the same as corn sensitivity.

Crop with missing data	Crop used as proxy
Hard Red Spring Wheat	Wheat
Hard Red Winter Wheat	Wheat
Soft Wheat	Wheat
Durham Wheat	Wheat
Barley	Wheat
Oats	Wheat
Silage	Corn
Oranges, fresh	Oranges
Oranges, processed	Oranges
Grapefruit, fresh	Oranges
Grapefruit, processed	Oranges
Tomatoes, processed	Tomatoes
Tomatoes, fresh	Tomatoes
Sugarcane	Rice
Sugarbeet	Hay

Cotton. Cotton yields were developed based on estimates from an NCAR study (Mearns, forthcoming) of the Southeastern US that included several additional crops. The study simulated yield effects using many of the same crop models used in our assessment and for several climate scenarios including the Hadley Center scenario. However, a climate representative of 2060 was used and no simulations were conducted based on the CCC modeled climate. Comparing yield effects among the NCAR crops showed no one of the other crops to respond similarly to cotton, offering evidence that no single crop would serve as a proxy for cotton. An attempt to statistically relate cotton yields using multiple regression analysis to the yields of all other crops verified the conclusion that no single crop nor any combination of crops explained the site level variation in yield impacts of cotton. The approach adopted instead, was to adapt the NCAR cotton yield sensitivity data directly as explained in the document underlying this section (McCarl, 2000). Operationally this involved extrapolating the 2060 spatial distribution of cotton yield and water use sensitivity from the NCAR study to 2030 and 2090 based directly on the climate in these years relative to the Hadley climate for 2060.

The intent of these assumptions is to avoid a particular bias of underestimating overall economic impacts of climate on the US agriculture economy by assuming no effect at all on these crops. Crop coverage has been an issue in all assessments of this type. Early agricultural assessments were often limited to the corn, wheat, rice, and soybeans. Recent assessments, including this one, have worked to provide broader crop coverage. Caution is obviously warranted in using the detailed crop results from the economic model where the crop yield effects were not simulated directly. These uncertainties also introduce uncertainties in the overall economic results. In a very limited way, we explored this uncertainty by simulating the economic model using the different approaches we developed for cotton.

With regard to extrapolation from site-level data, the ASM model includes 63 regions (Figure 3.3 with overlay of the USDA production regions). The crop simulations were done for no more than 46 sites with some crops simulated for only a subset of these sites. In some cases multiple sites were located in a single ASM region. When multiple simulation sites appeared in a region an unweighted average across those sites was used. Proxy regions were used for those regions in which no sites were located. Adjacent regions were used for proxies. The use of adjacent regions as proxies is discussed in greater detail the underlying supporting document (McCarl, 2000).

As with the use of crop proxies, the lack of direct estimates for a site within a region introduces considerable uncertainty in estimates for that region. Even for regions with site estimates, a sample of one or two sites may not be representative of the region. The PNNL crop yield model results were based on a far denser selection of representative sites. As a sensitivity analysis, we used the available PNNL results in the economic model as a substitute for the coordinated site-level results. While not a pure test of potential site selection bias because the crop models differed as well as other aspects of the scenarios, this comparison offers a check on the potential bias introduced by the limited number of sites.

These assumptions provide the basis for estimating yield impacts for all crops in each region of the ASM. The national average change in yields for dryland and irrigated crops with and without adaptation are given in Tables 3.3a,b and 3.4a,b. Table 3.5a,b gives the national results for changes in water use on irrigated crops. The national averages were constructed by weighting ASM regional estimates generated from the crop model results as described above by harvested acreage in each ASM region where the weights are based on data from the 1992 National Resource Inventory (NRI). McCarl (2000) provides additional details.

The estimates in Tables 3.3 through 3.5 are a summary of input into the ASM model. Actual national production depends on changes in the agricultural economy induced by these changes. The estimates are, however, a useful intermediate result that summarizes the crop modeling simulations. The site simulation results by themselves can give a misleading impression of overall impacts because crops were simulated at many sites where little of the crop is currently grown or at sites under dryland conditions where the crop is mainly grown only with irrigation. Weighting results for the site by area provides a better guide to how climate would affect production. These tables also provide input data for the ASM based on the PNNL crop results. PNNL modeled only corn, wheat, hay, and soybeans. Crops other than these (and those for which one of these crops were proxies) have identical changes as for the core results for the Hadley center climate scenario. These entries are shaded in the table.

