CROP YIELD RESPONSE TO DROUGHT IN ALABAMA Except where reference is made to the work of others, the work described in this thesis is my own or was done in collaboration with my advisory committee. This thesis does no include proprietary or classified information. __________________________ Tyler Leigh Kreps Certificate of Approval: __________________________ __________________________ Patricia A. Duffy Diane Hite, Chair Professor Professor Agricultural Economics Agricultural Economics and Rural Sociology and Rural Sociology __________________________ __________________________ Luke Marzen George T. Flowers Assistant Professor Dean Geology and Geography Graduate School CROP YIELD RESPONSE TO DROUGHT IN ALABAMA Tyler Leigh Kreps A Thesis Submitted to the Graduate Faculty of Auburn University in Partial Fulfillment of the Requirement for the Degree of Masters of Science Auburn, Alabama December 18, 2009 iii CROP YIELD RESPONSE TO DROUGHT IN ALABAMA Tyler Leigh Kreps Permission is granted to Auburn University to make copies of this thesis at its discretion, upon request of individuals or institutions and at their expense. The author reserves all publication rights. __________________________ Signature of Author __________________________ Date of Graduation iv VITA Tyler Leigh Kreps, daughter of Stan and Pam Kreps and Granddaughter of Martha Silver of Raleigh, NC, was born on September 22, 1985. She graduated from Booker T. Washington High School in Pensacola, FL in May 2003. She attended Troy University for her undergraduate career. She graduated with a degree in History in May 2007. She continued on at Auburn University for a Masters of Science in Agricultural Economics in the Fall of 2007. During her career at Auburn University she was the GTA for Aerial Photography and Remote Sensing, and Geographic Information Systems (GIS). v THESIS ABSTRACT CROP YIELD RESPONSE TO DROUGHT IN ALABAMA Tyler Leigh Kreps Masters of Science, December 18, 2009 (B.S., Troy University, 2007) 96 Typed Pages Directed by Diane Hite A model was used to estimate crop yield response to drought in Alabama for corn, cotton, hay, peanuts, and soybeans. The analysis includes Alabama County level data for crop yields and weather variables for the years 1986-2005, along with drought, hurricane, and policy variables. The results indicate that independent weather variables, extreme weather events, and government policy significantly affect crop yield per acre in Alabama. vi ACKNOWLEDGEMENTS The author would like to thank Dr. Diane Hite for her time and generosity with the thesis analysis and as her academic adviser. The author would also like to thank Dr. Luke Marzen and Dr. Patricia A. Duffy for their work as committee members. The author would like to thank the United States Geological Survey (USGS) and the Alabama Water Resource Research Institute (AWRRI) for their funding and for allowing the author to work on the grant titled ?Estimating Regional and Local Scale Surface Moisture as an Indicator of Drought and Crop Yield Using Thermal IR Remote Sensing.?. The author would like to thank AmericaView for providing data as well as funding. Also, the author would like to thank Donn Rodekohr for his time and generosity with GIS help and analysis. vii Style manual or journal used American Journal of Agricultural Economics Computer software used Microsoft Word, Microsoft Excel, Microsoft Access, SAS 9.1 (English), ArcGIS 9.2, ERDAS Imagine 9.1, and NSPECT 1.5 viii TABLE OF CONTENTS LIST OF TABLES............................................................................................................. ix I. INTRODUCTION .................................................................................................. 1 Drought: Definitions and Characteristics........................................................................ 3 Major US Droughts......................................................................................................... 5 Research Objectives........................................................................................................ 8 II. ALABAMA CROP PRODUCTION ...................................................................... 9 Corn............................................................................................................................... 10 Cotton............................................................................................................................ 12 Hay................................................................................................................................ 14 Peanuts .......................................................................................................................... 16 Soybeans ....................................................................................................................... 19 III. LITERATURE REVIEW ..................................................................................... 21 IV. DATA AND METHODS ..................................................................................... 30 Weather Variables......................................................................................................... 32 Temperature (Growing Degree Days) .......................................................................... 32 Precipitation .................................................................................................................. 35 Erosion .......................................................................................................................... 37 Evapotranspiration Variables........................................................................................ 38 Weather Dummy Variables........................................................................................... 39 Policy Dummy Variables.............................................................................................. 41 Models........................................................................................................................... 43 V. RESULTS ............................................................................................................. 48 Corn Model Results ...................................................................................................... 51 Cotton Model Results ................................................................................................... 53 Hay Model Results........................................................................................................ 56 Peanut Model Results ................................................................................................... 57 Soybean Model Results................................................................................................. 59 VI. CONCLUSION..................................................................................................... 61 BIBLIOGRAPHY............................................................................................................. 65 APPENDIX....................................................................................................................... 72 ix LIST OF TABLES Table 2.1 Alabama Corn Production ............................................................................ 10 Table 2.2 U.S. Corn Production.................................................................................... 11 Table 2.3 Alabama Cotton Production.......................................................................... 13 Table 2.4 U.S. Cotton Production................................................................................. 13 Table 2.5 Alabama Hay Production.............................................................................. 15 Table 2.6 U.S. Hay Production ..................................................................................... 15 Table 2.7 Alabama Peanut Production.......................................................................... 18 Table 2.8 U.S. Peanut Production................................................................................. 18 Table 2.9 Alabama Soybean Production....................................................................... 20 Table 2.10 U.S. Soybean Production............................................................................ 20 Table 4.1 Descriptive Statistics of the Data Used in the Estimation ............................ 31 Table 4.2 Base Temperatures for Selected Crops and Insects...................................... 33 Table 4.3 Alabama K Factors by County...................................................................... 38 Table 4.4 Independent Variables and Definitions......................................................... 45 Table 5.1 Results for Corn Model................................................................................. 51 Table 5.2 Results for Cotton Model.............................................................................. 53 Table 5.3 Second Unrestricted Cotton Model............................................................... 55 Table 5.4 Results for Hay Model.................................................................................. 56 Table 5.5 Results for Peanut Model.............................................................................. 57 Table 5.6 Results for Soybean Model........................................................................... 59 1 I. INTRODUCTION The United States is the world?s largest producer and leading exporter of agricultural products. The evaluation of fluctuations in crop production in the United States has become increasingly important in recent years, due to increased world population and increasing climate variability in all regions of the world (Kogan, 2000). If crop production in the U.S. experiences low yields for even a few successive years, a decrease in the quality of produce available and an increase in the cost of produce will be seen both domestically and internationally. Many factors shift crop yields and production from year to year, but the most important factor is commonly agreed to be extreme weather events (Luttrell and Gilbert, 1976). The 1990?s were named the ?International Decade for Natural Disaster Reduction? on December 11, 1987 by the United Nations General Assembly Resolution (CDC, 1994). Extreme weather events makeup almost eighty five percent of all natural disasters, and the most damaging environmental phenomenon is drought. During 1967- 91, 2.8 billion people were affected by weather-related disasters, and 51% of these people were affected by drought. Also 3.5 million people perished from weather-related disasters with 45% of the fatalities being related to drought (Obassi, 1994; Kogan, 2000). Severe and extreme droughts occurring in the years 1991-2000 claimed 50-150 million tons of grain, the main source of food for the world?s 6 billion people (a number that is likely to double by 2050). Satisfying the world demand for food and feed will become more 2 difficult due to increasing population and the negative effects of frequent droughts on agricultural production (Kogan, 2000). All parts of the United States are vulnerable to drought. The significant economic, social, and environmental costs are experienced at the local, state, and regional levels. Each year an average of 12% of the United States (excluding Alaska and Hawaii) experiences severe to extreme drought, and in 1934 a record breaking 65% of the population were affected (Wilhite and Svoboda, 2000; Wilhite and Buchanan-Smith, 2005). The U.S. National Drought Mitigation Center (NDMC) found that severe and extreme drought affected more than 25 % of the country in one out of four years (Wilhite and Svoboda, 2000). Annual losses attributed to drought were estimated, by the U.S. Federal Emergency Management Agency (FEMA), to be $6-8 billion (Wilhite and Svoboda, 2000; Wilhite and Buchanan-Smith, 2005, FEMA, 1995). However, the U.S. vulnerability to drought is much different than most developing countries, where major concerns are food security and nutritional needs, environmental degradation, and the development process (Wilhite and Svoboda, 2000; Wilhite and Buchanan-Smith, 2005). The economic, physical, and social effects of drought could be detrimental in developing countries, resulting in famine, abandonment of whole geographic regions, and even human suffering or death (Riebsame et al., 1990; Changonon, 1999; Kogan, 1997, 2000). In the last 50 years the United States has seen an agricultural technology boom, known as the Green Revolution. The advancing technologies associated with this revolution have increased crop yields per acre dramatically in a relatively short period of time (Kogan, 1997, 2000). Biological sciences have produced hybrid seed that are high- yielding and disease tolerant. Chemistry advancements have resulted in fertilizers and 3 insecticides that produce higher yields at a lower cost. More efficient machinery and agricultural equipment can be attributed to better engineering. Most important to increased crop production has been the technological advancements made in the field of agronomy. Agronomy has improved crop rotations, soil management, planting, cultivation, and harvesting by examining the specific needs of different crops and crop varieties (Luttrell and Gilbert, 1976). However, technological efforts of the Green Revolution are not enough to offset the negative impacts experienced when widespread drought occurs. Unpredictable weather fluctuations from year to year cannot be controlled with technology. It is even thought that the adoption of high-yielding and disease tolerant hybrid seed in combination with improved crop management techniques can have a negative effect on crop yields during drought years. The uniform planting practices and field operations can cause crop yields to become overly sensitive to weather conditions. Agricultural production should be adjusted to meet climatic conditions each year. This can be found difficult due to the unpredictable nature of weather and the lack of information available to farmers (Isik and Devados, 2006). Drought: Definitions and Characteristics Drought occurs in almost all climatic regions, and is a normal, temporary, and recurring feature of climate (Wilhite and Svoboda, 2000; Wilhite and Buchanan-Smith, 2005). A natural reduction in the amount of precipitation received in an area over an extended period of time causes drought to occur. The timing and effectiveness of the precipitation is also an important factor. Each drought year is unique in its climatic 4 characteristics and impacts. Drought can devastate even very productive lands and can go undetected. The severity of a drought can be aggravated by high temperatures, high winds, and low relative humidity. Severity of drought is normally determined by duration, intensity, and spatial extent of the drought episode. Also important to determine the severity of drought on a specific region, is to examine the demand placed on the water supply of that region by human activities and vegetation (Wilhite and Svoboda, 2000; Wilhite and Buchanan-Smith, 2005). Drought has three main differences from other weather-related natural hazards (Wilhite and Svoboda, 2000; Wilhite and Buchanan-Smith, 2005). The first difference is that the onset and end of a drought are difficult to determine. The impact of drought can not immediately be observed by eye or even ground data. The effects of drought may linger for years accumulating slowly over an extended period of time. Drought is referred to as a creeping phenomenon due to its cumulative impacts (Tannehill, 1947; Wilhite and Svoboda, 2000; Wilhite and Buchanan-Smith, 2005). Climatologists struggle to determine the onset of a drought and scientists and policy makers debate the criteria for declaring an end to a drought. This confusion leads to the second difference separating drought from other weather-related natural hazards; there is no universally accepted definition of drought. The existence and degree of severity of a drought must be regionally specific. It can occur in high as well as low rainfall areas and can affect even the most fertile lands. The principal season of occurrence, delays in the start of the rainy season and the occurrence of rains in relation to principal crop growth stages all play an important part in defining a drought in a specific region. Scientists and policy makers need definitions of drought that are 5 formulated with actual drought situations taken into consideration. There are numerous definitions for drought; many of them hold no meaningful content for declaring drought or specific impacts in key economic sectors (Wilhite and Svoboda, 2000; Wilhite and Buchanan-Smith, 2005). Finally drought impacts, in comparison to floods, hurricanes, tornadoes, or other natural hazards, are nonstructural (Wilhite and Svoboda, 2000; Wilhite and Buchanan- Smith, 2005). These impacts are felt over a much larger geographical area than the region where the drought actually occurred. An example of this is the U.S. drought of 1988 which was felt on the global scale and is discussed below. Because drought impacts are not structural or localized, government development of drought contingency plans are hindered by many complications. Accurate, reliable, and timely estimates of the impacts of drought are becoming increasingly important as world demand for food and feed increases (Wilhite and Svoboda, 2000; Wilhite and Buchanan-Smith, 2005). Major US Droughts Drought, a very common phenomenon in the United States, occurs on average in 12% of the nation every year (Wilhite and Svoboda, 2000), and agriculture is often seriously affected. Economic, environmental, and social impacts of drought are substantial in the United States, the world?s largest producer and leading exporter of agricultural products. The impact of the large-area severe drought in 1988 on the