Combining long-term ICP Forests monitoring data with the Yasso carbon cycling model at the European scale

Forests form a major organic carbon reservoir, both above- and belowground. In the course of global change, predicting possible changes in these carbon reservoirs is essential. To this end, the Horizon Europe PathFinder project aims to develop an innovative forest monitoring system allowing consistent EU greenhouse gas reporting of LULUCF (Land Use, Land Use Change & Forestry) in combination with advanced policy pathway assessments. Greenhouse gas reporting of soil organic carbon (SOC) stock changes in forests commonly relies on simulations by soil carbon cycling models, such as Yasso (Y20), which uses only climate data and soil carbon inputs that can be derived by country-specific approaches from National Forest Inventories. However, the agreement between measured versus simulated carbon stocks and changes at the European scale has not yet been established. Within the framework of this project, this study aims to derive European-wide harmonised soil carbon inputs and stock estimates since the 1990s and further develop the current estimation methodology. After exploration of the available data sets, the ICP Forests Level II forest condition monitoring database was found the most suitable to set the initial modelling conditions. It is the only harmonised data set at the European scale that comprises above- and belowground compartments and contains repeated assessments on a subset of about 200 plots across Europe. The pre-processing of the observed data on soil carbon stock, growth and litterfall from the central ICP Forests database was very labour-intensive. As part of the ICP Forests monitoring programme, carbon concentrations and bulk densities are measured down to a depth of 80 cm. Using mass-preserving splines, soil carbon stocks were estimated down to a depth of 100 m to make them comparable with Y20. Regression models were developed to estimate litterfall inputs based on forest inventory data. We simulated SOC stocks by Y20 in ICP Forests Level II plots with available stand inventory data and soil characterization. Soil carbon inputs were obtained using two approaches: an inventory approach, with litterfall estimated by the above-mentioned regression models, and root and coarse-woody inputs by allometric functions, and a satellite approach, with net primary production (NPP) from MODIS at 500 m resolution. The Y20-simulated SOC stocks were compared with the SOC stocks to 100 cm depth based on the soil inventory data. an inventory approach, with litterfall estimated by the above-mentioned regression models, and root and coarse-woody inputs by allometric functions, and a satellite approach, with net primary production (NPP) from MODIS at 500 m resolution. The Y20-simulated SOC stocks were compared with the SOC stocks to 100 cm depth based on the soil inventory data. On average, the satellite approach estimated higher soil carbon inputs than the inventory approach (+20%). The SOC stocks simulated by Y20 were overall in line with observed SOC stocks. The simulations for broadleaf-dominated stands agreed well with SOC measurements, with average deviations below 1 kg C m -2 using the satellite approach. In coniferous stands, Y20-simulated SOC stocks were lower than observed by 3-5 kg C m -2 . This is likely due to the intrinsic soil properties driving SOC storage and stabilization in highly acidic, coniferous forests (i.e. Podzols and Umbrisols), which are not accounted for in Y20.