In addition to the accuracy of the instruments and positioning errors, site conditions can influence correlations. They may have been variations between the different dates of measurements. In our case, repeated runs at the same date show correlations that are not substantially better than correlations between data collected at different dates. To explain this, we can think of a number of reasons. First, the statistical procedure we have chosen is robust. Minor deviations and outliers have little influence on t, and nonlinearities of relationships have no effect at all as long as monotonically increasing or decreasing models can describe them. For example, soil temperature is not influencing t because of the monotonic increase of ECa with rising temperature. Another reason can be the positive correlation of some of the most important factors influencing soil ECa on nonsaline mineral soils—namely, soil texture and water content (Luck and Eisenreich, 2001). Under the temperate climatic conditions of Germany, in addition to marshland, soil salinity is not an important factor. As long as mineral soils are not affected by spatiotemporal variations of groundwater level, lateral water flow or diverging water uptake by crops, and soil water content largely depends on soil texture. On these soils, organic matter content and cation exchange capacity can be correlated with soil texture as well (Luck and Eisenreich, 2001; Rogasik et al., 2002). Stability of ECa patterns over time was also observed by other authors (Sudduth et al., 2003).
Comparing the different soilscapes, correlations were best when the rate of fluctuation in ECa was low in the horizontal and vertical directions (Beckum). Vice versa, Bornim with its extremely heterogeneous soil, shows the worst correlation. When we try to compare correlations of different instruments, it is very important to bear in mind how electrode spacing influences the strength of correlation. None of the GCR instruments had exactly the same spacing as the GeoTom (even the ARP deviates because of the trapezoid arrangement of its electrodes). Best correlations were usually found when electrode configuration (or equivalent coil configuration with EMI) comes close to the electrode configuration of the GeoTom. Correlations could be better when configurations are matching exactly (e.g., Veris level 1 will show a higher correlation with GeoTom spacing between 20 and 30 cm). Therefore, one should not easily rate an instrument as superior by this kind of comparison. In our tests, most instruments produce good correlations regarding electrode spacing, depth of investigation, and soil variability. The CM-138 clearly is an exception. Correlations are generally lower and sometimes inconsistent. Repeated runs show low accuracy. There seem to be a number of reasons for this behavior; temperature influences the CM-138 substantially, it shows erratic drifts during stationary measurements, and it has very low dynamics in signal response (Dabas et al., 2004). These problems remained even after the instrument was returned to the manufacturer. The EM38 is subject to ambient influences and drifts as well (Sudduth et al., 2001), but it was much more reliable than the CM-138 (Dabas et al., 2004). To improve temperature stability of the EM38, Geon-ics Ltd. has been offering an update since 2004.
In most cases, ARP level 1, 2, and 3 correlated best with GeoTom level 1, 2, and 3. Veris levels 1 and 2 roughly coincide with GeoTom levels 1 and 2. EM38 (-DD) in the vertical mode can be related to GeoTom level 4. Two exceptions probably can be treated as "outliers": Veris level 2 does not show the best correlation with GeoTom level 2 in Kassow. This may be due to poor soil contact as a thick turf covered the soil surface. Also in Kassow there is an "irregular" best correlation of EM38 V mode with GeoTom level 3. We cannot explain this at the moment.
Correlation generally increased with depth of investigation. This can be due to a smoothing effect, which improves correlation. Because wider electrode spacing is integrating larger soil volumes, small-scale variability is smoothed out.
Comparing the correlations of EM38 horizontal mode in Beckum, Bornim, and Golzow, it is remarkable that highest t in Beckum was found at GeoTom level 4, and t in Golzow was highest at level 1 and in Bornim at level 3. We explain this by higher sensitivity to conductive structures which is characteristic for EMI methods (Dabas and Tabbagh, 2003). Golzow has layered soils where sometimes loam covers sand. In these cases, the EM38 reacts preferentially to the conductive structures in the topsoil, and there is less response to signals from deeper layers. Beckum has opposite conditions: highly conductive clay appears regularly at the bottom of the soil profile. It is interesting to find out whether the DC methods show an opposite behavior—that is, higher sensitivity to resistive structures. The results from Beckum seem to confirm this assumption. ARP 2 and Veris 2 showed the highest correlations with GeoTom 1 in Beckum, but correlations were better with Geo-Tom 2 at all the other sites. This explains the observations from earlier studies (Suddath et al., 2003), where correlation between Veris 3100 and EM38 was influenced by soil profile layering.
Evaluating depth sensitivity, it turns out that correlation over different GeoTom levels changes more gradually with the EMI methods than with the GCR methods. This means that EMI methods have less distinct depth sensitivity. Dabas and Tabbagh (2003) give theoretical reasons for this behavior.
Was this article helpful?