The possible impacts on wheat production of a recurrence of the 1930s drought in the U.S. Great Plains
out of 22
Post on 06-Jul-2016
THE POSSIBLE IMPACTS ON WHEAT PRODUCTION OF A RECURRENCE OF THE 1930s DROUGHT IN THE U.S. GREAT PLA INS RICHARD A. WARRICK* National Center for Atmospheric Research, Boulder, CO 80307, U.S.A. Abstract. What would be the impacts on wheat production if the U.S. Great Plains were to suffer another severe, prolonged drought? The 1930s drought is chosen as a worst-case scenario, and two sets of crop-yield regression models are employed to simulate yields using actual 1932-40 weather values and assuming constant 1975 technology. The results are first compared to normal or expected yields in each of 53 crop reporting districts in order to determine the range and spatial variation in yield departures over the nine-year period. Assuming a 1976 crop area, wheat production levels are then calculated and aggregated to give Plains-wide estimates for each year. It is found that the sequence of 1930s weather results in continuous, prolonged declines in expected production. Plains-wide yields are below normal (on average about 9-14%) for nine consecutive years. In the poorest years, the impacts are areally widespread with about nine-tenths of the Plains experiencing yield declines. The spatial variation in yields is substantial, however, ranging from over 100% to below 40% of expected even in the poorest years. In the worst year (1936), simulated production is sharply reduced by about -25%, or 9.6 million metric tonnes. The cumulative deficit over the nine-year period is roughly equivalent to a full year's wheat production. The major conclusion is that a return of a 1930s-type drought would still inflict widespread, heavy damage on wheat production in the Great Plains. 1. Introduction In the early 1970s, several events occurred that led to a growing concern about the re- lationship between climatic variability, agricultural production, and global food security. The quick depletion of grain reserves, the leap in grain prices, and the resultant food shortages in certain grain importing nations such as India during 1972-74 were dramatic signs that the world's food systems are not immune to sudden shocks, either from cli- matic, economic, or political origins. The climate-related famines in Bangladesh and the Sahel highlighted the fragile relationship between climate, food production, economics, politics, and human weU-being in many of the less developed countries. The sequence of deleterious weather events - excessive spring moisture, summer drought, early freeze - that resulted in poor U.S. grain harvests in 1974 demonstrated that even the high tech- nology systems of North America are not impervious to the vagaries of weather. After a relatively placid decade of the 1960s, the disruptive events of the 1970s underscored the increasing dependence of many developed and developing countries on imports of * Presently affiliated with the Climatic Research Unit, University of East Anglia, Norwich, NR4 7TJ, U.K. Climatic Change 6 (1984) 5-26. 0165-0009/84/0061-0005503.30. 9 1984 by D. Reidel Publishing Company. 6 Richard A. Warrick grain, and put into question the ability of our agricultural systems to meet global demands consistently. Arising from this concern is the fear that a major decrease in North American grain production - through drought, crop disease, or other natural calamities - could spell disaster abroad. Most grain exports originate from North America. In 1980, for example, the United States produced about 70% of the total grain and 45% of the wheat traded on the world market. About two-thirds of the nation's wheat is concentrated in the wheat belts of the Great Plains, a vast semi-arid region with an historical record punctuated by climatic extremes and crop failures. What would be theeffects on wheat yields through- out the Great Plains from an extremely disruptive climatic occurrence, such as a severe, prolonged drought? To what degree would aggregate wheat production be impaired? These are the questions addressed in this paper. The paper focuses on a potential climatic episode which lingers in the minds of those concerned with climate and agriculture: the occurrence of a decade-long, 1930s-magnitude drought in the Great Plains. Assuming 1975 levels of agricultural technology, two sets of crop-yield models are run to simulate spring and winter wheat yields under actual weather conditions that prevailed during the years 1932-40. The absolute and relative wheat yield declines throughout the Plains are calculated, and the spatial patterns of yields are depicted for several of the worst years. Assuming full wheat area under the plow, production is simulated for each areal unit of observation and then aggregated in order to ascertain the Plains-wide drought impacts for each of the nine weather-years. Model results are then compared and discussed. 2. The 1930s Drought: A 'Worst-Case Scenario' The 1930s - the decade of the Dust Bowl - is firmly implanted in the collective con- sciousness of Americans as a period of both economic and environmental disaster. In the agricultural heartland of the mid-continent, mild desiccation in the early 1930s gave way to severe droughts which lasted year after year until broken by the return of abundant precipitation in late 1940. For most areas of the semi-arid Great Plains, 1934 and 1936 were particularly severe. Giant dust storms swept the fragile topsoil from thousands of acres of farmland and dust settled as far away as Washington, D.C., providing tangible evidence of an ongoing environmental disaster to even East Coast residents. The plight of Great Plains' farmers under the double burden of drought and depression was do- cumented by extensive social studies conducted by the New Deal Administration (e.g., Great Plains Committee, 1937), and was popularized and immortalized by the photo- graphs of Dorothea Lange and Arthur Rothstein and by the literary talents of Steinbeck (1939) and Sears (1935). Today, the memory of the calamity of the 1930s is perpetuated by reexaminations of the disaster by interested scholars, as exemplified by the recent works of Worster (1979) or Bonnefield (1979). Seemingly, our lasting cultural perception portrays the drought of the '30s as an extreme climatic event. The popular perception of the severity and rarity of the '30s drought is supported by a number of climatological studies involving a variety of methodological approaches. The Possible Impacts on Wheat Production of a Recurrence 7 300- 250- 200- 150- I o: "~ 100- 50- 0 30 8b 3'5 4b 4i 5b 55 Year 6b 6i 7b 7i Fig. 1. Great Plains droughts, 1931-1977 (from Warrick, 1980). The graph depicts the number of climatic divisions exhibiting severe or extreme drought (~ -3.00) on the Palmer Drought Index scale (Palmer, 1965) summed over five months, April through August. In any given year, 320 division- months (64 divisions x 5 months) of drought are possible. Comparisons of drought years using Palmer Drought Indices a - calculated for 64 cli- matic divisions and aggregated over the entire Great Plains - indicates that the magnitude, areal extent, and duration of the 1930s drought was unsurpassed within the period 1931-77, as shown in Figure 1. Extension of the record of Palmer indices for individual states back to the 1890s - as, for example, for Kansas (Felch, 1978; Climate and Society Research Group, 1979) - confirms the primacy of the '30s drought within the time scale of a century. In general, these findings correspond with the conclusions of Stockton et al. (1981) who calculated Palmer indices from tree-ring data in order to reconstruct a 363-yr drought record for the western United States; they found that " . . . the am- plitude of the Dust Bowl drought of the 1930s stands out in this analysis as probably the most widespread drought event, as well as one of the most extreme events on the Palmer Index Scale since A.D. 1600" (p. 102). Further tree-ring studies by Stockton and Meko (1983) focusing specifically on areas within or adjacent to the Great Plains indicate that 1934, 1936, and 1939 rank among the driest 10 of 278 yr; when drought conditions are averaged over three or more years, the 1930s drought is equalled or surpassed in severity on only three occasions since 1700. 