Spatial and temporal variability in cotton yield in relation to soil apparent electrical conductivity, topography, and remote sensing imagery

Analysis of data from multiple fields over several years provides the ability to determine under what conditions precision agriculture may be suitable. The objectives of this study were to: a) evaluate the spatial variability in cotton (Gossypium hirsutum L.) yield; b) assess the temporal stability in cotton yield over different growing seasons; c) determine the spatial and temporal variability in cotton yield in relation to soil apparent electrical conductivity (ECa), terrain attributes, and bare soil brightness; d) delineate potential management zones based on ECa, terrain attributes, and bare soil brightness obtained from satellite images and evaluate the consistency of the management zones over different growing seasons.

This study was conducted on eight commercially managed cotton fields on the Southern High Plains of Texas from 2000 to 2003. Yield data were collected using harvesters equipped with yield monitors and global positioning systems (GPS). Digital elevation data were collected using a real time kinematic (RTK) GPS system. Elevation, slope, and curvatures were derived from the digital elevation data. The Normalized difference vegetation index (NDVI) was derived from multiple in-season Landsat remote sensing images. Bare soil brightness was obtained from two pre-season Landsat remote sensing images. Three potential management zones for each field in each year were delineated using the k-means and the fuzzy c-means methods.

Two fields with high spatial variability in yield and soil properties were temporally stable in relative yield distribution over the four years, while the other fields were not stable. Remote sensing images explained up to 70% of yield variability in fields with high variability in yield. The strongest relationship between yield and remote sensing images occurred in the middle of the growing seasons. Soil apparent electrical conductivity, terrain attributes, and bare soil brightness explained up to 81% of yield variability, which varied with fields and years. A greater amount of yield variability was explained in drier years than in wet years. Apparent electrical conductivity and bare soil brightness were more important in explaining yield variability than terrain attributes. Both k-means and fuzzy c-means were able to separate yield and the soil properties, but k-means tended to delineate more consistent and distinct management zones. Fields with higher variability in yield and soil properties tended to have more consistent management zones over different growing seasons. Based on the results from this study, soil apparent electrical conductivity and bare soil brightness appear to be the most important soil characteristics evaluated in this study for determining management zones in the Southern High Plains of Texas. Fields with high spatial variability in yield and soil conditions appear to be better suited for PA applications.