Tuesday, 8 January 2019
Hall 4 (Phoenix Convention Center - West and North Buildings)
Proactive QC is a fully flow-dependent quality control scheme that rejects detrimental observations based on EFSO impact in each DA cycle. Ota et al. (2013) and Hotta et al. (2017) found PQC alleviates GFS forecast skill dropout events in non-cycling fashion. Cycling PQC provides addition improvements through accumulating the corrections from the past PQC updates.
In this study, we explore cycling PQC using Lorenz '96 and GFS-T62 models both coupled with LETKF. With Lorenz ’96 model, several PQC update methods were compared, and we found that rejecting detrimental observations using the original Kalman gain performs much better than other methods including the standard data denial method. Also, it is found that rejecting the top 10% of the detrimental observations contributes most of the PQC improvement, and rejecting the rest of the detrimental observations slightly increase the improvement. We will also show the sensitivity of PQC performance to the suboptimal DA configurations and model error in the Lorenz ’96 model. We then confirm that PQC improves the analysis and forecast of the GFS-T62 model globally. Also, it is found that the accumulated PQC improvement dominates over the immediate improvement, meaning that PQC improves the forecast even without correcting the latest analysis and is free of the future data requirement.
- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner