1. Introduction
Information about agricultural usage of climate predictions was sought for three general purposes: to help guide research dealing with predictions, to guide efforts to improve services that provide predictions, and to enhance the usage of climate predictions in agriculture. Information about usage by those involved in agriculture is viewed as valuable since agriculture is the most weather–climate-sensitive sector of the nation's economy. Annual agricultural impacts resulting from weather average $10.4 billion (Changnon and Hewings 2001), a value that ranks ahead of weather effects on energy use ($8.6 billion) and on other sectors of the nation's economy. NOAA (1997) estimated that climate predictions applied in the agriculture and energy sectors produce savings estimated at $2–$4 billion a year. Climate predictions were defined herein as temperature and precipitation conditions predicted to be in one of three classes (above, near, or below average) for future months and seasons.
Seven studies conducted over a 20-yr period concerned the usage of climate predictions in agriculture. Each study involved extensive face-to-face structured interviews and/or extensive questionnaires with persons employed in various agricultural activities. Data collected from those involved in various sectors of agribusiness included information about usage, value of predictions, and factors affecting usage and nonusage. Assessment of these findings allowed identification of recommendations aimed at enhancing future usage.
2. Data and analysis
Each study collected information on each individual's type of firm, type of position, and types of tasks performed, including those that were weather sensitive. In all studies those interviewed were presented with definitions of weather and climate. Information was sought on each person's usage or nonusage of climate predictions. The types of uses were classed into categories including use as general background information and/ or use in making specific decisions. If the individual was a user, further information was sought. This information included 1) data about the annual value of using predictions, and if not known, an estimate of the value; 2) what changes in predictions or other actions had led to usage; and 3) what changes could be made to further enhance prediction usage and improve their value. If the respondent was not a user, information was sought on the reasons for nonusage, and what impediments had to be overcome to become a user.
Decision experiments were conducted with two large agribusiness firms. These presented growing season predictions set at different accuracy levels to decision makers, allowing them to identify potential decisions made under various levels of predictive accuracy for recent past growing seasons (Sonka et al. 1988). These various decisions were then translated into financial outcomes, which were compared with the actual fiscal outcomes in the years with and without any predictive information. These experiments allowed assessment of corporate value of prediction usage based on different accuracy levels.
Temporal analysis of the databases resulting from the seven studies performed between 1981 and 2001 were made in order to gain insight into several conditions important for research or for improving prediction services. These analyses include assessment of the temporal changes in usage, factors that had changed over time and that had led to usage, and what sectors of agriculture have become major users of climate predictions.
The results of this analysis of usage over 20 recent years should have applications in three areas: 1) for agribusiness managers to identify opportunities for use of climate predictions, 2) for atmospheric scientists who perform research on climate predictions, and 3) for government institutions and private firms that provide climate predictions. The results are presented in a chronological format to illustrate the shifts in usage and their causes.
3. Assessments of usage of climate predictions
a. Study 1: 1981–82
A major nationwide assessment of the uses and needs of climate information by the private agricultural sector was performed during 1981–82 (Lamb et al. 1984). The project, funded by government agencies and several agribusinesses, sought to assess climate information needs in the major sectors of agribusiness. After testing with 24 managers, the study involved a nationwide mail survey using a 10-page questionnaire responded to by 125 representatives from nine agribusiness sectors listed in Table 1.
This study identified several needs, including access to near-real-time climate data and information, and a strong need for climate predictions. The results regarding usage of climate predictions for general background information appear in Table 1. The values in Table 1 reveal that in 1981 climate predictions were extensively utilized by those in the grain trade, pest management consulting, and by farm management. Low use of predictions was reported by those in the chemical, seed-growing, and canning industries. The assessment also identified reasons for nonuse of climate predictions. This revealed that 28% of the respondents sampled expressed no need for predictive information, 73% considered the predictions not sufficiently accurate, and 13% indicated the predictions were either not available when needed and/or did not cover time periods of interest. In addition, 10% of the 125 reported uses in making specific decisions. This nationwide assessment provided information helpful in designing and pursuing future studies dedicated totally to prediction usage.
b. Study 2: 1986–88
This study involved extensive interviews with members of three major agribusinesses and also a decision experiment to assess the potential value of predictions having varying accuracies. The firms studied included a leading seed-corn firm, a major national food producer, and a large agricultural chemical firm (Changnon et al. 1988). Extensive face-to-face interviews were conducted in each firm with managers, planners, and those involved in operations, production, and sales. Those interviewed, a total of 37 (13 in one firm and 12 persons in the other two firms), all made major weather-sensitive decisions. Interviews revealed that only four (11%) were using climate predictions for making specific decisions. However, 18 (48%) used them for general information. Several potential uses of predictions were identified in all three firms. None of those interviewed could provide a measure of the value of their use of predictions, although the four using predictions in actual decision making estimated an annual value greater than $20,000 but less than $50,000.
