Keith Harris, Anthony O'Hagan and Shaun Quegan
University of Glasgow, UK and University of Sheffield, UK.
Publication details: Submitted to IEEE Geoscience and Remote Sensing Letters, 2009.
A methodology is proposed for consistently including uncertainty in land cover, as described by confusion matrices, with other sources of input uncertainty in quantifying the uncertainty in model-based estimates of carbon fluxes. The approach is illustrated by calculating for each model grid-cell the Bayesian estimate of net biospheric carbon flux, its overall uncertainty and the proportion of uncertainty due to land cover for England and Wales in 2000. The methodology is generally applicable to other models of land surface fluxes, such as those of energy and water, and to other types of environmental calculation where land cover is an input.
Keywords: Carbon cycle, Bayesian statistics, uncertainty, vegetation models, emulator, kriging.