Interpreting adjoint and ensemble sensitivity toward the development of optimal observation targeting strategies

Two general methods, adjoint or singular vector methods, and ensemble-based methods, have been previously investigated to identify locations where observations would have a significant positive impact on a numerical weather model forecast. In this paper, we perform a basic comparison of targeting regions chosen to reduce the expected variance of a chosen forecast response function within an ensemble Kalman filter (EnKF) based on both an adjoint method and an ensemble method. Ensemble sensitivity is defined by linear regressions of a chosen forecast response function onto the model initial-time state variables, and is used to calculate variance reduction fields to provide targeting guidance for the ensemble-based method. Adjoint sensitivity is used to provide targeting guidance for the adjoint-based method. 90 ensemble forecasts are considered over a 24-hour forecast period, and the response function is chosen to represent the sea-level pressure at a single point in the Pacifi