U.S. economy has been estimated to cause around $40 billion in damages to the environment, human health, and wildlife. This number can be compared to the $15 billion in damages from the 1989 San Francisco (Loma Prieta) earthquake. For the first time in the last half 6 century grain production fell below domestic consumption (Reibsame et al., 1990; Kogan 1997, 2000). The 1988 drought occurred in the most productive area of the Great Plains, the breadbasket of the United States (FAO, 2000; Kogan, 1997, 2000). By the end of June 1988 moisture and thermal stress, combined with the critical timing of the drought, had devastated corn and grain crop growth. U.S. corn production was reduced by 30%, a number that was felt globally. In 1988 total world grain production dropped 3% and corn production was 50 million tons less than in 1987 (FAO, 2000; Kogan, 1997, 2000). In 1989 the U.S. experienced a drought very early on and by the end of April winter wheat crops were affected. Vegetation stress continued on into the summer months, but only a few states of the central and northern Great Plains saw a reduced spring crop production (Kogan, 1997, 2000). The southwest and south central states experienced severe drought conditions in 1996, resulting in serious losses in crop and livestock production, increased wildfire and forest fires, and decreased public water supplies (Wilhite and Svoboda, 2000; Wilhite and Buchanan-Smith, 2005). Water-based tourism and recreational activities also took an economic loss due to the decreases in surface and ground water supplies. High temperatures also increased the demand for energy. Kansas, Oklahoma, Arizona, Utah, Nevada, New Mexico, Texas, and Colorado were among the states with the most substantial losses (Wilhite and Svoboda, 2000; Wilhite and Buchanan-Smith, 2005). Texas alone was estimated to have had nearly $5 billion in losses (Boyd, 1995). In 1998, the drought affected the southwest and south central region of the Great Plains again and expanded into the southeastern states (Wilhite and Svoboda, 2000; 7 Wilhite and Buchanan-Smith, 2005). Agricultural losses were felt in Texas, Oklahoma, Louisiana, South Carolina, and Georgia, and Florida experienced wildfires. Drought conditions in these areas returned in 1999. This drought also expanded into areas of the mid-Atlantic and northeast states, causing concerns about U.S. vulnerability to drought conditions. Widespread drought in the spring and summer months of 2000 resulted in severe impacts on agriculture and municipal water supplies in three regions of the country: southwest and south central states, southeastern and Gulf Coast states, and central and western Corn Belt states. The southeast region had the greatest devastation due to three years of reoccurring drought in Georgia, Florida, South Carolina, and Alabama (Wilhite and Svoboda, 2000; Wilhite and Buchanan-Smith, 2005). The impact of drought in the southeastern United States is likely to increase in magnitude as fresh water becomes increasingly scarce and as populations grow. Reliable and up to date information related to surface moisture conditions could be used to forecast crop yields, assess distressed areas for allocation of disaster relief funds, and could help resource managers and government officials plan ahead for difficult financial times. In the state of Alabama drought has caused serious problems for farming communities for the past several years and most recently in the summer of 2006. These conditions place farming communities in difficult financial situations, which also substantially affect other economic sectors within the state. In efforts to improve understanding of environmental weather factors associated with moisture stress, this study examines crop yield response to drought in Alabama. 8 Research Objectives This study has four objectives for the analysis of crop yield response to drought in Alabama. The first objective of this thesis is to identify a framework for analyzing crop yield response of major crops to climate fluctuations in Alabama. The second objective is to test and expand the weather variables examined by previous studies. The third objective is to econometrically use yield and climatic data from 1986 through 2005 to estimate the impacts of weather variables on crop yields. The fourth objective is to take into account government legislation affecting agricultural programs. This study looks at the Federal Agriculture Improvement and Reform Act of 1996, and the Farm Security and Rural Investment Act of 2002. 9 II. ALABAMA CROP PRODUCTION Crop production is an important aspect when considering the effects of drought on agriculture. Some crops produced in Alabama during the summer growing season include corn, cotton, hay, peanuts, and soybeans. Not all crops are grown in every county in Alabama and planting dates vary among these commodities. Also, which crops are produced in each county differ from year to year. However, agricultural producers in Alabama only have the option to plant one type of crop on each acre of land during a single April-September growing season. This study examines Alabama crop production for corn, cotton, hay, peanuts, and soybeans during the 1986-2005 summer growing seasons. Crop production statistics for each crop were collected from the United States Department of Agriculture?s (USDA) National Agriculture Statistic Serves (NASS) Quick Stats website (http://www.nass.usda.gov/QuickStats/) and are presented in tables (2.1) through (2.10). When examining crop production in Alabama it is also important to look at the same crops on a national level. Each major crop produced in Alabama is evaluated on both the state and national level in the following sections. Corn Corn is the primary feed grain making up 90% of the total feed grain production and use in the United States. Corn is produced for both human and animal consumption, and for industrial use. Most U.S. states grow corn, but the majority of corn production occurs in the Heartland region (USDAa, 2008). Alabama corn production from 1986- 2005 is found in Table 2.1. U.S. corn production from 1986-2005 is found in Table 2.2. On average 217,250 acres of corn are harvested and 17,989,000 bushels of corn are produced during the 20 year period. The average yield per acre for 1986-2005 is 84 bushels, which is approximately 33.5 % less than the U.S. average yield per acre for the same time period. Year Planted Acres Harvested Acres Yield (Bushels/Acre) Total Production (Bushels) 1986 340,000 270,000 57 15,390,000 1987 300,000 250,000 72 18,000,000 1988 240,000 170,000 44 7,480,000 1989 230,000 180,000 81 14,580,000 1990 290,000 240,000 58 13,920,000 1991 260,000 210,000 80 16,800,000 1992 330,000 295,000 94 27,730,000 1993 300,000 250,000 55 13,750,000 1994 290,000 260,000 96 24,960,000 1995 250,000 220,000 75 16,500,000 1996 300,000 270,000 82 22,140,000 1997 280,000 250,000 87 21,750,000 1998 300,000 200,000 63 12,600,000 1999 220,000 200,000 103 20,600,000 2000 230,000 165,000 65 10,725,000 2001 180,000 150,000 107 16,050,000 2002 200,000 180,000 88 15,840,000 2003 220,000 190,000 122 23,180,000 2004 220,000 195,000 123 23,985,000 2005 220,000 200,000 119 23,800,000 Table 2.1 Alabama Corn Production 10 Year Planted Acres (Thousand) Harvested Acres (Thousand) Yield (Bushels/Acre) Total Production (Thousand Bushels) 1986 76,580 68,907 119.4 8,225,764 1987 66,200 59,505 119.8 7,131,300 1988 67,717 58,250 84.6 4,928,681 1989 72,322 64,783 116.3 7,531,953 1990 74,166 66,952 118.5 7,934,028 1991 75,957 68,822 108.6 7,474,765 1992 79,311 72,077 131.5 9,476,698 1993 73,239 62,933 100.7 6,337,730 1994 78,921 72,514 138.6 10,050,520 1995 71,479 65,210 113.5 7,400,051 1996 79,229 72,644 127.1 9,232,557 1997 79,537 72,671 126.7 9,206,832 1998 80,165 72,589 134.4 9,758,685 1999 77,386 70,487 133.8 9,430,612 2000 79,551 72,440 136.9 9,915,051 2001 75,702 68,768 138.2 9,502,580 2002 78,894 69,330 129.3 8,966,787 2003 78,603 70,944 142.2 10,087,292 2004 80,929 73,631 160.3 11,805,581 2005 81,779 75,117 147.9 11,112,187 Table 2.2 U.S. Corn Production On average Alabama farmers harvest more than half a million acres of corn each year; making corn Alabama?s most important grain crop. Corn yields in Alabama can be attributed to several factors. Many factors which determine yield can be controlled with good management practices and technology, these include: soil fertilization and liming, tillage, hybrid selection, planting dates and depths, plant populations and row width, and weed and insect control. Uncontrollable weather factors are also a frequent problem in Alabama corn production. Insufficient moisture and high temperatures during the silking and tasseling stages of growth can dramatically reduce yields (Mask, 2009). The largest single variable cost in corn production is nitrogen fertilization, which produces consistently large increases in corn yields. However, corn produced during a season experiencing drought conditions can have elevated levels of nitrogen. This nitrogen toxicity in the grain can be harmful to humans and animals (McWilliams et al., 11 12 1999). Shortages of phosphorus, potassium, and sulfur in soil can also lead to reduced corn yields. Lime should be added to acid soils (low pH) to prevent poor corn yields. Low pH levels reduce root growth and nutrient availability, and can also result in toxicity of some elements and poor activity of herbicides. Alabama production guide for non- irrigated corn can be found on the Alabama Cooperative Extension System web site (http://www.aces.edu/dept/grain/cornpro.php). The suggested planting dates for Alabama corn in the different regions of the state are: in South Alabama: March 1-April 20; Central Alabama: March 15-April; and North Alabama: March 25-May 15 (Mask, 2009). Cotton The United States is one of the world?s four largest cotton-producing countries. Grown annually from seed each year, cotton is produced in 17 states in the U.S. and the cotton industry provides over 400,000 Americans with jobs. Cotton production in the U.S. has seen a decrease in planted acres and an increase in yield per acre due to technological advances and better production practices (USDAb, 2008). Alabama cotton production from 1986-2005 is found in Table 2.3. U.S. cotton production from 1986- 2005 is found in Table 2.4. On average 463,050 acres of cotton are harvested and 587,900 pounds of cotton are produced during the 20 year period. The average yield per acre for 1986-2005 is 605 pounds, which is approximately 8.5 % less than the U.S. average yield per acre for the same time period. Year Planted Acres Harvested Acres Yield (Pounds/Acre) Total Production (Bales) 1986 315,000 313,000 506 330,000 1987 335,000 333,000 572 397,000 1988 390,000 375,000 486 380,000 1989 328,000 322,000 571 383,000 1990 380,000 378,000 476 375,000 1991 410,000 405,000 655 553,000 1992 415,000 408,000 731 621,000 1993 443,000 430,000 524 469,000 1994 463,000 455,000 766 726,000 1995 590,000 578,000 409 492,000 1996 520,000 516,000 734 789,000 1997 535,000 442,000 597 550,000 1998 495,000 475,000 559 553,000 1999 565,000 561,000 535 625,000 2000 590,000 530,000 492 543,000 2001 610,000 605,000 730 920,000 2002 590,000 540,000 507 570,000 2003 525,000 510,000 772 820,000 2004 550,000 540,000 724 814,000 2005 550,000 545,000 747 848,000 Table 2.3 Alabama Cotton Production Year Planted Acres (Thousand) Harvested Acres (Thousand) Yield (Pounds/Acre) Total Production (Thousand Bales) 1986 9,933 8,357 547 9,525 1987 10,259 9,894 702 14,475 1988 12,325 11,759 615 15,077 1989 10,210 9,166 602 11,504 1990 12,117 11,505 632 15,147 1991 13,802 12,716 650 17,216 1992 12,977 10,863 694 15,710 1993 13,248 12,594 601 15,764 1994 13,552 13,156 705 19,324 1995 16,717 15,796 533 17,532 1996 14,395 12,632 700 18,414 1997 13,648 13,157 666 18,245 1998 13,064 10,449 619 13,476 1999 14,584 13,138 595 16,294 2000 15,347 12,884 626 16,799 2001 15,499 13,560 694 19,602 2002 13,714 12,174 652 16,530 2003 13,301 11,826 723 17,823 2004 13,409 12,809 843 22,505 2005 13,975 13,534 825 23,260 Table 2.4 U.S. Cotton Production 13 14 Hay More acres of hay are harvested in Alabama each year than any other commodity. There are few larger commercial producers of hay, and the majority of hay is produced and harvested by cattle and horse producers for their own use. Hay has often been overlooked as a major crop in Alabama, even though hay production is essential to Alabama?s multimillion-dollar cattle and horse industries. In addition, there is a large demand for hay from the construction and landscaping industry. The Alabama Department of Environmental Management (ADEM) now requires land left idle for more than 13 days to be reseeded by grass for erosion control, and hay mulch is needed to cover newly seeded and fertilized lands (Collins, 2005). Alabama hay production from 1986-2005 is found in Table 2.5. U.S. hay production from 1986-2005 is found in Table 2.6. On average 756,750 acres of hay are harvested and 1,675,200 tons of hay are produced during the 20 year period. The average yield per acre for 1986-2005 is 2 tons, which is approximately 27% less than the U.S. average yield per acre for the same time period. Year Harvested Acres Yield (Tons/Acre) Total Production (Tons) 1986 700,000 1.6 1,120,000 1987 700,000 2.1 1,470,000 1988 750,000 2 1,500,000 1989 700,000 2.2 1,540,000 1990 750,000 1.5 1,125,000 1991 780,000 2.1 1,638,000 1992 710,000 2.1 1,491,000 1993 720,000 2 1,440,000 1994 730,000 2.7 1,971,000 1995 720,000 2.1 1,512,000 1996 730,000 2.4 1,752,000 1997 770,000 2.25 1,733,000 1998 750,000 2.1 1,575,000 1999 800,000 2.3 1,840,000 2000 720,000 1.8 1,296,000 2001 920,000 2.6 2,392,000 2002 825,000 2.2 1,815,000 2003 780,000 2.6 2,028,000 2004 850,000 2.7 2,295,000 2005 730,000 2.7 1,971,000 Table 2.5 Alabama Hay Production Year Harvested Acres (Thousand) Yield (Tons/Acre) Total Production (Thousand Tons) 1986 62,334 2.49 155,385 1987 60,133 2.45 147,457 1988 64,771 1.94 125,736 1989 62,722 2.31 144,706 1990 61,030 2.4 146,212 1991 61,834 2.46 152,073 1992 58,903 2.49 146,903 1993 59,689 2.46 146,699 1994 58,815 2.55 150,136 1995 59,764 2.58 154,239 1996 61,169 2.45 149,779 1997 61,084 2.5 152,536 1998 60,006 2.52 151,387 1999 63,181 2.53 159,582 2000 60,355 2.54 153,603 2001 63,516 2.46 156,416 2002 63,942 2.34 149,467 2003 63,371 2.48 157,390 2004 61,944 2.55 158,122 2005 61,637 2.44 150,461 Table 2.6 U.S. Hay Production 15 16 Peanuts Climate and soil requirements for peanuts limit production to only a few states in the U.S., making it a relatively minor crop. In the Southeast peanuts are produced in Georgia, Alabama, Florida, and South Carolina. Farms that grow cotton, soybeans, corn, and wheat will typically grow peanuts on a 3 to 4 year rotation. The most common crop alternative for peanuts is cotton. During the 1996 Farm Act peanuts were only grown on about 12,000 farms in the U.S. and peanut revenues made up only one percent of the total U.S. crop production revenue, averaging $1 billion a year. During 1999-2000 about 60 percent of the U.S. peanut production occurred in the Southeast, and during 2000-2002 peanuts accounted for over 20 percent of total crop value in Alabama and Georgia (Dohlman et al., 2004). Peanuts were regulated by marketing quota systems, which limited the amount of peanuts that could be sold for food use in the U.S. domestic market. Under the 1996 Farm Act peanuts produced over the quota level were exported or sold for oil and peanut meal in the lower value crush markets. Producers with quota rights received a government- established loan rate of $610 per ton which gave growers strong incentive to attempt to produce the maximum amount of peanut production allocated to them. The loan rate of $132 per ton was given to peanut producers without quota rights (Dohlman et al., 2004). The 2002 Farm Act ended the peanut price support system. This dramatic change in policy resulted in two general observations that can be seen in both the U.S. and AL peanut production tables. First, there is a steep decrease in the planted acreage from 2001 to 2002. Planted acreage then sees an increase each year after 2002. Second, peanut producers took advantage of planting flexibility by shifting peanut production to higher 17 yielding areas. This shift can be attributed to the high level of crop revenue the peanuts provide. Even though peanut acres only make up a fifth of cropland on peanut growing farms, peanuts account for 30 % of the crop revenue (Dohlman et al., 2004). Alabama ranks third in peanut production in the United States, preceded by Georgia and Texas, generating more than $100 million per year for the Alabama economy. During 1986-2005 Alabama peanut production made up approximately 12.26% of the total peanut production in the United States. The top peanut producing counties in Alabama are Houston, Baldwin, Henry, and Geneva (Dixon, 2009). Alabama peanut production from 1986-2005 is found in Table 2.7. U.S. peanut production from 1986- 2005 is found in Table 2.8. On average 215,550 acres of peanuts are harvested and 482,198,000 pounds of peanuts are produced during the 20 year period. The average yield per acre for 1986-2005 is 2,241 pounds, which is approximately 13% less than the U.S. average yield per acre for the same time period. Year Planted Acres Harvested Acres Yield (Pounds/Acre) Total Production (Pounds) 1986 n/a 219,000 2,260 494,940,000 1987 n/a 220,000 2,115 465,300,000 1988 n/a 236,000 2,380 561,680,000 1989 n/a 239,000 2,250 537,750,000 1990 n/a 256,000 1,510 386,560,000 1991 n/a 277,000 2,305 638,485,000 1992 n/a 236,000 2,505 591,180,000 1993 n/a 239,000 1,980 473,220,000 1994 223,000 222,000 2,010 446,220,000 1995 213,000 212,000 2,280 483,360,000 1996 192,000 191,000 2,355 449,805,000 1997 194,000 193,000 1,930 372,490,000 1998 198,000 197,000 2,195 432,415,000 1999 207,000 206,000 2,175 448,050,000 2000 190,000 182,000 1,490 271,180,000 2001 200,000 199,000 2,675 532,325,000 