Impediments to prediction use included major concerns over their accuracy. An important finding was that those who saw themselves as potential users did not use predictions because they had no means to assess the internal economic value of using climate predictions, given their status of accuracy, within the complex operations of these firms.
An associated study included a major decision experiment involving 10 operational decision makers in the seed firm (Sonka et al. 1988). These were experienced decision makers who planned for the amount of seed types planted and type of seed carryover between years in dozens of areas in the Midwest. Each participant was given a series of seasonal predictions for different types of weather (based on recent years) in the numerous seed-growing areas across the Midwest. They were asked what decisions would have been made with each of the illustrated predicted values having probabilities set at different levels: 50%, 65%, and 100%, for growing season temperatures and precipitation values at below, near, or above normal levels. The results of this experiment, when tested against financial outcomes of the actual recent years and the decisions without predictions, revealed that had predictions with 50% probability levels been used, the firm would have averaged an increased income of $200,000 per year (1985 dollars). The increased annual income based on 65% probabilities of occurrence was $675,000, and with perfect (100%) predictions, the benefits would have averaged over $1.2 million per year to the firm (Sonka et al. 1988). This served as a useful illustration of the economic value of using uncertain climate predictions.
c. Study 3: 1990–91
An extensive assessment of uses of climate predictions in various agribusiness sectors was conducted during 1990–91 to obtain updated measures of uses, values of usage, potential uses, and impediments to uses of climate predictions across many sectors of agriculture (Changnon et al. 1991). Study 3 involved structured interviews with 47 weather–climate-sensitive decision makers from firms representing six agribusiness sectors including weather insurance, crop consultants, food processors, producers of agricultural chemicals, food producers, and seed producers. At least four persons were interviewed in each company, and each person represented a different area of corporate responsibility. Fifty-one percent of those sampled were found to be users of climate predictions for general background information, and 9 (20%) reported uses in making specific decisions.
The assessment of the value of using climate predictions was qualitative since specific financial values did not exist. The decision makers were asked to estimate the annual value of predictions to their firm as being in one of three classes: small (<$20,000), moderate ($20,000–$100,000), or large (>$100,000). Six of the individuals using predictions in corporate decisions estimated a moderate value and three estimated a high value. The 24 individuals using predictions as general background information classed the value as small. However, all 47 respondents indicated they thought the value of use would be large if predictions were more accurate than current levels. Potential applications of predictions were identified for the planning of budgets, operations, marketing, and sales (Changnon et al. 1991). Impediments to the use of climate predictions were also identified, and 10 identified appear in Table 2. These revealed a wide range of limitations facing usage (Changnon 1992).
d. Study 4: 1992–93
This study involved a mail survey using an extensive questionnaire sent to 181 decision makers in agribusinesses, and 114 responded. This was done to further define uses of climate predictions by decision makers in different types of agribusiness (Sonka et al. 1992). Table 3 presents the reported use of predictions during the 1991–92 period, showing the combined occasional users and the frequent users were 71% of those responding. The type of application reported by 47 of the occasional users was as background information without specific use in any decision. However, 34 respondents (40%) reported applications of predictions in making specific decisions. The good news was a relatively high percentage of the sample were users of climate predictions, but the bad news was that most usage was qualitative (nonspecific) and with low value.
Users in this 1992 study also reported on the monetary value of seasonal prediction usage as being large, moderate, or small, as used in all the studies. This revealed that 22% of the 32 users making specific decisions reported the estimated value as small, 48% reported them as moderate, and 30% reported them as having a high value. The value assessment also asked for ratings of the values of three seasonal predictions (all equal accuracy), and responses showed 83% rated a spring precipitation prediction as having a high financial value, 57% indicated a high value for summer temperatures, and only 22% considered a winter temperature or precipitation prediction as having a high value.
