Use of Soil and Remote Sensing Variables to Explain Spatial Differences in Corn Yield.

Spatial and temporal variability of soil and crop processes can be used to determine management zones for precision agriculture. In this study a set of direct and indirect measurements that include soil texture, electrical conductivity, NDVI, digital elevation model, and soil data from NRCS (including pH texture, organic matter, soil water content among others variables) was used and compared with corn yield on a silty loam soil located in Princeton, Kentucky. Soil texture was measured at 96 locations with a distance of 50x50 m at five depths, and was determined using the pipette method. NDVI was obtained from Landsat images at an early (May) and late (August) growth stages. DEM was obtain from a LiDAR 5 ft resolution image. Electrical Conductivity was measured using a contact sensor Veris 3150 at a shallow (≈30 cm) and deep (≈100 cm) depths. Data was analyzed using semivariograms, crossvariograms, cross-covariance, kriging, and cokriging. Results indicated an inverse relationship between corn yield and electrical conductivity. NDVI at late stage revealed a positive relationship with yield. Using a clay cokriging map derived from clay content at 0-20 cm and electrical conductivity at a shallow depth allowed us to identify areas with low yield and high clay content, and a positive relationship was observed with total silt and medium silt maps.  On the other hand, NRCS data did not allow to derive a detailed relationship with other variables due to its low spatial resolution at the field scale, but helped to identify areas with low yield. These results demonstrate the ability to detect spatial differences in the field using direct and indirect methods, which are related to biomass production and will help in the future to manage fields site specifically. By combining some of these variables we can improve the precision to develop management zones.

Key words: Precision agriculture, Geostatistics, Soil Physics.