Soil organic carbon (SOC) has been identified as the main global terrestria
l carbon reservoir, but considerable uncertainty remains as to regional SOC
variability and the distribution of C between vegetation and soil. We used
gridded forest soil data (8-km x 8-km) representative of Swiss forests in
terms of climate and forest type distribution to analyse spatial patterns o
f mineral SOC stocks along gradients in the European Alps for the year 1993
. At stand level, mean SOC stocks of 98 t C ha(-1) (N = 168, coefficient of
variation: 70%) were obtained for the entire mineral soil profile, 76 t C
ha(-1) (N = 137, CV: 50%) in 0-30 cm topsoil, and 62 t C ha(-1) (N = 156, C
V: 46%) in 0-20 cm topsoil. Extrapolating to national scale, we calculated
contemporary SOC stocks of 110 Tg C (entire mineral soil, standard error: 6
Tg C), 87 Tg C (0-30 cm topsoil, standard error: 3.5 Tg C) and 70 Tg C (0-
20 cm topsoil, standard error: 2.5 Tg C) for mineral soils of accessible Sw
iss forests (1.1399 Mha). According to our estimate, the 0-20 cm layers of
mineral forest soils in Switzerland store about half of the C sequestered b
y forest trees (136 Tg C) and more than five times more than organic horizo
ns (13.2 Tg C).
At stand level, regression analyses on the entire data set yielded no stron
g climatic or topographic signature for forest SOC stocks in top (0-20 cm)
and entire mineral soils across the Alps, despite the wide range of values
of site parameters. Similarly, geostatistical analyses revealed no clear sp
atial trends for SOC in Switzerland at the scale of sampling. Using subsets
, biotic, abiotic controls and categorial variables (forest type, region) e
xplained nearly 60% of the SOC variability in topsoil mineral layers (0-20
cm) for broadleaf stands (N = 56), but only little of the variability in ne
edleleaf stands (N = 91, R-2 = 0.23 for topsoil layers).
Considerable uncertainties remain in assessments of SOC stocks, due to unqu
antified errors in soil density and rock fraction, lack of data on within-s
ite SOC variability and missing or poorly quantified environmental control
parameters. Considering further spatial SOC variability, replicate pointwis
e soil sampling at 8-km x 8-km resolution without organic horizons will thu
s hardly allow to detect changes in SOC stocks in strongly heterogeneous mo
untain landscapes.