The impediments to use of climate predictions assessed in this study were identical to those shown in Table 2. Many respondents emphasized the need for improvements in the format and timeliness of predictions. This fourth study also led to the identification of the characteristics that agribusiness users desired in climate predictions (Table 4). Several characteristics leading to improved use did not require research to enhance the skill of predictions. Importantly, several called for prediction presentations in more understandable formats, and the inclusion of additional information about past and future conditions based largely on analysis of climatological data.
e. Study 5: 1996–97
A fifth project was conducted during 1996–97 to measure usage after new predictions had been issued since January 1995 by the Climate Prediction Center (CPC) of the National Weather Service (Changnon 1997). These were quite different from the prior climate outlooks CPC had issued, and several changes were in response to some of the characteristics users desired (Table 4). The new predictions had longer lead times, covered periods farther into the future, and provided more regional information than did the prior outlooks.
Structured interviews were conducted with 39 agribusiness managers in six sectors. Sampling included three or four persons in each firm, and two firms from each agricultural sector were sampled. Twenty of those interviewed were persons previously interviewed during the 1990–91 study, study 3, which was done to help assess changes in usage and reasons. The individuals interviewed were in middle- to upper-level management positions involving 1) long-range budget and operational planning, 2) scheduling of operations, 3) determining production quotas, 4) acquiring supplies, and/or 5) advising others in the firms about conditions.
Table 5 presents an analysis of the responses obtained in the 1996–97 interviews. It reveals that 28 of the 39 sampled (72%) were routinely using the new predictions in making decisions, and 32 (82%) were also using them for general information. The value of the prediction usage was identified within one of three categories, and a third of the 28 users making prediction-related decisions (10) reported the value of usage exceeded $100,000, while 50% indicated a moderate value ($20,000–$100,000), and 18% reported a small value of use (<$20,000).
The number of sources of the predictions, as being from CPC and/or private firms, exceeded the sample of 39 since some respondents used predictions from both sources. Importantly, 16 were using private firms, whereas in 1992, only 5 of the 47 persons sampled reported using private sector firms for their predictions. The primary impediments to usage reported were 1) the predictions were seen as having little economic value, 2) prediction accuracy was too low, 3) the format of the information was inadequate to meet internal needs, and 4) internal corporate obstacles to usage. Wilks (2000) noted the potential for high economic value from use of climate predictions, but concluded that the information supplied in the CPC predictions did not provide the type of information needed by many decision makers. The improvements in the CPC-issued predictions that began to be issued in 1995 had led to the elimination of several impediments including those relating to formats, detailed information, longer predictive periods, and timeliness. Many respondents reported that these improvements had led to the use of the predictions in making specific decisions.
A decision experiment also was conducted in 1997, as part of this study, with one of the nation's major food-producing firms (Changnon 1997). This firm contracts for crops to be grown across seven regions in the central United States, and several staff make pre–growing season decisions about regional allocations of three crops to be grown (peas, green beans, and sweet corn). Fourteen staff who were involved in such decisions participated in the experiment. They were each given maps based on probabilistic predictions of precipitation and temperature conditions (above, near, or below average) for different recent growing seasons. They were asked to identify what, if any, decisions would be made for each of the different levels of forecast probabilities including 50%, 65%, and 100% accurate. The results of this decision experiment revealed that use of the 50% probability forecasts in recent years would have provided an annual average increase in the firm's income of $16 million (1997 dollars) related to spatial contractual decisions. Decisions based on the 65% level would have increased annual benefits by $31 million; and the 100% level (total accuracy) would have yielded average profits of $86 million per year to the firm. These results, along with those of the decision experiment done in 1987 (Sonka et al. 1988), revealed that use of climate predictions of conditions at the 50% level could provide sizable benefits to firms that make decisions involving hedging around weather variations.