2002 185,000 180,000 2,110 379,800,000 2003 190,000 185,000 2,750 508,750,000 2004 200,000 199,000 2,800 557,200,000 2005 225,000 223,000 2,750 613,250,000 Table 2.7 Alabama Peanut Production Year Planted Acres (Thousand) Harvested Acres (Thousand) Yield (Pounds/Acre) Total Production (Thousand Pounds) 1986 1,564.7 1,535.2 2,408 3,697,085 1987 1,567.4 1,547.4 2,337 3,616,010 1988 1,657.4 1,628.4 2,445 3,980,917 1989 1,665.2 1,644.7 2,426 3,989,995 1990 1,846.0 1,815.5 1,985 3,603,650 1991 2,039.2 2,015.7 2,444 4,926,570 1992 1,686.6 1,669.1 2,567 4,284,416 1993 1,733.5 1,689.8 2,008 3,392,415 1994 1,641.0 1,618.5 2,624 4,247,455 1995 1,537.5 1,517.0 2,282 3,461,475 1996 1,401.5 1,380.0 2,653 3,661,205 1997 1,434.0 1,413.8 2,503 3,539,380 1998 1,521.0 1,467.0 2,702 3,963,440 1999 1,534.5 1,436.0 2,667 3,829,490 2000 1,536.8 1,336.0 2,444 3,265,505 2001 1,541.2 1,411.9 3,029 4,276,704 2002 1,353.0 1,291.7 2,571 3,321,040 2003 1,344.0 1,312.0 3,159 4,144,150 2004 1,430.0 1,394.0 3,076 4,288,200 2005 1,657.0 1,629.0 2,989 4,869,860 Table 2.8 U.S. Peanut Production 18 19 Soybeans Soybeans account for about 90% of oilseed production in the U.S., and soybean oil accounts for 55-65% of U.S. vegetable oils and animal fats consumed. Nationally, soybean acreage has increased in recent years due to planting flexibility, narrow-rowed seeding practices, increased corn-soybean crop rotations, and the adoption of herbicide- tolerant varieties. Soybean yields per acre can be higher for areas that have a more concentrated planting practice (USDAc, 2008). Planted acres in Alabama have seen a decreasing trend since 1986, Table 2.11 illustrates soybean production in Alabama from 1986-2005. U.S. soybean production from 1986-2005 is found in Table 2.12. On average 315,750 acres of soybeans are harvested and 7,665,250 bushels of soybeans are produced during the 20 year period. The average yield per acre for 1986-2005 is 26 bushels, which is approximately 28.6% less than the U.S. average yield per acre for the same time period. Year Planted Acres Harvested Acres Yield (Bushels/Acre) Total Production (Bushels) 1986 650,000 610,000 23 14,030,000 1987 600,000 580,000 18 10,440,000 Table 2.9 Alabama Soybean Production 1988 590,000 570,000 25 14,250,000 1989 600,000 570,000 21 11,970,000 1990 470,000 440,000 17 7,480,000 1991 360,000 350,000 23 8,050,000 1992 290,000 270,000 29 7,830,000 1993 310,000 295,000 24 7,080,000 1994 310,000 295,000 31 9,145,000 1995 240,000 225,000 24 5,400,000 1996 320,000 305,000 34 10,370,000 1997 350,000 340,000 25 8,500,000 1998 340,000 320,000 22 7,040,000 1999 240,000 200,000 16 3,200,000 2000 190,000 160,000 18 2,880,000 2001 140,000 135,000 35 4,725,000 2002 170,000 155,000 24 3,720,000 2003 170,000 160,000 36 5,760,000 2004 210,000 190,000 35 6,650,000 2005 150,000 145,000 33 4,785,000 Year Planted Acres (Thousand) Harvested Acres (Thousand) Yield (Bushels/Acre) Total Production (Thousand Bushels) 1986 60,405 58,312 33.3 1,942,558 1987 58,180 57,172 33.9 1,937,722 1988 58,840 57,373 27.0 1,548,841 1989 60,820 59,538 32.3 1,923,666 1990 57,795 56,512 34.1 1,925,947 1991 59,180 58,011 34.2 1,986,539 1992 59,180 58,233 37.6 2,190,354 1993 60,085 57,307 32.6 1,869,718 1994 61,620 60,809 41.4 2,514,869 1995 62,495 61,544 35.3 2,174,254 1996 64,195 63,349 37.6 2,380,274 1997 70,005 69,110 38.9 2,688,750 1998 72,025 70,441 38.9 2,741,014 1999 73,730 72,446 36.6 2,653,758 2000 74,266 72,408 38.1 2,757,810 2001 74,075 72,975 39.6 2,890,682 2002 73,963 72,497 38.0 2,756,147 2003 73,404 72,476 33.9 2,453,845 2004 75,208 73,958 42.2 3,123,790 2005 72,032 71,251 43.1 3,068,342 Table 2.10 U.S. Soybean Production 20 21 III. LITERATURE REVIEW The evaluation of fluctuations in crop production in the United States has become increasingly important in recent years (Luttrell and Gilbert, 1976). If crop production in the U.S. experiences low yields for even a few successive years, a decrease in the quality of produce available and an increase in the cost of produce will be seen. Many factors shift crop yields and production from year to year, but the most important factor is commonly agreed to be weather fluctuations. The evaluation of patterns associated with crop yields is important when determining agricultural policy. This evaluation includes national accumulation of crop reserves needed to maintain satiability during low crop yielding years (Luttrell and Gilbert, 1976). This chapter reviews published literature on crop production response to climate change which is applicable to drought and possibly could be used to estimate drought?s impacts on crop yield. Mendelsohn et al. (1994) use agricultural land prices to measure the economic impact of climate. They used the Ricardian approach to examine the impact of climate and other economic factors on farmland values and revenues in the U.S. for 1978 and 1992. Nearly 3,000 counties in the U.S. were used as cross-sectional data in this study. The Ricardian approach was compared to the traditional production-function approach for estimating economic impacts of climate change. By varying input variables such as precipitation, temperature, and carbon dioxide levels these traditional studies rely on underlying production functions to predict environmental damage caused by global 22 warming. Mendelsohn et al. criticizes production-function studies for having inherent bias and overestimated results for the impact of global warming on reduced crop yields, even though their estimates rely on calibrated crop yield models. Important traditional production-function studies referred to by Mendelsohn et al. include Callaway et al. (1982), W. Decker et al. (1986), Adams et al. (1988, 1990), Adams (1989), D. Rind et al. (1990), and Rosenzweig and Perry (1994). The bias discussed is referred to as the ?dumb-farmer scenario? and implies farmers will not adjust their practices in response to changing environmental or economic conditions. Overestimations are said to be caused by the failure of production-function models to allow for complete adjustments such as new crop production decisions, major technology advancements, or conversion of agricultural land to other land uses. Mendelsohn et al. address this bias in their model by replacing the dependent variable from crop yields with the net value of farmland. The Ricardian approach allows for direct measurement of the effects climate change has on different crops, while also allowing for the indirect effects to be measured. Adjustments in farmland value take into account land user?s management decisions for both changes in nutrient inputs as well as crop selections. They also include urban measurements to account for urban development that might have replaced agricultural lands. Weather variables include January, April, July, and October temperature and precipitation averages. Soil factors for erosion, soil type, moisture capacity, and permeability are also included. Mendelsohn et al. find that climate change has high degrees of nonlinear effects on agriculture. These effects are found to vary season to season. Also, the effect of climate change on farmland value was found to be dramatically different than the effect 23 of climate change on farmland revenue. The impact of global warming was found to have a dramatic negative impact of farmland values. Different results were found for the climate impact on farmland revenue. The revenue model suggested global warming actually had a positive effect on crop revenue. Cline (1996) comments on Mendelsohn et al. (1994) stating that even though their study provided an important service for analysis of global warming impacts for the U.S., their study also understated the damaging impacts. He explains three reasons for the underestimation. First Cline questions the conceptual framework behind applying cross- sectional analysis to future climate (global-warming) impacts based on current land values. He calls their approach ingenious for trying to capture farmer responses to changing climate, but also critiques the investigation for not recognizing other methods of the production-function approaches such as those in the studies by Easterling et al. (1993) and Rozenzweig et al. (1993). Cline (1996) further questions the meaningfulness of the results of Mendelsohn et al. in comparison with direct estimates from the improved production-function models listed above. Second, Mendelsohn et al. assumed implicitly that the price of irrigated water was infinitely elastic. The problem that arises from the partial-equilibrium model or Ricardian approach with respect to water availability for irrigation is also discussed by Cline. He quickly points out that experts are speculating on water scarcity in the U.S. and this will increase in years to come. Finally, to measure global warming effects in the U.S. Mendelsohn et al. used the global mean for precipitation and warming effects. Cline states that there is evidence suggesting actual damages in the U.S. would be more severe than global data implies and that regional data would be more appropriate. Schlenker and Roberts (2006) examine the reduced-form relationship between daily weather records and county level corn yields in eastern United States for 1950- 2004. Their study combines broad aggregate weather measures used in reduced-form regressions and detailed nonlinear weather interaction used in crop-simulation models to estimate weather effects on corn yields. Schlenker and Roberts? detailed set of daily weather records and corn yields from almost 2,000 U.S. counties allow them to accurately estimate nonlinear impacts of weather on yields. Equation (3.1) below is the model proposed by Schlenker and Roberts to identify the appropriate temperature bounds as well as utilize a large data set. Yield growth is a nonlinear function of heat , so that the log yield , in county and year is ()hg it y i t (3.1) () () itiitit h h it czdhhhgy ??? +++= ? Where ()h it ? = the time distribution of temperatures over the season in each county h and h = the lower and upper bounds of temperature observed it z = other factors, such as precipitation and technological change i c = a time-invariant county-fixed effect In their study, growing degree days (GDDs) are incorporated to measure the effects of temperature. They define degree days as the sum of truncated degrees of a given day that occur between the lower and upper bounds of temperature. Degree days are then summed over the entire growing season producing the GDDs. Schlenker and Roberts explain their flexible functional form allows them to observe possible negative effects of extreme heat above the upper temperature threshold on yield. Their model also treats time as being dimensionless, implying that temperature is perfectly substitutable over time. 24 Equation (3.2) is given by Schlenker and Roberts to discretize the integral over heat by using 1?C intervals ranging from -5 ?C to 50 ?C. The 1? heat interval ()1, +hh becomes ()(hh itit )?? ?+1 and replaces ( )h it ? from Equation (1.1).They also approximate the nonlinear function by using a mth order Chebyshev Polynomial for j = 1,..., m , and evaluate the interval at midpoints. ()hg () j T (3.2) ()()()[] itiit x itit h j m j jit czhhhTY jit ???? +++?++= ?? ?== 4444434444421 , 15.0 49 51 Where = the exogenous variable obtained by summing the jth order Chebyshev Polynomial evaluated at each temperature interval midpoint multiplied by the time spent in each temperature interval tji x , Schlenker and Roberts found that yield increases as temperatures increase from 12 ?C to 25 ?C and then quickly becomes negative for temperatures that exceed 30 ?C. They found a significant nonlinear relationship between temperature and corn yields, with an R? of 0.76 for the regression that included year and county fixed effects. Schlenker et al. (2006) estimate potential impacts of global warming on farmland values for U.S. counties east of the 100 th meridian. Mendelsohn et al.?s (1994) Ricardian approach was used for the bases of their model. The study area, east of the 100 th meridian, represents the boundary where agricultural production is possible without the use of irrigation. This was done to eliminate discrepancies between previous studies. Schlenker et al.?s model is unique in the way it links climatic, soil, and socioeconomic factors to farmland values. The growing season months April through September were selected to represent U.S. crop response to ambient weather conditions. Precipitation monthly averages for 25 April-September were collected for the 30 years proceeding each census year. Degree days were chosen to represent the effects of temperature on plant growth. Schlenker et al. explain that much agronomic literature represents plant growth as being linear within a certain range. They use temperatures between 8 and 32?C for the linear range, and temperatures of 34?C to represent the harmful effects of temperature on crop growth. Schlenker et al. experimented with both linear and quadratic specifications for the degree days with base 34?C, and found the square root to be the best fit. Explaining that the square root function best approximated the dramatic negative effects high temperatures have on plant growth. They also found the soil K factor to be significant at the 10% level in three of the five regression runs. Soil K factor represents erodibility of top soil, which can be harmful to the productivity of agricultural lands. Linking climatic, soil, and socioeconomic to farmland values resulted in robust estimates that remained robust even after various specification tests. Isik and Devadoss (2006) used historical data on crop yields and climatic variables to develop an econometric model of stochastic production functions. Their analysis focused on the impacts of climate change on crop yields and yield variability. They quantify the impacts of climate variables on the mean, variance, and covariance for wheat, barley, potatoes, and sugar beet yields in Idaho. Also, the estimated production function parameters and their elasticities are used to show the projected long-term temperature and precipitation on Idaho agricultural yields. Isik and Devadoss use the Just-Pope (1978) production function, equation (3.3), to represent crop yields. (3.3) () ( ) 2/1 ;; ??? itititit xhxfy += 26 Where = crop yield for region i and year t it y = weather variables for region i and year t it x it ? = the stochastic term with mean zero and variance 2 ? ? ? and? = the production function parameters to be estimated The effects of the independent weather variables on mean crop yield is given by the estimation of expected crop yield ( ( )?; it xf ). The effect of the independent weather variables on the variance of yields is given by the estimation of ( )?; it xh . Isik and Devadoss use weather variables that include a constant, precipitation, temperature, and a trend. They choose the Just-Pope function because it does not impose a priori restriction of the risk effects of these inputs, and provided accommodation to both increasing and decreasing risk effects on production outputs. Isik and Devadoss looked toward Saha et al. (1997) for an estimation of the Just-Pope production function, equation (3.4), interpreted as estimation with heteroscedastic errors. (3.4) () ititit uxfy += ?; Where = it u () 2/1 ;?? itit xh Var( ) = it u ()?? ? ; 2 it xh In order to ensure positive out-variance Isik and Devadoss assumed that the exponential form for the variance of crop yield was Var( )= it u ( ) it x?exp with =1 (i.e. 2 ? ? it ? ~N(0,1)). According to Saha et al., the Just-Pope production function has been widely used in applied economics. Just and Pope proposed a theoretical model that allowed the effects of inputs on the stochastic component of production to be analyzed separately from the deterministic component, and offer two alternative methods of estimating the stochastic production function. Just and Pope?s 1978 article examines a maximum likelihood (ML) 27 procedure for estimating the stochastic production function and their 1979 article provided a production function estimation by feasible generalized least squares (FGLS) under heteroscedastic disturbances. Saha et al. (1997) compares the Just-Pope production function estimated by FGLS, which is the traditional estimation method, to the ML estimator. Their main objective is to investigate the small-sample properties of FGLS and ML estimators in heteroscedastic error models. They found that the maximum likelihood approach was more effective for estimation. In order to take into account region-specific effects the log-likelihood function for panel data estimation was given by equation (3.5), where N is the number of observations. (3.5) () ( )( ) () ? ? ? ? ? ? + ? +?= ?? == N i n i it it itit x x xfy NL 11 2 exp ; 2ln* 2 1 ln ? ? ? ? Isik and Devadoss (2006), under the assumption it ? ~N(0,1), use the maximization of equation (3.2) to estimate the parameters of ? and? . They estimated both a linear and quadratic to represent the relationship between the mean of Idaho crop yields and the weather variables, but only use the linear functional form for estimation of the relationship between the variance and covariance of crop yield and the weather variables. Their econometric model found varying effects of temperature and precipitation on wheat, barley, potatoes, and sugar beet yields. The effect of climate on the mean of yields was found to be modest. For most of the crops the variance and covariance of yield was significantly reduced by climate change. 