Effect of the representative coordinate of the aggregation on apparent electrical conductivity data and its relationship with measures of spatial dependence

Abstract

Estimating soil resources at a different scale on which observations are made is a major problem that continues to generate related research. An increase in scale means an increase in the variation of the parameter, and this can cause problems when interacting with non-linearity in a process or model. Changing the spatial resolution by adding or disaggregating data carries the risk of conflicting results. To demonstrate this fact, Apparent Electrical Conductivity data were taken with the EM38-MK2 sensor in a vertical position to the ground simultaneously with the two dipoles at two relative depths (0.75m and 1.5m), associated with the same coordinate. Spatial aggregation sizes were evaluated from a fishnet of 5m´5m to 70m´70m, with an arithmetic ratio of 5m. Representative coordinates were used to generate the matrix of spatial weights based on the: i) center of the grid, ii) mean value of the coordinates that spatially intercept each cell, and iii) value of the centroid of the points added by each cell. To analyze the spatial autocorrelation pattern, the Moran Montecarlo index was used for the residuals of the adjusted model. The results showed that as the size of the grid is increased, the univariate spatial dependence begins to decrease for all representative coordinates, with the coordinate of the center of the cell being the most affected. For a specific sensor depth, the use of the centroid coordinate and in aggregations that exceed 20m is recommended to maintain the spatial dependency structure that could be natural in this variable and convenient in spatial regression modeling processes.