Soil property maps for Europe at 100 m resolution will inform forest management and policies
Forest soils provide important ecosystem services, such as biodiversity support, nutrient cycling, carbon sequestration, water regulation and erosion control. The sustainability of forest ecosystems depends on the health and functioning of forest soils themselves, making their conservation and management essential for the well-being of both natural ecosystems and human societies. Nonetheless, many forest soils face degradation thereby enhancing climate change and biodiversity loss.
To address these complex issues the EU H2020 HoliSoils project (2021-2025), led by the National Resources Institute Finland (Luke), is furthering research, developing models and tools, and conducting field experiments on forest soils to support holistic management practices and monitoring for European forest soils. Within the project, ISRIC – World Soil Information is focusing its efforts on improving the spatial information on forest soils for Europe.
Collating and standardising forest soil data
To prepare for detailed mapping, forest soil datasets, identified and collated by HoliSoils partners Luke and Thünen Institute, have been cleansed, standardised and ingested in a large relational soil database called WoSIS, as previously described by Niels Batjes – senior soil science expert and Luis Calisto database management expert at ISRIC. This was achieved using refactored “extract transform load” procedures. These quality-assessed data provided the basis for the subsequent soil property mapping activities, as described below.
Developing procedures for mapping forest soil properties
Forest soil data for Europe collected in WoSIS were identified through a GIS-overlay with the EU Corine Land Cover dataset (for 2018), yielding about 28,000 profiles. These training data were used to map soil properties at a resolution of 100 m. To achieve this, the digital soil mapping approach employed featured model tuning, environmental covariate selection and regression random forest (Figure 1), building on procedures developed for SoilGrids250m. Special attention was paid to key soil properties such as pH, organic carbon content, bulk density, and soil texture, as identified earlier by HoliSoils partners.
The above selection of profiles does not provide a balanced overview of forest soils in Europe (about 12.3/1000 km2). Inherently, there are gaps in the geographic distribution of the soil data shared for consideration in WoSIS, and this is partly due to license constraints. Soil properties at “unvisited sites” were predicted from a set of co-variates for Europe representing soil forming factors, such as climate, elevation and parent material, while accounting for the associated uncertainties (Figure 2).
The time demanding calculations ― which included iterations and output evaluation ― were carried out in Wageningen using a high-performance computer. Figure 3 shows the predicted distribution of soil organic carbon in the upper 30 cm of forest soils in Europe as an example; note that the area of forest soils has been “masked” using Corine, hence the somewhat patchy geographic patterns.
Inherently, uncertainties are attached to the predictions; these can be due to errors in the soil data and covariates, and to mistakes associated with the model structure itself. To provide modellers and policy makers with a measure for the “fitness-for-intended-use” of the maps, procedures were developed to quantify this uncertainty (Figure 4).
Evaluation of the predictions
The initial maps were presented by Laura Poggio – senior digital soil mapping and remote sensing expert at ISRIC – during a session on “Digital (soil) mapping as a suitable approach to generate spatial forest site and biodiversity data at different scales” at the 2024 IUFRO conference in Stockholm.
Comparisons with soil organic carbon (SOC) maps generated in the framework of several other EU-scale projects, such as CUP4SOIL and ESA World Soils, indicated that there are still quite some differences in the estimated SOC stocks as well as in their spatial distribution. The work showed the importance of having sufficient quality-assessed harmonised profile data when mapping and modelling across countries at continental scale to support decision making towards climate and sustainability goals. Possibly, data fusion of the different SOC maps could lead to better and more reliable estimates in the future.
“The HoliSoils generated soil property maps can be used for decision making and integrated in vulnerability modelling and climate change scenarios at continental level”, said Laura Poggio, “but the inherent uncertainty associated with the various predictions should always be taken into consideration by the user community”.
How will the maps be distributed by HoliSoils?
In first instance, ISRIC will make the soil property maps freely available (CC-BY-4.0) via simple webservices from where they can be “harvested” by HoliSoils’ partner at Thünen Institute for incorporation and enhanced visualisation and querying facilities through the TISDAR portal.
Long-term preservation of the soil profile data and property maps, as well as possible updates after completion of the HoliSoils project in 2025, will be ensured by ISRIC in its capacity of ICS World Data Centre for Soils.
CONTACT EMAIL:
Niels Batjes [] and Laura Poggio []
RELATED LINKS:
- HoliSoils project: ‘Working together for Forest Soils’
- World Data Centre for Soils
- FAQ – SoilGrids
- FAQ – WoSIS
This article was written by Niels Batjes (ISRIC).
