Observation and Diagnostics

p1-20 January 21 15:10-16:10

On the Properties of Ensemble Forecast Sensitivity to Observations

S. Kotsuki (RIKEN), K. Kurosawa (RIKEN), T. Miyoshi (RIKEN)

Abstract

Forecast sensitivity to observations (FSO) is an important method to evaluate the impact of assimilated observations on forecast skill in numerical weather prediction (NWP). This study investigates the sensitivity of FSO impact estimates to the choice of the verification reference. We implemented the ensemble-based forecast sensitivity to observation (EFSO) method proposed by Kalnay et al. (2012) with a global atmospheric data assimilation system NICAM-LETKF (Terasaki et al. 2015), which comprises the Nonhydrostatic ICosahedral Atmospheric Model (NICAM) and Local Ensemble Transform Kalman Filter (LETKF). We evaluate the impact of observations with the moist total energy norm verified against the NICAM-LETKF’s own analysis and the ERA Interim reanalysis. In addition, we implemented an observation-based verification metric.

The results suggest that the observational impact be overestimated in 6-h forecasts if the NICAM-LETKF analysis is used for the verification reference. However, no overestimation is observed if we use the ERA-Interim reanalysis and radiosonde observations for the reference. The results imply that the impact of observations at the analysis time would persist in the analysis 6-h later. In the observation-based verification metric, each type of observations mainly contributes to the improvement of the observed variable. This poster also includes additional properties on EFSO such as the treatment of covariance relaxation.

Contact information

Dr. Shunji Kotsuki RIKEN Center for Computational Science