Wednesday, 1 August 2001
Quantifying the impact of observations using ensembles
Suppose an agency (the National Hurricane Center, a private corporation, etc.) wanted to take extra observations to improve a forecast. The ensemble transform Kalman filter (ET KF) can be used to make a prediction of the impact of these observations. If the ET KF could be statistically corrected to make a quantitatively accurate prediction of the forecast improvement, those charged with deciding whether or not to make the flight would be able to perform a cost benefit analysis of sending out the plane (or other data collection device). Not only would they know the best place to fly, they would know whether or not it would be worth it to make the flight, if the improvement would be great or not.
Results will be presented from experiments on a barotropic vorticity model. For each analysis cycle, the ET KF will be used to predict the expected reduction in global vorticity error as well as the optimal location for taking two additional observations. The impact these observations make at both the analysis time and at the forecast time will then be compared with the predicted reduction in error variance, with the belief that a linear correlation between them exists. Having formed a statistical correlation, the ET KF predicted signal will be adjusted, and then compared to the actual reduction in error to see if the ET KF can be used to predict the impact of observations.