28 29 A study done by Luttrell and Gilbert (1976), investigated whether or not weather changes have random effects on crop yields. According to Luttrell and Gilbert the belief in weather cycles dates back to the Old Testament and since then several studies have been done to evaluate if crop yields are random, cyclical, or bunchy. These studies include Jevons (1884), Jevons (1909), Moore (1923), Fulmer (1972), and Lin et al. (1963). Luttrell and Gilbert used United States crop yield data from 1866-1932 for wheat, corn, rye, barley, and oats, and crop yield data from 1933-1974 for wheat, corn, rye, barley, oats, and cotton. According to Luttrell and Gilbert the empirical results of the 1866-1932 tests primarily reflect the influences of weather on crop yields, based on the natural fertility of the land. They found little evidence of positive autocorrelation, or bunchiness for crop yields during this time period. The test results for 1933-1974 also found no evidence of bunchiness. They did however find some evidence of positive autocorrelation indicating that the trend of yields was misspecified by regressing the natural log of yield. They attributed crop yields variation from trend to increased technology, such as hybrid crops and fertilizer, and government policy. One main finding of their statistical evaluation of crop yields was that they found little evidence for nonrandom, cyclical, or bunchy yields in either the national average yields or the weighted average for major agricultural producing areas in the United States. Another important finding was that government acreage control programs created major changes in cotton production. According to Luttrell and Gilbert acreage removed from production during years of government policy were probably less fertile than the acreage that remained in production. They found that years prior to acreage restriction, cotton yields were below trend, and years during acreage restriction were above trend. 30 IV. DATA AND METHODS This analysis used panel data to evaluate the effects of weather variables on average crop yields for Alabama during the years 1986-2005. This study used the statistical analysis software SAS 9.1. Data for corn, cotton, hay, peanuts, and soybeans were collected at the county level. Average crop yields for each commodity from 1986- 2005 were collected from the United States Department of Agriculture?s (USDA) National Agriculture Statistic Serves (NASS) Quick Stats website (http://www.nass.usda.gov/QuickStats/) and can be found in the Appendix. Weather variables including precipitation, minimum and maximum temperature, wind speed, solar radiation, and relative humidity were collected from Texas A&M Blackland Laboratory . Descriptive statistics for yield and weather variables are found in Table 4.1. N Mean Minimum Maximum Std. Dev Corn Data Corn Yield Bushels Per Acre 960 79.41 9.34 178.00 27.95 Growing Degree Days (50-86?F) in Thousands 960 4.24 3.47 4.89 0.25 Temperature Stress Degree Days 960 262.41 17.03 763.61 114.26 Coefficient of Variation for Precipitation 960 2.86 1.99 5.77 0.47 Average Precipication (mm) 960 3.85 1.37 9.02 1.20 Average Daily Solar Radiation (MJ m-2d-1) 960 21.82 19.01 24.44 0.93 Average Wind Speed (m/sec) 960 2.79 2.28 3.87 0.30 Average Relativ e Humidity 960 68.51 58.25 77.07 3.84 Cotton Data Cotton Yield Hundred Pounds Per Acre 620 5.97 1.72 12.72 1.75 Growing Degree Days (60-86?F) in Thousands 620 2.60 2.03 3.13 0.19 Temperature Stress Degree Days 620 266.87 31.08 645.12 111.52 Coefficient of Variation for Precipitation 620 2.85 1.99 5.43 0.48 Average Precipication (mm) 620 3.92 1.37 9.02 1.25 Average Daily Solar Radiation (MJ m-2d-1) 620 21.84 19.30 23.95 0.95 Average Wind Speed (m/sec) 620 2.83 2.28 3.87 0.32 Average Relativ e Humidity 620 68.33 58.25 76.52 3.80 Hay Data Hay Yield Tons Per Acre 1340 2.23 0.80 3.89 0.52 Growing Degree Days (40-86?F) in Thousands 1340 5.97 5.14 6.71 0.27 Temperature Stress Degree Days 1340 256.55 17.03 763.61 114.72 Coefficient of Variation for Precipitation 1340 2.85 1.99 5.77 0.46 Average Precipication (mm) 1340 3.83 1.37 9.02 1.16 Average Daily Solar Radiation (MJ m-2d-1) 1340 21.87 19.01 24.44 0.93 Average Wind Speed (m/sec) 1340 2.80 2.28 3.87 0.29 Average Relativ e Humidity 1340 68.01 57.59 77.07 4.02 Peanut Data Peanut Yield Thousand Pounds Per Acre 240 2.26 1.13 4.08 0.44 Growing Degree Days (56-86?F) in Thousands 240 3.32 2.85 3.76 0.15 Temperature Stress Degree Days 240 291.37 67.58 645.12 116.00 Coefficient of Variation for Precipitation 240 2.84 2.00 5.43 0.51 Average Precipication (mm) 240 4.04 1.86 7.38 1.28 Average Daily Solar Radiation (MJ m-2d-1) 240 22.00 19.50 23.88 0.89 Average Wind Speed (m/sec) 240 2.57 2.32 2.98 0.18 Average Relativ e Humidity 240 70.87 64.81 77.07 2.49 Soybean Data Soybean Yield Bushels Per Acre 440 25.54 6.00 46.54 7.74 Growing Degree Days (50-86?F) in Thousands 440 4.13 3.47 4.89 0.25 Temperature Stress Degree Days 440 236.42 17.03 645.12 104.92 Coefficient of Variation for Precipitation 440 2.82 1.99 4.88 0.43 Average Precipication (mm) 440 3.83 1.37 7.87 1.12 Average Daily Solar Radiation (MJ m-2d-1) 440 21.47 19.30 23.95 0.87 Average Wind Speed (m/sec) 440 2.86 2.28 3.33 0.26 Average Relativ e Humidity 440 68.16 58.25 75.86 4.00 Table 4.1 Descriptive Statistics of the Data Used in the Estimation 31 32 Weather Variables The total effect environmental conditions have on plant yield can only be observed after the crops are harvested. Plant stress results from a number of factors and often is not visually noticeable. Soil moisture, atmospheric conditions, nutrients, diseases, insects, and weeds all interact on a daily basis to create numerous different kinds of crop stress (Shaw et al., 2009). Disease, insect, and weed control, nutrient application are all environmental factors that are under control of management practices. These factors are not considered by this study, and are held constant under the ceteris paribus assumption. This study will focus on the environmental factors that cannot be controlled (i.e. soil moisture and atmospheric conditions), and their relationship with crop yield. Moisture stress occurs when soil moisture and atmospheric conditions become unbalanced. Soil moisture is determined by both the amount of precipitation and soil characteristics. Atmospheric conditions can be defined as the combination of air temperature, the amount of energy available (solar radiation), dryness of the air (humidity), and movement of evaporation from plant surfaces (wind) (Shaw et al., 2009). The following sections explain how each weather variable is measured and how each is posited to affect crop yields. Temperature (Growing Degree Days) In agriculture it is important to find a way to measure the impact of temperature on crop development, and also the potentially negative effect of weeds and insects on crop yields. Agricultural producers use a concept called growing degree-days (GDD) to measure the impacts of air temperature on plant growth, development, and maturity. GDD is based on the idea that plants have a minimum temperature at which plant growth will start to occur, and a maximum temperature at which growth will shut down. Each crop and variety has its own minimum developmental threshold temperature ( ) (Fraisse et al., 2007, 2009). Table 4.2 lists for selected crops and insects provided by the University of Florida?s Institute of Food and Agricultural Sciences Extension. base T base T Table 4.2 Base Temperatures for Selected Crops and Insects Crop Base Temperature Corn, Sorghum, Rice, Soybeans, Tomato 50 ?F Cotton 60 ?F Peanuts 56 ?F Potato, Sunflower 45 ?F Wheat, Barley, Rye, Oats, Flaxseed, Lettuce, Asparagus 40 ?F Insect Base Temperature Alfalfa Weevil 48 ?F Black Cutworm, European Corn Borer 50 ?F Corn Rootworm 44 ?F Green Cloerworm 52 ?F GDD is calculated for each 24-hour day during plant development. The accumulation of GDD throughout a growing season can provide useful information on how daily air temperatures and plant development are related. Equation (4.1) illustrates the standard calculation for GDD. (4.1) GDD = base MINMAX T TT ? ? ? ? ? ? ? + 2 Where = the daily maximum reported temperature MAX T = the daily minimum reported temperature MIN T = the crop specific minimum temperature required for growth base T 33 The standard way of calculating GDD can be modified in several ways. One of the most common modifications for agricultural studies, shown in equation (4.2), is to set the daily minimum temperature ( ) equal to if < . MIN T base T MIN T base T (4.2) GDD = base baseMAX T TT ? ? ? ? ? ? ? + 2 Where = the daily maximum reported temperature MAX T base T = the crop specific minimum temperature required for growth Equation (4.2) can also be modified to consider the maximum temperature at which plant growth will start to shut down is known as the upper developmental threshold ( ). The upper cutoff is commonly assumed to be equal to 86 ?F. Equation (4.3) illustrates the upper developmental threshold modification for GDD by setting the daily maximum temperature ( ) equal to if > . cutoff T MAX T cutoff T MAX T cutoff T (4.3) GDD? = base basecutoff T TT ? ? ? ? ? ? ? ? ? + 2 Where = the upper developmental threshold temperature (86 ?F) cutoff T base T = the crop specific minimum temperature required for growth (Fraisse et al., 2007, 2009). This study uses the modified equation (4.3) to calculate the GDD for corn, cotton, hay, peanuts, potatoes, and soybeans for each county in Alabama for the years 1986- 2005. Each crop?s GDD represents the range between the minimum temperature and maximum temperature at which growth occurs, and should represent the positive linear effects of temperature on yield. However, if a crop is experiencing stress, high temperatures even within the range of each crop GDD could have a negative effect. 34 Temperatures experienced over the upper developmental threshold ( ) have a negative effect on yield. The accumulation of degree days above are known as temperature stress degree days (TSDD). Calculation of the TSDD is represented by equation (4.4). cutoff T cutoff T (4.4) TSDD = GDD ? GDD? Where GDD = growing degree days without an upper developmental threshold GDD?= growing degree days with an upper developmental threshold (Danneberger and Street, 1985, Walker and Hatfiled, 1979) Crops have different GDD accumulations based on their specific . In this study each county has the same TSDD accumulation for each crop, because is assumed to be the same for all crops. Even though TSDD is calculated beyond the point were plants are negatively effected by heat, there is a point where the incremental damage from further exposure to temperatures over is dramatically reduced. To capture this affect the square root of TSDD has been placed in the model, as applied by Schlenker et al, 2006. base T cutoff T cutoff T Precipitation Crop moisture stress originates from a deficiency of precipitation over a period of time. The occurrence, and severity, of moisture stress is unique to each crop and crop variety. The timing of precipitation is as important to crop yield as the amount of precipitation received in an area. Effectiveness of precipitation for crop production comes from a direct relationship between rainfall intensity and the occurrence of precipitation during principle development stages (NDMC, 2006). The studies of Schlenker et al. 35 36 (2006) and Isik and Devadoss (2006) both use the average precipitation over the growing season to measure the effects of precipitation on yield. However, precipitation is a dynamic process, and the uses of data collected at traditional ground based weather stations have a major problem with accuracy. As rainfall moves through an area, its form and intensity can change dramatically. The data may only be accurate for the actual amount of rain for that particular weather station and a small area around the weather station. Not only the amount of rain, but the occurrence of rain is in question. Precipitation acquired at weather stations may not have been received by agricultural lands, and vice versa (Jensen and Pedersen, 2005). The data used in this study come from one weather station for each county to measure atmospheric conditions. In order to account for measurement errors in the precipitation amounts and frequency, the coefficient of variation (CV) was used to measure precipitation. Because crop moisture stress originates from a deficiency of precipitation over a period of time, it is important to look at the variability of precipitation over time as a function of the amount of precipitation received. CV is the standard deviation divided by the mean (Jensen and Pedersen, 2005). For this study the CV of precipitation was calculated by dividing the standard deviation of average precipitation for each county in each year by the average precipitation for each county in each year. The CV for precipitation can be considered a measurement of risk and has a negative relationship with yield. As the coefficient of variation increases, risk for planting crops also increases. Variables for the average precipitation from each station as well as averaged precipitation squared were included in the model to capture both the positive and negative effects of precipitation. 37 Erosion Soil moisture is determined by both the amount of precipitation and soil characteristics. One way to characterize soil is by its erosion properties. Erosion is an important factor when looking at agricultural yields for an area. The four soil properties that determine the erodibility of soil are; particle size, structure, organic content, and permeability. Based on the four properties, soil?s potential for erosion by water can be interpreted by a number known as the K factor (IWR-MSU, 2002). County level soil data is collected and managed by the USDA Natural Resource Conservation Service. The Soil Survey Geographic (SSURGO) database can be found at the USDA Soil Data portal (http://soildatamart.nrcs.usda.gov/). K factor is the soil erodibility factor. Medium textured soils, such as the silt loam, are moderately susceptible to detachment. K factor values of 0.25 to 0.4 produce moderate runoff. Values of 0.4 or greater are highly erodible (IWR-MSU, 2002). The average K factor for each county in Alabama, collected from SSURGO is shown in Table 4.3. In this study the K factor was used as an interactive variable with average precipitation. Table 4.3 Alabama K Factors by County County Name K Factor County Name K Factor County Name K Factor Autauga 0.21 Dallas 0.26 Marion 0.30 Baldwin 0.21 De Kalb 0.26 Marshall 0.26 Barbour 0.19 Elmore 0.24 Mobile 0.17 Bibb 0.27 Escambia 0.22 Monroe 0.26 Blount 0.29 Etowah 0.31 Montgomery 0.30 Bullock 0.22 Fayette 0.30 Morgan 0.30 Butler 0.23 Franklin 0.25 Perry 0.28 Calhoun 0.31 Geneva 0.20 Pickens 0.30 Chambers 0.28 Greene 0.28 Pike 0.21 Cherokee 0.33 Hale 0.28 Randolph 0.25 Chilton 0.27 Henry 0.16 Russell 0.23 Choctaw 0.27 Houston 0.17 Shelby 0.33 Clarke 0.24 Jackson 0.27 St. Clair 0.33 Clay 0.26 Jefferson 0.32 Sumter 0.29 Cleburne 0.30 Lamar 0.29 Talladega 0.30 Coffee 0.19 Lauderdale 0.35 Tallapoosa 0.25 Colbert 0.30 Lawrence 0.31 Tuscaloosa 0.30 Conecuh 0.20 Lee 0.24 Walker 0.28 Coosa 0.27 Limestone 0.37 Washington 0.26 Covington 0.18 Lowndes 0.28 Wilcox 0.29 Crenshaw 0.21 Macon 0.22 Winston 0.28 Cullman 0.28 Madison 0.33 Dale 0.16 Marengo 0.28 Evapotranspiration Variables Evapotranspiration can be defined as water lost from the surface of plants into the atmosphere (USGS, 2009). There are six factors that affect transpiration rates from plants: the type of plant, soil-moisture available, temperature, relative humidity, radiation, and wind. Evapotranspiration is part of how plants ?breathe?. Like people plants have to breathe to stay alive. Some types of plants transpire less than others and need less water to stay alive. Also if soil-moisture isn?t available plants whither and transpire less. High temperatures also increase evapotranspiration rates. When relative humidity is high in the air surrounding the plant evapotranspiration will happen less quickly. Humidity in the air allows for plants to keep moisture in their leaves. One thing 38 39 to point out is that 10% of all moisture in the air is due to moisture lost during the plant transpiration process (USGS, 2009). High levels or humidity could indicate higher rates of plant transpiration. A plant?s photosynthesis and transpiration is also affected by the amount of radiation intercepted by the plant. Leaf area development throughout the growing season determines the amount of radiation intercepted by the crop (NeSmith, 1997). Solar radiation is necessary for plant development until a certain level. After that level of radiation is received by the plant radiation can have a negative impact of growth and development. High levels of solar radiation can increase the evapotranspiration process and create plant stress. Not so obvious are the effects of wind on crop yield production. Wind plays a critical role in the growth and development of crops in many ways. Fast winds can significantly reduce crop yield. The most damaging effect wind has on plants is breakage, or greensnap. Wind can also cause abrasions and tear leaves. Root or stem lodging can also occur due to high winds. A secondary physical effect wind plays in crop growth is increased transpiration and crop water use. However there are some positive effects of wind. The shaking of crops caused by wind can increase the plants mechanical strength and root to shoot ratio. These stronger plants have thicker and wider leaves, and may be less affected by moisture stress (Elmore, et al. 2005). Weather Dummy Variables A dummy variable for years with extreme drought was created to help capture the effects of drought on yield. The ?tree-ring? reconstruction of the Palmer Drought Severity Index (PDSI) was used to select drought years for Alabama. This reconstruction 40 was done for North America for all years before and including 2003 and can be found on the National Climatic Data Center (NCDC) website (http://www.ncdc.noaa.gov/cgi- bin/paleo/pd04plot.pl). Each composite image of the summer months June-August was examined for 1986-2003. The years 1986, 1988, 1998, 2000, and 2002 were selected as drought years. These years all had negative PDSI values which indicate dry conditions. For years 2004 