Efficient Estimation of the Impact of Observing Systems using EFSO

Massive amounts of observations are being assimilated every day into modern Numerical Weather Prediction (NWP) systems. This makes difficult to estimate the impact of a new observing system with Observing System Experiments (OSEs) because there is already so much information provided by existing observations. In addition, the large volume of data also prevents monitoring the impact of each assimilated observation with OSEs. We demonstrate in this study how effectively the use of Ensemble Forecast Sensitivity to Observations (EFSO) can help to monitor and improve the impact of observations on the analyses and forecasts. In the first part, we show how to identify detrimental observations within each observing system using EFSO, which has been termed as Proactive Quality Control (PQC). The withdrawal of these detrimental observations leads to improved analyses and subsequent 5-day forecasts, which also serves as a verification of EFSO. We display the feasibility of PQC towards operational implementation. In the second part, it is found that in the estimated impact of MODIS polar winds, one of the contributors of detrimental observations, a positive u-component of the innovation, is associated with detrimental observations, whereas negative u-innovations are generally associated with beneficial impacts. Other biases associated with height, and other variables when the net impact is detrimental were also found. By contrast, such biases do not appear in systems using similar cloud drift wind algorithm, such as GOES satellite winds. The finding provides guidance towards improving the system and gives a clear example of efficient monitoring observations and testing new observing systems using EFSO. The potential of using EFSO to efficiently improve both observations and analyses is clearly shown in this study.