1 The Palmer Drought Index (PDI) is a relative measure of soil moisture and groundwater conditions with respect to long-term means (see Palmer, 1965). Derived empirically from data for the U.S. Great Plains, the PDI is based on precipitation and temperature values which are used to calculate potential evapotranspiration. The index varies from greater than +4.00 for extremely wet conditionsto less than -4.00 for extreme drought. 8 Richard A. Warrick Thus, society's perception of the rare, extreme nature of the '30s drought appears to be borne out by scientific inquiry. The severity of the drought - in terms of its mag- nitude, areal extent, and duration - can serve well as a 'worst-case scenario' for assessing the potential impacts of a major climatic perturbation to our agricultural systems. It is sobering to note, however, that climatic studies of the influence of increasing atmos- pheric carbon dioxide have led the scientific community to predict a global warming in the decades ahead (NAS, 1979, 1982; Clark, 1982). Since temperature and precipitation appear to be inversely related in most of the U.S. Midwest, a warming trend in this region may be accompanied by an increase in the frequency of such severe droughts in the future. This accords with scenarios based on instrumental records (Wigley et al., 1980; Jffger and Kellogg, 1983) or paleoclimatic data (Kellogg and Schware, 1981), and is compatible with global circulation model results (Manabe et al., 1981). If, indeed, this is the case, the 'worst-case scenario' may become more probable as time passes. 3. Assessing the Impact A number of past studies have examined the impacts of climate fluctuations on North American grain production, but more often than not such studies were concerned with the effects of slow climate change (like regional warming or cooling) rather than extreme occurrences like the '30s drought. For example, in response to concerns over possible climate changes arising from stratospheric ozone depletion from SSTs, the Climate Impact Assessment Program (CIAP) used crop-yield models to investigate the net effects on Great Plains wheat production of alterations in average temperature and precipitation (Ramirez et al., 1975). Similarly, the National Defense University made crop impact estimates under various scenarios of average global cooling and warming, drawing upon expert opinion regarding likely crop responses to variations in key climatological variables (NDU, 1980). One major study did focus explicitly on the possible agricultural impacts of a 1930s drought recurrence. That study, entitled the Impacts of Climatic Fluctuations on Major North American Food Crops by the Institute of Ecology (TIE, 1976), sought to deter- mine, among other things, what the impacts would be if the weather of 1933-36 oc- curred at modern levels of grain production. The method involved calculating the actual percent deviations of wheat yields from 'expected' yields during selected 'scenario' years (where 'expected' yields were defined by a curvilinear regression fit to USDA national wheat yield data for the years 1866-1975)). These percent deviations were then simply applied to a constant U.S. wheat production assuming current (1973) tech- nology and constant (1975) crop area in order to estimate the production deviations should comparable climatic conditions occur again. As shown in Figure 2, production in the four selected years of the 1930s falls short of expectations with 1933 being the worst year. There are two major drawbacks of the study. First, actual historical crop data were used directly to calculate relative yield deviations. In so doing, the analysis included the effects of important short-term non-weather 'events' - prices, policies, diseases, pests - The Possible Impacts on Wheat Production o f a Recurrence 9 3000 ~_ .~ EXPECTED PRODUCTION 2500 . 1973, TECHNOLOGY 1975 CROP AREA 2000 1500 I000 U.S. CONSUMPTION LEVEL 1975-76 500 O ~ 9 9 9 to to to ~ ~o ~ to ~ ta r .~ O~ - - IO ta m IO ta Fig. 2. The Institute of Ecology scenario production of U.S. wheat, in million bushels (from TIE, 1976). For each scenario year, proportional yield deviations were calculated from the residuals of a curvilinear regression line fit to actual USDA national yield data for the period 1866 to 1975. These proportional deviations were then applied directly to a constant production adjusted to 1973 technol- ogy and 1975 crop area to give the scenario production levels shown. which can have a disproportionately high impact on average yields in a given year. 2 These were particularly influential during the 1930s. In short, the direct effects of weather were not isolated. Second, the study employed average crop yield and acreage data aggregated at a national level. Thus, the analysis did not consider the regional or local impacts of drought, which can be quite severe and which can display considerable spatial variation. This paper partially deals with these problems by scaling the analysis to subregions within the Great Plains itself, and by simulating yields from weather data through the use of crop-climate models, as described below. 4. Crop-Climate Models Crop-climate models can be divided roughly into crop-growth simulation (or 'process') models and crop-yield regression models (Stewart, 1975; Baler, 1977). Crop-growth simulation models are based on knowledge of the physical and biological processes in- volved in plant-environment interaction, drawing upon basic research in soil physics, 2 For example, in Figure 2 the year 1974 shows a simulated total wheat production decline which is greater than the worst year of the 1930s. But rather than worse weather, it is likely that the 1974 yield decline resulted in large part from increased planting on marginal acres in response to soaring wheat prices, which also encouraged less abandonment of poorly producing fields at harvest time (wheat prices reached a national average of $4.09 per bushel in 1974 - a tripling of 1971 prices - and planted and harvested acreage was over 30%higher nationwide than just three years earlier). Both these effects - planting of marginal land and less land abandonment - tend to lower the figures on average yields per harvested acre. That these factors came into play in 1974 is evidenced by the fact that even though wheat yields were notably low, total 1974 national wheat production broke all previous records because of expanded acreage - in contrast to the scenario portrayed in Figure 2. 