and 2005 the USDA U.S. Drought Monitor Archives were used (http://drought.unl.edu/dm/archive.html). Drought conditions were not found in either year. Also a dummy variable was created to capture hurricane damage to crops in southern Alabama. Because hurricanes make landfall on the Gulf Coast and then lose strength as they move inland only southern counties were selected to represent damaging hurricane winds and rains. Counties selected included Autauga County and Elmore County as well as all counties located in the Gulf, Costal Plain, and Prairie climate divisions. The years 1995, 2004, and 2005 were selected to illustrate damages done by Hurricane Opal, Ivan, and Katrina. Hurricane Opal made landfall on October 4 th 1995. In 1995 the harvesting of corn began in late July, by October there was little corn for Opal to damage. Cotton, Peanuts, and Soybeans harvest began in September, and when Opal arrived in Alabama they were all vulnerable to her heavy rains and high winds. Cotton was severly damaged (Hudson, 1995). Hurricane Ivan arrived in Alabama on September 16 th , 2004. Ivan?s effect on crop production was a critical concern of the USDA?s State Statistical Office. They increased the monthly crop report surveys in September, November, and again in December to asses the impacts of Ivan. The crop production effects for Alabama were similar to that of Opal. Cotton suffered the most of all 41 Alabama?s row crops. Un-harvested fields of corn also suffered minimal to severe damage. Peanut and soybean fields had little to no damage (Vanderberry, 2004). The very next year on August 29 th , 2005 Hurricane Katrina hit Alabama?s crop production hard. Katrina?s extreme winds and heavy rains damaged cotton and corn fields similarly to Ivan (Schnepf and Chite, 2005). A dummy variable for Opal, Ivan, and Katrina was created for Corn and Cotton. Policy Dummy Variables It is important when looking at crop yield to take into account government legislation affecting agricultural programs. These programs can make major changes to agriculture policy, and in return change the planting and land use decisions of agricultural producers. This study looks at the Federal Agriculture Improvement and Reform Act of 1996, and the Farm Security and Rural Investment Act of 2002. By creating a separate dummy variable for each policy this study hopes to capture any significant effects either policy may have had on Alabama crop production during the years of 1996-2005. The Federal Agriculture Improvement and Reform Act of 1996 made significant changes in U.S. agricultural policies. Signed into law in April of 1996, the Act was designed to guide agricultural programs from 1996-2002. Title I of the 1996 Act, known as the Agricultural Market Transition Act (AMTA), provided the most significant change to long-standing U.S. agricultural policies. Farmers who participated in any of the 1991- 1995 programs for wheat, feed grains, cotton, or rice were given a series of predetermined annual contract payments as long as they adhered to a Production Flexibility Contract (PFC); drastically changing the approach used for making direct 42 payments to farmers. The 1996 Act also eliminated target prices, deficiency payments, and production adjustment programs. Most importantly the AMTA lifted restrictions on the use of cropland enrolled in commodity programs allowing for more flexibility in farmers planting decisions. However some programs involving fruits and vegetables were excluded (Nelson, 1996). The AMTA continued to base nonrecourse commodity loan rates on a moving average of recent past market prices, setting maximum commodity loan rates equal to 1995 levels. Marking loan provisions for wheat, feed grains, oilseeds, rice, and upland cotton were made available to producers if market prices fell below commodity loan rates. The 1996 Act also modified the price support program for quota peanuts. It held the nonrecourse loan rates for quota peanuts constant for 1996 through 2002 at $610 per ton, and the cost-of-production estimates were no longer used as the basis for support. The loan rate for additional peanuts ensured the Commodity Credit Corporation (CCC) no losses from the sale or disposal of additional peanuts (Nelson, 1996). By 2002 U.S. agricultural producers witnessed depressed prices for all major agricultural commodities and frail outlooks for short-term price recovery. Record government payment in recent years also contributed to the passage of the Farm Security and Rural Investment Act of 2002. The Farm Act of 2002 had a six year lifespan from 2002-2007 and was similar to the Farm Act of 1996. The market loan program continued with slightly higher rates than the Farm Act of 1996, and provisions for fixed annual payments were continued at similar rates. It also retained planting flexibility by allowing for updating base acreages. One new program introduced by the Farm Act 2002 was the counter-cyclical program (CCP) which established direct payment to producers for 43 supplemental and disaster payments (Tiller et al., 2009). The Farm Act of 2002 also ended the marketing quota program for peanuts. Peanut quota owners were given a buyout payment and made peanut producers eligible for crop commodity programs (Dohlman et al., 2004). Models One of the main objectives of this study is to expand independent weather variables used by previous studies to improve regression results explaining crop yield response to moisture stress. In order to test this hypothesis both an unrestricted model and a restricted model were tested. Included in the unrestricted model were additional independent weather variables, weather dummy variables, and policy dummy variables. This analysis used the framework from Isik and Devadoss (2006), Schlenker et al. (2006), and Peiris and McNicol (1996) to form five unrestricted models to illustrate crop yield response to drought in Alabama. How each previous study contributed to the formation of the models is explained below, and a review of each of these studies can be found in the previous chapter. Isik and Devadoss (2006) performed both a linear and quadratic regression to find the impact of precipitation and temperature on average crop yields. Their use of crop yields as the dependent variable was applied for this study. This study used a semi- logarithmic model, and the natural log of crop yield per acre was used as the dependent variable. Schlenker et al. (2006) examined the impact of global warming on U.S. Agriculture by regressing county level data for U.S. counties located east of the 100 th Meridian. They used growing degree days (GDD), and GDD squared, square root of 44 stress degree days, average precipitation, and average precipitation squared as the weather variables in their model. The five independent weather variables (Schlenker et al., 2006) were used as a guideline for this study, and are nested in the five crop models. Also used as a guideline for this study was from Schlenker et al. was the inclusion of soil property K factor. Peiris and McNicol (1996) also expanded weather variables further to include solar radiation, wind speed, and relative humidity in their model. These additional weather variables can also be found in the five crop models: equation (4.5) is the corn model, equation (4.6) is the cotton model, equation (4.7) is the hay model, equation (4.8) is the peanut model, and equation (4.9) is the soybean model. The restricted models for corn, cotton, hay, peanuts, and soybeans including only the five independent weather variables presented by Schlenker et al. are given by equations (4.10) through (4.14) respectively. Variables for each model and their definitions are listed in Table 4.4. Table 4.4 Indepentent Variables and Definitions Variable Definitions Y Crop Yield Per Acre GDDcorn Corn Growing Degree Days ( 50-86 ?F) in Thousands GDDcorn? Corn GDD ( 50-86 ?F) Squared GDDcotton Cotton Growing Degree Days ( 60-86 ?F) in Thousands GDDcotton? Cotton GDD ( 60-86 ?F) Squared GDDhay Hay Growing Degree Days ( 40-86 ?F) in Thousands GDDhay? Hay GDD ( 40-86 ?F) Squared GDDpeanut Peanut Growing Degree Days ( 56-86 ?F) in Thousands GDDpeanut? Peanut GDD ( 56-86 ?F) Squared GDDsoy Soybean Growing Degree Days ( 50-86 ?F) in Thousands GDDsoy? Soybean GDD ( 50-86 ?F) Squared SqrtTSDD Square Root of Temperature Stress Degree Days Precip Average Precipitation in mm Precip? Average Precipitation Squared CVprecip Coefficient of Variation for Precipitation Rad Average Solar Radiation Humidity Average Relative Humidity Wind Average Wind Speed K*precip K Factor and Average Precipitation Interaction Drought Drought Dummy for Years 86, 88, 98, 00, & 02 Hurricane Hurrican Dummy for Years 95, 04, & 06 Time Time Intercept Policy96 Farm Act 1996 Dummy Policy02 Farm Act 2002 Dummy (4.5) lny = a + b Time + c GDDcorn + d GDDcorn? + e SqrtTSDD + f it Precip + it it it it it it g it Precip ?+ h CVprecip + i it Rad + j Wind + k Humidity + l K*precip + it it it it m Drought + n Hurricane + o Policy96 + p Policy02 + it it it it it ? (4.6) lny = a + b Time + c GDDcotton + d GDDcotton? + e it SqrtTSDD + f Precip + g Precip ?+ h CVprecip + i Rad + j Wind + k it Humidity + l K*precip + m Drought + n Hurricane + o it Policy96 + p Policy02 + it it it it it it it it it it it it it it it ? 45 (4.7) lny = a + b Time + c GDDhay + d GDDhay? + e SqrtTSDD + f it Precip + it it it it it it g it Precip ?+ h CVprecip + i it Rad + j Wind + k Humidity + l K*precip + it it it it m Drought + n Policy96 + o Policy02 + it it it it ? (4.8) lny = a + b Time + c GDDpeanut + d GDDpeanut? + e it SqrtTSDD + f Precip + g Precip ?+ h CVprecip + i Rad + j Wind + k it Humidity + l K*precip + m Drought + n Policy96 + o Policy02 + it it it it it it it it it it it it it it it ? (4.9) lny = a + b Time + c GDDsoy + d GDDsoy? + e SqrtTSDD + f Precip + it it it it it it it g it Precip ?+ h CVprecip + i it Rad + j Wind + k Humidity + l K*precip + it it it it m Drought + n Policy96 + o Policy02 + it it it it ? (4.10) lny = a + b Time + c GDDcorn + d GDDcorn? + e SqrtTSDD + it it it it it it f Precip + g Precip? + it it it ? (4.11) lny = a + b Time + c GDDcotton + d GDDcotton? + e it SqrtTSDD + it it it it it f Precip + g Precip? + it it it ? (4.12) lny = a + b Time + c it GDDhay + d GDDhay? + e SqrtTSDD + it it it it it f Precip + g Precip? + it it it ? 46 47 it it it it it it it it it (4.13) lny = a + b Time + c GDDpeanut + d GDDpeanut? + e SqrtTSDD + f Precip + g Precip? + ? (4.14) lny = a + b Time + c GDDsoy + d GDDsoy? + e SqrtTSDD + it it it it it it f Precip + g Precip? + it it it ? 48 V. RESULTS Average crop yields for each commodity from 1986-2005 were collected from the United States Department of Agriculture?s NASS website. We then excluded the counties that had more than three years of missing yield data for each commodity. The yield and weather data was combined for each county that remained. The multiple imputation (MI) procedure was chosen to estimate the incomplete cases. In SAS a substantial number of statistical analysis procedures exclude observations with any variables that have missing values (SAS, 2000). For Proc Panel to work correctly a complete and balanced panel data set is required. The Proc MI procedure draws a random sample of the missing values from its distribution, and results in a valid statistical inference that reflects the uncertainty due to the missing values. First the missing data are filled in m times to generate m complete data sets. Then the m complete data sets are analyzed using standard statistical analyses. Finally the results from the m complete data sets are combined to produce the results. MI does not attempt to estimate the missing values through simulation (SAS, 2000). For data sets where there are incomplete observations the multiple imputation is a well-established technique for estimation (Carpenter et al., 2006). Multiple imputation reduces the bias that may occur in studies that delete incomplete data, by allowing researchers to complete the missing values. In recent years there has been increased literature on multiple imputation and how researchers should deal with missing data (Penn, 2007). These studies include Schafer, 1997; Vriens and Melton, 2002; Schafer and 49 Graham, 2002; Raghunathan, 2004; Little and Rubin, 2002; Carpenter et al, 2005; and Penn, 2007. Once each of the commodity data were complete, models for each crop and their weather variables were selected. This study used panel data to examine group effects and time effects by using the fixed effects model. Each model was then run in SAS 9.1 using Proc Panel. The models were first run using a random effects model. The Hausman m statistic for all models was statistically significant at the 5% or 1% level indicating that random effects were not present. In order to include dummy variables the one way model was selected and the intercepts across counties in Alabama were examined. In order to capture time effects, an annual time variable was created within the model. Group differences in the intercepts are examined by the one-way fixed least squares dummy variable model (LSDV). The fixed effect model creates as many dummy variables as the number of cross sections, and then drops the last dummy for a reference point. LSDV can be problematic when there are a large number of cross sections in the panel data set. Another option for fixed effects models with large number or cross sections is the within effects model, which does not use dummy variables and provides identical parameter estimates of the LSDV model. However the within effects model no intercept, has small MSE and incorrect standard errors and R2 values (Park, 2008). The PANEL procedure in SAS was chosen because it uses the LSDV model and allows users to fit the within effects model and still reports correct MSE, SEE, R2, and standard errors for the LSDV1 model without the creation of dummy variables for each cross section. Also, this procedure uses the F test for the fixed group effects. In order to accomplish this, the data must be sorted by variables to appear in the ID statement of the PANEL procedure (Park, 2008). For this study the data were sorted by county and the intercept for each model is the last county with crop yield for that commodity. The intercept for the corn model is Wilcox County, the cotton model is Tuscaloosa County, the hay model is Winston County, the peanut model is Russell County, and the soybean model is Talladega County. An F test was then used to test the significance of the additional variance explained by the unrestricted model. The restricted model (Model 1) has independent variables with variance explained by . The unrestricted model (Model 2) has + independent variables with variance explained by . The F test statistic shown in equation (5.1) can be compared to the critical values of the F distribution with and n ? ( + ) ? 1 degrees of freedom, where n is the number of observations: 1 k 2 1 R 1 k 2 k 2 2 R 2 k 1 k 2 k (5.1) )1)(/()1( /)( 21 2 2 2 2 1 2 2 ?+?? ? = kknR kRR F The null hypothesis for this test statistic is that all the coefficients for the additional independent variables included in the unrestricted model are zero. The alternative hypothesis is that at least one of the additional independent variables is non-zero. For this study the hypotheses for the corn and cotton models are as follows: H : h CVprecip = i Rad = j Wind = k it Humidity = l K*precip =m it Drought = n it Hurricane = o Policy96 = p Policy02 = 0 O it it it it it it H : h CVprecip ? 0 or i Rad ? 0 or j Wind ? 0 or k Humidity ? 0 or l K*precip ? 0 or m Drought ? 0 or n it Hurricane ? 0 or o Policy96 ? 0 or p Policy02 ? 0 A it it it it it it it it 50 The hypotheses for the hay, peanut, and soybean models are as follows: H : h CVprecip = i Rad = j Wind = k it Humidity = l K*precip = m it Drought = n it Policy96 = o Policy02 = 0 O it it it it it H : h CVprecip ? 0 or i Rad ? 0 or j Wind ? 0 or k Humidity ? 0 or l K*precip ? 0 or m Drought ? 0 or n it Policy96 ? 0 or o Policy02 ? 0 A it it it it it it it The results generated by the previously described models, for both the restricted model and the unrestricted model are reported below. The results for the F test are also evaluated. Corn Model Results Restricted Model Unrestricted Model Variable Coeff. Estimates (Std. Err.) Coeff. Estimates (Std. Err.) Intercept 0.855 (2.264) 4.343 (2.184) ** Time 0.023 (0.002) *** 0.001 (0.003) GDD ( 50-86 ?F) 1.528 (1.072) 0.952 (0.677) GDD? ( 50-86 ?F) -0.143 (0.127) -0.073 (0.081) SqrtTSDD -0.061 (0.004) *** -0.041 (0.005) *** Precip 0.089 (0.045) * 0.159 (0.056) *** Precip? -0.007 (0.005) -0.010 (0.004) ** CVprecip -0.072 (0.023) *** Rad -0.026 (0.030) Humidity -0.023 (0.010) ** Wind 0.013 (0.145) K*precip -0.255 (0.122) ** Drought -0.371 (0.038) *** Hurricane 0.057 (0.020) *** Policy96 0.163 (0.028) *** Policy02 0.325 (0.032) *** Observations 960 960 R? 0.551 0.673 The astrisks indicate a 1%, 5%, and 10% significance different from zero by ***, **, *, respectively. Table 5.1 Results for Corn Model In the two models, = 9 and n ? ( + ) ? 