The results vary across crops, time periods and climate scenarios but some broad patterns emerge.

• Even without adaptation the weighted average yield impact for many crops grown under dryland conditions across the entire US is positive under both the Canadian and Hadley Center climate models. In many cases, yields under the 2030 climate conditions are improved compared with the control yields under current climate and improve further under the 2090 climate conditions. These generally positive yield results are observed for cotton, corn for grain and silage, soybeans, sorghum, barley, sugar beet, tomatoes, and citrus fruit. The yield results are mixed for other crops (wheat, rice, oats, hay, sugar cane, and potatoes) showing yield increases under some conditions and declines other conditions.
• Changes in irrigated yields, particularly for the grain crops, were more often negative or less positive than dryland yields. This reflected the fact that under these climate scenarios precipitation increases were substantial. Precipitation increases do not provide a yield benefit to irrigated crops because no water stress occurs because all the water needed is provided through irrigation. Higher temperatures sped development of crops and reduced the grain filling period thereby reducing yields. For dryland crops the negative effect of higher temperatures was counterbalanced by the positive effect of more moisture.
• Water demand by irrigated crops dropped substantially for most crops. The faster development of crops due to higher temperatures reduced the growing period and thereby reduced water demand more than offsetting increased evapotranspiration due to higher temperatures while the crops were growing. To a large extent the reduced water use thus reflects the reduced yields on irrigated crops. Increased precipitation also reduced the need for irrigation water.
• Adaptation contributed small additional gains in yields of dryland crops, particularly for those with large yield increases due to climate change. Adaptation options were considered for both sites with losses and those with gains but, for the most part, had little additional benefit where yields increased from climate change. This suggests that adaptation may be able to partly offset changes in comparative advantage across the US that results under these scenarios. Other strategies for adaptation such as whether to switch crops or to irrigate or not are part of the economic model. The decisions to undertake these strategies are driven by economic considerations: i.e. whether they are profitable under market conditions simulated in the scenario. We did not consider adaptation for several crops because the measures we considered such as planting date were not applicable to many perennial and tree fruit crops. Adaptation studies were conducted for only a limited number of sites.
• Adaptation contributed greater yield gains for irrigated crops. Shifts in planting dates are able to reduce some of the heat-related yield losses. With higher yields than in the not adaptation case, water demand declines were not as substantial. Again, this reflected the fact that the adaptations considered extended the growing (and grain-filling period) and this extension meant a longer period over which irrigation water was required.

The PNNL results for dryland crop yields are very similar in most cases to those estimated using the more detailed site-level crop models. PNNL did not consider adaptation. PNNL also only considered irrigation for corn and alfalfa. The PNNL results for these irrigated yields differ substantially from the site-level models. Whereas the site-level models show yield losses and reductions in irrigation water use, the PNNL results show yield gains. In the site-level models, higher temperatures speed development of the crop and reduce yield and water demand. The EPIC model on which the PNNL results are based do not show this negative effect of temperature, instead temperature is increasing yield.

3.3.1 Irrigation water supply

Water supply for irrigation is also an important consideration. The ASM includes a description of agricultural water supply that is allocated to crops. An estimate of the change in water supply under the climate scenarios considered was derived from simulated total water supply changes developed in the Water Sector Assessment effort of the National Assessment. The critical assumption made was that the change in water supply to agriculture was proportional to the change in total water supply; i.e. that agriculture and non-agricultural users faced the same proportional change in water supply. More detail on the specific changes and how they were derived from the estimates developed by the Water Sector Assessment are provided in McCarl (2000).

3.3.2 Crop input usage

Yield changes can also imply changes in some inputs such as chemical inputs and those related to crop harvesting, drying, and storage. A larger (or smaller) yield will require more (or less) of these other inputs. This association between yield and input use can be seen over time. As technical progress that increased yield has been accompanied by increases in input usage. On the other hand, yield, by definition, is per unit of land and other inputs such as labor and water are more closely related to area than to yield. As part of the earlier EPRI study (Adams et al., 1999) using the ASM model, a yield-input relationship was estimated. Land, labor, and water inputs were excluded from the estimation. For most crops the increase in use of these other inputs was 40 percent of the yield change. Thus if yield went down by one percent crop input use went down by 0.4%. Similarly a two percent yield increase would be matched by a 0.8% input usage increase. This relationship was included in the simulations. It has the effect of making yield improvements less economically beneficial than they otherwise would be because to obtain the increases requires purchase of these other inputs. Conversely, yield losses are as economically costly because purchase of material inputs is reduced. This type of adjustment is appropriate for the consideration of ongoing climate change, for which technical change, the basis for the estimate, provides a good analogy.

