Geostatistical filtering for improved soil water content estimation from electrical resistivity data

In order to take best advantage of the electrical resistivity data, kriging spatial component technique was applied to separate the small and large-scale structures of resistivity data. Then the spatial structures in resistivity data which are poorly correlated with soil water content are filtered out prior to integrating resistivity data into water content estimation in the soil of a 10 ha area located in the Picardie region (Northern of France). The soil water content was measured until 1.2 m depth in 81 sites with 0.3 m increments by gravimetric method. The resistivity was exhaustively measured for three different depths: 0.5 m, 1 m, and 2 m. The resistivity data represents a set of 5656 measurement points for each of the three depths over the 10 ha study area. The methodology involves successively: (1) a principal component analysis (PCA) on the electrical measurements for the three depths of soil; (2) a geostatistical filtering of the local component and noise in the first component (PC1) of PCA which account for 90% of the variance in the electrical measurements for the three depths of soil. Results have shown that the correlation between water content of soil profile (W) and PC1 is highly improved when the local component and noise were filtered out. Finally, the influence of the smoothness of the external drift function on the quality of estimates of a target variable was empirically demonstrated. Indeed, the estimates of W were highly improved when the external drift function used in kriging is much smoother: the large scale of PC1 instead of PC1.

► The geostatistical filtering improves the quality of soil's electrical resistivity.
► The structures poorly correlated with a target should be filtered prior to estimation.
► The smoothness effect of the external drift in the estimation was demonstrated.