Analyzing Long-term Historical Peanut Yield in Georgia with a Crop Simulation Model: the Southeast Climate Consortium Experience

Abstract

It is common practice to use crop simulation models and long-term historical weather datato study the impact of climate variability on agricultural production. The variation in simulated yieldusing this approach reflects the inter-annual and intra-annual weather variability while cropimprovement through breeding and management practices are not reflected. Therefore, historicalyield data cannot be readily used for evaluation of crop simulation models. The objectives of thisstudy were to analyze long-term historical peanut yields in Georgia from a dynamic crop simulationmodel and to assess the use of long-term county average yield from the USDA-National AgriculturalStatistics Services (NASS) for evaluation of simulated yield. Yield data for Burke, Sumter, and Tiftcounties from 1934 to 2003 were obtained from the USDA-NASS. The CSM-CROPGRO-Peanutmodel was used to simulate yield by grouping the period 1934-2003 in three technological periods(TP). For each TP, three soil types, three planting dates, one to three peanut varieties, and irrigatedand/or rainfed conditions were used for the simulations. A unique cropping season yield for eachperiod was obtained with a weighted average based on the acreage of the soil type, the peanut variety type, and the proportion of rainfed and irrigated land in each county. Then, observed andsimulated yield, total rainfall, and air temperature of the cropping season were grouped with respectto the climatological, El Nio Southern Oscillation (ENSO) phases, and TPs data sets. Each set ofdata was standardized using the Z-score, which converts all values into compatible units with adistribution that has an average of 0 and a standard deviation of 1. Summary statistics were obtainedand the Pearsons coefficient of correlation was used as a measure of similarity between observedand simulated yields. Linear regressions were also calculated to assess the relationship betweenrainfall and yield patterns. The inter-annual variation of peanut yield, mainly due to climate variabilitywas clearly observed in the simulated series. We also found that the use of observed and simulatedyields provided a better understanding of the historical peanut production. The impact of the climatevariability on observed yield was low from 1934 to 1964 but high from 1974 to 2003. Thetechnological periods provided an improved characterization for peanut production in Georgia. The1934-1954 period was characterized by low and stable yields. A significant increase in yield due tonew technologies occurred during the 1955-1978 period but yields were generally stable during the1979-2003 period. The results from this study showed that crop models can be useful tools forunderstanding the historical variation in yield due to climate variability if appropriate adjustments aremade to account for changes in agrotechnology.