1= 896. Using equation (5.1) and the R? values from Table (5.1) the empirical F value was calculated to be F = 37.009. For the 2 k 1 k 2 k 51 52 corn model the empirical F value exceeds the critical value of F = 2.339, assigned at the 1% level. We reject the null hypothesis, and find the unrestricted model for corn is a better fit than the restricted model. The Time variable was found to be positive and significant at the 1% level in the restricted model, but in the unrestricted corn model the Time variable is no longer significant. This could indicated that increasing yield trends from 1986-2005 are captured by other variables not included in the restricted model such as government policy. The SqrtTSDD is negative and significant at the 1% level indicating that an increased stress degree day during the growing season reduces corn yields. Precipitation is found to be positive and significant at the 1% level and Precip? is negative and significant at the 5% level. More informing however is that the variable CVprecip is negative and significant at the 1% level. As the variation of precipitation increases moisture stress will increase, causing yield to reduce. Also humidity is found to be negative and significant at the 5% level. Because humidity slows down evapotranspiration this result seems odd. However humidity can be related to cloud cover, and reduced sunlight may have a negative effect on yield. The K*precip variable is negative and significant at the 5% level. This shows that the interaction of soil erodibility and precipitation reduces crop yield, most likely due to higher runoff and nutrient loss. During drought years corn yield reduction is significant at the 1% level. Another interesting result to notice is the hurricane dummy. As explained in the previous chapter some corn damage was seen after all three hurricanes. However this model found that the hurricane dummy was positive and significant at the 1% level. This positive effect of hurricanes could be attributed to the increased precipitation late in the season. Also in the unrestricted model both the policy dummy variables, Policy96 and Policy02 are positive and significant at the 1% level. The 1996 Farm Act lead to a large acreage shift to corn as a result of increased planting flexibility and high prices that favored planting corn. Cotton Model Results Restricted Model Unrestricted Model Variable Coeff. Estimates (Std. Err.) Coeff. Estimates (Std. Err.) Intercept 1.718 (1.510) 13.359 (2.217) *** Time 0.007 (0.002) *** 0.000 (0.004) GDD ( 60-86 ?F) 3.669 (1.163) *** 3.877 (1.030) *** GDD? ( 60-86 ?F) -0.630 (0.225) *** -0.681 (0.193) *** SqrtTSDD -0.047 (0.004) *** -0.037 (0.008) *** Precip 0.098 (0.049) ** -0.031 (0.081) Precip? -0.008 (0.005) 0.001 (0.006) CVprecip -0.098 (0.026) *** Rad -0.226 (0.042) *** Humidity -0.082 (0.018) *** Wind -0.482 (0.142) *** K*precip 0.171 (0.113) Drought -0.077 (0.028) *** Hurricane -0.129 (0.049) *** Policy96 0.017 (0.043) Policy02 0.131 (0.058) ** Observations 620 620 R? 0.434 0.507 The astrisks indicate a 1%, 5%, and 10% significance different from zero by ***, **, *, respectively. Table 5.2 Results for Cotton Model In the two models, = 9 and n ? ( + ) ? 1= 573. Using equation (5.1) and the R? values from Table (5.2) the empirical F value was calculated to be F = 9.470. For the cotton model the empirical F value exceeds the critical value of F = 2.339, assigned at the 1% level. We reject the null hypothesis, and find the unrestricted model for cotton is a better fit than the restricted model. 2 k 1 k 2 k 53 54 The Time variable in the unrestricted cotton model saw similar results as the corn model. Also the SqrtTSDD was negative and significant at the 1% level. In contrast with the corn model, the cotton model GDD was positive and significant at the 1% level GDD? was negative and significant at the 1% level. The significance of all three temperature weather variables in the cotton model suggests that cotton yield is more sensitive to temperature than precipitation. CVprecip was negative and significant at the 1% level. Also in the cotton model it is interesting to note that all three of the evapotranpiration variables radiation, humidity, and wind speed were negative and significant at the 1% level. As expected the drought and hurricane dummy variables were negative and significant at the 1% level. However the Policy variables are not as easy to interpret with cotton. Because the Farm Act of 1996 directly changed policy on cotton you would expect to see significance there. The Farm Act of 2002 was passed at a critical time for U.S. cotton producers. Cotton prices were 60% lower than they had been prior to the Farm Act of 1996. Due to low cotton prices, total government payment for assistance to cotton producers exceeded $15.5 billion during 1998-2003 (Tiller and Brown). Because the Farm Act of 2002 increased the financial situations of cotton growers in Alabama as well as ended the quota peanut program it is not shocking that the Policy02 variable would be positive and significant at the 5% level. If the Policy02 dummy is removed from the cotton model the R? increases to 0.522, shown in Table (5.3). Table 5.3 Second Unrestricted Cotton Model Variable Coeff. Estimates (Std. Err.) Intercept 13.863 (2.086)*** GDD ( 60-86 ?F) 3.433 (1.008)*** GDD? ( 60-86 ?F) -0.594 (0.188)*** SqrtTSDD -0.033 (0.008)*** Precip 0.004 (0.074) Precip? -0.004 (0.005) CVprecip -0.080 (0.023)*** Rad -0.244 (0.043)*** Humidity -0.085 (0.017)*** Wind -0.314 (0.137)** K*precip 0.167 (0.113) Drought -0.071 (0.031)** Hurricane -0.243 (0.044)*** Time 0.017 (0.003)*** Policy96 -0.130 (0.035)*** Observations 620 R? 0.522 The astrisks indicate a 1%, 5%, and 10% significance different from zero by ***, **, *, respectively. This second unrestricted cotton model is still a better fit than the restricted model with an F value calculated to be F =13.78. Also all variables other than Policy96 have the same effects at the same significant level as the unrestricted model in Table (5.2). This suggests that the second model is a more complete representation of cotton yields during 1986-2005. The cotton model that included Policy 02 was overestimating the effects of that farm act and failed to capture the effects of Policy96. In the second cotton model the Farm Act 1996 dummy variable is negative and significant at the 1% level. After the Farm Act of 1996, Alabama saw a reduced cotton acreage reduction due to more attractive corn and soybean prices in comparison with cotton prices. These lower cotton prices in combination with lifted acreage restrictions may have lead to cotton being planted on less fertile land resulting in lower yields (USDA 1997). This study finds that 55 cotton yield per acre was reduced due to government policy, and our results are similar to the findings of Luttrell and Gilbert (2001). Hay Model Results Restricted Model Unrestricted Model Variable Coeff. Estimates (Std. Err.) Coeff. Estimates (Std. Err.) Intercept -4.816 (2.152) ** -0.249 (1.985) Time 0.012 (0.001) *** 0.009 (0.002) *** GDD ( 40-86 ?F) 1.649 (0.724) ** 1.271 (0.566) ** GDD? ( 40-86 ?F) -0.128 (0.061) ** -0.096 (0.049) ** SqrtTSDD -0.023 (0.002) *** -0.017 (0.002) *** Precip 0.216 (0.026) *** 0.311 (0.040) *** Precip? -0.021 (0.003) *** -0.024 (0.004) *** CVprecip -0.027 (0.013) ** Rad -0.062 (0.021) *** Humidity -0.028 (0.007) *** Wind -0.100 (0.102) K*precip -0.293 (0.124) ** Drought -0.088 (0.018) *** Policy96 0.003 (0.016) Policy02 0.040 (0.023) ** Observations 1340 1340 R? 0.486 0.521 The astrisks indicate a 1%, 5%, and 10% significance different from zero by ***, **, *, respectively. Table 5.4 Results for Hay Model In the two models, = 8 and n ? ( + ) ? 1= 1258. Using equation (5.1) and the R? values from Table (5.4) the empirical F value was calculated to be F = 11.417. For the hay model the empirical F value exceeds the critical value of F = 2.339, assigned at the 1% level. We reject the null hypothesis, and find the unrestricted model for hay is a better fit than the restricted model. 2 k 1 k 2 k Weather variables are unique to hay production because hay is cut several times during the growing season. In the unrestricted model we see that all variables except for 56 the Intercept, Wind, and Policy96 are statistically significant. GDD is positive and significant at the 5% level, Precip and Policy02 are positive and significant at the 1% level. All three of the evapotraspitation variables are negative and significant at the 1% level. It is important to notice the CVprecip, Precip?, and K*precip are all negative and significant at the 1% level. Similar to other crops as CVprecip and K*precip increases yield will decrease. Interesting to note too much rain can also play a large role when discussing hay production. If it rains or is too humid after hay is cut, the hay cannot dry enough to be bailed. Hay left cut for too long can rot which would reduce yield per acre. Peanut Model Results Restricted Model Unrestricted Model Variable Coeff. Estimates (Std. Err.) Coeff. Estimates (Std. Err.) Intercept 3.552 (3.397) 8.258 (2.618) *** Time 0.003 (0.002) -0.002 (0.004) GDD ( 56-86 ?F) 1.944 (2.052) 2.590 (1.595) GDD? ( 56-86 ?F) -0.236 (0.310) -0.339 (0.233) SqrtTSDD -0.018 (0.004) *** -0.019 (0.006) *** Precip 0.236 (0.056) *** 0.196 (0.109) * Precip? -0.022 (0.006) *** -0.022 (0.010) ** CVprecip -0.004 (0.017) Rad -0.087 (0.024) *** Humidity -0.044 (0.009) *** Wind -0.249 (0.169) K*precip 0.135 (0.354) Drought -0.021 (0.022) Policy96 -0.003 (0.033) Policy02 0.121 (0.047) *** Observations 240 240 R? 0.462 0.494 The astrisks indicate a 1%, 5%, and 10% significance different from zero by ***, **, *, respectively. Table 5.5 Results for Peanut Model 57 In the two models, = 8 and n ? ( + ) ? 1= 213. Using equation (5.1) and the R? values from Table (5.5) the empirical F value was calculated to be F = 1.701. For the peanut model the empirical F value exceeds the critical value of F = 1.612, assigned at the 10% level. We reject the null hypothesis, and find the unrestricted model for peanut is a better fit than the restricted model. 2 k 1 k 2 k Peanuts go through a rapid vegetative growth stage where adequate moisture is available. Precipitation is very important early in the growing season due to this vegetative growth. In the unrestricted model we see Precip is positive and significant at the 10% level. The Precip? is negative and significant at the 5% level. This may be because of heave rains over saturating peanuts in the ground. Also hurricane damage could play a role in the estimation of negative affects attributed to precipitation. Temperatures in the GDD range of 56-86 ?F shouldn?t affect peanut yield because peanut flowers open and night and are self-fertilized before high afternoon temperatures can damage the flowers. However the SqrtTSDD is negative and significant at the 1% level due to vegetative stress that would occur. Rad and Humidity are negative and significant at the 1% level. Wind also doesn?t affect peanut pollination because that is done internally inside the flower and flower stalk (Extension 2009). The Policy02 was positive and significant at the 1% level. After the Farm Act of 2002 was passed peanut producers took advantage of planting flexibility by shifting peanut production to higher yielding areas. Also producers that may have not planted peanuts in previous years were now allowed to do so with the end of the peanut quota program. 58 Soybean Model Results Restricted Model Unrestricted Model Variable Coeff. Estimates (Std. Err.) Coeff. Estimates (Std. Err.) Intercept 5.628 (2.781) ** 13.232 (1.863) *** Time 0.013 (0.002) *** 0.016 (0.004) *** GDD ( 50-86 ?F) -1.342 (1.346) -1.651 (0.779) ** GDD? ( 50-86 ?F) 0.184 (0.164) 0.200 (0.094) ** SqrtTSDD -0.048 (0.005) *** -0.042 (0.005) *** Precip 0.230 (0.062) *** 0.219 (0.079) *** Precip? -0.021 (0.007) *** -0.026 (0.006) *** CVprecip 0.000 (0.035) Rad -0.147 (0.033) *** Humidity -0.047 (0.012) *** Wind -0.115 (0.151) K*precip 0.181 (0.134) Drought 0.117 (0.024) *** Policy96 -0.159 (0.033) *** Policy02 0.061 (0.045) Observations 440 440 R? 0.506 0.566 The astrisks indicate a 1%, 5%, and 10% significance different from zero by ***, **, *, respectively. Table 5.6 Results for Soybean Model In the two models, = 8 and n ? ( + ) ? 1= 403. Using equation (5.1) and the R? values from Table (5.6) the empirical F value was calculated to be F = 7.075. For the soybean model the empirical F value exceeds the critical value of F = 2.356, assigned at the 1% level. We reject the null hypothesis, and find the unrestricted model for soybean is a better fit than the restricted model. 2 k 1 k 2 k Soybeans are similar to peanuts in there growth and development. In this unrestricted model SqrtTSDD is negative and significant at the 1% level and Precip? is negative and significant at the 1% level. GDD is negative and significant at the 5% level and GDD? is positive and significant at the 5% level, indicating that soybeans are significantly affected by high temperatures. Also Precip is positive and significant at the 59 60 1% level. The same as the peanut model, the soybean model sees no significance with CVprecip , because both crops are considered somewhat drought tolerant and their roots extend deep into the soil. The drought dummy is positive and significant at the 1% level. Radiation and Humidity are both negative and significant at the 1% level as well. The Policy96 dummy is negative and significant at the 1% level. This decrease in yield per is similar to the effect of the Farm Act of 1996 on Cotton. 61 VI. CONCLUSION This study identified a framework for analyzing crop yield response of major crops to climate fluctuations in Alabama. This analytical framework expanded the weather variables examined by previous studies to improve regression results for crop yield response to moisture stress. In order to test this hypothesis both an unrestricted model and a restricted model for corn, cotton, hay, peanuts, and soybeans were tested. Included in the unrestricted model were additional independent weather variables, weather dummy variables, and policy dummy variables. An F test was then used to test the significance of the additional variance explained by the unrestricted model. The unrestricted models for corn, cotton, hay, and soybeans were found to be a better fit than the restricted model, with empirical F values exceeding the critical F value, at the 1% level. The unrestricted peanut model was found to be a better fit that the restricted model, at the 10% level. Because each drought year is unique in its climatic characteristics and impacts drought can devastate even very productive lands and can go undetected. To estimate the impacts of weather variables on crop yields this study econometrically use the additional weather variables CVprecip, Radiation, Humidity, Wind, K*precip, Hurricane, Drought, Policy96, and Policy02. All were found to be significant in at least one or all the models. The onset and end of a drought are difficult to determine and its impacts can not immediately be observed by eye or even ground data. It is easy to determine drought if 62 plants are showing visible signs of distress. What is difficult to determining is how unobservable drought damage could be affecting crop yields. The crop yield study for corn, cotton, hay, peanuts, and soybeans for the years 1986 through 2005 found the impacts of unobservable weather variables (i.e. evapotranspiration and erosion) to significantly reduce crop yields. Although drought is frequent in the United States there is no national policy to mitigate the impacts. The U.S. Department of Agriculture explains that the National Drought Policy Act was passed through Congress and thus creating the National Drought Policy Commission (NDPC) to advise on developing a comprehensive national policy to mitigate the impacts of drought to improve public awareness, and to achieve federal/nonfederal partnerships for better coordination and response to drought (Motha 2001). Motha explained that in the survey conducted for the final report, the NDPC found that 30 of the 50 states in the United States had drought plans, with most oriented toward relief rather than preparedness. This survey also revealed that in most states, drought responsibilities are normally located in the agencies that are responsible for the functions of agriculture, natural resources, water management, environment, or emergency management. Fewer than five states have independent, designated drought coordinators, while more than 20 states have drought task forces (Motha 2001). Government legislation affecting agricultural programs has been passed in recent years to increase planting variability and establish more secure financial situations for agricultural producers. These agricultural programs give producers more control over production decisions based on their available land, inputs, and market prices. This study looked at the Federal Agriculture Improvement and Reform Act of 1996, and the Farm 63 Security and Rural Investment Act of 2002. Both programs were designed to allow producers more flexibility with their land. Additional land was brought into production that was previously idled due to restriction programs. Producers could now respond to signals from the market, which resulted in an economically more efficient agricultural production. This study found both Policy dummy variables to have significance in both the addition and reduction of yield per acre for all major commodities in Alabama: illustrating the important role government policy plays on agriculture. Both the Federal Agriculture Improvement and Reform Act of 1996, and the Farm Security and Rural Investment Act of 2002 brought idle agricultural lands into production and allowed producers more flexibility in their farm management decisions. However flexible planting and price support programs don?t protect producers from reduced yields caused by drought. Government incentives for better management practices and government subsidies for drought tolerant hybrid seed could reduce the impact of drought. Because drought impacts are not structural or localized, government development of drought contingency plans are hindered by many complications. Accurate, reliable, and timely estimates of the impacts of drought are becoming increasingly important as world demand for food and feed increases. The impact of drought in the southeastern United States is likely to increase in magnitude as fresh water becomes increasingly scarce and as populations grow. Reliable and up to date information related to surface moisture conditions could be used to forecast crop yields, assess distressed areas for allocation of disaster relief funds, and could help resource managers and government officials plan ahead for difficult financial times. In the state of Alabama drought has cause serious problems for farming 64 communities for the past several years and most recently in the summer of 2006. These conditions place farming communities in difficult financial situations which also substantially affects other economic sectors within the state. BIBLIOGRAPHY Adams, Richard, Bruce McCarl, Daniel Dudek, and David Glyer. ?Implications of Global Climate Change for Western Agriculture.? Western Journal of Agricultural Economics 13, no.2 (1988): 348-56. Adams, Richard. ?Global Climate Change and Agriculture: An Economic Perspective.? American Journal of Agricultural Economics 71, no.5 (1989):1272-79. Adams, Richard, Cynthia Rosenzweig, Robert Pearl, Joe Ritchie, Bruce McCarl, David Glyer, Bruce Curry, James Jones, Kenneth Boote, and Hartwell Allen. ?Global Climate Change and U.S. Agriculture.? Nature 345, no.6272 (1990): 219-24. Boyd, J. Southwest Farmers Battle Record Drought. United Press International, 1995. Callaway, John, F. Cronin, J. Currie, and J. Tawil. ?An Analysis of Methods and Models for Assessing the Direct and Indirect Economic Impacts of CO -Induced Environmental Changes in the Agricultural Sector of the U.S. Economy.? Pacific Northwest Laboratory (Richland, WA) Working Paper No. PNL-4384, 1982. 