10 Richard A. Warrick plant physiology, and micrometeorology (Stewart, 1975). The models attempt to simulate the actual physiological growth of the plant. Examples include SIMAIZ (Duncan, 1975) and SPAM (Stewart and Lemon, 1969; Shawcroft et al., 1974) for corn, and TAMW (Maas and Arkin, 1980) for wheat. The conaplexity of the models, as well as the fact that they attempt to describe physical and biological processes, give them the potential for considerable accuracy in crop yield assessments. Regression models commonly employ multiple regression techniques for the purpose of estimating yields as the sum (or product) of linear and quadratic terms involving weather predictors, such as monthly or weekly temperature and precipitation. In contrast to simulation models, these statistical-empirical methods draw upon historical climate and yield data series for the determination of model parameters. Wallace (1920) was among the first to demonstrate the use of regression techniques for crop prediction in the United States. The best known representatives of models of this type are Thompson's models, the first of which were constructed in the 1960s (Thompsori, 1962; 1969a, b; 1975). These models, based on state level data, were developed to assess the influence of selected weather factors and technology in the production of specific grain crops in the U.S. Midwest. In multiple regression crop-yield models the observed long-term increases in average yields due to 'technology' (or, more accurately, time-dependent, non-weather factors) are typically treated as linear, piecewise linear or quadratic time trends. As evidenced by several applications of the models, it is possible to hold this factor constant in order to assess the effects of historical weather conditions on crop yields under different assumed levels of technology. For example, McQuigg et al. (1973), using Thompson's state-level regression models, followed such a procedure in order to determine whether better technology or better weather accounted for the high, consistent grain yields observed in the U.S. Midwest during the late 1950s and 1960s. By simulating an 81-yr series of yields using actual weather observations with a technology fixed at 1945 levels, it was found that yields (particularly corn) still displayed high, consistent levels in recent years, which strongly indicates that unusually good weather was responsible. This ability to manipulate the long-term technology (or time trend) component of the regression models is of direct interest to our study of the potential effects of a return to the 1930s drought, as we shall see. The short-term, year-to-year effects of non-weather factors can have considerable impact on yields as well. Actual wheat yields and production during the 1930s drought were influenced by low prices, depression, severe soil erosion, pest and disease damage, and a host of government relief programs. Since the 1930s, price support programs, market prices, and conservation efforts have alternately encouraged and discouraged the use of marginal lands, and have raised or lowered the threshold at which planted acres are harvested or abandoned in years of low yields. All these factors affect yields and are mixed into state and U.S. Department of Agriculture statistics, which makes it some- what difficult to determine the effects of weather on yields from agricultural statistics alone (one of the problems with the TIE (1976) study described above). In current crop simulation models, these non-weather factors do not exist. In the regression models, The Possible Impacts on Wheat Production of a Recurrence 1 1 the influence of these factors still exists because the models are derived from empirical yield data, but the influence is probably small, as evidenced by the fact that state-of-the- art models are able to predict area-averaged yields accurately using weather variables alone (as indicated by model validation procedures). Thus, both simulation and regression models are potentially useful for isolating the specific effects of climate on yields, which is the major reason why the crop-climate modeling approach is adopted here. For our purposes, there are distinct drawbacks to each kind of model as well. Simula- tion models have yet to produce reliable predictions of crop yields observed in the field - nor should they, since the models are designed more for management, rather than pre- dictive, purposes. In addition, as input the models require detailed data on agroclima- tological variables such as solar radiation, maximum and minimum temperatures, and available soil moisture on a weekly, daily, and sometimes hourly basis in order to simulate plant growth. These detailed data are often not available for large areas and historic time periods. The regression models, too, have their disadvantages. First, the regression approach itself creates 'black-box' models, the physical structure of which may be inappropriate to the realities of plant-environment relationships. Herein lies the generic problem of regression models and the potential advantage of crop-simulation models. While regression models simply lump all the factors together in a 'black box' fashion and assume that all of the climate-related interannual variability in yields can be extracted by purely statistical means, the crop-simulation models attempt to build agronomically realistic structures based on plant physiological processes. Second, substantial errors may arise from violation of the assumptions of linearity and of independence of predictor variables, as, for example, between weather and tech- nology variables. (See Katz, 1977, for a comprehensive discussion.) Third, in using regression models constructed from historical data over long periods of record, it is assumed that no new technologies or management practices have been adopted recently which substantially alter the relationship between weather and yields. (For example, if widespread adoption of drought-resistent wheat varieties had occurred within the last ten years, the effects would not be adequately incorporated into the model.) One can find evidence both in support of and in opposition to this assumption in the literature (Warrick, 1980). Finally, regression techniques tend to underestimate the most extreme departures of yields from expected values (e.g. Michaels, 1978, p. 13). This means that all other things being equal, the predictions of yield during severe drought years like the 1930s are likely to be overly optimistic. Despite the problems, however, the regression modelsutilize climatic data collected routinely over space and time and are convenient for predicting average yields over large agricultural regions. State-of-the-art regression models commonly explain a large amount of the interannual variation in observed yields due to climate, and have proven accurate in predicting yields for independent years during validation procedures and applications. For these reasons and bearing in mind the above caveats, two sets of multiple regression models were selected to simulate Great Plains wheat yields under 1930s drought con- 12 Richard A. Warrick ditions, as described below. (For reviews of crop-climate models, see CIAP (1975) or Baier (1977).) 