Data assimilation to leverage diverse datasets for improved CO2 & CH4 flux estimation and future observing system design

The surface and aircraft records from NOAA�??s Global Greenhouse Gas Reference Network (GGGRN) provide a wealth of information about the spatial distributions and trends of CO2 and CH4, but key regions such as the Amazon and the Arctic are not adequately sampled due to expense and logistical challenges. Fortunately, complementary data records are available that will provide strong new constraints on flux estimates and simulated atmospheric transport. Newly available profile data from the Japanese CONTRAIL commercial aircraft sampling program and from the IPEN Brazilian regional aircraft network will provide especially useful constraints on previously under-sampled regions, and ground-based networks have greatly expanded over the past 10 years. Meanwhile, intensive sampling campaigns such as the global-scale HIPPO and ATom experiments and the Arctic-focused CARVE and ABoVE programs provide detailed snapshots of poorly observed regions. Satellite sensors such as GOSAT, OCO-2 and TROPOMI are providing new information in regions where other types of atmospheric data are sparse, but retrievals are complicated and susceptible to systematic errors that can complicate or confound flux estimation. Consequently, the best flux estimates will come from a rigorous combination of a variety of data types, and targeted research is needed to develop objective approaches for appropriately weighting and combining diverse data streams. Methods are needed that reliably reveal errors in simulated atmospheric transport, which can lead to biases in estimated fluxes if not corrected. 1. We will apply a novel data assimilation strategy developed at the University of Toronto that optimizes the 4-dimensional CO2 and CH4 distributions instead of, or together with, the surface fluxes. When sufficiently dense and accurate data are available, simulated atmospheric trace gas distributions can be locally adjusted to mitigate errors in the model transport. We will integrate aspects of this approach into NOAA�??s CarbonTracker. 2. We will incorporate GGGRN and additional in situ data into the Harvard methane inversion framework. This system has already been used to estimate global CH4 emissions and OH fields using GOSAT retrievals. The Harvard effort will complement and extend GMD inhouse work to identify factors responsible for increased CH4 growth over the past decade. 3. We will conduct a limited set of observing system simulation experiments to investigate what optimal combinations of surface, aircraft, and satellite sampling are needed to reliably detect potential future changes in the global carbon cycle such as releases of CO2 and/or CH4 from permafrost thaw or from changes in the uptake of tropical forests. We aim to extract the maximum benefit from the GGGRN dataset by augmenting it with complementary data and using state-of-the science modeling tools to produce optimized estimates of the global CO2 and CH4 distributions and their uncertainties, along with optimized estimates of emissions and removal processes. We expect that the combined dataset will provide sufficient constraints to finally resolve long-standing questions about the relative influences of tropical versus northern land CO2 sinks and the distribution of sinks within the northern extratropics. We will estimate trends in methane emissions and removals, with particular emphasis on anthropogenic emissions in North America and Arctic emissions. Thus, our proposed work will powerfully demonstrate the value of sustained observation and directly responds to the request for proposals focused on methane.