3.4 Livestock performance and grazing and pasture usage

Much of the work on climate change impacts on agriculture considers mainly impacts on crops and only indirectly impacts on the livestock sector through changes in crop yields. Temperature change can also cause livestock to achieve altered rates of gain. As part of the earlier EPRI study (Adams et al., 1999) changes in livestock performance due to temperature change were estimated. These estimates were used as a basis for developing temperature-related declines in livestock performance. McCarl (2000) provides the assumed changes in livestock production on a per head basis.

Altered livestock performance in terms of altered ending weights of sale animals or sales of livestock products means that animals need different amounts of feedstuffs to produce that ending weight or volume of products. In this study we assumed that feedstuff usage was strictly proportional to the volume of products, although changes in climate could change this proportion. Thus, if 10 percent more milk were produced, then 10 percent more feedstuffs had to be consumed. When the livestock unit produced multiple products then a weighted average of the percentage change in output is used to adjust the feedstuff usage. The feed usage quantities for which we applied these adjustments included not only traditional grains but also the number of animal unit months required of grazing and the acreage of pasture required. Similar to the case of crops and with the same rationale, nonfeed input use was assumed to increase by 40 percent of the production increase.

Grazing and pasture use are also important assumptions in the ASM model and are influenced by climate change. The crop modeling component of the agriculture sector assessment included estimates of changes in grass and pasture growth due to climate change. Alterations in the growth rate of grass changes the available feed supply from a given area of pasture. Pasture use and grazing land availability are represented in the ASM and were changed to reflect the change in grass and pasture growth. Pasture use was adjusted by the change in grass growth. Thus if grass growth increased by 10 percent then livestock pasture use was multiplied by 0.9 (1/1.1). This adjustment was done after changing the pasture required as a result of any change in body weight directly due to temperature.

Grazing on western rangelands was addressed in a manner similar to the adjustment for pasture, however, the availability of such lands traditionally has been measured in terms of animal unit months (AUMs) of grazing. An estimate of the AUM supply sensitivity to climate change was developed by assuming the change in AUM supply was the same as the change in grass supply. Thus if grass growth increased by 10 percent then the AUMs available increased by 10 percent.

This combination of climate effects on livestock includes most of the primary effects of climate on the livestock sector. The principal omissions are direct losses of livestock due to extreme storms. More or fewer floods or extreme winter weather events are potentially additional changes to the livestock sector not considered here.

3.5 Pesticide Costs

A change in the incidences and range of agricultural pests is another likely effect of climate change. Most insects, weeds, and diseases are sensitive to climate; climatic factors are an important determinant of the range of many important agricultural pests. No previous assessment of agricultural impacts of climate change has integrated this affect fully into an economic assessment. To consider how climate could affect agriculture through its affect on pests we conducted a statistical analysis relating pesticide expenditures to climate. This analysis was conducted on cross-section data and is explained in greater detail in Chapter 6. The change in pesticide expenditures for corn, cotton, soybeans, wheat and potatoes due to a percentage change in precipitation and temperature was estimated.

Based on these statistical relationships, a change in pesticide costs under each climate scenario was estimated. The limitations and advantages of such cross-section evidence applied to time-series phenomena such as climate change have been discussed in the context of other such efforts, the broadest such effort being the Ricardian rent method developed by Mendelsohn, Nordhaus, and Shaw (1994). A main additional limitation in the context used here is that, as applied, this approach implicitly assumes that any additional potential damage due to pest range and incidence expansion is fully controlled by the use of additional pesticides. Thus, the only economic loss to farmers is the additional pesticide expenditures. If, even with additional pesticide expenditures, there were greater crop losses the lost revenue from the reduced sale of crops would be an additional loss.