5. Impacts of a 1930s Drought - Yields Effects and Spatial Patterns One set of regression models - one model of spring wheat (Climate/Food Research Group, 1975), and the other of winter wheat (Michaels, 1978) developed at the Institute of Environmental Studies at the University of Wisconsin - predict yields per planted hectare within the major U.S. wheat belts. The area covered by the models (hereafter referred to as the IES models) represents about two-thirds of all U.S. wheat. Unlike many other crop-yield models which employ state- or regionally-averaged data, the IES models use the comparatively small USDA crop reporting districts (CRD) as units of observation and are thus able to provide a reasonable spatial representation of interannual variability of yields (Michaels, 1979). There are a total of 53 CRDs (twenty-two CRDs for spring wheat and 31 for winter wheat) for which predicted yields (~z) are generated. The general form of the models is as follows where y = /, rb Yt Y c e Y= J* +e= Yb + Yt + Yc+e actual yield per planted hectare, -- predicted yield per planted hectare, = yield component attributable to crop reporting district habitat conditions, = yield component attributable to technological change, = yield component attributable to climate, = unexplained residual. The habitat component (Yb) primarily reflects local environmental factors like soil differences and thus captures the spatial, cross-district variations in observed yields. The product of a regression coefficient (estimated during multiple regression) and a dummy variable with a value of one, the term approximates base level yields in each CRD in the period prior to the postwar yield increases. 3 The technological component (Yt) is a linear time trend which increases yields after 1945 (for winter wheat, 1951 for spring wheat), presumably reflecting the effects of fertilizers, pesticides, and other technological inputs on wheat production after World War II. 4 The climate component (Yc) is comprised of 3 For the multiple regression, Yb was represented by a set of dummy variables, one for each of N CRDs (Yb], for J = 1, N). The value of Ybj is set to unity for all observations in thej-th CRD, while values for all observations in other CRDs are set to zero. Proceeding in this manner, separate regression coefficients are estimated for each CRD. These account for the constant spatial differences in yield levels which are not attributable to technological trends (Yt) or climate (Yc). (See Climate/Food Research Group, 1975.) 4 Yt is also the product of a regression coefficient and a dummy variable. For the multiple regression, a set of dummy variables was created, one variable per CRD. For thej.th CRD, Yt] is equal to zero for all observations prior to 1945 (for winter wheat, 1952 for spring wheat) and increases by one for each year thereafter, while all observations for other CRDs are set to zero. Thus, in effect a separate linear technology trend is estimated for each CRD. (See Climate/Food Research Group, 1975.) It should be The Possible Impacts on Wheat Production of a Recurrence 13 weather terms roughly matched to the phenological growing periods of the two major kinds ~ of wheat; average monthly temperature and precipitation for each CRD provide input data. For each model, regression weights were estimated from CRD observations over the period 1932-1975 for winter wheat and 1932-1971 for spring wheat. Thus, there were slightly less than 31 x 44 'district-years' of observation for winter wheat, and 22 x 40 district-years for spring wheat, after accounting for missing data. Each model accounts for about 95% of the total cross-district and interannual variability in observed yields; after accounting for Yt and Yb' about half of the residual variance is explained by the climate component (Ye) al~ There are several important assumptions specific to the IES models. First, using yields per planted - rather than harvested - land unit, the assumption is that land abandon- ment (i.e. land planted but not harvested) is mainly a function of climate, whereas we know that other factors can also come into play, as noted above. Second, since the climate ,component coefficients are determined simultaneously from data pooled over all CRDs within each wheat belt, it is assumed that the weather-yield relationships are spatially uniform, whereas other statistical studies (e.g., Sakamoto, 1978) suggest that subregional differences in such relationships exist. Keeping the above assumptions in mind, the IES models were run with several modifi- cations to simulate a recurrence of the 1930s drought. 6 Yield figures were calculated by setting the technology (Yt) at a constant 1975 level. Yields were calculated on the basis of actual climatic data recorded in each of the 53 CRDs from 1932-40. Thus, with 'modern' technology, the effects of severe drought on yields can be estimated, both in absolute terms and as a percent of 'expected' yields (where expected yields in each CRD are equal to the sum of district base level and 1975 trend values). 7 Figure 3 shows the spring and winter wheat yields that would be expected with 1975 technology. The values in Figure 3 are assumed to correspond to agriculturalists' perceived expected yields per planted acre at the time of planting; large dr from these planting time expectations can be considered either 'crop losses' or 'bumper crops', depending upon the direction of departure. The spatial patterns of drought impacts on wheat yields for a replication of 1930s weather at 1975 technology can now be examined. Using weather data for two of the noted that in many of the eastern CRDs of the winter wheat region, the trend flattens prior to i975; in general there is a southeast-to-northwest gradient in the time at which the trend flattens (Michaels, 1978). s It should be noted that simply because the model explains only about 50% of the non-technological yield variability, this does not necessarily mean that it is explaining about 50% of the potential climate signals. Even a 'perfect' climate component may explain only a little more than this 50%. Thus, since we are interested in determining the climatic influence on yields only, using a model of this sort is a valid approach. For a more thorough discussion of the models, the reader is referred to the Climate/ Food Research Group (1975) and Michaels (1978; 1979). 6 A modified version of the spring wheat model was run that contains cross-seasonal interactive terms to allow for the nonlinear climate/yield interaction between growth stages, as per Michaels (1979). 7 Generally, this approach is Similar to that employed by McQuigg et al. (1973) to quantify the in- fluence of weather variability on U.S. grain yields. 14 Richard A. Warrick 20.3 ~ 22.1 I~ 18.8 16.2 23.2 21.8 29.7 KEY ~ = PREDICTED YIELDS IN CRD (BUSHELS/ PLANTED ACRE) I0.1 25.5 23.2 13.3 25.6 24.7 28.6 15.3 Fig. 3. Simulated wheat yields: expected yields, 1975 technology. worst droughts of the 1930s, the deviations (expressed as a percent of expected yields) are depicted spatially in Figures 4 and 5, along with simulated absolute yields (in bushels per planted acre). In 1936 - the worst weather-year - 87% of the 53 CRDs experienced yields which fell short of expectations. As shown in Figure 4, about three-fourths of the CRDs suffered yield deviations of at least -10% - a threshold frequently cited as con- stituting 'drought loss' (e.g., see McQuigg et al., 1973; or NAS, 1976). Only 5 CRDs produced yields that were higher than expected, and these ,were located in the eastern, 12.5 The Possible Impacts on Wheat Production of a Recurrence 60*/ . "~~'~15.9 ~ J 80% 15 8.2 40% _ ~ / 40% I00% 60% 40% ~ 40% 15.8 84 5.1 20.6 55.0 22.9 27.9 I00% ,100% KEY 1 = PREDICTED YIELDS IN CRD ( BUSH ELS / PLANTED ACRE ) = PERCENT OF EXPECTED YIELDS 17.3 18.0 Fig. 4. Simulated wheat yields: 1936 weather, 1975 technology. more humid margins of the Plains states. 