f. Study 6: 1998–99
A sixth study assessed the effects on usage of the highly accurate seasonal predictions issued as a result of the massive El Niño of 1997–98 (Changnon et al. 2000). Atmospheric conditions associated with the highly publized 1997–98 El Niño event, which developed in the summer of 1997, allowed CPC to issue unusually accurate seasonal predictions for various regions in the United States for the fall, winter, and spring seasons of 1997–98. Assessment of the usage of these climate predictions was made during 1998–99, based on a series of interviews with 17 weather–climate-sensitive decision makers in agribusiness firms (Changnon 2000). The responses of 13 of the agribusiness decision makers did not reveal many specific applications of the highly accurate forecasts, primarily because the El Niño–based predictions were for the fall, winter, and spring seasons and, thus, did not apply to the growing season of 1998. However, prediction-based operational decisions were made by all four agribusinesses that faced cold season issues. All four cases of usage provided a beneficial outcome identified as being >$100,000. Fifteen agribusiness individuals sampled (88%) indicated that they used the predictions as general information and all reported being impressed with the high accuracy.
g. Study 7: 2000–01
In March 2000, the National Oceanic and Atmospheric Administration (NOAA) issued for the first time a series of drought predictions for three different regions of the United States. One of the regions was the Midwest where five states (Illinois, Indiana, Iowa, Missouri, and Nebraska) were predicted to have an intensifying drought through the spring and summer (NOAA 2000). Unlike past climate predictions, no levels of uncertainty were expressed in the prediction, just that the droughts were going to continue and intensify. During June and July heavy rains fell throughout most of the Midwest, quickly ending the drought and making the prediction incorrect for much of the forecast area (Changnon 2002). Agricultural interests reported in the media that the incorrect prediction led many to make decisions that led to sizable financial losses.
An assessment of the use of the drought prediction was made by sampling 809 agricultural producers (farmers) and 208 agribusiness managers in a five-state area (Changnon 2002). Four Midwestern states had drought and were predicted to experience continued drought through 2000, and one state, Ohio, had no drought and was not predicted to have one. Table 6 shows the state distributions for three main actions taken by producers as a result of the drought prediction. As a result of the drought prediction, 144 of the 208 agribusiness staff sampled (69%) used the prediction in some form of decision, and farmers made decisions relating to production choices, purchases of crop insurance, and/or marketing options.
More than 75% of those buying insurance received no payoff (normally only 55% fall into this class), and more than half of all farmers sampled reported financial losses from their production-related decisions. Eighty percent of the farmers who made marketing decisions reported a sizable loss in crop sales revenues, and the principal outcome for prediction users was a regional financial loss estimated at $1.1 billion. The primary lesson for users is to not use a climate prediction that does not offer a measure of its uncertainty, and providers should not present 100% certain forecasts.
4. Summary and recommendations
Seven studies concerning the use of climate predictions in U.S. agribusiness have been conducted over the past 20 years. Review and assessment of their results suggest temporal increases in the amount of usage, improved values from their usage, and in understanding among users as to how to employ predictions in their specific decisions. The usage of climate predictions, as measured in the seven studies conducted during 1981– 2001, was evaluated using the percent of each sample of decision makers (Table 7). One cannot draw firm conclusions from the Table 7 percentages because the seven studies differed 1) in the data-gathering process (face to face versus mail interviews), 2) the number and types of agribusinesses sampled, and 3) types of persons sampled (managers versus farmers). Results suggest that usage of climate predictions for general background information grew from 1981 to 1999, as did usage for making specific decisions.
Usage was assessed in three studies sampling the same sectors of the agribusiness industry, and Table 8 presents results for the six sectors sampled. The values, when examined over the three time periods, show that the largest increases in usage occurred in crop consultants/farm managers, food producers, and seed producer sectors. Slight growth occurred in the agricultural chemical industry, weather insurance, and the food processing industry.
The apparent rapid growth of usage in the seed producing and food producing sectors was largely attributed to the fact these firms can easily hedge their weather risk and adapt to uncertainty in the predications. Firms in both sectors have numerous weather-sensitive regions where seed/food growth is performed and this geographical diversification allows hedging. A failed prediction for one region is not a major loss. The hedging value was further illustrated by the outcomes of the two decision experiments, which were conducted in two large firms having such regional diversification and where several individuals make similar weather-related decisions. The results revealed that use of predictions with accuracy levels of only 50% would have had considerable financial value, and use of forecasts of 65% and 100% accuracy offered quite sizable corporate benefits. Importantly, the benefits occurred in circumstances where the firms can hedge their weather risk. For example, they can use the predictions when the probabilities of the predicted conditions are relatively high and ignore the prediction when probabilities for each outcome (above, near, or below) are similar and thus low.
