Improving Cloud Water Initialization with DFI

Using the cloud water and cloud ice from the previous day's 24-hour forecast as initialization and using the new digital filter initialization (DFI), we can improve the cloud-water initialization and spin up.
• This loop compares the initial fields of the clouds in the lowest 3000 feet of the model using the DFI approach and in a standard cold-start model.
• This loop compares the lowest clouds at forecast hour 6.
• This loop compares the lowest clouds at forecast hour 12.

The Effect of PBL Schemes on Low Clouds

• Though PBL schemes can improve the low-clouds, none of the new PBL schemes is a huge improvement. The older ACM2 scheme produces the most clouds as can be seen in this loop at 12 UTC 20 July 2009 (24-hour fcst) and this loop at 0 UTC 21 July 2009 (36-hour fcst). Compare those forecasts to these satellite images. or these satellite images. Note the clouds extending up to Astoria.
• For the 2008091800 case, an extensive marine layer covered the interior of Western Washington, as can be seen in this satellite loop and this 1-km satellite loop. Again, the ACM2 (Pleim) scheme does the best job of representing this layer. Pay particular attention to the clouds that should be present in the region from 30N-45N and 140W-150W in these loops comparing the forecasts at 00 UTC 19 September 2008 and 12 UTC 19 September 2008. Here too is a 12-km loop valid at 00 UTC 19 September 2008 showing that none of the PBL schemes is able to produce enough low clouds on its own.
• Temperature statistics for 2-month tests in 2 seasons are in progress.
• Conclusion: The ACM2 (Pleim) PBL scheme increases low clouds over the ocean. Verification statistics should be gathered for it for 2 months and 2 seasons.

The Effect of PBL Schemes on PBL Heights for these Low Cloud Cases

• The ACM2 has lower PBL heights than the YSU. The QNSE scheme gives higher PBL heights.
• Compare PBL heights for the 2008091800 case with this loop valid at 00 UTC 19 September 2008 and this loop valid at 12 UTC 19 September 2008.
• Compare PBL heights for the 2009071912 case with this loop valid at 00 UTC and this loop valid at 12 UTC.
• Question: Are the ACM2 PBL heights too low?

Some PBL Scheme Statistics July-August 2008

 00 UTC (5 pm PDT) 12-km verification statistics for SA stations
Field            ME   MAE   RMS   STDEV  COUNT  PBL Scheme
dew2m           0.57  4.80  6.43  6.40   17408  YSU w/new nudging
dew2m           0.55  4.81  6.44  6.42   17408  YSU
dew2m           1.88  5.33  7.03  6.77   17408  ACM2 w/new nudging
dew2m           1.88  5.34  7.04  6.79   17408  ACM2
dew2m           2.27  5.37  7.12  6.74   17408  QNSE
dew2m           8.66  9.29 11.83  8.06   17408  YSU without LSM

dir10m          2.93 45.19 62.61 62.54   17127  YSU without LSM
dir10m          3.45 45.52 62.90 62.81   17127  YSU
dir10m          3.00 45.63 62.92 62.85   17127  YSU w/new nudging
dir10m          4.98 46.13 63.21 63.01   17127  QNSE
dir10m          3.49 46.86 64.34 64.24   17127  ACM2
dir10m          3.43 47.05 64.53 64.44   17127  ACM2 w/new nudging

pcp6            3.02 15.51 86.36 86.31   10482  YSU w/new nudging
pcp6            2.48 15.42 86.64 86.61   10482  YSU
pcp6            1.76 15.08 86.67 86.65   10482  YSU without LSM
pcp6            2.58 15.54 87.26 87.23   10482  ACM2
pcp6            3.09 16.00 87.97 87.92   10482  ACM2 w/new nudging
pcp6            6.85 18.76 92.02 91.77   10482  QNSE

