Optimizing the design of the Arctic GHG monitoring tower network using synthetic data experiments.

Climate change is affecting the High Northern latitudes severely. Many drastic changes are expected to the environment in these regions e.g. increases in tundra fire, greening of the tundra, methane release from degrading Yedoma. A key method of monitoring the carbon cycle is through atmospheric tall tower observations, which have large fields of views covering extensive areas. In this study we perform a series of case studies to assess the networks capabilities to detect signals associated with expected disturbance processes.

This research follows a 4-step approach: transporting known flux estimates in a 4D atmospheric transport model, generation of synthetic observations, signal detection, and inverse modeling. Synthetic observations as well as the baseline nature runs are produced through the Goddard Earth Observing System, Version 5 model (GEOS 5). Noise is added to synthetic observations within the acceptable measuring range for each GHG species as set by WMO.

Signal detection functions by several metrics where pairwise t-tests are conducted on temporally binned data to assess the networks and individual towers ability to differentiate between the baseline run and prescribed scenarios at varying levels of signal amplification. Here we are able to ascertain lower detection limits both during the peak perturbation as well as the decreasing ability to detect these over time and with increased distance from the source. For one of the scenarios we tested (abrupt thaw in ice-rich permafrost) we find that for 10% of the network to detect an flux amplification of 12% is required, whereas for 90% of the network to detect this a 103% percent increase is required. A year after the perturbation these values rise to 100% and 573% increase.

Finally, during the inversion step the synthetic observations are run through the Jena inversion system (http://www.bgc-jena.mpg.de/CarboScope/). The inverse flux estimates are compared to the GEOS fluxes both in magnitude and spatial extent to identify the network's capabilities and make recommendations on network improvements.