Uncertainty Specifications for GlobalSoilMap (Lincoln, Aug.2012)

Google Earth image of the river Geul valley (the Netherlands) with bubble plot of soil pollution observations. Measurement and interpolation errors cause uncertainty in the resulting soil pollution map.

Any map suffers from measurement, classification, interpolation and generalisation errors, which may limit its usability. The GlobalSoilMap community is engaged in the production of high-resolution soil property maps for the whole world.Hence it is important to quantify the uncertainty associated with the maps produced. In order to get familiar with concepts of error, accuracy, uncertainty and spatial uncertainty, ISRIC provided a one-day course as an introduction to the GlobalSoilMap uncertainty workshop held at the NRCS National Soil Survey Center in Lincoln, Nebraska (USA) from 27 to 30 August, 2012.

The course started with exploring and comparing error, uncertainty and accuracy and explaining how these concepts can be defined in statistical terms by means of probability distributions. Next it extended the probabilistic model of uncertainty to variables that vary in space, in which case use is made of geostatistics. It also looked at how these modelling approaches may be incorporated in digital soil mapping.

The course was attended by about 30 people, mainly from the USA and Australia, but also from Canada, New Zealand and France. It provided a very useful introduction to the main theme of the workshop and helped formulate uncertainty-related details of the specifications of the GlobalSoilMap products.