Soil organic carbon-stock changes in Flemish grassland soils from 1990 to 2000
Faculty of Sciences. Bioscience Engineering
Journal of plant nutrition and soil science. - Weinheim
, p. 24-31
Total soil organic-carbon (SOC) stocks for grassland soils in Flanders (N Belgium) were determined for the Kyoto Protocol reference year 1990 and 2000 in order to investigate whether these soils have been CO2 sinks or sources during that period. The stocks were calculated by means of detailed SOC datasets, which were available at the community scale for the whole of Flanders. The total SOC stocks for Flemish grassland soils (1 m depth) were estimated at 38 Mt SOC in 1990 and 34 Mt SOC in 2000. The loss of SOC resulted from a decrease in the SOC content of grassland soils (71%) and could also partly (29%) be explained by a decline in grassland area. Significant decreases in %SOC for the 0-6 cm depth layer were found for the 1990s for the coarser-textured soils with SOC losses ranging between -0.3% and -0.5% over the 10 y period. Specific management practices that disturb the SOC balance such as conversion to temporary grassland and a reduction of animal-manure application are hypothesized to have contributed to the observed loss of SOC stocks. We furthermore conducted an analysis of uncertainty of the 1990 and 2000 grassland SOC-stocks calculation using Monte Carlo analysis. Probability-distribution functions were determined for each of the inputs of the SOC-stock calculation, enabling us to assess the uncertainty on the 1990 and 2000 SOC stocks. The frequency distributions of these simulated stocks both closely approached lognormal distributions, and their 95%-confidence intervals ranged between 150% and 50% of the calculated mean SOC stock. The standard error on the measured decrease in SOC stocks in Flemish grassland soils during the 1990s was calculated to be 7-8 Tg SOC, which is equivalent to twice this decrease. This clearly shows that large-scale changes in SOC stocks are uncertainty-ridden, even when they are based on detailed datasets.