Factors affecting the uncertainty of sinks and stocks of carbon in Finnish forests soils and vegetation

Abstract

Monitoring and transparent reporting of forest carbon sinks are currently needed under the Climate Convention. From 2005 onwards, national GHG inventories should also provide uncertainty estimates of the reported emissions and removals. Comprehensive uncertainty analysis and key category analysis of the carbon inventory can provide guidance for prioritizing efforts in further development of the inventory. In this study, the estimates of the forest carbon stock and carbon sink were obtained by combining forest inventory data, models of biomass and turnover, and a dynamic decomposition model for SOM and litter, Yasso. To study the decisive factors affecting uncertainties of forest carbon sink and stock estimates, we conducted a Monte Carlo analysis for the calculation of the forest carbon budget of Finnish forests for the period 1989–2004. Uncertainty of the vegetation carbon sink was affected mostly by input data on growth variation and drain. Uncertainty of the soil carbon sink was dominated by the soil model initialization, but the effect decreased with time. After few years, the effect of initialization leveled with the effect of temperature and drain, both of which were given as input data to the system and which varied inter-annually. The contribution of these variables was less important to uncertainty of stocks in vegetation and soil than the contribution of model parameters. The most influential parameters for vegetation C stock were carbon density and conversion factors for tree and ground vegetation biomass, and for soil C stock, they were soil model parameters, and biomass conversion factors and turnover rates of fine roots and ground vegetation. After short initialization period for soil C, uncertainty of soil sink can be reduced by improving the precision of input data (harvests on upland soils, annual temperature). Precision of vegetation sink can be improved mainly by improving the quality of input data on growth variation and harvests. There is an unknown error source related to inter-annual variability of the forest ecosystems, which cannot be represented with the system. Vegetation sink was compiled with biomass models that are based on long-term averages and they do not support year-to-year variations which may occur in forest ecosystems. Averaged biomass models with averaged turnover models, produce highly auto-correlated series of litter input, which result in relatively precise annual soil sink estimates. Due to these reasons, the current inventory-based approach is more justified for the estimation of average sinks for longer periods than 1 year.