In the three studies during the 1990s, users applying predictions to actual corporate decisions were asked to estimate the value of using predictions in one of three classes. In study 3, 67% of the respondents reported moderate value ($20,000–$100,000) and 33% reported a high value (>$100,000). In study 4, 48% reported moderate values and 30% high values, and study 5 found moderate values in 50% of those sampled and high values in 32% of those sampled. All three studies found high values reported by 30%–33% of the respondents.
The key factors identified by users as having improved usage since 1981 and the value of usage are listed in Table 9. The major improvement in the formats of predictions made by CPC in 1995 was an important factor. The highly successful El Niño–based predictions of 1997–98, which engendered huge national media attention, were important factors enhancing usage. Predictions based on El Niño and La Niña conditions bring a new dimension of skill and opportunity for use. The development of new technologies, and new public institutions and private firms providing climate services, enhanced usage by making access to expertise much easier by 2002 than in 1981 (Changnon and Kunkel 1999). The search for information to provide corporate advantages in the highly competitive world of agribusiness also has led firms to use predictions more often than 10–20 yr ago (Peters and Waterman 1982).
Assessment of the reasons for nonusage of predictions revealed there is still considerable potential for improving usage. I have identified five recommendations for improving use of climate predictions in U.S. agriculture. A 2003 conference that focused on improving responses to climate forecasts (AMS 2003) identified six findings for improvements in use, and five are among those that I found in assessing these seven agriculturally based studies. Three of the recommendations relate to the atmospheric sciences and producers of climate predictions, and two relate to actions that users and potential users in agriculture need to take.
Atmospheric science–related recommendations include the need to improve the accuracy of predictions. Most importantly, when improvements are achieved, news about these achievements needs to be widely circulated. However, providing a near-perfect prediction does not guarantee wise or beneficial usage. NOAA (1997) noted that improved predictive accuracy may not lead to improvements in usage, recognizing the existence of limitations within decision making institutions to incorporate predictive information. NOAA further noted the need to develop decision making tools based on climate predictions (see the fifth recommendation).
Second, there are ways that climate predictions can be improved by furnishing additional climate information based on analysis of historical information relevant to different agricultural operations. For example, temperature and precipitation predictions need to be related to other conditions such as growing degree days or the incidence of extremes, and to identifying recent years with similar weather to the predicted conditions. Relating predicted conditions to similar past conditions helps decision makers identify past corporate outcomes and decisions when such conditions occurred and thus integrate this information into current decisions.
The third recommendation is for education and outreach by government atmospheric agencies and private sector firms to illustrate to the agricultural industry the many successes and benefits being accrued from use of climate forecasts, and to build a bridge between agribusiness firms and providers of predictions. This addresses the major gap between what is being provided by the government and the information needed by many in agricultural sectors. One could argue that this gap could be filled by private companies involved in weather and climate predictions. One reason for the gap is that many managers in agribusiness have limited knowledge of weather and climate. Hence, they remain uncertain about employing such information in their decisions. A recent assessment identified ways to make climate forecasts more user friendly (National Research Council 1999; Pielke 2000).
My fourth recommendation pertains to the user community and relates to the fact that many agricultural firms do not have a useful measure of the potential value of using uncertain climate forecast information, as revealed by the two decision experiments. The two decision experiments concerning use of climate predictions at relatively low accuracy levels (50%) revealed that relatively low accuracy levels can provide sizable benefits, particularly in firms that make decisions involving hedging around weather variations. Many firms lack decision-oriented economic models that allow assessment of the use of probabilistic weather information, and I strongly recommend such firms to develop such models.
My fifth recommendation calls for access to and use of atmospheric expertise by agricultural firms, either through adding corporate staff educated in atmospheric sciences, or with private companies hired to serve the specific needs of the firm. Successful usage of climate predictions requires a close interaction between the producer and the user (Vogelstein 1998). They must interact, and understand and trust each other to achieve successful usage.
Acknowledgments
This assessment was partially supported by the Midwestern Regional Climate Center.
REFERENCES
AMS, 2003: Improving responses to climate predictions. Atmospheric Policy Program Rep., Amer. Meteor. Soc., 34 pp.
Changnon, D., 2000: Who used and benefitted from the El Niño forecasts? El Niño 1997–1998: The Climate Event of the Century, S. Changnon, Ed., Oxford University Press, 109–135.
