Uncertainty of Carbon Changes for Agricultural Land in Canada.

Uncertainties are required for reporting national greenhouse gas emissions, including changes in carbon stocks on cropland and grassland, under the United Nations Framework Convention on Climate Change and the Kyoto Protocol.  Uncertainty analysis of soil C change is relatively more complicated than that for other agricultural greenhouse gas emissions (CH₄ and N₂O) because C change is affected by changes in land use and agricultural activities over many decades prior to a particular inventory year whereas uncertainty of emissions of CH₄ and N₂O are mostly affected by agricultural activities in the inventory year.   Uncertainty of C stock changes for agricultural land in Canada involves about one million uncertain quantities and was estimated using error propagation methods largely assuming negligible spatial, temporal, or relational covariances among uncertain quantities.  The 95% confidence interval for the C change on cropland in Canada in 2005 was a sink of 9.3 to 14.3 Pg.  The effect of different types of probability distributions of values of uncertain quantities and of significantly non-zero covariances was tested using subsets of data with Monte Carlo simulation.  The latter indicated that most plausible probability distributions and covariances do not affect the final uncertainty greatly.