Zone Delineation Based On Soil Fertility and Terrain Attributes with Factorial Cokriging.

There is an information revolution occurring in agriculture creating massive amounts of data at different spatial and temporal scales; therefore, efficient techniques for processing and summarizing data is crucial for management.  The objective of this study was to apply multivariate geostatistical techniques to create efficient zones for summarizing soil and landscape data.  For a 50 ha Central Kentucky field, LIDAR data collected in November of 2010 were used to calculate terrain attributes (i.e., elevation, slope) on a 1-m grid. Deep and shallow EC measurements were made in 1999 along transects with approximately 10-m between passses. Soil samples (n= 540) collected in 2000 were tested for pH, SMP buffer pH, P, Ca, K, Mg, Zn, total carbon.  Data fusion techniques were used to combine the data from the three different scales of sampling.  A linear isotropic model of coregionalization (LMC) was fit for multicollocated cokriging and factorial cokriging and included three basic variances: nugget, spherical (range=140m), and spherical (range=500m) components. The nugget variance was large and represented approximately 1/3 of the overall variance with remaining variability split evenly between the other two structures.  Only the first factors for both the 140-m and 500-m spatial scales had eigenvalues greater than 1 indicating that these were the only significant factors in the dataset. These two factors efficiently described and synthesized most of the complex multivariate spatially correlated information observed in the fertility and terrain data.  The first factor for the 140-m structure with was positively and primarily associated with EC, Ca, and Mg. The first factor for the 500-m structure was negatively and mostly associated with elevation, P, and Zn.  The analyses suggests that multivariate techniques can be used to efficiently summarize multiple spatial data variables collected at different spatial scales into a small number of agronomically meaningful indices.