Sensitivity and climate-driven uncertainty in the Holos model for agricultural greenhouse gas emissions

Holos is the whole-farm greenhouse gas model developed by Agriculture and Agri-Food Canada. For cropping systems, it accounts for direct and indirect nitrous oxide, energy-related carbon dioxide, and change in soil carbon. Neither its sensitivity to individual inputs nor the uncertainty of its emission estimates have been fully quantified. This thesis provides an assessment of Holos modeling results for Ontario cropland in two stages: a screening stage that ranks the model inputs, and a propagation stage that quantifies the uncertainty contributed by the inputs it identifies as dominant. The first stage is a one-factor-at-a-time local sensitivity analysis of a synthetic grain-corn farm, replicated across Soil Landscapes of Canada polygons in 39 Ontario ecodistricts. Twenty-six climate, soil, management, crop, and energy inputs are perturbed individually and ranked by elasticity and by nominal-range sensitivity. Potential evapotranspiration ranked first in every ecodistrict and nitrogen fertilizer rate second under both measures. The propagation stage therefore treats only the climate inputs as random. The second stage fits a trivariate normal model of monthly temperature, precipitation, and potential evapotranspiration for two ecodistricts with contrasting climate regimes, and propagates 500 Monte Carlo climate scenarios through Holos. Because soil texture emerges as a source of variation comparable to climate, the two effects are separated by refitting at a texture-matched polygon pair, where a larger simulation of 5,000 scenarios supports a degree-two polynomial chaos decomposition of the remaining climate-driven uncertainty. Two of the four emission components are unresponsive to climate: energy-related carbon dioxide is computed independently of climate inputs in Holos, and the reported soil carbon change is likewise insensitive by construction. The whole of the propagated uncertainty therefore falls on the two nitrous oxide components, and mainly on the direct one, with temperature the leading driver of direct nitrous oxide and of total emissions. Weather variation alone can thus shift a Holos estimate substantially, which places the nitrogen cycle at the center of agricultural emissions budgeting and of any claim that a mitigation practice has worked.