|
The message from Steve Silberberg a couple of weeks ago about PBL heights being unreasonable led to a series of investigations regarding both data assimilation and behavior of the 20km RUC model near the surface. The event that he noted with unreasonably large PBL depth in Manitoba was basically from assimilation of METAR temps from grid points with much too cool background (1h fcst) temp, but with bkg values at other nearby grid points being quite reasonable. The result was that since the analysis increment (correction) is smoother than the local variation in sfc temp in the bkg (for various reasons, some mentioned below), then the analyzed value of sfc temp at adjacent grid points could end up being much too high. This resulted in the large PBL heights over MB that Steve S. noted that day.
The code changes noted last week so that the background grid points used for assimilation of sfc obs (land or buoy) would match the land or water land-use were part of the solution. However, we also found that there were abnormally cool/moist soil model points in the 20km RUC that were contributing to this problem. This was traced to some soil parameters for certain soil types, which were subsequently corrected. Two days ago, the soil temp/moisture fields were reset without these problem areas (including western Ontario, other points in Canada, an area in ND, isolated points in MN and WI, and others in the US). Some of you may have previously noticed these cold points in sfc temp (and other fields - CAPE, visibility, etc.) in 20km RUC fcsts before this fix was made.
With this final change, the sfc fields from the 20km RUC have improved substantially in verification against METARs, and they look much more reasonable in these areas. Thanks again to Steve for his posting, and we encourage additional comments on the 20km RUC products from readers of this forum. These observations on the behavior of the 20km RUC can really make a difference for what gets implemented, as was true in this case.
|