Changnon, S. A., 1992: Contents of climate predictions desired by agricultural decision makers. J. Appl. Meteor, 31 , 1488–1491.
Changnon, S. A., 1997: Assessment of uses and values of new climate forecasts. Changnon Climatologist, CRR-43, Mahomet, IL, 35 pp.
Changnon, S. A., 2002: Impacts of the Midwestern drought forecasts of 2000. J. Appl. Meteor, 41 , 1042–1052.
Changnon, S. A., and Kunkel K. E. , 1999: Rapidly expanding uses of climate data and information in agriculture and water resources: Causes and characteristics. Bull. Amer. Meteor. Soc, 80 , 821–830.
Changnon, S. A., and Hewings G. D. , 2001: Losses from weather extremes in the U.S. Nat. Hazards Rev, 2 , 113–123.
Changnon, S. A., Sonka S. , and Hofing S. , 1988: Assessing climate information use in agribusiness. Part I: Actual and potential use and impediments to usage. J. Climate, 1 , 757–765.
Changnon, S. A., Sonka S. , and Hofing S. , 1991: Assessment of uses of climate predictions in agribusiness and the implications for research and presentation of predictions. Preprints, Fifth Conf. on Climate Variations, Denver, CO, Amer. Meteor. Soc., 443–446.
Changnon, S. A., Pielke R. A. Jr., Changnon D. , and Wilkins L. , 2000: Summary: Surprises, lessons learned, and the legacy of El Niño 1997–98. El Niño 1997–1998: The Climate Event of the Century, S. Changnon, Ed., Oxford University Press, 197–209.
Lamb, P. J., Sonka S. , and Changnon S. A. , 1984: The present and potential use of climate information by the U.S. agricultural sector. Illinois State Water Survey, Contract Rep. 344, Champaign, IL, 138 pp.
National Research Council, 1999: Making climate forecasts matter. National Academy Press, 68 pp.
NOAA, 1997: Implement seasonal and interannual climate forecasts. National Oceanic and Atmospheric Administration Strategic Plan 1995–2005, Washington, DC, 15 pp.
NOAA, cited 2000: Drought grips nearly half the U.S. NOAA News, 13 March. [Available online at http://www.noaanews.noaa.gov/stories.].
Peters, T. J., and Waterman R. H. , 1982: In Search of Excellence. Harper and Rowe, 160 pp.
Pielke R. A. Jr., , 2000: Policy responses to El Niño 1997–1998: Implications for forecast values and the future of climate services. El Niño 1997–1998: The Climate Event of the Century, S. Changnon, Ed., Oxford University Press, 172–196.
Sonka, S. T., Changnon S. A. , and Hofing S. , 1988: Assessing climate information use in agribusiness. Part II: Decision experiments to estimate economic value. J. Climate, 1 , 766–774.
Sonka, S. T., Changnon S. A. , and Hofing S. , 1992: How agribusiness uses climate predictions: Implications for climate research in provision of predictions. Bull. Amer. Meteor. Soc, 73 , 1999–2008.
Vogelstein, F., 1998: Corporate America loves the weather. U.S. News and World Report, 13 April, 48–49.
Wilks, D. S., 2000: On interpretation of probabilistic climate forecasts. J. Climate, 13 , 1965–1971.
Intrasectoral variation in extent of use of climate pre dictions by agribusiness in the United States during 1981–82. The number of respondents for each parameter are in parentheses: pre cipitation and then temperature
Impediments to use of climate predictions identified by 47 managers in six agribusinesses sampled during 1991–92, with the number of firms reporting each impediment
Use of climate predictions during 1991–92, as reported by 114 survey respondents from nine different agribusiness sectors
Characteristics desired in climate forecasts based on responses by 114 decision makers in the nine sectors of agribusiness sampled in 1992
Usage and values of new climate predictions and sources of predictions used in 1996–97 survey of 39 individuals
Actions of agricultural producers in the Midwest in response to the Mar 2000 drought prediction
Use of climate predictions for general background in formation or for specific decisions in agribusinesses, as found in seven studies during 1981–2001
Usage of climate predictions in various agribusiness sectors, as measured in three studies, and the number of users in each sector is expressed as a percent of total sampled
Major factors that led to growth in the use of climate predictions in agriculture from 1981 to 2001
