RESEARCH PAPER: Carbon sequestration rates in organic layers of boreal and temperate forest soils - Sweden as a case study

Aim: The aim of this work was to estimate C sequestration rates in the organic matter layer in Swedish forests. Location: The region encompassed the forested area (23 x 10 super(6) ha) of Sweden ranging from about 55 degree N to 69 degree N. Methods: We used the concept of limit values to estimate recalcitrant litter remains, and combined it with amount of litter fall. Four groups of tree species were identified (pine, spruce, birch and 'other deciduous species'). Annual actual evapotranspiration (AET) was estimated for 5 x 5 km grids covering Sweden. For each grid, data of forested area and main species composition were available. The annual input of foliar litter into each grid was calculated using empirical relationships between AET and foliar litter fall in the four groups. Litter input was combined with average limit values for decomposition for the four groups of litter, based on empirical data. Finally, C sequestration rate was calculated using a constant factor of the C concentration in the litter decomposed to the limit value, thus forming soil organic matter (SOM). Results: We obtained a value of 4.8 x 10 super(6) metric tons of C annually sequestered in SOM in soils of mature forests in Sweden, with an average of 180 kg ha super(-1) and a range from 40 to 410 kg ha super(-1). Norway spruce forests accumulated annually an average of 200 kg C ha super(-1). The pine and birch groups had an average of 150 kg ha super(-1) and for the group of other deciduous trees, which is limited to south Sweden, the C sequestration was around 400 kg ha super(-1). Conclusions: There is a clear C sequestration gradient over Sweden with the highest C sequestration in the south-west, mainly corresponding to the gradient in litter fall. The limit-value method appears useful for scaling up to a regional level to describe the C sequestration in SOM. A development of the limit value approach in combination with process-orientated dynamic models may have a predictive value.