slp             0.26  1.42  1.90  1.88   14700  YSU w/new nudging
slp             0.30  1.42  1.90  1.88   14700  ACM2 w/new nudging
slp             0.14  1.45  1.92  1.92   14700  QNSE
slp             0.36  1.43  1.92  1.88   14700  YSU
slp             0.41  1.44  1.92  1.88   14700  ACM2
slp             0.53  1.53  2.00  1.93   14700  YSU without LSM

t2m            -0.96  3.90  5.04  4.95   17747  YSU w/new nudging
t2m            -1.01  3.90  5.04  4.93   17747  YSU
t2m            -0.57  3.94  5.13  5.10   17747  ACM2 w/new nudging
t2m            -0.64  3.93  5.13  5.09   17747  ACM2
t2m            -1.25  4.04  5.17  5.02   17747  QNSE
t2m            -1.26  4.24  5.32  5.17   17747  YSU without LSM

wsp10m         -0.44  3.44  4.50  4.48   17938  YSU
wsp10m         -0.47  3.45  4.52  4.49   17938  YSU w/new nudging
wsp10m         -0.12  3.48  4.55  4.55   17938  ACM2
wsp10m         -0.15  3.49  4.56  4.56   17938  ACM2 w/new nudging
wsp10m         -0.44  3.50  4.56  4.54   17938  YSU without LSM
wsp10m          1.20  3.90  5.02  4.88   17938  QNSE

 12 UTC (5 am PDT) 12-km verification statistics for SA stations
Field            ME   MAE   RMS   STDEV  COUNT  PBL Scheme
dew2m           0.04  3.79  5.17  5.17   17009  ACM2 w/new nudging
dew2m           0.13  3.80  5.19  5.19   17009  ACM2
dew2m          -0.62  3.85  5.19  5.15   17009  YSU w/new nudging
dew2m          -0.63  3.86  5.21  5.18   17009  YSU
dew2m           1.27  3.90  5.22  5.06   17009  QNSE
dew2m           5.61  6.15  8.05  5.78   17009  YSU without LSM

dir10m          4.10 62.30 79.77 79.67   17401  ACM2
dir10m          3.82 62.33 79.78 79.70   17401  ACM2 w/new nudging
dir10m          5.38 62.66 80.39 80.22   17401  YSU without LSM
dir10m          1.07 63.03 80.65 80.64   17401  QNSE
dir10m          7.28 63.48 81.13 80.81   17401  YSU
dir10m          7.45 63.46 81.15 80.81   17401  YSU w/new nudging

slp             0.26  1.31  1.72  1.70   14476  YSU w/new nudging
slp             0.39  1.33  1.75  1.71   14476  YSU
slp             0.30  1.35  1.76  1.73   14476  ACM2 w/new nudging
slp             0.43  1.37  1.79  1.74   14476  ACM2
slp             0.42  1.39  1.82  1.77   14476  QNSE
slp             0.54  1.40  1.82  1.74   14476  YSU without LSM

t2m            -0.14  4.10  5.34  5.34   17392  QNSE
t2m             1.60  4.12  5.36  5.11   17392  YSU without LSM
t2m            -1.69  4.40  5.70  5.44   17392  ACM2
t2m            -1.78  4.45  5.76  5.48   17392  ACM2 w/new nudging
t2m             2.05  4.62  6.06  5.70   17392  YSU w/new nudging
t2m             2.15  4.62  6.06  5.67   17392  YSU

wsp10m          1.45  3.37  4.28  4.02   17539  ACM2 w/new nudging
wsp10m          1.48  3.37  4.28  4.01   17539  ACM2
wsp10m          1.32  3.35  4.29  4.09   17539  YSU without LSM
wsp10m          1.70  3.55  4.51  4.18   17539  YSU w/new nudging
wsp10m          1.74  3.57  4.53  4.18   17539  YSU
wsp10m          1.68  3.57  4.56  4.24   17539  QNSE

