Evaluation of soybean yield variation for sites in Tennessee with landscape and soil classification using GPS/GIS technologies

Crop yields vary spatially across field landscapes. This variation in production can be quantified on a site-specific basis with current technologies like global positioning systems (GPS) and yield monitoring equipment. However, the factors affecting yield are not as easily identified due to spatial differences in topography and inherent soil properties. This research focused on identifying areas of varying production and examining topographic factors affecting yield variation. Detailed elevation data were used to distinguish the major topographic factors affecting soybean yield. Furthermore, intensive soil classification data available for each field was combined with topography information in an effort to increase explanation. The objective of this study was to determine to what degree detailed topographical information, alone or together with intensive soil mapping data, explains yield variability on a field scale. High resolution field topography data was collected using Real-Time Kinematic Differential GPS (RTK-DGPS). Topographical variables were used to model hydrological parameters within selected fields in Tennessee. Soil classification and topographical variables were compared with soybean yield variability. Topographical variables for each field used in this study were combined into classes based on hydrological parameters, landscape position, and percent slope. These landscape classes were compared with 1997 and 1999 soybean yields in an effort to explain yield variation. Significant explanation of yields varied from field-to-field and year-to-year. Results showed that landscape classification explained a maximum of 27.6% of yield variation for all sites and years in this study. In-field soybean yield variability for 1999 was better explained by landscape positions than in 1997. Yields in 1999 were lower than normal due to lack of precipitation. Significant yield explanations were obtained when landscape classes identified landscape areas that were not conducive to crop production under droughty conditions, such as ridgetops and sideslopes with greater slopes. It was determined that 1997 in-field yield variability was better explained when landscape classification delineated areas of excessive water accumulation. In these areas, yields were significantly lower due to stand reduction and stunted growth caused by excessive water accumulation. Landscape classification and soil mapping unit data increased the amount of soybean yield variation explained in this study. Models that contained both variables explained as much as 33.89% of the yield variation present in fields examined. Soil mapping units provided a majority of the yield explanation produced from models with landscape and soil variables. Statistical analysis was conducted to determine how well landscape classes derived from topographic data described soil mapping units within each field. Results showed that significant relationships existed between the two variables, but the amount of correlation was not consistent among the fields. However, further investigation revealed that landscape classes were useful in identifying soil unit polygons. Visual interpretation of the two data layers in a GIS format offered a better understanding of the relationships between landscape position and soil units. The ability of landscape classification to aid in determining specific soil breaks within fields depends on the subjective and knowledgeable adjustment of data sets to fit unique field characteristics. Soybean yields for 1997 and 1999 at the Franklin County sites were compared statistically in an effort to determine correlations between multiple year soybean yield data. Statistical analysis was conducted on normalized yield classes for all fields. Significant relationships existed between the two years of soybean yield for each field, but the amount of relation varied from field-to-field. Results showed the areas of average yield remained constant from 1997 to 1999. However, the below average areas in 1997 were frequently part of the average yielding areas in 1999. The soybean yields for areas of excessive water accumulation were not as adversely affected in 1999 compared to 1997. When investigating multiple year yield differences, consideration must be placed on the potential yield of an area depending on the topographic location within a field and the effect of different yearly weather conditions to that specific area. Areas of differing soybean production were identified by landscape classifications. Results from this study show that landscape classification based on topographic and hydrological features provides insight into varying yields in different fields. The amount of yield explanation depends on individual field characteristics and varies accordingly. With this in mind, further investigation may need to be focused on using multiple year data to identify, adjust and interpret classification techniques adaptive to individual field characteristics