Errors in soil maps: The need for better on-site estimates and soil map predictions

Authors: BUENEMANN, MICHAELA - New Mexico State University; COETZEE, MARINA - Non ARS Employee; KUTUAHUPIRA, JOSEPHAT - Non ARS Employee; MAYNARD, JONOTHAN - University Of Colorado; Herrick, Jeffrey - Jeff

Submitted to: PLOS ONE
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 6/10/2022
Publication Date: 1/11/2023
Citation: Buenemann, M., Coetzee, M.E., Kutuahupira, J., Maynard, J.J., Herrick, J.E. 2023. Errors in soil maps: The need for better on-site estimates and soil map predictions. PLOS ONE. 18(1). Article e0270176. https://doi.org/10.1371/journal.pone.0270176.
DOI: https://doi.org/10.1371/journal.pone.0270176

Interpretive Summary: Natural resource managers, including farmers and land use planners, all need soil information to make good decisions. However, the accuracy of soil map information not only varies widely but is often unknown. To address this issue, we used field soil data to assess the accuracy of seven spatial soil databases (Digital Soil Map of the World, Namibian Soil and Terrain Digital Database, Soil and Terrain Database for Southern Africa, Harmonized World Soil Database, SoilGrids1km, SoilGrids250m, and World Inventory of Soil Property Estimates) using topsoil texture as an example soil property and Namibia as a case study area. In addition, we visually compared topsoil texture maps derived from these databases. We found that the maps showed the correct topsoil texture in only 13% to 42% of all test sites. These and other analyses led to three conclusions. First, existing soil maps are often insufficient to support local land management and problematic for the use in global models of climate change, biodiversity, and ecosystem services. Second, to address many of today’s issues in a sensible manner, we need soil data that are up-to-date, sufficiently accurate for their intended purpose, associated with uncertainty information, three-dimensional, available at multiple spatial resolutions, spatio-temporally explicit and continuous, affordable, easily integrated with other digital spatial data, and readily available for interested stakeholders. Third, to generate soil data that meet the requirements of the diverse stakeholders in soil resources research and management, we need to improve on-site estimates and map predictions of soil properties.

Technical Abstract: High-quality soil maps are urgently needed by diverse stakeholders, but errors and uncertainties in existing soil maps are often unknown, particularly in countries with limited soil surveys. To address this issue, we used field soil data to assess the accuracy of seven spatial soil databases (Digital Soil Map of the World, Namibian Soil and Terrain Digital Database, Soil and Terrain Database for Southern Africa, Harmonized World Soil Database, SoilGrids1km, SoilGrids250m, and World Inventory of Soil Property Estimates) using topsoil texture as an example soil property and Namibia as a case study area. In addition, we visually compared topsoil texture maps derived from these databases. We found that the maps showed the correct topsoil texture in only 13% to 42% of all test sites, with substantial confusion occurring among all texture categories, not just those in close proximity in the soil texture triangle. Visual comparisons of the maps moreover showed that the maps differ greatly with respect to the number, types, and spatial distribution of texture classes. The topsoil texture information provided by the maps is thus sufficiently inaccurate that it would result in significant errors in a number of applications, including irrigation system design and predictions of potential forage and crop productivity, water runoff, and soil erosion. Clearly, the use of these existing maps for policy- and decision-making is highly questionable and there is a critical need for better on-site estimates and soil map predictions. We propos that mobile apps, citizen science, and crowdsourcing can help meet this need.