Some PBL Scheme Statistics January-February 2009

 00 UTC (4 pm PST) 12-km verification statistics for SA stations
Field            ME   MAE   RMS   STDEV  COUNT  PBL Scheme
dew2m           2.83  4.84  6.45  5.80   17022  YSU w/new nudging
dew2m           2.89  4.90  6.53  5.86   17022  YSU
dew2m           4.00  5.39  7.09  5.86   17022  QNSE
dew2m           4.13  5.39  7.09  5.77   17022  ACM2 w/new nudging
dew2m           4.26  5.44  7.14  5.74   17022  ACM2
dew2m           5.27  5.97  7.86  5.83   17022  YSU without LSM

dir10m          4.72 57.25 76.59 76.45   17387  QNSE
dir10m          7.41 57.57 76.89 76.54   17387  YSU w/new nudging
dir10m          6.06 57.46 77.03 76.80   17387  ACM2 w/new nudging
dir10m          7.33 57.93 77.25 76.90   17387  YSU
dir10m          6.37 57.83 77.42 77.16   17387  ACM2
dir10m          8.13 58.21 77.63 77.21   17387  YSU without LSM

pcp6            7.37 20.34 68.17 67.77   10030  YSU w/new nudging
pcp6            7.15 19.99 68.47 68.10   10030  YSU without LSM
pcp6            7.50 20.85 68.97 68.56   10030  ACM2
pcp6            6.78 20.57 69.01 68.68   10030  ACM2 w/new nudging
pcp6            7.96 20.76 69.25 68.79   10030  YSU
pcp6            8.34 21.30 69.83 69.34   10030  QNSE

slp            -2.03  2.66  3.59  2.96   14289  YSU w/new nudging
slp            -2.00  2.67  3.61  3.00   14289  ACM2 w/new nudging
slp            -2.09  2.70  3.63  2.97   14289  YSU
slp            -2.03  2.70  3.64  3.02   14289  ACM2
slp            -2.04  2.71  3.64  3.01   14289  QNSE
slp            -2.20  2.75  3.72  3.00   14289  YSU without LSM

t2m            -0.80  4.89  6.40  6.35   17340  QNSE
t2m            -0.04  5.02  6.68  6.68   17340  YSU w/new nudging
t2m             0.10  5.04  6.71  6.71   17340  YSU
t2m             1.62  4.99  6.80  6.61   17340  YSU without LSM
t2m            -1.91  5.39  7.26  7.01   17340  ACM2
t2m            -2.06  5.42  7.31  7.01   17340  ACM2 w/new nudging

wsp10m          0.54  3.66  4.89  4.86   17593  ACM2 w/new nudging
wsp10m          0.58  3.69  4.93  4.89   17593  ACM2
wsp10m          1.37  3.91  5.27  5.09   17593  YSU without LSM
wsp10m          1.27  3.95  5.31  5.15   17593  QNSE
wsp10m          1.44  3.98  5.36  5.16   17593  YSU w/new nudging
wsp10m          1.51  4.02  5.41  5.20   17593  YSU


 12 UTC (5 am PDT) 12-km verification statistics for SA stations
Field            ME   MAE   RMS   STDEV  COUNT  PBL Scheme
dew2m           2.69  5.15  6.95  6.41   16078  QNSE
dew2m           3.31  5.40  7.30  6.51   16078  ACM2 w/new nudging
dew2m           3.32  5.40  7.31  6.51   16078  ACM2
dew2m           2.28  5.54  7.51  7.15   16078  YSU w/new nudging
dew2m           2.24  5.57  7.55  7.22   16078  YSU
dew2m           5.32  6.21  8.60  6.76   16078  YSU without LSM

dir10m          7.99 60.58 79.17 78.76   16473  QNSE
dir10m         10.60 61.91 80.48 79.78   16473  ACM2
dir10m         10.80 61.95 80.58 79.85   16473  ACM2 w/new nudging
dir10m         13.12 63.75 82.28 81.23   16473  YSU without LSM
dir10m         12.47 63.91 82.32 81.37   16473  YSU w/new nudging
dir10m         12.68 64.06 82.48 81.50   16473  YSU