3.6 International trade

As reviewed in Chapter 2, studies that have considered global impacts of climate change have demonstrated that the economic impact on a country can be heavily affected by how climate change affects agriculture production in major agricultural exporting and consuming countries. The ASM includes an international sector and thus, in all scenarios, climate impacts on the US affect US competitiveness in export markets. However, in the base scenarios, while US agricultural production is affected by climate there is no climate impact elsewhere in the world. It was, however, beyond the scope of Agricultural Sector Assessment to conduct a full assessment of the rest of the world. This is roughly equivalent to assuming that, while there may be positive and negative impacts of climate change on agriculture elsewhere in the world, the net impact is to balance out to no change. In fact, in the global studies that have been conducted, the net global effect is often relatively small due to a combination of gainers and losers around the world. To consider the sensitivity of our results to the implicit assumption of no impact elsewhere in the world, we constructed 3 sensitivity scenarios for potential climate impacts on the rest of the world based on previous global assessments. Two scenarios were developed from the earlier work based on an economic modeling analysis of international yield changes based on climate scenarios of the GISS and UKMO climate scenarios (Reilly, et al., 1993). The production changes in other regions is given in Tables 3.6a, b. Another scenario was developed based on a global modeling exercise using the Hadley center climate scenario (Darwin, personal communication) although the analysis was not conducted directly as part of the National Assessment. This scenario is based on a model developed at the Economic Research Service (Darwin, et al., 1995). The GISS/UKMO climate scenarios are fairly old, the impact analysis dating to the early 1990s, and are doubled CO₂ equilibrium scenarios. The advantage of these scenarios is that the underlying approach used for the crop studies are similar to the approach used in this assessment and the study provides details on the major crops and world regions represented in the ASM. For the Darwin scenario we based adjustments on changes in net exports from the US.

None of these scenarios are completely consistent with the analysis of the US we conducted but they provide a useful way to demonstrate the sensitivity of the economic estimates we obtained to different assumptions about how climate change could affect the rest of the world. The GISS/UKMO scenarios were chosen, in part, because in the study from which they were taken they represented the mildest (GISS) and the most severe (UKMO) scenarios considered among those that considered both adaptation and the CO₂ fertilization effect.

3.7 Economic Results

We discuss below the main economic results. We first discuss the results from the core scenarios. We then consider sensitivity cases. These include the trade sensitivity results, the alternative Hadley Center Scenarios based on the PNNL crop modeling, and a set of miscellaneous sensitivities. We report here the major results. Altogether we ran 43 scenarios representing different impact combinations (e.g. with and without adaptation or pest effects) and alternatives (e.g. alternative trade and crop yield effects), producing results for aggregate economic effect, regional production, and resource use for each scenario. In most cases, the general pattern of change across regions and resource use closely reflects differences in the aggregate economic effect across scenarios. We have tried to highlight here the broad pattern of results. Complete tables of results are provided in McCarl (2000).

3.7.1 Results from the Core Scenarios

A value of an economic model like the ASM is that it can summarize the net impact of a combination of many different changes. The model also provides the ability to consider distributional and resource use effects, reported below.

3.7.1.1 Aggregate Economic Impacts

We report the aggregate results in terms of a change in welfare, here measured as the sum of producer and consumer surplus. Welfare is preferred as a measure of economic impact over measures such as change in agricultural production or consumption because it includes consideration of the fact that with less production fewer inputs are used and that consumers, in shifting consumption away from agricultural goods, are able to substitute consumption of other goods. Figure 3.4a displays the results based on the Canadian Center (CC) climate model and Figure 3.4b displays results based on the Hadley Center (HC) model. Included in these figures are changes of consumer and producer surplus in the US as well as a change in total surplus. The difference between the two is the economic impact on producers and consumers outside the US. The scenarios reported in Figures 3.4a,b do not include any direct climate impact on agriculture outside the US but impacts on foreign producers and consumers occur because of changes in prices of internationally trade commodities. The figures provide results for 2030 and 2090 under 3 different scenarios. The first is the impact of climate change, including crops, livestock, and water demand and supply effects without adaptation. The second series adds adaptation and the third adds, in addition, the effects of climate on pesticide expenditures.