2 Carpenter, James, and Michael G. Kenward. ?A Comparison of Multiple Imputation and Doubly Robust Estimation for Analyses with Missing Data.? J.R. Statistic Society 169, no.3 (2006): 571-584. Center for Disease Control (CDC). ?Notice to Readers International Decade for Natural Disaster Reduction.? Morbidity and Mortality Weekly Report 43, no.17 (1994), http://www.cdc.gov/mmwr/preview/mmwrhtml/00030685.htm (accessed June 1, 2009). Changnon, S.A. ?Impacts of 1997-98 El Nino-generated Weather in the United States.? Bulletin of the American Meteorological Society 80 (1999):1,819-1,827. Cline, William R. ?The Impact of Global Warming on Agriculture: Comment.? The American Economic Review86, no.5 (1996): 1309-1311. Collins, Matt. ?Hay Days In Alabama.? Alabama Farmers Federation Publication, 2005. http://www.alfafarmers.org/neighbors/neighborsStory.phtml?id=4309 (accessed February 26, 2009). 65 66 Danneberger, T.K., and J.R. Street. ?Climatic Adaptability of Annual Bluegrass in Ohio Using Growing Degree-Days.? Ohio Journal of Science 85, no.3 (1985): 108-111. Decker, W., V. Jones, and R. Achutuni. The Impact of Climate Change from Increased Atmospheric Carbon Dioxide on American Agriculture, DOE/NBB-0077. Washington, D.C: U.S. Department of Energy, 1986. Dixon, Rob. ?Peanut Production in Alabama.? Encyclopedia of Alabama, 2009. http://www.encyclopediaofalabama.org/face/article.jsp?id=h-2016 (accessed February 26, 2009). Dohlman, Erik, Edwin Young, Linwood Hoffman, and William McBride. ?U.S. Peanut Sector Adapts to Major Policy Changes.? U.S. Department of Agriculture (USDA), Economic Research Service, 2004. http://www.ers.usda.gov/amerwaves (accessed July 20, 2009). Easterling, William E., Pierre R. Crosson, Norman J. Rosenberg, Mary S. McKenney, Laura A. Katz, and Kathleen M. Lemon. ?Agricultural Impacts of and Response to Climate Change in the Missouri-Iowa-Nebraska-Kansas (MINK) Region.? Climate Change 24 (1993): 23-61. Elmore, Roger W., David B. Mars, Ralph G. Klein, and Lori J. Abendroth. ?Wind Effect on Corn Leaf Azimuth.? Crop Science 45 (2005): 2598-2604. Federal Emergency Management Agency (FEMA). National Mitigation Strategy. Washington, D.C.: FEMA, 1995. Food and Agriculture Organization (FAO). Crop Production Statistics, 2000. http://www.fao.org. Fraisse, Clyde W., John Bellow, and Charles Brown. ?Degree Days: Heating, Cooling, and Growing.? University of Flordia IFAS Extension, ABE 381, 2007. Fraisse, Clyde W., Norman E. Breuer, David Zierden, and Keith T. Ingram. ?From Climate Variability to Climate Change: Challenges and Opportunities to Extension.? Journal of Extension 47 (2009). Fulmer, J.L., and L. ?Relationship of the Cycle in Yields of Cotton and Apples to Solar and Sky Radiation.? Quarterly Journal of Economics 56 (1972):305-405. Hudson, Mark E. ?Crops Review.? Alabama Agricultural Statistics Service, (1995). http://www.aces.edu/department/nass/bulletin/1995/pg10.htm (accessed June 20, 2009). 67 Institue of Water Research, Michigan State University (IWR-MSU). ?RUSLE On Line Soil Erosion Assessment Tool.? USDA-NRCS State Office of Michigan, 2002. http://www.iwr.msu.edu/rusle/ (accessed July 8, 2009). Isik, Murat, and Stephen Devadoss.?An Analysis of the Impact of Climate Change on Crop Yield and Yield Variability.? Applied Economics 38 (2006): 835-844. Jensen, N.E., and L. Pedersen. ?Spatial Variability of Rainfall: Variation within a Single Radar Pixel.? Atmospheric Research 77 (2005): 269-277. Jevons, H. Stanley. ?The Causes of Unemployment.? Contemporary Review 96 (1909): 501-22. Jevons, W. Stanley. Investigations in Currency and Finance, edited by H.S. Foxwell. London: Macmillan and Co., 1884. Kogan, F.N. ?Global Drought Detection and Impact Assessment from Space.? In Drought: A Global Assessment, edited by D.A. Wilhite, 196-210. New York: Routledge Publishers, 2000. Kogan, F.N. ?Global Drought Watch from Space.? Bulletin of the American Meteorological Societ. 78 (1997): 621-636. Lin, Ying-shiang, R.J. Hildreth, and K.R. Tefertiller. ?Non-Parametric Statistical Tests for Bunchiness of Dryland Crop Yields and Reinvestment Income.? Journal of Farm Economics 45 (1963): 592-98. Little, Roderick J.A., and Donald B. Rubin. Statistical Analysis with Missing Data, 2 nd ed. New York: John Wiley & Sons, Inc., 2002. Luttrell, Clifton B., and R. Alton Gilbert. ?Crop Yields: Random, Cyclical, or Bunchy?? American Journal of Agricultural Economics August (1976): 521-531. Mask, Paul L., and Charles C. Mitchell. ?Corn Production: Alabama Production Guide For Non-Irrigated Corn.? Alabama Cooperative Extension System (ACES), 2009. http://www.aces.edu/dept/grain/cornpro.php (accessed June 8, 2009). McWilliams, D.A., D.R. Berglund, and G.J. Endres. ?Corn Growth and Management Quick Guide.? North Dakota State University Agriculture and University Extension, 1999. http://www.ag.ndsu.edu/pubs/plantsci/rowcrops/a1173w.htm (accessed February 26, 2009). Mendelsohn, Robert, William D. Nordhaus, and Daigee Shaw. ?The Impact of Global Warming on Agriculture: A Ricardian Analysis.? The American Economic Review 84, no.4 (1994): 573-771. 68 Moore, Henry Ludwell. Generating Economic Cycles. New York: Macmillan Co., 1923. Motha, Ray. Recommendations on Drought Monitoring by the U.S. National Drought Policy Commission. U.S. Department of Agriculture, Washington, DC, 2001. National Climatic Data Center (NCDC). ?Tree-Ring Reconstructions of Palmer Drought Severity Index Across North America Over The Last 2000 Years? WDC for Paleoclimatology. http://www.ncdc.noaa.gov/cgi-bin/paleo/pd04plot.pl (accessed June 3, 2009). National Drought Mitigation Center (NDMC). ?What is Drought? Understanding and Defining Drought.? (2006). http://drought.unl.edu/whats/concept/ (accessed July 6, 2009). Nelson, Federick J., and Lyle P. Schertz, editors. Provisions of the Federal Agriculture Improvement Act of 1996. U.S. Department of Agriculture. Agriculture Information Bulletin, no. 729. Washington, DC, 1996. NeSmith, D. Scott. ?Summer Squash (Cucurbita pepo L.) Leaf Number as Influenced by Thermal Time.? Science Horticulture 68 (1997): 219-225. Obasi, G.O.P. ?WMO?s Role in the International Decade for Natural Disaster Reduction.? Bulletin of the American Meteorological Society 75 (1994):1,655- 1,661. Park, Hun Myoung, ?Linear Regression Models for Panel Data Using SAS, Stata, LIMDEP, and SPSS?. IndianaUniversity, 2008. http://www.indiana.edu/~statmath/stat/all/panel/panel.pdf (accessed January 3, 2009). Peiris, Ramanee, and James McNicol. ?Modeling Daily Weather With Multivariate Time Series.? Agricultural and Forest Meteorology 79 (1996): 219-231. Penn, David. ?Estimating Missing Values From the General Social Survey: An Application of Multiple Imputation.? Social Science Quarterly 88, no.2 (2007): 573-584. Raghunathan, Trivellore E. ?What Do We Do with Missing Data? Some Options for Analysis of Incomplete Data.? Annual Review of Public Health 25(2004): 99-117 Riebsame, W.E., S.A. Changnon, and T.R. Karl. Drought and Natural Resource Management in the United States: Impacts and Implications of the 1987-1989 Drought. Boulder, Colorado: Westview Press, 1990. 69 Rind, D., R. Goldberg, J. Hansen, Cynthia Rosenzweig, and R. Ruedy. ?Potential Evapotranspiration and the Likelihood of Future Drought.? Journal of Geophysical Research 95, no. D7 (1990): 9983-10004. Rosenzweig, Cynthia, M. Perry, K. Frohberg, and G. Fisher. Climate Change and World Food Supply. Oxford: Oxford University Press, 1993. Rosenzweig, Cynthia, and Martin L. Perry. ?Potential Impact of Climate Change on World Food Supply.? Nature 367, no. 6459 (1994): 133-38. Saha, A., A. Havenner, and H. Talpaz. ?Stochastic Production Functions Estimation: Small Sample Properties of ML versus FGLS.? Applied Economics 29 (1997): 459-69. SAS Institute Inc. Chapter 9 The MI Procedur. SAS OnlineDoc Version 8, 2000. http://support.sas.com/rnd/app/papers/miv802.pdf (accesed June 15, 2009). Schafer, Joseph L. Analysis of Incomplete Multivariate Data. New York: Chapman & Hall/CRC, 1997. Schafer, Joseph L., and John W. Graham. ?Missing Data : Our View of the State of the Art.? Psychological Methods 7(2002): 147-77. Schlenker, Wolfram, and Michael J. Roberts. ?Nonlinear Effects of Weather on Corn Yields.? Review of Agricultural Economics 28, no.3 (2006): 391-398. Paper Present at the Principal Paper session, ?Applications of Quasi-Experimental Methods in Agricultural Economics,? Allied Social Science Association annual meeting, Boston, January 6-8, 2006. Schlenker, Wolfram, W. Michael Hanemann, and Anthony C. Fisher. ?The Impact of Global Warming on U.S. Agriculture: An Econometric Analysis of Optimal Growing Conditions.? The Review of Economics and Statistics 88, no.1 (2006): 113-125. Schnepf, Randy, and Ralph M. Chite. ?U.S. Agriculture After Hurricane Katrina: Status and Issues.? Congressional Research Service, The Library of Congress (2005). http://digital.library.unt.edu/govdocs/crs/permalink/meta-crs-7808:1 (accessed June 20, 2009). Shaw, R.H., J.E. Newman, D.G. Baker, R.V. Crow, R.F. Dale, Rich Feltes, D.G. Hanway, P.R. Henderlong, and J.D. McQuigg. ?Weather Stress in the Corn Crop.? Oklahoma Cooperative Extension Service, CR-2099. http://www.pods.dasnr.okstate.edu/docushare/dsweb/Get/Document-2627/CR- 2099web.pdf (accessed July 5, 2009). 70 Tannehill, I.R. Drought: Its Causes and Effects. Princeton, NJ: Princeton University, 1947. Tiller, Kelly, J. Brown, J. Sartwelle, and J. Richardson. ?Impacts of the 2002 Farm Bill on Southeastern Representative Cotton Farms.? http://www.agpolicy.org/pubs/beltwidefb.pdf (accessed July 6, 2009). United States Deparment of Agriculture (USDA). ?Drought Monitor Archives.? U.S. Drought Monitor. http://drought.unl.edu/dm/archive.html (accessed June 20, 2009). U.S. Department of Agriculture (USDAa), Economic Research Service. ?Corn.? Briefing Rooms, 2008. http://www.ers.usda.gov/briefing/corn/ (accessed May 3, 2009). U.S. Department of Agriculture (USDAb), Economic Research Service. ?Cotton.? Briefing Rooms, 2008. http://www.ers.usda.gov/briefing/cotton/ (accessed May 3, 2009). U.S. Department of Agriculture (USDAc), Economic Research Service. ?Soybeans and Oil Crops.? Briefing Rooms, 2008. http://www.ers.usda.gov/briefing/soybeansoilcrops/ (accessed May 3, 2009). U.S. Department of Agriculture (USDA). ?USDA National Agricultural Statistics Service-Quick Stats U.S. & All States County Data.? http://www.nass.usda.gov/QuickStats/ (accessed February, 2009). United States Geological Service (USGS). ?The Water Cycle: Evapotranspiration.? Water Science for Schools, (2009). http://ga.water.usgs.gov/edu/watercycleevapotranspiration.html (accessed June 20, 2009). Vanderberry, Herb. ?October 1 Crop Production.? Alabama Farm Facts, (2004). http://www.aces.edu/department/nass/farmfact/ff0411.pdf (accessed June 20, 2009). Vriens, Marco, and Melton Eric. ?Managing Missing Data.? Marketing Research 14 (2002): 12-17. Walker, G.K., and J.L. Hatfield. ?Test of the Stress-Degree-Day Concept Using Multiple Planting Dates of Red Kidney Beans.? Agronomy Journal 71 (1979): 967-971. 71 Wilhite, Donald A., and Mark D. Svoboda. ?Drought Early Warning Systems in the Context of Drought Preparedness and Mitigation.? In Early Warning Systems for Drought Preparedness and Drought Management, edited by D.A. Wilhite, M.V.K. Sivakumar, and D.A. Wood, 1-16.Geneva, Switzerland: World Meteorological Organization, 2000. Proceedings of an Expert Group Meeting, Lisbon, Portugal, September 5-7, 2000. Wilhite, Donald A., and Margie Buchanan-Smith. ?Drought as Hazard: Understanding the Natural and Social Context.? In Drought and Water Crises: Science, Technology, and Management Issues, edited by D.A. Wilhite, 3-29. U.S.: CRC Press, 2005. 72 APPENDIX The Data Used in Thesis Alabama County Corn Yields 1986-2005 Year Autauga Baldwin Barbour Blount Butler Calhoun Cherokee 1986 38.5 76.5 38.2 66.7 73.3 59.4 47.7 1987 47.7 91.1 77.0 86.9 74.1 90.0 71.0 1988 20.0 60.9 29.4 54.3 45.0 54.5 32.7 1989 62.5 103.3 65.4 85.0 73.5 106.3 73.3 1990 51.3 68.7 51.0 65.3 63.9 76.9 42.9 1991 90.0 93.0 85.2 96.3 87.5 84.4 65.5 1992 71.7 100.8 72.1 123.9 83.8 118.4 101.6 1993 29.4 78.9 51.4 57.1 62.8 64.0 56.1 1994 80.0 112.2 80.0 95.8 85.4 118.6 113.6 1995 50.0 96.0 77.0 61.0 88.0 58.0 52.0 1996 38.0 102.0 54.0 94.0 68.0 91.0 80.0 1997 75.0 99.0 88.0 94.0 77.0 90.0 73.0 1998 33.0 68.0 70.0 83.0 31.0 84.0 65.0 1999 76.0 112.0 110.0 119.0 85.0 107.0 2000 31.0 58.0 53.0 17.0 47.0 2001 71.0 99.0 127.0 66.0 119.0 2002 52.0 94.0 88.0 94.0 40.0 96.0 78.0 2003 120.0 120.0 115.0 128.0 178.0 117.0 132.0 2004 83.0 120.0 93.0 139.0 144.0 143.0 2005 103.0 124.0 114.0 135.0 Year Coffee Colbert Conecuh Covington Crenshaw Cullman Dale 1986 42.7 59.7 62.1 45.0 50.0 62.5 33.8 1987 58.9 78.0 66.9 71.0 72.7 85.8 57.0 1988 42.9 55.0 35.3 22.1 46.9 41.2 25.5 1989 79.4 109.1 80.3 87.9 81.4 82.4 82.9 1990 34.2 51.5 64.4 47.8 62.1 55.8 80.5 1991 73.5 50.0 65.0 84.0 71.8 72.9 83.0 1992 93.3 103.3 97.6 83.3 85.3 116.4 71.3 1993 59.6 55.9 61.5 63.5 64.6 65.4 40.0 1994 87.9 121.6 88.7 89.8 90.2 112.6 60.3 1995 66.0 74.0 68.0 78.0 82.0 82.0 65.0 1996 74.0 82.0 70.0 75.0 84.0 97.0 62.0 1997 88.0 100.0 80.0 87.0 91.0 92.0 86.0 1998 42.0 79.0 57.0 32.0 52.0 78.0 40.0 1999 99.0 126.0 94.0 92.0 122.0 110.0 94.0 2000 38.0 99.0 21.0 25.0 18.0 60.0 18.0 2001 145.0 147.0 68.0 2002 88.0 118.0 53.0 76.0 50.0 93.0 54.0 2003 123.0 161.0 120.0 109.0 121.0 113.0 2004 111.0 138.0 97.0 122.0 130.0 90.0 2005 127.0 147.0 98.0 66.0 125.0 94.0 Corn Yields Continued 73 Year Dallas De Kalb Elmore Escambia Etowah Fayette Geneva 1986 44.1 49.3 40.8 62.6 53.1 60.0 52.5 1987 60.7 79.8 51.3 91.1 56.7 78.1 67.8 1988 45.5 41.8 51.0 38.1 43.3 50.0 32.3 1989 71.5 90.0 100.0 88.3 88.9 78.6 55.6 1990 53.7 60.6 58.2 89.8 45.6 79.3 53.9 1991 65.4 85.6 88.2 93.0 85.0 78.8 66.7 1992 70.7 123.7 77.3 100.0 106.7 90.7 78.8 1993 41.4 53.8 43.3 70.1 52.4 58.8 48.4 1994 80.0 109.9 97.5 126.1 104.7 71.3 64.7 1995 65.0 46.0 96.0 60.0 83.0 1996 70.0 92.0 41.0 101.0 83.0 83.0 70.0 1997 75.0 86.0 79.0 109.0 87.0 70.0 85.0 1998 66.0 74.0 36.0 50.0 77.0 54.0 40.0 1999 81.0 101.0 100.0 123.0 98.0 91.0 88.0 2000 48.0 80.0 33.0 38.0 43.0 30.0 53.0 2001 94.0 117.0 95.0 107.0 2002 76.0 84.0 92.0 81.0 87.0 81.0 96.0 2003 105.0 100.0 117.0 127.0 114.0 104.0 108.0 2004 129.0 136.0 132.0 90.0 104.0 101.0 115.0 2005 129.0 113.0 86.0 132.0 91.0 108.0 Corn Yields Continued Year Greene Hale Henry Houston Jackson Lamar Lauderdale 1986 30.0 64.5 34.0 45.0 53.5 54.2 70.0 1987 45.5 82.7 65.8 72.2 65.5 62.5 77.9 1988 38.3 58.9 40.8 60.9 50.5 54.2 34.5 1989 64.0 81.7 86.8 74.5 79.5 87.1 79.3 1990 43.3 56.4 49.9 60.6 60.3 37.8 35.9 1991 74.2 78.4 87.0 91.5 82.9 90.0 54.5 1992 70.0 90.4 67.5 80.0 120.5 84.2 104.0 1993 40.0 49.2 46.8 44.5 54.7 52.9 54.7 1994 68.0 98.5 81.8 82.2 117.4 75.7 96.2 1995 62.0 54.0 63.0 92.0 74.0 95.0 1996 82.0 50.0 67.0 99.0 73.0 80.0 1997 86.0 81.0 91.0 97.0 79.0 59.0 76.0 1998 61.0 48.0 76.0 69.0 52.0 61.0 1999 97.0 121.0 89.0 86.0 98.0 81.0 100.0 2000 25.0 57.0 57.0 80.0 54.0 95.0 2001 88.0 128.0 2002 50.0 95.0 73.0 83.0 85.0 92.0 99.0 2003 84.0 107.0 129.0 127.0 121.0 104.0 145.0 2004 85.0 106.0 107.0 130.0 135.0 2005 112.0 118.0 124.0 83.0 112.0 Corn Yields Continued 74 Year Lawrence Limestone Lowndes Macon Madison Marengo Marion 1986 75.0 69.3 56.3 42.2 63.3 54.5 52.3 1987 65.7 58.2 77.2 57.3 71.2 50.0 76.7 1988 31.1 30.3 40.0 30.0 45.9 53.8 47.6 1989 83.3 82.2 96.4 40.0 80.0 62.7 77.5 1990 44.3 41.2 51.3 23.3 49.3 31.1 41.4 1991 78.6 70.2 57.5 40.0 82.0 50.6 53.0 1992 109.2 119.1 64.7 61.3 98.9 64.0 92.0 1993 65.6 66.7 30.0 30.0 50.0 37.9 52.1 1994 110.9 93.9 75.0 58.8 116.7 75.4 95.4 1995 92.0 94.0 38.0 104.0 47.0 68.0 1996 82.0 90.0 93.0 81.0 74.0 1997 108.0 86.0 99.0 94.0 83.0 51.0 65.0 1998 68.0 71.0 34.0 55.0 68.0 57.0 57.0 1999 136.0 112.0 105.0 108.0 121.0 61.0 2000 99.0 86.0 24.0 40.0 88.0 59.0 55.0 2001 138.0 146.0 84.0 118.0 106.0 2002 115.0 107.0 77.0 79.0 102.0 93.0 2003 153.0 150.0 121.0 125.0 65.0 2004 151.0 150.0 137.0 119.0 2005 136.0 139.0 143.0 100.0 Corn Yields Continued Year Marshall Mobile Monroe Morgan Pickens Pike Randolph 1986 61.4 77.8 71.0 53.3 55.0 36.2 47.5 1987 86.8 83.0 75.3 55.2 70.0 84.7 57.3 1988 42.3 35.4 45.2 37.1 35.0 69.5 56.3 1989 94.3 91.0 94.8 88.5 55.0 85.0 76.3 1990 64.7 92.9 87.0 60.6 39.1 61.8 43.3 1991 79.8 89.3 95.2 65.4 66.0 93.4 57.1 1992 109.7 89.8 85.2 109.4 71.4 88.1 86.7 1993 43.7 77.6 65.1 53.9 50.0 61.6 50.0 1994 99.5 118.5 89.6 95.3 66.5 97.7 71.4 1995 67.0 87.0 84.0 74.0 90.0 49.0 1996 80.0 95.0 92.0 83.0 55.0 85.0 66.0 1997 78.0 91.0 101.0 88.0 90.0 94.0 66.0 1998 64.0 58.0 54.0 51.0 27.0 74.0 63.0 1999 108.0 89.0 110.0 122.0 63.0 100.0 74.0 2000 70.0 68.0 25.0 78.0 38.0 45.0 2001 103.0 103.0 