8 In this weather-year, the gradient of yield reductions is largely east-west, with extensive areas in Montana, western Dakotas and Nebraska, Colorado, and the panhandles of Oklahoma and Texas suffering extremely low yields of 60% or less of expected. The weather-year 1934, on the other hand, contains a strong latitudinal gradient in 8 In fact, for the most part these CRDs actually fall just outside the boundaries of the Great Plains proper, as defined by the Great Plains Committee (1937) or Saarinen (1966), for example. 16 Richard A. Warrick 18.3 80% I00% 40% 20% IOO*/c~]~_ 80% i4.5 60%' - - - ~ l i .3 ----z_.__ 7.7 17.0 14.5 11.5 18.2 )0% )% 6o~176 22.6 6O% 8O% f _ too% 20% 23.4 80% KEY ~ = PREDICTED YIELDS IN CRD ( BUSHELS / PLANTED ACRE ) -- PERCENT OF EXPECTED YIELDS 17.:5 14.6 Fig. 5. Simulated wheat yields: 1934 weather, 1975 technology. 100% yield reductions, with a depression of low yields occurring in the Central Plains states. Colorado, western Kansas, Nebraska, and South Dakota obtain yields averaging 80% or less of expected; the southern and extreme northern Plains fare better. In terms of the areal extent of wheat yield losses, the 1934 weather-year is comparable to that of 1936: 87% of the total CRDs fall below expected yields, and 74% achieve yields which dip below the - 10% 'drought loss' threshold. A cluster of CRDs in SouthDakota and Colorado The Possible Impacts on Wheat Production of a Recurrence 17 (representing about one-eighth of all CRDs) experience astonishingly low yields averaging below 50% of expected yield - a devastating harvest. The aggregate impact of the drought for the entire 1932-40 period is reflected in the frequency distribution of yield departures, as observed in each CRD for each year of the decade (53 x 9 = 477 'district years'). As shown in Figure 6, approximately 78% of all observed departures were negative. The mean yield for the period amounts to only about 86% of expected. (The lowest yearly mean yield is 75% of expected, occurring in both 1936 and 1934; only two years, 1932 and 1938, have means greater than 90% of expected, and none exceeds 100%.) More than half of all district-years fall below the 'drought loss' threshold. Clearly, the results of the IES models suggest that a replication of the 1930s Great Plains weather would have spatially widespread and temporally persistent impacts. A second set of crop-yield regression models, first developed at the Center for Climatic and Environmental Assessment (CCEA, 1975) andtater revised (Sakamoto, 1978), were also run in order to obtain an additional (though not necessarily independent) set of estimates of the effects of a 1930s-type drought (hereafter, these models are referred to as the CCEA models). Like the IES models, the CCEA spring and winter wheat models are multiple regression equations in which yields are regressed against a time trend variable (as a surrogate for technological factors affecting yields) and a set of weather variables measuring the influence of weather. Unlike the IES models, however, the model's spatial units of observation are larger, comprising aggregates of CRDs: for the Great Plains, yields are predicted for 12 subregions (5 for spring wheat and 7 for winter wheat). Moreover, multiple regression models were developed separately for each subregion. This means that A I'-- I',.- ~" 160 I I Z 9 ~" 140 (f) n," 12.0 w >- I00 I I-- 8O (..) ~: so F-- m 4o Q 2O m 0 I I I I I I I I I I I I I I _ MEAN -- 86.4% STD DEV= 19.8% EXPECTED OR BELOW EXPECTED ABOVE 371 DISTRICT - I06 DISTRICT- YEARS (?7.5('/(,) I YEARS (22.5%) 1251 ,~ ! 1510 4 i -~ I0 20 50 40 50 60 70 PERCENT OF 1 1 1 96 72 60 I - - 20 _ 80 90 I00 I10 120 130 140 150 EXPECTED YIELDS Fig. 6. Frequency distribution of simulated wheat yields as percent of expected yields (at constant 1975 technology) for weather-years 1932-40. 18 Richard A, Warrick the number of distinct weather variables and their regression coefficients vary from subregion to subregion and capture subregional differences in climate-yield relationships existing within each of the two major wheat belts. The separate regressions were based on independent examinations of wheat production and weather influences in each subregion (Sakamoto, 1978). Thus, while the CCEA models do not allow for the detailed spatial representation of yield differences afforded by the IES models, they make a step toward constructing more realistic model structures, statistically and agronomically. As with the IES models, the CCEA models were run to estimate yields (per harvested land unit, however) given actual 1932-40 weather at 1975 levels of technology. In contrast to the IES models, yield deviations were calculated as departures from 'normal' yields (that is, the ] 975 yield values obtained with meteorological nonnals derived from the data period 1932-76). 9 The resultant spatial patterns of yield deviations are roughly comparable to those of the IES models, despite model differences and assumptions, so the spatial analysis will not be repeated here. 1~ Rather, taken together the IES and CCEA model runs now provide yield results from which two separate estimates of total wheat production can be derived. 6. Impacts of a 1930s Drought on Total Wheat Production In order to obtain Plains-wide wheat production figures from both the CRD and sub- regional yield estimates, some assumptions must be made about land in wheat production. For the purposes of this study, a high production scenario was adopted;that is, a situation in which due to policy and price influences, a large proportion of formerly idle land is brought into production. 1976 is an example of such a year. The actual planted and harvest areas n for each CRD and subregion for the 1976 high production year were obtained and multipled by the simulated yields to obtain total wheat production for each areal unit. The production figures were then summed to derive total Plains-wide pro- duction for each of the nine weather-years. Average Plains-wide yield for a given year was calculated simply by dividing total production for that year by the sum of all land area in production. The land area under production was held constant for the nine weather-years for which yields were simulated. The results from both sets of crop-yield models are displayed in Table I. An additional set of estimates (based on a constant 1973 technology) by McQuigg et al. (1973) simulations 0 For Great Plains wheat, average yields fall below the yields obtainable under normal weather con- ditions by about 2.3% due to nonlinear effects of weather variables (McQuigg et al., 1973). m In order to determine the degree of correspondence between the two model results, a correlation analysis was performed. The 53 CRDs of the IES model were weighted according to wheat acreage in production and aggregated to form 10 districts which matched those of the CCEA models (two CCEA districts had to be eliminated). After aggregation 90 yield estimates were available for each set of models (10 districts x 9 yr). Analysis of these data produced a correlation coefficient of 0.73, which is significant at the 0.001 level. Thus, the degree of spatial/temporal 'overlap' amounts to about 53% of the variance. While not great, this figure must be interpreted in light of differences in data, model structure and assumptions. 