Given the differences in the climate models and the intermediate crop modeling results, the economic results are generally as expected. Overall, the effects on total surplus are generally positive, much more so for the HC scenario. Net economic benefits range from about -0.5 to 3.5 billion dollars (year 2000$) in the CC scenario and between 6 and 12.5 billion dollars for the HC scenario. In both climate scenarios the total and domestic surplus increases between 2030 and 2090, indicating that the general overall beneficial effects of climate change continue at least through 2100. A number of analysts have suggested that at more extreme levels of climate change one should expect losses. Since we have not conducted a full transient crop model/economic analysis we cannot be sure whether by 2090 benefits are declining from some peak experienced between 2030 and 2090 or whether benefits are continuing on a general upward trend. As illustrated by the CC climate scenario, however, the time path of impact may not be easily described by a simple function. In 2030, at least for the no adaptation case and for US Surplus, the net effect is an economic loss that turns to gains by 2090 for all but the no adaptation case. Given the multitude of effects across many different regions it is not possible to trace these results conclusively to a specific aspect of the climate scenarios. The pattern of results very likely reflects the fact that in the CC climate scenario, precipitation decreases in the US in 2030 and then increases by 2090. Care must be taken in over-interpreting this time path or any of the specific results. Climate models produce variability from year-to-year and decade-to-decade. Even for specific models such as the CC or HC models, a particular decade of climate drawn from a particular scenario must be considered only one possible draw from a distribution of possibilities. By 2030, the additional greenhouse gas forcing beyond that of current climate is relatively smaller compared with 2090 and so the natural variability on a decade-scale can have a large effect relative to the signal due to greenhouse gas forcing.

The distribution of benefits between foreign and domestic is notably different in the two scenarios, with much of the benefit going abroad in the CC scenario and relatively little flowing abroad in the HC scenario. This difference occurs because of the differential effects on crops where exports are important versus those that are mainly consumed domestically.

As observed for the intermediate crop yield results, adaptation is considerably more important when the impacts are adverse than when they are beneficial. While this shows up in the comparison of the two climate scenarios, more research is required to assess the robustness of this result. It is possible that a more expansive exploration of adaptation options such as double cropping could reveal further gains in northern regions.

The net effect on pesticide expenditures is an increase, thereby reducing total economic surplus. This effect is quite small. The size of the effect is not surprising, however, given that pesticide expenditures account for only a few percent of total costs. As previously noted, however, this estimate may understate losses because it does not include any increase in damage that cannot be eliminated through the increased use of pesticides.

3.7.1.2 Distributional Effects

Both the distribution of economic effects between producers and consumers and among regions can vary. Figures 3.5a and 3.5b display the distribution of effects between domestic US producers and consumers. Across all the scenarios, consumers generally gain from lower prices whereas these lower prices cause producer losses despite the fact that climate change has improved productivity. The CC climate scenario produces an approximate balance in terms of domestic consumer gains and producer losses in most scenarios. In contrast, the HC climate produces large consumer gains. The productivity gains are so substantial, however, that the volume and output and export gains to producers nearly offset the price declines. While the absolute level of change is comparable between producer and consumers, in percentage terms the changes to producers are much more substantial. For comparison purposes, the total economic benefit derived from food consumption in the base is estimated at approximately 1.1 trillion dollars whereas total producer surplus is on the order of 30 billion dollars. Thus, the 4 to 5 billion dollar surplus losses in the CC scenario represent 13 to 17 percent loss of surplus to producers whereas the gains of 12 to 14 billion dollars of consumer surplus in the HC scenarios represent only a 1.1 to 1.3 percent gain to consumers. We would expect producer losses to ultimately be realized as changes in the value of land. A 13 to 17 percent loss in this asset value is substantial but, to place this in context, agricultural land values fell on the order of 50 percent between 1980 and 1983.

Figures 3.6a and 3.6b display the regional differences. We report an aggregate index of the production across crops. The plotted values are percentage change from base production. The figures show substantial regional differences in both scenarios. The basic regional pattern is similar in both scenarios. The Lake States, Pacific, Mountain, and the Corn Belt regions, in that order, show large increases in production, generally between 50 and 150 percent increases in output. The pattern of absolute (or relative) losers varies more across the scenarios. The Southeast, Southern Plains, and Delta States lose absolutely in the CC scenario or show the smallest increases in production in the HC scenario. Appalachia is also more negatively affected. Impacts on the other regions vary substantially across the two climate scenarios or over the two time periods.

In the HC climate scenario no region shows a production decline but with substantial overall producer losses in the US due to declining commodity prices farmers in those regions that show only modest increases in production are clearly suffering substantial economic loss. In these cases we expect economic losses to show up as declines in the value of assets located in these regions--primarily agricultural land. In the CC climate scenario, several regions show absolute declines in production. This adjustment process over the longer term explains why production can continue to increase even though the region experiences economic loss. Owners of land may be forced out of business and the resulting price of land would reflect the reduced production potential due to degrading climatic conditions. A new buyer, paying the lower price could then profit because the asset cost was lower. Thus, production continues despite the fact that owners of farmland take a significant economic loss. In the CC scenarios, regions with production losses are also suffering from price declines, though not as severe as in the HC case.

