Rapid Refresh (RAP) and High-Resolution Rapid Refresh (HRRR) Forum
changes to 3DVAR made today
Posted By: Geoff Manikin
Date: 24-September-03 1818Z
A couple of modifications to the 3DVAR initialization were implemented at 15Z today (24 Sep):
1) Data from land observational sites located at model water grid points will now be flagged. An example of problems associated with this data being assimilated by the model is shown at http://wwwt.emc.ncep.noaa.gov/mmb/gmanikin/ruccase/ruc_2mtcheck.gif In this example, data is available at Hougton, Michigan (located on the Keweenaw Peninsula, extending from the Upper Peninsula into Lake Superior). This peninsula is not resolved by the RUC, so the first guess for that location is colder and drier (consistent with a water point in that region) than the observation. A positive increment (difference between observation and first guess) is therefore created by the analysis code and is spread inland, causing unreasonably high temperature and especially dew point values inland over the U.P. It was determined that data from approximately 30 such land stations located at model water grid points will need to be flagged.
2) The second change is to tighten the thresholds in the quality control checks. Some bad data, particularly moisture observations, has been getting into the analysis, and limiting the permissible difference between the observed value and first guess will reduce the frequency of occurrence.
Rapid Refresh (RAP) and High-Resolution Rapid Refresh (HRRR) Forum is maintained by RR/RAP Forum Administrators with WebBBS 5.12.