Correction to the  Parallel SREF Physics Diversity Eta BMJ-SAT and KF-DET memberS

Brad Ferrier, Jun Du, Geoff Manikin, Binbin Zhou and Jeff McQueen

Implemented in Parallel runs on : June 10, 2004, 21 Z

Summary of change:

A bug was found and corrected in the two BMJ-SAT members of the SREF system which caused strong overprediction of precipitation and occasional execution failures.  In addition, an improvement is proposed for the two  KF-DET members to reduce a precip bias noted during the parallel evaluation.

 Description of Problem .

  1. A logical flag used to define the experimental Betts-Miller-Janjic (BMJ) convective scheme using moistened reference profiles (referred to as BMJ_SAT) was not passed into the Eta driver, resulting in convection not being called.  This resulted in strong overprediction of precipitation and occasional execution failures of the 2 BMJ-SAT members. The impact of the change will be positive as a result of stabilization by convection occurring in the revised runs.  The change will be manifested in improved QPF with fewer grid-point bombs and spurious precipitation maxima, better pre-convective forecast soundings, and overall model performance.  The reference profiles are not saturated but rather are half way between saturation and the moisture profiles currently used in the operational Eta.
  2. Slight modifications to the Kain-Fritsch (KF) convective scheme will implement the change intended to be made in April, in that half of the rain and snow in the convective updrafts are detrained back onto the grid with the remaining convective precipitation reaching the surface.  The impact of this "half detrainment" change is to produce some convective precipitation, replacing the "full detrainment" version where no convective precipitation is produced.  This change is also expected to reduce both the low bias in precipitation, and the cool, moist bias that has formed in association with these two Eta members.

Analysis of Change

Below are a series of SREF member ETA-BMJ-SAT and KF-DET runs using the same (perturbed) ICs and LBCs obtained from the bred n4 (SAT) run that failed on 20040602 at 21 Z..  For reference, here is the 24-h CPC/RFC verification.


Below is a table of images showing 24-h total (2nd column), convective (3rd), and grid scale (last) precipitation for each run (described in 1st column).  You can see from this table the following:

  1. The total absence of convective precip in the original SAT run (1st row), which then shows up in the corrected versions (rows 2-4).
  2. The sensitivity of the (corrected, modified) SAT runs to the different moisture profiles (rows 2-4).  The largest area of precip is predicted using the operational moisture profiles (2nd row), progressively shrinking with more grid-scale and less convective precip as the reference profiles moisten (3rd, 4th rows; saturated profiles assumed in 4th row).
  3. A drawback in assuming moister profiles, which I only realize now, is that it exacerbates the low bias in precip out west.
  4. The 5th row shows the impact of the current KF detrainment run, showing no convective precip and only grid-scale precip, even though we intended to have only half of the convective rainfall detrained onto the grid scale and the rest fall to the surface as rain.  The last row shows that this was fixed, in which most of the precip is still convective despite having half of the condensate in the updrafts detrained back onto the grid.


Total Precip (15-39 h)
Conv Precip (15-39 h)
Grid Precip (15-39 h)
Original SAT
sattp satcp satgp
Corrected SAT,
operational moisture profiles
csattp csatcp csatgp
Corrected SAT,
moistened moisture profiles (half way to saturation)
chsattp chsatcp chsatgp
Original DET,
full detrainment in KF 
(what we intended to be "half detrainment")
kfdtp kfdcp kfdgp
Modified DET, 
half detrainment in KF
hkfdtp hkfdcp hkfdgp
Technical contact on These Changes:
Brad Ferrier < >
EMC/Mesoscale Modeling Branch