3.7.1.3 Resource Use

Overall, measures of resource use generally decline across all categories, both climate scenarios, and both time periods (Figure 3.7). Irrigated land and water use decline most, reflecting both the overall increase in production and decline in prices and the relative yield effects between irrigated and dryland. Overall, the results for these scenarios suggest considerable less pressure on resources, a result of the overall increase in productivity.

3.7.2 Trade Scenarios

Table 3.7 provides the aggregate economic results for the three different trade scenarios for 2030 and 2090. All three foreign trade scenarios were run against both the CC and HC domestic US scenarios. The trade scenarios generally do not lead to a substantial change in total surplus or total US surplus. This result reflects the fact that the US is a substantial commodity exporter but also a substantial food consumer. Hence, global price changes have roughly offsetting effects -- consumers gain from price decreases while producers lose. With price increases these effects go in the opposite direction but again roughly offset one another. The biggest effect of the trade scenarios is thus a reallocation of the total domestic effect between producers and consumers. The Darwin scenario creates somewhat greater losses for US producers, the implication being that the impact on production in the rest of the world for those goods in which the US trades is positive with generally lower world prices than in the comparable cases where world production was left unchanged. For the GISS and UKMO scenarios, the effect is the opposite. World prices increase, very modestly in the GISS case and more substantially in the UKMO case thereby shifting some of the gains from US consumers to US producers.

These trade scenarios, as previously noted, were not developed consistently with the domestic impacts. If the results obtained for the US with these climate scenarios, generally more positive yield effects than in past assessments, would be observed across the world then we would expect world prices to generally decline. The result would be further gains by US consumers and losses by producers as observed in the Darwin scenario rather than in the GISS or UKMO scenario. On the other hand, a factor that is no doubt important in moderating the climate impacts on the US is the cooling effect of sulfate aerosols in the Northern Temperate regions. Earlier assessments used climate scenarios that did not include the sulfate aerosol effects. Often these showed warming benefits in more northerly regions and losses in tropical regions. Sulfate aerosol effects could produce a regional pattern of climate change that reduces benefits to some northern regions compared with earlier assessment while leaving unchanged the losses in the tropical regions. If such a result would hold, the implication would be perhaps world price increases and a shift of benefits from US consumers to US producers. More complete global studies with newer climate scenarios are required to resolve this effect.

3.7.3 The Alternative PNNL Crop Scenarios

Table 3.8 provides the aggregate economic results for the alternative PNNL crop simulations. These were produced only for the Hadley scenario and did not include adaptation. We did not include the pest changes in this comparison, the purpose here being primarily to evaluate scenarios for PNNL crop simulations versus the core crop simulations for a comparable set of scenarios. The overall conclusion of this comparison is that the PNNL scenarios show very similar results to those obtained with the detailed site simulations. The total economic welfare gain is somewhat higher in 2030 and somewhat lower in 2090. While not reported here, the regional effects differ somewhat. In the PNNL scenarios the Southeast does not show up as a particularly severely affected region and the Southern Plains and Northeast are considerably more positive than in the core scenarios. The Northern Plains appears as the more negatively affected region in the PNNL scenario. The Lake States, Corn Belt, and Pacific regions are among the more positively affected regions in both scenarios.

Overall, this comparison is reassuring in the sense that the limited site selection in the core scenarios does not appear to have created a substantial bias in aggregate estimates. The aggregate effects offer a relatively weak test, however, as several crops were left unchanged between the core and PNNL results because the crops were not simulated by PNNL. Clearly, some differences do occur at the regional level, emphasizing the uncertainties in producing consistent predictions at the regional level.

3.7.4 Other Scenarios and Sensitivities

As indicated previously, we were unable to generate yield changes for cotton using a cotton crop model. Instead we adapted results from another study. We also simulated results using soybeans as a proxy for cotton. Soybean results were generally quite negative in the South in the CC scenarios whereas the alternative cotton scenarios showed more positive effects. As a result, this alternative assumption produced quite different results. Notably, under the CC scenarios the .6 billion dollar total surplus loss in 2030 doubled and the approximately 1 billion dollar gain in 2090 was changed to a 1 billion loss. Most of this change accrued to domestic and foreign consumers. Producers losses were actually slightly reduced in 2090 due to higher cotton prices. The negative production effects were most substantial in the Delta regions.

