Data Assimilation in the NCAR Community Atmosphere Model

The Community Atmosphere Model (CAM) has several well known biases that have important dynamical implications for research. Some examples of these biases include a cold temperature bias near the tropopause at high latitudes, systematic biases in the mean zonal wind, and transient momentum fluxes that are too strong. One method for investigating the impacts that these biases have is to use data assimilation techniques to reduce the bias and observe how the system reacts. The cold polar tropopause temperature bias found in CAM is common to climate models. For the Northern Hemisphere, the bias is generally observed in the summer months. By assimilating a July 2003 set of observations in the region of this bias, we observe that reducing this bias forces the zonal jet and storm tracks to shift poleward. This finding mirrors the results of several earlier studies where the height of the polar tropopause influences the location of storm tracks and zonal jets by moving them equatorward as the height decreases, and poleward as the height increases as is predicted to occur with increased atmospheric concentrations of greenhouse gasses. With the recent CHAMP/COSMIC missions, GPS radio occultation observations have become available. These observations have many desirable qualities including their global nature and accuracy in polar regions where quality observations are rare. In this research, we determine how strongly these observations can constrain CAM through data assimilation. We also begin to quantify the process by which these assimilated observations impact the model. We use an ensemble adjustment Kalman filter (EAKF) provided by the Data Assimilation Research Testbed (DART) to assimilate COSMIC data into CAM for January 2007. We compare a case where only COSMIC observations are assimilated into CAM with cases where we assimilate observations from multiple sources. We find that GPS observations do improve our analysis. Further, when taken by themselves, the assimilation of GPS observations is capable of constraining CAM to a substantial degree. However, we find that water vapor, which is has a strong relationship to refractivity is poorly constrained and assimilated, while dynamical quantities are more accurately assimilated. Finally, we take a detailed look into the mechanism by which GPS observations impact the state of CAM’s modeled atmosphere. By linearizing the refractivity forward operator, we use multivariate linear regression to piece apart the components of the operator. Through this process we determine how the assimilation of a GPS observation will influence the state of our system differently depending on its location in Earth’s atmosphere. In particular large cancellations due to correlated temperature and pressure variations in the midtroposphere of the midlatitudes seem to reduce the potential impact of assimilated GPS observations in this location.