C-Store: An Australian remote-sensing and observation-driven carbon storage assessment system

[摘要] The terrestrial carbon (C) cycle involves the biological capture of atmospheric CO2 through photosynthesis, the temporary storage of C in plant-derived matter, and the eventual re-release of C into the atmosphere. Whilst the C cycle affects almost all terrestrial processes, its dynamics are extremely difficult to measure both spatially and temporally. As a consequence, quantitative estimates of C stores and fluxes are typically made indirectly, via models. This report describes the development and evaluation of C-Store, an Australian remote-sensing and observation-driven carbon assessment system.The desired characteristics that drove the development of C-Store were that the model be national in scope but with a fine spatial resolution, that it be data-driven (especially remotely sensed data), that it be temporally dynamic and produce estimates of model uncertainty. Maximum model simplicity and computational efficiency were also essential. In response, the C-Store system was developed, which is the full assessment mechanism that produces C mass estimates, uncertainty estimates and minimises prediction uncertainty through data assimilation techniques. The system contains the C-Store model, which is a simple, biophysical model that produces dynamic estimates of the C mass contained within 11 separate storages (such as, for example, tree leaves, grass roots, and soil C).The C-Store model uses remotely sensed estimates of tree and grass gross primary productivity (GPP) as the main inputs, as well as temperature and remotely sensed landscape wetness and fire event data. It runs at a monthly time-step, from January 2001 and December 2012, at a 9 second (~250 m) spatial resolution. Estimates of model uncertainty are made within the C-Store system using ensemble model predictions, currently driven by the known uncertainty in the GPP input data.Improvement of the C-Store model is being conducted in three sequential stages: model validation, model calibration, and data assimilation. Each stage is undertaken only after the previous is substantially complete. So calibration is performed only after the model has been rigorously validated, and assimilation is performed only on a well calibrated model. Currently C-Store development is in the validation stage, with model outputs being validated against field-based estimates and with an initial examination of the parameter sensitivities.Validation of C-Store outputs is done using a database of field-based C mass estimates. At this stage of model development, only general agreement between model and field estimates is expected. Compared to the field-based measures, C-Store generally over-estimates above-ground (AG) biomass and AG tree mass, but shows a reasonable agreement with total soil C mass. Continentally, C-Store outputs are on par with other modelled estimates for AG biomass but generally overestimate the total soil C mass.The age, distribution and availability of field data currently prohibit a rigorous assessment of C-Store outputs and place an inherent limit on both the validation and calibration of the model. Due to lack of corresponding field samples some C store outputs remain completely untested (for example, sapwood, grass leaf and grass root C masses). Field-based estimates of C stores themselves have high uncertainties, which further complicates the evaluation of model results.There are four priority areas for the continuing development of C-Store:1) substantial expansion of the field-based C dataset to increase the effectiveness of model validation and calibration, including quantifying observation errors;2) maximising the accuracy, through model calibration, of the predictions of the heartwood, grass leaf, AG litter, and soil C stores (as these arguably have the most important applications);3) expanding error estimation to encompass parameter uncertainties; and4) continuing the progression of C-Store towards full, sequential data assimilation capability.

[发布日期] 2015-08-01 [发布机构] CSIRO

[学科分类] 地球科学(综合)