11 Harvested area is assumed to be 88% of planted, as in 1976. This figure is uniformly applied across subregions and remains constant over time. TABLE I: Simulated Great Plains wheat production: 1930's weather at 1975 technology and 1976 acreage levels University of Wisconsin IES models CCEA models Weather- Predicted yield Production Percent deviation Predicted yield year (bushels/planted (million metric tons) from expected a (bushels/ acre) harvested acre) Production (million metric tons) Percent deviation from normal b Normal or expected 24.9 37.0 30.4 38.1 1932 23.8 35.3 -4.6 29.7 37,2 1933 19.1 29.0 -21.6 27.5 34.5 1934 20.2 30.0 18.9 26.6 33,3 1935 20.7 30.7 -17.0 27.4 34,3 1936 18.5 27.4 -25.9 23.0 28.8 1937 22.3 33.1 -10.5 28.2 35.4 1938 23.3 34.5 -6.8 29.3 36.8 1939 22.6 33.5 -9.5 28.4 35.7 1940 23.4 34:7 -6.2 28.0 35.1 -2.4 -9.4 -12.6 -10.0 -24.4 -7.1 -3.4 -6.3 -7.9 Drought decade 21.5 Average 32.0 -13.5 27.6 Average 34.6 Cumulative deficit 44.7 Cumulative deficit 31.8 -9.3 a Expected yields are equated to 1975 technology trend values. b Normal yields are those model values derived from meteorological normals (calculated from the data period 1932-76) at 1975 technology. ga 20 Richard A. Warrick using Thompson's state-level regression models (discussed previously) is presented in the Appendix. The most striking overall characteristic of all these data is that simulated production falls below normal year after year, consecutively, for the entire span of nine weather-years. Moreover, the worst production years, according to both simulations in Table I, are consecutive from 1933-36. The general patterns of drought impact are roughly comparable to the actual variations in yield departures (per harvested acre) experienced during the real decade of the '30s, as shown in Table II. TABLE II: Actual Great Plains wheat yield departures from expected yields, 1932-40 a Year % Departure 1932 -6% 1933 -32% 1934 -19% 1935 -26% 1936 -24% 1937 -14% 1938 -16% 1939 -23% 1940 --4% a 'Expected' yield for any given year is defined by a polynomial curve fit to Great Plains wheat yields per harvested acre, 1890-1977 (see Warrick, 1980). The major difference between the actual and the simulated yields is quite evident as well: the actual 1930s yield departures are considerably larger. There are several possible reasons for this, including: severe soil erosion, different technology and management practices, different land use patterns, pests and disease, or even methodological reasons, such as differences in the definitions of normal and expected yields or the assumed technology trends. But the year-by-year picture of negative departures is probably more important than the absolute values themselves, for they show that a return of the unusual 1930s weather could once again result in continuous, prolonged declines in expected wheat production. At no time since 1890 have Great Plains wheat yields suffered such an extended calamity, although the early to middle 1950s came close as wheat, inflicted with rust and drought, produced disappointing yields year after year. Nevertheless, the magnitudes of the simulated departures are still considerable. The poorest production year in both simulated series is 1936, in which wheat production declines by about -25%. 12 Other than in the 1930s, actual Great Plains wheat production has fallen that far below expectations only once since the turn of the century (in 1911, a 12 Of course, such Plains-wide yield departures mask the local effects, which could be far more extreme. As we saw earlier, pockets of extremely poor yields - below 50% of expected - are to be found among the CRDs of the Plains. Within these hard-hit CRDs at individual county or farm level, the impacts could be catastrophic. The Possible Impacts on Wheat Production ofl a Recurrence 21 severe drought year). In terms of absolute production at 1975 technological levels, the loss in weather-year 1936 represents about 9.6 million metric tons (mmt) of wheat - or an amount equivalent to the total U.S. exports to the Soviet Union in 1972-73. During the nine-year period, wheat production averages 32.0 to 34.6 mmt - or 13.5% to 9.3% below normal - depending upon the set of models. The cumulative deficit over the decade resulting from these yield departures amounts to 31.8 mint (CCEA Models) or 44.8 mmt (IES models) - roughly a full year's total wheat production on the Great Plains (or, on the demand side, two years' worth of U.S. domestic wheat consumption). This indeed is a sizeable quantity of wheat. But whether such a long-term cumulative deficit per se would have long-term, cumulative effects on prices, consumption, reserve levels, domestic consumption, export, and the like is difficult to say, and perhaps un- answerable. It is, however, likely that the agricultural and economic systems would respond and adjust to production perturbations on a year-to-year basis, in which case the possibility of severe shortfalls - particularly the back-to-back ones like those of 1933-36 simulated weather-years - and their potential consequences deserve close attention by policy-makers and agricultural decision-makers. 7. Model Differences The two sets of models were run to provide two estimates of what could happen under 1930s drought conditions. There was no attempt to estimate a likely range of possible impacts, since the models were not systematically chosen nor turned to reflect 'pessimistic' or 'optimistic' assumptions. 13 One set of models does not necessarily provide a 'check' on the other; they are simply two estimates from basically similar multiple regression equations, derived largely from similar climate and yield data over the same period of record. Thus, one should expect to find close similarities in the results. And for the most part, this is the case: both show negative departures for all simulated weather-years, with the greatest departures falling in the same years, and a similar estimate - about -25% - for the worst year. Nevertheless, the reasons for the dissimilarities deserve elaboration. With the exception of weather-year 1940, the CCEA model estimates of yield deviations are consistently less than the IES model estimates. In several weather years, the departures are twice as large in the CCEA model. These differences could be due in part to the implicit assump- tions about 'abandoned' acreage. The CCEA models predict yields per 'harvested' land unit, while the IES models are built from yield data per 'planted' land unit. In the latter case, it is assumed that land abandonment occurs primarily because of poor weather rather than price or policy influences, and this weather dependence is absorbed into the regression model Actual observed yields per planted land unit, however, are always lower (with greater negative departures during poor years) than yields per harvested acre. Thus, models built from them should predict lower average yield figures and larger relative declines during 13 Confidence bands could be constructed around each set of model estimations. Although not at- tempted here, such a procedure would at least provide a range of estimates around the separate results of each model 22 Richard A. Warrick severe drought than models constructed from yields per harvested acre. The simulated yields in Table I reflect these differences. In this regard, no one set of results is 'better' than the other. Much depends upon what one assumes about the relative roles of weather, technology, prices, and policy in farm management decisions. These factors are inextricably intertwined and notoriously difficult to separate (see Haigh, 1977, for instance). Another major difference between the IES and CCEA models that can shape the nature of the predicted yields is the size and number of observational units (53 CRDs vs 12 subregions) used during multiple regression. The smaller areal units are apt to pick up greater variability in both yield and weather data. 14 In addition, the separate sub-regional formulations of weather-yield relationships present in the CCEA models should be ex- pected to produce results which differ from the uniform, belt-wide formulations of the IES models. Finally, some of the differences in model results could be attributable to errors of statistical estimationl Otherwise, the models are basically very similar in structure and derivation. 