Upgrading a 1/20,000 soil map with an apparent electrical conductivity survey

Abstract

Despite their shortcomings, choropleth soil maps remain the most widespread source of information on soil resources. Since most nationwide soil surveys were conducted in the second half of the previous century, a need for upgrading emerges. We evaluated the potential of detailed observations made by a mobile, non-invasive proximal soil sensor to upgrade a part of the 1/20,000 choropleth soil map of Belgium. This study was conducted on a 14 ha area which had been mapped twice in the 1950s: first, during the national soil survey yielding a 1/20,000 soil map, and second, during a detailed investigation resulting in a 1/5000 map. The first map failed to identify the presence of a Tertiary clay substratum at variable depths, while the second map indicated this substratum to be present within 1.2 m below the soil surface for about a third of the area. A recent survey with the EM38DD soil sensor provided 9192 measurements of the apparent electrical conductivity (EC a) within the study area. The depth of the substratum ( D ts) was noted at 60 calibration locations by augering and the relationship between EC a and D ts was modelled by an exponential curve with an R 2 of 0.80. This allowed the detailed mapping of D ts by regression kriging. The predictions were validated using 46 independent observations of D ts indicating a reasonable average error of 0.24 m and a very good correlation coefficient between observed and predicted values of 0.94. A map accuracy assessment indicated that even after classification, the D ts classes were better predicted by the sensor data than the 1/5000 map which was based on many more auger observations. Finally an upgraded 1/20,000 soil map was presented, illustrating the potential of combining existing soil maps with proximal soil sensing technology.