The results derived by projecting the agricultural economy forward to 2030 and 2090 were not qualitatively different. The specific quantitative results depend crucially on highly uncertain forward projections. The two basic aspects of these projections are yield growth and demand. Projecting ahead historical yield growth and increases in demand due to population growth increases the absolute size of the agricultural economy. If we consider yield changes in percentage terms as operating on the new higher yields, the percentage effect is similar. Differences can arise due to different assumptions about yield and demand growth for different crops and differences in yield impacts among crops.

We also jointly considered the impacts of changes in agriculture and forestry. Because of the long growth cycle of forests, there is a far greater need to look forward and consider the present value of changes over a number of years. The forest sector assessment was conducted under as part of the National Assessment (Joyce et al., 2000; Ireland et al., 2000; McCarl, 2000; Alig and Adams, 1997). Forest yield scenarios were derived based on the Canadian and Hadley climate models and two ecological process models. Results suggest that consideration of both sectors suggests generally beneficial effects for the US economy. Increased supplies from forests lead to reductions in log prices that in turn, decreases producers' welfare (profits) in the forest sector. At the same time, lower forest product prices mean that consumers generally benefit. This pattern of distributional impacts on forestry producers and consumers is similar to results obtained in the agricultural sector. Increases in the net present value of total economic welfare (combined forestry and agriculture) ranged between 0.9 and 1.2 percent, with the higher positive impacts under the Hadley climate change scenarios. More details on these results are provided in McCarl (2000).

Land use changes between forestry and agricultural uses are an important avenue of adjustment to climate-induced shifts in production, and there are notable differences in these adjustments across climate change scenarios. Over the full projection period the Base and Canadian GCM cases project a net shift of land from agriculture to forests, the latter at about half the rate of the former, while the Hadley GCM scenarios project a net loss of forest land to agriculture. Yields from the land generally increase in both the forest and agricultural sectors in all four scenarios. In the Canadian scenarios these shifts are relatively more favorable for forestry profits compared to agriculture, while the opposite is true in the Hadley scenarios.

3.7.5. Summary of the Main Economic Results

The main results of the economic analysis are:

1. Climate change as modeled under the climate scenarios considered is mostly beneficial for total society in terms of agricultural impacts particularly if adaptation is considered. This differs from the results of previous scenario analysis where results have been mixed and generally negative in the absence of adaptation.
2. Climate change uniformly shows increases in crop production and exports with decreases in crop prices. Livestock production and prices are mixed.
3. Climate change is found to be largely detrimental for producers. Climate changes are also found to be beneficial for foreign surplus and for consumers. These results reflect the overall positive effect on production which leads to decreasing prices.
4. There are substantial shifts in regional production with gainers and losers. The Lake states, Mountain states and Pacific region show gains in production while the Southeast, the Delta, Southern Plains and Appalachia generally lose. Results in the Corn Belt are generally positive. Results in other regions are mixed depending on the climate scenario and time period. The regional results show broadly that climate change favors northern areas and can worsen conditions in southern areas, a result shown by many previous studies.
5. Our analysis suggests increases in pesticide expenditures due to climate change, a partial offset to the overall benefits. The magnitude of this effect is relatively small.
6. The overall benefits of climate change are greater in 2090 than in 2030 for the US as a whole and, even for regions with losses, these are generally less in 2090 than in 2030. Changes in precipitation are likely the source of this result.
7. Climate change largely causes a decrease in resource usage due to expanded productivity. In particular dryland, total crop land, pasture land, and water usage declines.
8. Farm-level adaptation increases the climate change benefits to total society by about a billion dollars. Producer losses are generally reduced by adaptation.
9. Consideration of climate effects in other countries did not greatly alter the climate change benefits to total society. It can have substantial distributional assumptions depending on how climate affects the rest of the world.
10. Changing the base year does not alter the sign of the climate change benefits to total society.
11. The results obtained from using two different crop yield simulation approaches were quite similar in overall magnitude.
12. Jointly considering forest and agricultural changes due to climate does not change the impacts substantially. The net effect on society of both changes is positive and the distribution effects are similar, with producers suffering surplus losses due to declining prices while consumers benefit.