8. Summary and Conclusions In a hungry world with an uncertain future, the question of whether a rare climatic event would have crippling impacts on the world's major wheat producing regions is an in- creasingly pressing issue. To explore this question, this paper examined the effects of an unusually severe climatic event - the 1930s drought - on U.S. Great Plains wheat pro- duction. The method involved the use of two sets of multiple regression crop-yield models for spring and winter wheat. Wheat yields were simulated throughout the Plains for nine years, using actual 1930s weather and assuming 1975 levels of technology and 1976 acreage levels. In sum, if the weather-years of the 1930s were to recur, the study suggests that Pervasive, long-duration wheat production declines shouM still be expected. Plains-wide yields averaged below normal for all nine consecutive weather-years of the analysis. In the poorest years, severe impacts on wheat yields wouM be areally widespread. In each of the four worst back-to-back weather years, about nine-tenths of the Plains experienced yields below expected. In the worst weather-years of 1934 and 1936, about half of the entire Plains suffered severe yield declines greater than -25%. The spatial variation in wheat yieM loss wouM be substantial. Even in the worst weather- years of the simulation, yield declines ranged from over 100% to below 40% ofexpected throughout the Plains. In a single, poor drought year, Great Plains wheat production couM be sharply reduced. In the worst weather-year, 1936, simulated production suffered a drop of -25% - a loss of approximately 9.6 million metric tons. This would reduce national wheat production la This advantage may be offset in the IES aggregate model since the model assumes the same weather- yield relationship for each CRD. The Possible Impacts on Wheat Production of a Recurrence 23 by about -15%, assuming average production elsewhere in the United States. The cumulative, decade-long wheat production impacts would be large in relation to total expected production. Yearly yields averaged about 9-14% below normal for the nine weather-years of the simulation. The cumulative loss over the decade is roughly equivalent to a full year's production in the Great Plains. These results should be interpreted cautiously. If any new technologies, management practices, or wheat varieties have been adopted recently which help to buffer yields from poor weather, they are only partially taken into account insofar as they are reflected in the long-term record of wheat yields upon which the models are constructed, is If this is the case, the models may overestimate the yield losses. On the other hand, it is sobering to note that the regression techniques used in the models tend to underestimate the most extreme departures in yield from expected values. The degree to which these factors might balance out is unknown. In general, given the relative accuracy of the models in predicting yields for independent sets of years during model verification procedures (Michaels, 1979; Sakamoto, 1978), one can be confident that, at the very least, the patterns, directions, and rough magnitudes of yield departures - if not the actual figures themselves - indicated by the 1930s simulation are plausible estimates. The major conclusion is that a severe drought could once again inflict heavy damage on Great Plains wheat production. A larger issue, however, is whether drought-inflicted wheat production would result in hardship at local, national, and/or international levels. There is some evidence to suggest that at the local and regional level, the societal impacts of Great Plains droughts have lessened over time, although our agricultural systems have not been sufficiently tested by a long-duration drought in the last 25 years (Bowden et al., 1981 ;Warrick and Bowden, 1981). What we need to know is whether the potential declines in U.S. wheat (and other grains) production, caused by climatic fluctuations, could have serious, disruptive con- sequences in an increasingly interdependent global food system. This is the next research step. Acknowledgements Early drafts were reviewed by Maurice Blackmon, Tom Downing, John Firor, Michael Glantz, Richard Heede, Richard Katz, William Kellogg, Maria Krenz, Clarence Sakamoto, Stephen Schneider, Robert Schware, Tom Stewart, and T. M. L. Wigley. The author is grateful to James McQuigg and Patrick Michaels for making necessary model runs. Thanks are also due to Beverley Chavez, Jan Stewart, and Tracey Dunthorne who typed various drafts of the manuscript. The research was supported by the National Sciences Foundation, through the National Center for Atmospheric Research and through NSF Grant No. ATM-8006635; however, the opinions and conclusions contained in this paper are those of the author solely. 15 Whether such 'buffering' technologies exist remains a highly controversial issue (Schneider and Bach, 1981; Warrick, 1980). 24 Richard A. Warrick APPENDIX: Simulated Great Plains wheat production using Thompson's models: 1930s weather at 1973 technology and 1976 harvested acreagea Weather-year Predicted yield b Production Percent deviation (bushels per (million from 'normal' harvested acre) metric tons) 'Normal' weather 29.8 28.5 1932 28.9 27.7 - 2.8 1933 26.0 24.9 -12.6 1934 24.8 23.7 -16.8 1935 27.5 26.3 - 7.7 1936 25.9 24.8 -13.0 1937 27.7 26.5 - 7.0 1938 27.5 26.3 - 7.7 1939 26.7 25.6 -10.2 1940 26.4 25.3 -11.2 Drought Decade 26.8 Average 25.7 - 9.8 Cumulative deficit 25.4 a Data derived from McQuigg et al. (1973). b Based on five-state weighted average (North Dakota, South Dakota, Nebraska, Kansas and Oklahoma). References Baier, W.: 1977, 'Crop-weather Models and Their Use in Yield Assessments', World Meteorological Organization Technical Note No. 151, Geneva, WMO. Bonnefield, P.: 1979, Dust Bowl, University of New Mexico Press, Albuquerque, New Mexico. Bowden, M., Kates, R., Kay, P., Riebsame, W., Warrick, R., Johnson, D., Gould, H., and Weiner, D.: 1981, 'The Effect of Climatic Fluctuations on Human Populations: Two Hypotheses', in T. M. L. Wigley, M. J. Ingrain, and G. Farmer (eds.), Climate and History: Studies in Past Climates and Their Impacts on Man, Cambridge University Press, Cambridge. Center for Climatic and Environmental Assessment (CCEA) Staff: 1975, 'Wheat Yield Models for the United States', Technical Note 75-1, CCEA, Environmental Data Service, Columbia, Missouri. Clark, W. C. (ed.): 1982, Carbon Dioxide Review: 1982, Oxford University Press, New York. Climate and Society Research Group: 1979, 'The Effects of Climatic Fluctuations on Human Popula- tions', Progress Report No. 2, Clark University Center for Technology, Environment and Develop- ment, Worcester, Massachusetts. Climate/Food Research Group: 1975, 'A Detailed Model of the Production and Consumption of Spring Wheat in the United States', IES Report 49, Center for Climatic Research, Institute of Environmental Studies, University of Wisconsin, Madison, Wisconsin. Climatic Impact Assessment Program (CLAP): 1975, Monograph 5, Impacts of Climatic Change on the Biosphere, U.S. Department of Transportation, Washington, D.C. Duncan, W. G.: 