Soil data harmonisation and geostatistical modelling efforts in support of improved studies of global sustainability

Future Earth and other large international research and development programmes aim to provide the scientific evidence base required for developing into a sustainable future. Soil, which is an important provider of ecosystem services, remains one of the least developed data layers in global land models and uncertainties are large. In this context, there is a pressing need for improved, qualityassessed soil information at multiple scale levels. ISRIC – World Soil Information, in its capacity of World Data Centre for Soils within the ICSU World Data System, is developing inter-operable web-based facilities aimed at facilitating collaborative soil mapping. The Global Soil Information Facilities (GSIF) provide a global spatial framework for collating, standardising resp. harmonising, and analysing soil data profile obtained from disparate sources. At present, the facility includes a 3D soil information services for the world at 1 km resolution (SoilGrids1km), which draws on analytical data for some 100,000 soil profiles and over 70 co-variate layers representing soil-forming factors. Global regression models were used to predict values (mean and 90%-confidence interval) for selected soil attributes (e.g. soil pH, clay content, bulk density, and organic carbon content) for six depth intervals up to a depth of 2 meter. Cross-validation for the initial run showed prediction accuracies of 23%-51%, which is promising. Being based on reproducible automated procedures, the geo-statistical predictions are improved on a regular basis. New releases will consider a larger complement of harmonised soil profiles for the World, as collated and shared for example within the broader collaborative framework of the Global Soil Partnership (GSP), as well as more advanced geo-statistical approaches that may be targeted at specific agro-ecological regions. Confidence limits generated by the SoilGrids model may be used to assess the impact of uncertainty in soil property predictions (means) during scenario/model testing — data are freely available for visualization and download at http://soilgrids.org. The SoilGrids procedure has already been applied at various resolutions, depending on specified user needs. For example, a 250m product in support of agricultural planning in Africa versus a 50 km (or 0.5 by 0.5 arc degree) product for Global Land Models that underpin IPCC-related assessments. Further, development of the overall system is already catalysing institutional collaboration and data sharing. Capacity building and collaboration with (inter)national soil institutes around the world on data collection and sharing, data screening and harmonisation, mapping and the subsequent dissemination of the derived information will be essential to create ownership of the newly derived soil information as well as to create the necessary expertise and capacity to further develop and test the system worldwide. The system can also be used as the basis for a distributed system, where national soil institutes build and provide standardised databases and digital soil maps for their respective regions, which can then be ‘combined’ with the SoilGrids-derived information to arrive at a product with global coverage and local ownership, possibly