Soil maps of all scales are needed to help solve global issues such as food security, climate change and water shortages. Yet, nearly all soil maps are inaccurate, which means that decisions regarding agriculture and nature management based on these maps may be incorrect. In his inaugural lecture at Wageningen University & Research, Professor Gerard Heuvelink, Professor by special appointment of Pedometrics and Digital Soil Mapping, explains the importance of the quantification of the accuracy of soil maps. His chair is financed by ISRIC – World Soil Information.
Global soil maps provide important information for studies investigating the increase of food production and for implementing measures against climate change and environmental degradation, such as erosion, loss of biodiversity and the increasing scarcity of clean drinking water. According to Professor Gerard Heuvelink, it therefore comes as no surprise that soil information forms part of eight of the seventeen Sustainable Development Goals of the United Nations. “Hydrologists use soil maps to determine the infiltration of rainwater, agronomists use soil information to formulate a prediction for the potential yield of farmland, ecologists use soil physical and chemical information to determine and interpret the soil biodiversity of an area. These are all soil-based services that might hit a limit in the future if soil maps do not have the required level of accuracy.”
Five years ago, ISRIC developed SoilGrids to address the global need for soil information. SoilGrids is a system that uses existing soil profiles and smart algorithms to map soil type and soil characteristics across the world. “Now we can make global predictions with a 250-metre resolution about the carbon content of the soil, the clay content and the porosity or the soil water storage capacity, which is particularly important for flood studies and crop production,” explains Professor Heuvelink.
Limited accuracy of maps
Despite their importance, the accuracy of many soil maps leaves much to be desired. For example, there might be an enormous discrepancy between the percentages of organic carbon in the soil on two different maps of the same area, says Professor Heuvelink in his inaugural lecture titled On the quality of soil information. How pedometrics can help us. “Such inaccuracies are then propagated in the calculations on which governmental bodies, such as the Intergovernmental Panel on Climate Change (IPCC), base models for environmental issues, such as climate change. And that can lead to incorrect decisions being made,” warns Professor Heuvelink.
As a result, he focuses his research on mathematical and statistical methods to quantify and reduce the uncertainties of soil data and soil maps, which can be summed up under the term ‘pedometrics’ (pedon is the Greek word for ‘soil’). “The most obvious way to determine uncertainties in soil maps would be to compare them with independent soil samples. While these are valuable validation metricsthey still tell us nothing about how accurate the soil prediction for each location in the area is. In pedometrics we therefore always indicate the uncertainty in soil mapping. So you actually make several maps of an area. The organic matter content of the topsoil in Tanzania that we recently mapped serves as an example of this. In addition to a map of the predicted value, we also made maps of the lower and upper limits of a 90% confidence interval of the organic matter content. We can then say with 90% certainty that the actual value is between these two limits,” explains Professor Heuvelink.
“The quantification of uncertainties helps us to further improve soil maps, as these analyses can also identify the weakest link in the mapping process. For example, it allows us to determine how much improvement can be expected from more intensive use of satellite images and ‘crowd-sourced’ soil data. I will continue to work with my team in the coming years to develop tools to determine the accuracy of soil maps and to continue to improve them,” concludes Professor Heuvelink.