Assessing Limitations and Areas of Improvement to an Ensemble-based Observation Impact Estimate Metric

The goal of this project is to develop and enhance an ensemble-based sensitivity calculation within the EnKF which is able to calculate the impact that assimilating different observations have on a forecast. While generally assimilating new observations improves the accuracy of a short-range forecast, determining the exact value added by each observation or observing platform is important in evaluating the true impact. Observation impact depends on the instrument type, observation type, observation locations, as well as the availability of other observations. This sensitivity calculation is a more general approach to a recently-developed and applied sensitivity calculation for the LETKF (Liu and Kalnay 2008; Li et al. 2010; Kunii et al. 2011). This study employs a simple modeling approach, the isentropic primitive two-layer model, to allow for reliable results without complications of large model errors, as has been done in several other EnKF studies. The first goal of this project is to assess in rigorous detail the strengths and weaknesses of this sensitivity technique and determine the limit to which it can be applied, as well as which areas can be modified and improved upon. One main area of testing is associated with sampling errors which cause spurious correlations and possibly spurious sensitivities. The application of the group filter technique developed by Anderson (2007) is evaluated for possible improvements to the ensemble sensitivity technique, with the regression confidence factor used as a mathematical tool to measure the amount of sampling error in the ensembles.