Soil cores taken from each subplot of each plot (i.e. 9 per plot), if possible down to 40 cm (Poland), but in other countries it was only possible to 20 cm (except Spain only 10 cm). Forest floor samples were taken separately from the soil cores. The cores were dividied into layers of 10 cm, and respective layers of each plot were pooled together, rendering one forest floor samples and 1 to 4 mineral samples per plot. Living fine roots (< 2 mm) were extrtacted from these samples, and classified as tree or understorey roots. After drying at 40 degrees the roots were weighed and milled. Subsequently the samples were analysed with a FTIR Spectrometer (Shimadzu IR Prestige-21, Kyoto, Japan), equipped with an integrating sphere (NIR IntegratIR, PIKE Technologies, WI, USA). Each spectrum was recorded as the average of 64 scans over the range between 3800 and 10000 cm-1 (i.e. 1000–2600 nm) with a resolution of 8 cm-1. Three spectra were taken from each sample and averaged before further statistical analyses. About 30% of the samples were left for external validation. Data analyses of the NIRS spectra were performed by partial least squares (PLS) regression using The Unscrambler X (version 10.1, Camo Software AS, Oslo, Norway), using PLS1 and PLS2. Mathematical pretreatments comprised , 1/x weighting of the original x-data, first derivative of Savitzky-Golay smoothing, mean and peak normalisation, as well as multiplicatice scatter correction (MSC) and standard normal variate (SNV). The best models for each species were selected based on lowest root mean square errors (RMSE) of both regression and validation, and the highest R2 values.

Samples are identical with the fine root biomass samples, which were collected from all FunDiv plots.

Samples taken between May 2012 and October 2012