Improving Ensemble Reanalyses of the Earth System using a Quantile-Conserving Ensemble Filtering Framework

Scientists at the NSF National Center for Atmospheric Research (NCAR) have produced several reanalyses using versions of atmospheric models that are part of the Community Earth System Model. These include a pair of decade long reanalyses using the standard tropospheric Community Atmosphere Model (CAM) and shorter reanalyses using the CAM model with chemistry (CAM-Chem) and the Whole Atmosphere Community Climate Model with thermosphere and ionosphere extension (WACCM-X). These previous reanalyses were performed using the state-of-the-art Ensemble Adjustment Kalman Filter (EAKF) available in NCAR’s Data Assimilation Research Testbed (DART). The CAM reanalyses have also been used as forcing fields to produce DART reanalyses with other CESM components including the Parallel Ocean Program (POP) and the Community Land Model (CLM).

The most recent release of DART includes a Quantile-Conserving Ensemble Filtering Framework (QCEFF) that implements general but affordable nonlinear and non-Gaussian extensions to the EAKF. While the EAKF fits a normal continuous distribution to the prior ensemble before computing the posterior continuous distribution, the QCEFF allows any distribution family to be used. In particular, nearly non-parametric distributions like the Bounded Normal Rank Histogram distribution can be used when the appropriate distributions are not well-known. This is particularly important for data assimilation for bounded quantities like tracer concentrations but can also lead to significant improvements for fields that are more nearly normal. The QCEFF can also explicitly represent many types of nonlinear relations between observations and state variables leading to improved analysis estimates for all kinds of variables.

Results are presented from applying both the traditional EAKF and the QCEFF to reanalysis applications in the troposphere with CAM and CAM-Chem and to the deep atmosphere with WACCM-X. The results for CAM-Chem include ensemble reanalysis estimates of global CO as well as surface sources for CO. Forecast fits to observations are shown to be consistently better with the QCEFF method, especially for the CO tracers. There are many societal benefits of better estimates of atmospheric chemical and pollutant tracers and their sources, especially under a changing climate. Space weather also has potentially enormous impacts and the WACCM-X reanalysis provides a novel dataset for better understanding predictability and prediction of the upper atmosphere and ionosphere.