1975, 'SIMAIZ, a Model for Simulating Corn Growth and Yield', in D. N. Baker, R. G. Creech, and F. G. Maxwell (eds.), The Application of Systems Methods to Crop Production, Departments of Agronomy and Entomology, Mississippi State University, Mississippi. Felch, R.: 1978, 'Drought: Characteristics and Assessment', in N. J. Rosenberg (ed.), North American Droughts, Westview Press, Boulder. Great Plains Committee: 1937, The Future of the Great Plains, 75th Cong., 1st Sess., H.R. Doc. 144, U.S. Government Printing Office, Washington, D.C. Haigh, P.: 1977, 'Separating the Effects of Weather and Management on Crop Production', James D. McQuigg, Certified Consulting Meteorologist, Columbia, Missouri. The Possible Impacts on Wheat Production of a Recurrence 25 Institute of Ecology (TIE): 1976, 'Impact of Climatic Fluctuations on Major North American Food Crops', Charle s F. Kettering Foundation, Dayton, Ohio. J/iger, J. and Kellogg, W. W.: 1983, 'Anomalies in Temperature and Rainfall During Warm Arctic Seasons', Climatic Change 5, 39. Katz, R.: 1977, 'Assessing the Impact of Climatic Change on Food Production', Climatic Change 1, 85. Kellogg, W. W. and Schware, R.: 1981, Climate Change and Society: The Consequences of Increasing Atmospheric Carbon Dioxide, Westview Press, Boulder, Colorado. Maas, S. J. and Arking, G. F.: 1980, 'TAMW: A Wheat Growth and Development Simulation Model', Texas Agricultural Experiment Station, Texas A&M University, College Station, Texas. Manabe, S., Wetherald, R. T. and Stouffer, R. J.: 1981, 'Summer Dryness Due to an Increase of Atmospheric CO~ Concentration', Climatic Change 3,347. McQuigg, J., Thompson, L., LeDuc, S., Lockard, M., and McKay, G.: 1973, 'The Influence of Weather and Climate on United States Grain Yields: Bumper Crops or Droughts', Report to the Adminis- trator, National Oceanic and Atmospheric Administration, Washington, D.C. Michaels, P.: 1978, 'A Predictive Model for Winter Wheat Yield in the United States Great Plains', IES Report 94, Center for Climatic Research, Institute of Environmental Studies, University of Wis- consin-Madison, Madison. Michaels, P.: 1979, 'A Simple Large-Area Crop/Climate Model for United States Winter Wheat', pre- print Volume, Fourteenth Conference on Agriculture and Forest Meteorology and Fourth Con- ference on Biometeorology, April 2-6, 1979, Minneapolis, Minn., American Meteorological Society, Boston, Mass. National Academy of Sciences (NAS): 1976, Climate and Food: Climatic Fluctuation and U.S. Agri- cultural Production, Report of the Committee on Climate and Weather Fluctuation and Agri- cultural Production, NAS, Washington, D.C. National Academy of Sciences (NAS): 1979, Carbon Dioxide and Climate: A Scientific Assessment, Report of ad hoc study group on Carbon Dioxide and Climate, Woods Hole, Massachusetts, Na- tional Academy of Sciences, Washington, D.C. National Academy of Sciences (NAS): 1982, Carbon Dioxide and Climate: A Second Assessment, National Academy of Sciences, Washington, D.C. National Defense University (NDU): 1980, Crop Yields and Climate Change to the Year 2000: Vol- ume 1, Government Printing Office, Washington, D.C. Palmer, W.C.: 1965, 'Meteorological Drought', U.S. Department of Commerce Weather Bureau Re- search Paper No. 45, U.S. Government Printing Office, Washington, D.C. Ramirez, 1., Sakamoto, C., and Jensen, R.: 1975, 'Agricultural Implications of Climatic Change', in Climatic Impact Assessment Program (CIAP), Monograph 5, lmpacts of Climatic Change on the Biosphere, U.S. Department of Transportation, Washington, D.C. Saarinen, T.: 1966, Perception of Drought Hazard on the Great Plains, Research Paper No. 106, Department of Geography, University of Chicago, Chicago. Sakamoto, C.: 1978, 'Reanalysis of CCEAI, U.S. Great Plains Wheat Yield Models', Center for Climatic and Environmental Assessment Technical Note 78-3, National Aeronautics and Space Administra- tion, Houston. Schneider, S. H. and Bach, W.: 1981, 'Interactions of Food and Climate: Issues and Policy Consider- ations', in W. Bach, J. Pankrath, and S. Schneider (eds.), Food-Climate Interactions, D. Reidel Publ. Co., Dordrecht, Holland. Sears, P.: 1935, Deserts on the March, University of Oklahoma Press, Norman. Shawcroft, R., Lemon, E., Allen, L., Stewart, D., and Jensen, S.: 1974, 'The Soil-plant-atmosphere Model and Some of its Predictions', Agricultural Meteorology 14 (1/2), 287. Steinbeck, J.: 1939, The Grapes of Wrath, Viking Press, New York. Stewart, D.: 1975, 'Modeling Plant Atmosphere Systems', in Climatic Impact Assessment Program, Monograph 5, lmpacts of Climatic Change on the Biosphere, U.S. Department of Transportation, Washington, D.C. Stewart, D. and Lemon, E.: 1969, 'The Energy Budget at the Earth's Surface: A Simulation of New Photosynthesis of Field Corn', Tech. Rep. ECOM 2-68, 1-6. National Tectinical Information Service, Springfield, Virginia. 26 Richard A. Warrick Stockton, C. W., Mitchell, J. M., and Meko, D. M.: 1981, 'Tree-Ring Evidence of a Relationship Between Drought Occurrence in the Western United States and the Hale Sunspot Cycle', in M. P. Lawson and M. E. Baker (eds.), The Great Plains: Prospeetives and Prospects, Center for Great Plains Studies, University of Nebraska-Lincoln, Lincoln, Nebraska. Stockton, C. W. and Meko, D. M.: 1983, 'Drought Recurrence in the Great Plains as Reconstructed from Long-term Tree-ring Records', JournalofClimate and Applied Meteorology 22, 17. Thompson, L. M.: 1962, 'Evaluation of Weather Factors in the Production of Wheat in the United States', Journal of Soil and Water Conservation 17,149. Thompson, L. M.: 1969a, 'Weather and Technology in the Production of Wheat in the United States', Journal of Soil and Water Conservation 23, 219. Thompson, L. M.: 1969b, 'Weather and Technology in the Production of Corn in the U.S. Corn Belt', Agronomy Journal 61,453. Thompson, L. M.: 1975, 'Weather Variability, Climatic Change and Grain Productions', Science 188, 535. Wallace, H. A.: 1920, 'Mathematical Inquiry into the Effects of Weather on Corn Yields in Eight Corn Belt States', Monthly Weather Review 48,439-446. Warrick, R.: 1980, 'Drought in the Great Plains; A Case Study of Research orl Climate and Society in the U.S.A.', in J. Ausubel and A. K. Biswas (eds.), Climatic Constraints and Human Activities, Pergamon Press, Oxford. Warrick, R. and Bowden, M. J.: 1981, 'Changing Impacts of Drought in the Great Plains', inM. Lawson and M. Baker (eds.), The Great Plains: Perspectives and Prospects, Center for Great Plains Studies, University of Nebraska-Lincoln, Lincoln, Nebraska. Wigley, T. M. L., Jones, P. D., and Kelly, P. M.: 1980, 'Scenario for a Warm, High-CO2 World',Nature 283,205. Worster, D.: 1979, Dust Bowl: The Southern Plains in the 1930s, Oxford University Press, New York. (Received 8 July, 1982; in revised form 18 July, 1983)
View more >
Causes of Drought in the Great Plains - of Drought in the Great Plains ... No one definitive definition exists - drought is in the eyes of ... evaporative stress
The drought that began in the early 1930s wreaked havoc on the Great Plains. During the previous decade, farmers from Texas to North Dakota-the region.
The impacts of parent material and landscape position on drought and biomass production of wheat under semi-arid conditions
Assessment of the possible drought impact on farm production in the SE of the province of Buenos Aires, Argentina
Drought and Heat Wave of 2012 Midwest and Great Plains Worst drought since 1956 with ~60% of contiguous U.S. under drought, worst agricultural drought.