119.0 98.0 2002 98.0 75.0 66.0 104.0 63.0 78.0 52.0 2003 116.0 107.0 120.0 126.0 119.0 2004 113.0 105.0 117.0 116.0 2005 108.0 128.0 106.0 Corn Yields Continued 75 Year Russell Sumter Talladega Tuscaloosa Washington Wilcox 1986 35.0 40.0 72.2 60.0 73.3 50.0 1987 65.8 59.1 76.7 75.2 87.1 65.8 1988 21.4 61.5 65.5 58.9 22.7 40.7 1989 55.6 70.0 49.1 75.0 63.3 78.1 1990 66.9 45.0 37.2 53.7 64.5 82.6 1991 65.7 70.0 90.0 80.0 63.6 85.7 1992 65.0 79.1 88.5 87.4 81.9 79.1 1993 38.3 35.8 70.4 67.6 56.8 58.1 1994 95.0 95.8 97.8 90.9 91.4 90.6 1995 50.0 71.0 55.0 87.0 84.0 90.0 1996 58.0 67.0 95.0 82.0 75.0 61.0 1997 75.0 76.0 117.0 99.0 84.0 89.0 1998 40.0 26.0 65.0 60.0 38.0 40.0 1999 80.0 75.0 123.0 112.0 95.0 97.0 2000 29.0 35.0 23.0 53.0 2001 80.0 75.0 117.0 85.0 2002 63.0 73.0 105.0 81.0 39.0 53.0 2003 76.0 111.0 120.0 117.0 95.0 2004 134.0 122.0 99.0 2005 56.0 131.0 108.0 99.0 Corn Yields Continued 76 Alabama County Cotton Yields 1986-2005 Year Autauga Baldwin Barbour Blount Calhoun Cherokee Coffee 1986 667 640 508 417 792 253 348 1987 609 476 450 502 895 543 449 1988 579 471 539 437 608 519 420 1989 588 529 585 497 328 593 283 1990 531 605 483 408 300 564 336 1991 767 634 934 688 572 770 685 1992 720 907 830 609 630 691 584 1993 413 798 438 388 350 224 509 1994 620 888 688 713 953 816 539 1995 172 494 421 393 303 473 1996 653 889 723 826 825 852 608 1997 660 760 479 617 347 1998 430 478 567 710 748 333 1999 523 841 616 686 438 384 424 2000 355 571 540 458 248 2001 755 813 652 1047 872 843 524 2002 537 396 438 792 610 580 270 2003 873 722 783 890 907 790 624 2004 713 647 665 1094 900 395 2005 781 661 830 1010 847 706 Cotton Yields Continued Year Colbert Conecuh Covington Cullman Dale Dallas 1986 491 240 546 584 345 612 1987 600 489 750 553 287 605 1988 448 523 661 463 208 621 1989 502 754 614 398 383 545 1990 434 375 541 411 419 512 1991 533 457 1015 609 600 730 1992 629 571 791 460 630 692 1993 396 580 651 338 490 573 1994 700 743 631 646 661 576 1995 215 507 522 181 1996 821 572 672 632 412 633 1997 528 483 688 410 550 1998 612 308 520 716 243 467 1999 427 560 674 682 451 607 2000 592 356 486 557 381 398 2001 722 637 780 920 600 626 2002 586 595 465 789 314 589 2003 833 738 781 664 788 2004 773 579 536 1024 462 669 2005 623 630 864 779 683 77 Cotton Yields Continued Year Elmore Escambia Etowah Fayette Geneva Henry 1986 420 793 348 576 541 293 1987 616 659 449 537 507 470 1988 526 602 331 352 504 375 1989 589 601 291 637 425 419 1990 522 544 403 600 488 426 1991 655 686 494 533 656 779 1992 679 733 686 590 754 696 1993 501 724 233 419 642 454 1994 592 811 888 649 638 683 1995 210 463 339 453 499 509 1996 753 884 681 755 536 499 1997 670 853 577 687 425 1998 593 482 758 642 217 444 1999 576 725 408 411 443 443 2000 421 513 324 360 198 413 2001 879 753 781 700 674 581 2002 561 442 517 458 371 2003 837 858 706 589 693 658 2004 795 567 1178 436 517 2005 735 710 852 640 816 Cotton Yields Continued Year Houston Lauderdale Lawrence Lee Limestone Macon 1986 404 478 509 398 554 532 1987 432 583 599 398 559 469 1988 490 356 457 433 436 465 1989 468 462 642 403 576 419 1990 402 361 435 362 372 429 1991 625 495 574 737 555 822 1992 706 681 756 534 801 645 1993 447 410 494 485 512 443 1994 619 792 804 715 865 572 1995 592 526 243 361 382 274 1996 427 847 833 739 907 785 1997 428 599 591 675 491 1998 313 718 788 590 741 544 1999 422 342 438 591 548 626 2000 234 615 543 460 520 309 2001 518 713 703 730 814 785 2002 347 516 598 647 561 496 2003 619 922 773 800 834 764 2004 599 840 826 779 857 763 2005 712 726 701 790 725 78 Cotton Yields Continued Year Madison Mobile Monroe Pike Shelby Tuscaloosa 1986 534 832 592 518 408 632 1987 571 790 872 480 576 706 1988 538 647 791 495 399 591 1989 755 1272 772 435 626 581 1990 595 602 628 534 523 671 1991 726 559 957 894 674 718 1992 844 661 902 626 687 713 1993 670 757 775 467 508 537 1994 957 872 928 686 792 735 1995 471 605 574 478 246 457 1996 871 886 924 567 718 775 1997 478 643 864 431 800 1998 711 467 593 383 724 641 1999 557 713 713 433 643 525 2000 758 736 451 213 459 458 2001 857 791 817 508 688 750 2002 624 530 386 331 516 2003 844 805 723 756 934 2004 1080 703 593 503 795 848 2005 855 571 711 765 800 765 Alabama County Hay Yields 1986-2005 Year Autauga Baldwin Barbour Bibb Blount Bullock Butler 1986 2.00 1.65 2.00 1.50 2.00 1.63 1.62 1987 2.80 2.56 2.91 2.02 2.21 2.00 2.00 1988 2.75 2.35 1.86 1.91 1.54 2.00 2.75 1989 2.83 2.54 2.00 1.75 2.00 2.18 2.22 1990 2.20 2.44 0.82 1.33 1.91 2.13 2.38 1991 2.54 2.00 2.41 2.05 2.21 2.70 2.35 1992 2.10 2.00 2.69 1.78 2.10 2.17 2.05 1993 2.02 2.16 2.19 2.13 2.00 2.05 2.13 1994 2.35 2.91 3.44 3.02 3.05 3.07 2.58 1995 1.30 2.70 1.70 1.80 2.20 1.80 3.60 1996 1.50 2.30 2.50 2.70 2.70 2.50 3.20 1997 2.20 2.20 2.60 2.60 2.40 2.30 2.70 1998 1.70 1.70 1.90 2.30 2.40 1.90 2.00 1999 2.10 2.70 2.90 2.60 2.80 2.30 2.80 2000 1.40 2.80 1.70 1.40 2.60 1.50 1.40 2001 2.90 2.90 3.20 2.40 2.90 2.90 2.70 2002 1.70 2.50 3.00 2.30 2.50 2.20 2.40 2003 3.30 2.50 3.20 2.40 2.90 2.20 3.50 2004 2.90 2.40 3.30 2.70 3.40 2.90 3.10 2005 2.10 2.80 3.60 2.40 2.80 2.70 2.50 79 Hay Yields Continued Year Calhoun Chambers Cherokee Chilton Choctaw Clarke Clay 1986 1.33 1.12 1.09 1.83 1.88 1.82 1.25 1987 1.60 2.14 1.80 1.68 2.29 3.23 1.67 1988 2.67 2.50 2.00 1.71 2.00 2.88 1.46 1989 1.78 1.80 2.50 1.60 2.29 3.43 2.00 1990 1.00 1.85 1.67 1.11 1.73 1.67 1.57 1991 2.07 2.00 2.13 2.34 2.00 1.67 1.81 1992 2.24 2.15 2.04 2.16 2.08 2.00 2.22 1993 1.95 2.11 2.12 2.19 2.03 1.80 2.24 1994 2.75 2.32 2.54 3.19 2.39 3.79 2.59 1995 2.00 1.40 2.00 1.60 2.20 2.20 1.70 1996 2.40 2.00 2.80 2.50 2.50 2.60 2.50 1997 2.20 2.00 2.30 2.60 2.30 2.40 2.30 1998 1.90 2.00 2.20 2.00 2.00 2.20 2.10 1999 2.50 2.00 2.40 2.30 2.20 2.40 2.20 2000 1.90 1.80 1.70 1.50 1.50 1.10 1.60 2001 2.80 2.60 2.60 3.00 2.00 3.00 2.50 2002 2.00 1.90 2.60 2.30 1.80 1.80 2.40 2003 2.70 2.30 3.10 2.50 2.90 3.30 2.30 2004 2.70 3.20 3.20 3.40 3.00 2.70 2.40 2005 2.80 2.60 2.20 2.40 2.90 2.30 2.80 80 Hay Yields Continued Year Cleburne Coffee Colbert Conecuh Coosa Covington Crenshaw 1986 1.20 1.25 1.50 1.64 1.45 1.67 1.49 1987 1.70 2.35 1.60 2.71 1.90 3.05 2.86 1988 1.50 3.00 1.78 2.00 2.00 2.36 2.83 1989 1.60 2.89 2.50 2.00 2.00 2.00 2.50 1990 1.25 1.71 1.89 1.63 2.17 0.80 1.40 1991 1.95 2.82 1.57 1.82 1.85 2.74 3.00 1992 2.09 2.17 1.79 2.28 2.20 2.56 2.28 1993 2.09 2.08 1.77 2.14 2.29 2.16 2.08 1994 2.50 3.08 2.34 2.96 2.78 3.70 3.11 1995 2.40 2.30 1.90 2.60 1.60 2.30 2.80 1996 2.40 3.30 2.20 3.20 1.90 3.20 2.00 1997 1.90 2.60 2.10 2.90 1.90 2.90 3.10 1998 2.00 2.50 1.80 2.20 1.70 2.30 2.50 1999 2.40 2.40 2.20 2.60 2.00 3.10 2.80 2000 1.60 1.50 2.30 1.70 1.30 1.90 2.00 2001 2.70 2.30 2.40 3.00 2.10 3.00 3.30 2002 2.40 2.60 2.10 2.40 1.80 2.30 2.40 2003 3.00 3.00 2.40 2.80 2.10 2.80 2.70 2004 2.30 2.70 2.40 2.70 2.50 2.70 3.20 2005 2.80 3.70 2.10 2.70 2.00 3.40 3.20 Hay Yields Continued Year Cullman Dale Dallas De Kalb Elmore Escambia Etowah 1986 1.82 1.30 1.94 1.42 1.93 1.66 1.18 1987 1.96 2.00 2.29 2.09 2.00 2.44 2.17 1988 1.62 1.47 2.88 1.74 2.00 2.00 1.75 1989 1.91 2.14 3.33 2.63 2.00 2.80 2.00 1990 1.31 1.09 2.17 1.58 1.47 2.33 1.07 1991 2.29 2.59 2.24 2.42 2.50 2.57 1.81 1992 2.17 2.13 2.17 2.38 2.17 2.67 2.20 1993 2.23 2.00 2.03 2.26 1.75 2.27 2.14 1994 3.03 3.06 2.48 2.83 2.58 3.13 2.45 1995 2.50 2.20 2.10 2.50 1.40 2.90 2.10 1996 2.70 3.40 3.30 2.70 2.30 2.10 2.60 1997 2.40 3.00 2.50 2.40 2.40 2.20 2.20 1998 2.40 2.00 2.30 2.40 2.30 2.40 2.40 1999 2.50 2.80 2.20 2.50 2.30 2.90 2.70 2000 2.20 1.10 1.00 2.40 1.50 1.70 1.70 2001 3.00 2.30 2.90 2.90 2.50 2.90 2.60 2002 2.60 2.50 2.20 2.30 2.10 2.00 2.90 2003 3.00 2.70 2.30 3.10 3.00 2.40 2.70 2004 3.30 2.80 2.50 3.00 3.20 2.50 3.20 2005 2.80 3.20 2.40 3.10 2.60 2.30 3.00 81 Hay Yields Continued Year Fayette Franklin Geneva Greene Hale Henry Houston 1986 1.64 1.53 1.33 1.56 1.50 1.40 1.38 1987 1.70 1.92 2.80 2.63 1.81 2.80 2.15 1988 2.00 2.45 2.00 2.00 2.25 1.63 2.00 1989 2.25 2.29 1.83 2.43 2.50 2.00 2.22 1990 1.64 1.26 1.43 1.11 1.00 1.20 1.00 1991 1.86 1.52 2.32 1.72 2.64 2.41 2.12 1992 1.89 1.85 2.58 1.80 2.13 2.31 2.28 1993 2.12 1.85 2.22 1.56 2.00 1.84 1.69 1994 2.97 2.58 2.63 3.05 2.12 2.50 2.62 1995 2.50 2.50 2.20 2.80 2.00 1.70 2.00 1996 2.50 2.10 2.40 2.10 2.10 1.90 1.80 1997 2.60 2.10 2.60 2.20 2.20 1.90 2.10 1998 2.40 2.10 2.50 1.90 2.20 2.00 2.20 1999 2.50 2.30 2.40 2.00 2.20 2.20 2.40 2000 1.60 2.30 1.00 1.30 1.60 1.10 1.00 2001 2.50 2.50 3.00 1.80 1.80 2.10 1.70 2002 2.20 2.50 2.20 1.90 2.00 2.50 2.50 2003 2.70 2.70 2.10 2.90 2.20 2.00 2.10 2004 2.60 2.80 2.50 2.20 2.30 2.80 2.10 2005 2.80 2.60 2.30 2.70 2.70 2.30 2.30 Hay Yields Continued Year Jackson Jefferson Lamar Lauderdale Lawrence Lee Limestone 1986 1.57 1.68 1.42 1.66 1.60 1.88 1.50 1987 1.61 1.78 1.70 1.58 1.82 2.71 1.71 1988 1.88 2.10 2.00 1.78 1.53 1.44 1.40 1989 1.55 1.63 2.25 2.06 1.94 2.14 1.40 1990 1.71 1.20 2.40 0.91 1.50 1.22 1.20 1991 2.07 1.89 1.73 1.50 1.93 2.53 1.57 1992 2.02 2.00 1.96 1.73 1.85 2.26 1.76 1993 2.01 2.08 2.00 1.62 1.81 2.04 1.64 1994 2.52 2.79 2.59 2.18 2.63 3.17 2.26 1995 2.20 2.10 2.30 1.90 2.20 2.10 1.60 1996 2.30 2.10 2.70 1.80 1.90 2.60 1.90 1997 2.20 2.20 2.30 1.80 1.90 2.70 1.80 1998 2.30 2.40 2.20 2.00 1.80 2.20 1.90 1999 2.50 2.60 2.10 1.90 2.00 2.70 1.50 2000 2.30 1.50 1.60 2.10 1.80 1.30 1.70 2001 2.80 2.10 2.80 2.00 2.50 3.20 2.20 2002 2.10 1.90 2.40 1.80 2.10 2.90 1.70 2003 2.30 2.20 2.50 2.00 2.30 2.90 2.30 2004 2.90 2.30 2.40 2.20 2.40 3.50 2.30 2005 2.70 2.80 3.10 2.30 2.60 3.20 2.20 82 Hay Yields Continued Year Lowndes Macon Madison Marengo Marion Marshall Mobile 1986 1.55 1.50 1.59 1.90 1.89 1.90 2.22 1987 2.33 2.13 1.76 2.00 1.73 2.31 2.82 1988 1.93 2.00 1.83 1.74 2.63 2.00 3.00 1989 1.86 2.00 2.40 2.53 2.20 2.75 2.00 1990 1.45 1.47 1.18 1.94 1.50 1.83 1.55 1991 1.90 2.66 2.00 1.75 1.63 1.89 2.91 1992 1.81 2.14 1.69 2.16 2.11 2.30 2.70 1993 1.77 1.76 1.70 1.88 1.95 2.25 2.63 1994 2.19 2.54 2.35 2.29 2.53 2.89 3.77 1995 2.00 1.40 2.10 2.10 2.60 2.40 2.10 1996 2.50 2.00 2.10 2.20 2.30 2.80 2.30 1997 2.10 2.20 1.70 1.80 2.10 2.40 2.40 1998 2.00 1.80 1.80 1.90 2.30 2.20 2.50 1999 2.90 1.70 1.80 1.60 1.80 2.70 2.10 2000 1.10 1.70 2.00 1.30 2.00 2.50 2.30 2001 3.50 3.40 2.00 2.10 2.80 2.70 2.30 2002 2.10 2.60 1.90 1.90 2.10 2.50 2.60 2003 2.20 3.00 2.30 2.00 2.60 3.00 2.90 2004 2.40 2.60 2.30 2.30 2.80 3.10 2.50 2005 3.10 3.00 2.40 2.40 2.40 3.00 3.30 Hay Yields Continued Year Monroe Montgomery Morgan Perry Pickens Pike Randolph 1986 1.95 1.44 1.65 1.91 1.60 1.55 1.21 1987 3.00 1.76 2.26 2.36 1.80 2.96 1.93 1988 3.60 1.52 2.36 2.08 1.81 2.00 3.13 1989 2.50 2.72 2.10 1.79 1.45 2.00 2.31 1990 1.14 0.82 1.42 1.00 1.05 2.32 2.00 1991 1.75 1.98 1.85 2.10 2.27 3.10 1.91 1992 2.22 2.01 1.91 1.62 2.22 3.46 1.88 1993 2.19 1.75 1.89 1.70 2.16 2.67 2.11 1994 2.67 2.21 2.42 2.13 2.72 3.76 2.69 1995 2.60 1.40 2.10 1.50 2.40 2.80 1.90 1996 2.80 1.80 2.40 2.10 2.80 3.30 2.10 1997 2.60 2.10 2.00 1.90 2.60 3.50 2.00 1998 1.70 1.60 2.10 2.00 2.10 2.50 2.10 1999 2.50 2.00 2.00 1.60 2.70 3.30 2.30 2000 2.00 1.10 2.00 1.50 1.30 1.90 1.90 2001 2.20 2.20 2.40 2.20 2.90 3.30 2.50 2002 2.40 1.50 2.00 1.50 2.10 2.60 2.30 2003 3.20 2.30 2.60 2.40 3.20 3.10 2.80 2004 2.50 2.20 2.40 2.20 3.30 3.60 2.50 2005 3.50 2.50 2.50 2.50 3.30 3.70 2.40 83 Hay Yields Continued Year Russell Shelby St. Clair Sumter Talladega Tallapoosa Tuscaloosa 1986 1.88 1.16 1.32 1.58 1.19 1.45 1.32 1987 2.29 1.62 2.20 2.00 1.89 2.43 2.70 1988 3.29 1.88 1.60 2.36 1.76 1.56 1.67 1989 3.00 2.27 2.50 2.75 1.83 1.75 3.15 1990 2.22 1.00 1.08 1.20 1.95 2.29 2.00 1991 2.86 1.77 1.71 2.22 2.02 2.14 2.27 1992 2.65 1.71 2.22 2.10 1.77 2.38 2.21 1993 2.13 1.89 2.25 1.95 1.91 2.44 2.12 1994 3.47 2.92 3.00 2.28 3.89 2.94 2.48 1995 2.10 1.40 2.20 2.60 1.50 1.60 1.60 1996 2.70 2.10 1.80 2.20 2.50 2.50 2.50 1997 2.90 2.00 2.30 2.20 2.20 2.60 2.00 1998 2.00 1.90 2.30 1.70 1.80 2.30 2.30 1999 3.30 2.00 2.00 2.40 1.80 2.60 2.00 2000 2.70 1.70 1.90 1.80 1.60 2.00 1.70 2001 3.20 2.60 2.60 1.70 2.70 3.20 2.10 2002 2.50 1.80 2.00 1.70 2.00 2.50 2.30 2003 2.60 2.10 3.10 2.20 2.20 2.50 2.70 2004 2.90 2.30 3.20 2.50 2.60 2.30 2.50 2005 2.90 2.70 3.30 2.00 2.50 2.50 2.90 Hay Yields Continued Year Walker Washington Wilcox Winston 1986 1.43 2.01 1.64 1.67 1987 1.90 2.75 1.68 1.75 1988 1.00 1.82 2.95 2.00 1989 1.86 2.20 2.15 2.55 1990 1.50 1.11 1.94 1.78 1991 1.90 2.45 2.30 2.24 1992 2.16 2.13 2.26 2.13 1993 2.02 2.16 2.07 2.00 1994 2.56 3.02 2.86 2.78 1995 1.50 2.60 2.30 2.10 1996 2.00 2.60 2.40 2.20 1997 1.90 2.20 2.10 2.30 1998 1.80 2.40 2.40 1.90 1999 2.80 2.80 1.80 2.00 2000 1.40 1.60 1.30 1.90 2001 2.30 2.80 2.40 3.10 2002 2.40 2.60 2.00 2.30 2003 2.90 3.30 2.40 2.50 2004 2.60 2.50 2.50 2.40 2005 2.80 2.80 2.20 2.80 84 Alabama County Peanut Yields 1986-2005 Year Barbour Coffee Conecuh Covington Crenshaw Dale 1986 2515 2680 2105 2375 2645 2325 1987 1545 2285 2050 2430 2250 2105 1988 2135 2570 2160 2680 2315 2300 1989 2110 2120 2255 2585 2015 2265 1990 1150 1255 1295 1720 2255 1285 1991 2375 2340 2130 2525 2265 2400 1992 2660 2580 2160 2680 2445 2675 1993 1835 2250 2190 2360 2370 2030 1994 2250 1855 1830 2110 2100 2030 1995 2210 2205 2405 2565 2220 2315 1996 2710 2510 2305 2930 2850 2225 1997 2025 1760 2015 2290 1935 1695 1998 1805 2105 1840 2620 1960 2140 1999 2620 2125 2195 2470 2060 2190 2000 1390 1130 1325 1585 1660 1220 2001 2605 2350 2700 2700 2580 2570 2002 2420 2070 1500 2390 2410 1980 2003 2830 2585 2920 3000 2410 2004 2620 2430 2565 2420 2005 2810 2445 2805 2975 Peanut Yield Continued Year Escambia Geneva Henry Houston Pike Russell 1986 2665 2485 1985 1895 2260 1940 1987 2665 2280 2050 2165 1855 1660 1988 2450 2480 2155 2570 2190 1520 1989 2045 2330 2275 2390 2040 2000 1990 2000 1770 1200 1685 1605 1585 1991 2330 2285 2390 2235 2155 2115 1992 2415 2390 2620 2410 2250 2645 1993 2475 2180 1425 2050 1850 1300 1994 2610 1755 2290 1935 1885 2360 1995 2715 1900 2445 2385 2350 1650 1996 2925 2275 2185 2100 2380 2990 1997 2940 1980 1900 1960 1885 2180 1998 2920 2130 2265 2265 2255 2675 1999 3165 1965 1890 2100 2305 2185 2000 2600 1355 1250 1525 1465 1725 2001 4075 2525 2570 2415 2720 2905 2002 1960 1995 1910 2055 2380 2485 2003 3335 2190 2620 2755 2955 2004 3425 2380 2610 2530 2660 2005 2965 2485 2685 2400 2640 85 Alabama County Potato Yields 1986-2005 Year Baldwin Cullman De Kalb Jackson 1986 150.00 125.00 140.00 130.00 1987 135.00 170.00 156.00 150.00 1988 140.00 115.00 88.00 85.00 1989 245.00 180.00 174.00 155.00 1990 150.00 180.00 160.00 149.00 1991 120.00 146.00 141.00 133.00 1992 155.00 152.00 177.00 163.00 1993 157.00 123.00 83.00 96.00 1994 176.00 170.00 173.00 167.00 1995 161.00 197.00 168.00 170.00 1996 160.00 140.00 140.00 177.00 1997 172.00 156.00 148.00 160.00 1998 154.00 85.00 136.00 111.00 1999 172.00 140.00 178.00 250.00 2000 197.00 107.00 158.00 160.00 2001 163.00 150.00 161.00 164.00 2002 205.00 135.00 213.00 148.00 2003 143.00 267.00 234.00 182.00 2004 139.00 94.00 246.00 187.00 2005 114.00 113.00 160.00 189.00 Alabama County Soybean Yields 1986-2005 Year Baldwin Blount Calhoun Cherokee Colbert Cullman Dallas De Kalb 1986 25.60 26.50 24.00 17.90 23.00 21.00 18.50 18.00 1987 25.00 19.00 15.00 16.50 15.00 19.50 16.50 17.00 1988 27.70 25.10 23.30 26.70 21.20 25.00 24.10 25.00 1989 22.00 27.30 23.00 25.00 24.00 27.10 17.00 26.80 1990 19.00 16.80 18.20 15.70 12.60 14.70 12.60 21.90 1991 25.90 25.00 23.80 23.50 19.10 25.10 21.80 22.00 1992 30.00 35.00 35.70 28.10 32.10 34.10 23.60 32.80 1993 30.10 28.20 19.70 14.70 24.90 28.30 23.90 24.70 1994 28.60 43.50 39.00 36.40 34.80 43.10 26.10 37.10 1995 29.00 21.00 23.00 18.00 22.00 19.00 25.00 1996 35.00 37.00 35.00 36.00 30.00 40.00 29.00 36.00 1997 25.00 28.00 27.00 23.00 27.00 19.00 26.00 1998 22.00 20.00 25.00 22.00 31.00 27.00 18.00 20.00 1999 28.00 20.00 18.00 16.00 11.00 13.00 23.00 15.00 2000 20.00 20.00 15.00 17.00 9.00 16.00 12.00 25.00 2001 34.00 35.00 35.00 35.00 35.00 38.00 2002 27.00 27.00 21.00 24.00 28.00 31.00 20.00 19.00 2003 31.00 41.00 40.00 36.00 39.00 2004 31.00 39.00 40.00 35.00 42.00 32.00 39.00 2005 31.00 41.00 31.00 36.00 40.00 86 Soybean Yield Continued Year Escambia Etowah Fayette Geneva Houston Jackson Lauderdale 1986 27.50 18.50 22.00 19.50 16.70 25.00 21.80 1987 23.50 17.50 21.40 18.40 16.40 16.50 16.50 1988 29.00 25.90 22.30 29.30 29.60 27.10 17.70 1989 17.10 21.60 20.20 19.00 19.30 23.00 21.00 1990 17.40 16.20 21.20 12.50 10.10 20.60 13.60 1991 25.00 18.90 23.20 20.90 21.40 24.00 25.00 1992 33.40 32.10 27.90 26.70 25.00 27.70 30.50 1993 25.10 22.70 23.30 18.80 17.90 19.20 23.90 1994 31.00 34.60 32.50 28.90 27.20 31.90 27.20 1995 19.00 25.00 27.00 27.00 1996 36.00 36.00 33.00 24.00 26.00 34.00 36.00 1997 26.00 24.00 23.00 20.00 18.00 26.00 27.00 1998 33.00 20.00 23.00 23.00 17.00 19.00 27.00 1999 33.00 18.00 19.00 18.00 14.00 13.00 6.00 2000 13.00 18.00 12.00 15.00 26.00 9.00 2001 35.00 35.00 38.00 2002 33.00 25.00 22.00 26.00 32.00 21.00 25.00 2003 33.00 41.00 29.00 33.00 32.00 33.00 37.00 2004 33.00 41.00 37.00 36.00 29.00 34.00 31.00 2005 39.00 41.00 32.00 32.00 28.00 Soybean Yield Continued Year Lawrence Limestone Madison Marion Marshall Morgan Talladega 1986 20.60 25.00 29.00 28.00 20.50 22.00 25.00 1987 14.00 13.00 17.00 15.00 16.50 18.00 11.00 1988 18.30 23.10 24.20 21.30 20.50 23.50 26.20 1989 23.50 23.20 22.30 22.60 18.00 23.20 18.70 1990 12.60 15.70 22.00 18.80 14.60 20.90 20.90 1991 17.60 26.50 28.20 15.10 17.00 16.50 23.40 1992 31.50 29.90 31.00 29.20 29.40 30.70 24.50 1993 26.30 28.20 25.60 19.60 17.40 24.00 20.50 1994 28.70 32.60 36.30 30.70 35.40 27.90 30.70 1995 26.00 24.00 24.00 26.00 20.00 22.00 17.00 1996 37.00 36.00 37.00 34.00 35.00 37.00 37.00 1997 29.00 30.00 27.00 24.00 25.00 24.00 23.00 1998 25.00 22.00 22.00 29.00 18.00 21.00 21.00 1999 12.00 8.00 12.00 16.00 18.00 16.00 2000 13.00 17.00 20.00 13.00 20.00 13.00 2001 35.00 35.00 34.00 33.00 31.00 43.00 2002 27.00 25.00 23.00 21.00 26.00 22.00 21.00 2003 37.00 39.00 36.00 40.00 27.00 35.00 35.00 2004 36.00 40.00 35.00 38.00 30.00 33.00 37.00 2005 30.00 31.00 31.00 33.00 31.00 33.00 87