Quantifying and visualizing uncertainties in upscaled regional carbon budgets
Quantifying and diagnosing sources of uncertainty in upscaling CO2 fluxes is critical for the design and monitoring of climate change mitigation and adaptation. Decisions regarding the type of input data (spatial resolution of land cover data, spatial and temporal length of flux data), representation of landscape structure (land use vs. disturbance regime), and type of modeling framework all influence the estimates CO2 fluxes and the associated uncertainties, but are rarely considered together. Here, we present a synthesis of past and present efforts for upscaling CO2 fluxes and associated uncertainties in the ChEAS (Chequamegon Ecosystem Atmosphere Study) region in northern Wisconsin and the Upper Peninsula of Michigan. Previous upscaling work used land cover type to map mean and uncertainty in CO2 flux and quantified uncertainty in CO2 fluxes due to spatial resolution of land cover data.
The current work presents a new modeling approach and quantifies uncertainty in upscaling CO2 fluxes due to tower location, spatial extent of flux data, temporal length of flux data, and landscape disturbance patterns derived from historical Landsat imagery. We analyze whether inclusion of this data constrains uncertainty and improves yearly to decadal terrestrial CO2 flux hindcasts compared to previous efforts. We comment on two key future research needs. First, the characterization of uncertainties due to all of the abovementioned factors reflects only a (hopefully relevant) subset the overall uncertainties. Second, interactions among these factors are likely critical but are poorly represented by the tower network at landscape scales. Yet, results indicate significant spatial and temporal heterogeneity of uncertainty in CO2 fluxes which can inform carbon management efforts and prioritize data needs.