Welcome, Guest
Username: Password: Remember me

Reply Topic: Abnormal HIRLAM control members in recent GLAMEPS

Enlarge /  Shrink
Enter code here

Topic History of: Abnormal HIRLAM control members in recent GLAMEPS

Max. showing the last 6 posts - (Last post first)
7 years 1 week ago #1313

Xiaohua Yang

Xiaohua Yang's Avatar

It was lately observed peculiar features by met.no duty forecasters that somehow the HIRLAM control members show abnormal behavoir both as seen from MSLP forecast chart and also in terms of variability between consecutive cycles.

We've now identified two major problems in the GLAMEPS scripts, one on filtering of ECMWF bufr data which was implemented since early December in association with changes of the ECMWF observation data flow. The deficiency effectively disabled data assimilation for the HIRLAM control members. This caused gradual deteriation of the HIRLAM control forecasts in the recent period. Ironically, a second script deficiency had resulted in unintended use of purturbed ECMWF analysis instead of HIRLAM analysis for the purturbed HIRLAM components, which has thus made the above mentioned first problem largely invisible. The second script error enabled the GLAMEPS probabilistic forecast to stay roughly normal despite of the problem in the control forecast, by mistake!

We have now corrected preliminary the first error which hopefully will return the control member of the HIRLAM components back to normal starting from the 18 UTC run as of Jan 7 2015. We will implement the remaining needed correction later.

While the above configuration problem did have major impact on control member of the HIRLAM components (STRACO and KF schemes), the overall impact on GLAMEPS probabilistic forecast is estimated to be minor.

Acknowledgement goes to met.no duty forecasters for detecting the anomaly. Inger-lise and Kai have both been involved detection of the problem and follow-up solutions.
Time to create page: 0.060 seconds