Manuel Pondeca, John Derber, Jacob Carley, Jim Purser, Steve Levine, Runhua Yang, David Parrish
The NOAA Real-Time Mesoscale Analysis (RTMA) creates analyses of near-surface sensible weather elements and cloud fields on domains that match those of the National Digital Forecast Database. An important drive guiding the development of the RTMA towards a gold standard Analysis of Record is the continued feedback from the NWS forecast community on the quality of the analysis. This work addresses forecasters concerns regarding the unacceptably warm RTMA 2m-temperatures often seen in valley cold pool events, especially in the mountainous western CONUS.
The RTMA uses a one-hour model forecast --currently from the HRRR, but previously from the RUC and RAP-- as the background for its 2DVar analysis. During valley cold pool events, it is common to see background temperatures that are much warmer than the observations, at times by as much as 20 C. A recent RTMA upgrade that relaxes the so-called observation gross-error check in and around valleys and implements a buddy-check quality control now ensures that the cold-pool resolving observations are no longer erroneously flagged. However, one is still faced with the fact that the analysis is not drawing close enough to those observations. The deviation of the background from the observations is simply too large for the current RTMA configuration to overcome. This work seeks to improve the fit of the analysis to the observations by replacing the current constant background error variances with ones that are modelled by a constructed valley map and a functional of the standard-deviation map of the background field itself. By design, the background error variances in and around valleys are significantly increased to force the analysis to weigh more towards the observations in those areas. This change can be regarded as a natural extension of the RTMA background error model, which uses structure functions that follow the terrain to a controlled degree.
Results for the 2m-temperature will be presented at the meeting, and the planned extension to the analysis of the other RTMA parameters will discussed. The beneficial impact of the change on the Lanczos-based estimate of the RTMA analysis uncertainty will also be discussed.