pcp6            6.43 21.67 69.54 69.25   10096  QNSE
pcp6            7.92 21.55 69.74 69.29   10096  YSU
pcp6            6.68 21.75 69.75 69.43   10096  ACM2
pcp6            8.09 21.76 70.46 70.00   10096  YSU without LSM
pcp6            6.25 21.10 71.34 71.07   10096  YSU w/new nudging
pcp6            4.93 21.08 71.59 71.42   10096  ACM2 w/new nudging

slp            -2.18  2.83  3.86  3.18   13574  YSU w/new nudging
slp            -2.19  2.86  3.89  3.21   13574  ACM2 w/new nudging
slp            -2.19  2.87  3.90  3.23   13574  QNSE
slp            -2.28  2.89  3.91  3.18   13574  YSU
slp            -2.27  2.91  3.94  3.22   13574  ACM2
slp            -2.38  2.97  4.02  3.23   13574  YSU without LSM

t2m             1.55  5.56  7.48  7.32   16392  ACM2 w/new nudging
t2m             1.65  5.56  7.50  7.32   16392  ACM2
t2m             2.43  5.60  7.50  7.10   16392  QNSE
t2m             4.27  6.30  8.43  7.27   16392  YSU w/new nudging
t2m             4.40  6.38  8.51  7.29   16392  YSU
t2m             4.76  6.46  8.91  7.53   16392  YSU without LSM

wsp10m          1.71  4.07  5.32  5.04   16602  ACM2 w/new nudging
wsp10m          1.74  4.09  5.36  5.07   16602  ACM2
wsp10m          2.36  4.41  5.81  5.31   16602  YSU without LSM
wsp10m          2.44  4.44  5.83  5.30   16602  QNSE
wsp10m          2.58  4.53  5.96  5.37   16602  YSU w/new nudging
wsp10m          2.63  4.56  6.00  5.39   16602  YSU

The New RRTG Radiation Scheme

• At 36-km, the new RRTG shortwave radiation scheme decreases the areal coverage of the low clouds and appears to make them thicker than our standard Dudhia scheme as can be seen in this loop valid at 00 UTC 19 September 2008 and this this loop valid at 12 UTC 19 September 2008.
• For the same case, the 12-km domain, however, shows that daytime highs are higher over land and there are more low clouds over the ocean, as can be seen in this loop valid at 00 UTC 19 September 2008. The overnight lows decrease and there are more clouds seen in this loop valid at 12 UTC 19 September 2008.
• Temperature statistics for 2-month tests in 2 seasons

   0O UTC (highs) for Jan-Feb 2009
         ME   MAE   RMS   STDEV  COUNT  Domain and Radiation Scheme
        0.19  4.64  6.23  6.23    9455  12-km RRTM+Dudhia
        1.27  4.83  6.42  6.29    9455  12-km RRTG
   12 UTC (lows) for Jan-Feb 2009
         ME   MAE   RMS   STDEV  COUNT  Domain and Radiation Scheme
        3.75  5.76  7.67  6.70    9000  12-km RRTM+Dudhia
        3.18  5.65  7.57  6.87    9000  12-km RRTG
  
   0O UTC (highs) for Jul-Aug 2008
         ME   MAE   RMS   STDEV  COUNT  Domain and Radiation Scheme
       -1.58  3.87  4.99  4.73    9263  12-km RRTM+Dudhia
       -0.68  3.68  4.88  4.83    9263  12-km RRTG
   12 UTC (lows) for Jul-Aug 2008
         ME   MAE   RMS   STDEV  COUNT  Domain and Radiation Scheme
        1.60  4.36  5.78  5.56    9331  12-km RRTM+Dudhia
        2.17  4.58  6.06  5.66    9331  12-km RRTG
  

• For the 2009071912 case, the new RRTG shortwave radiation scheme decreases the areal coverage of the low clouds in this 36-km loop valid at 12 UTC 21 July 2009 but it increases the areal coverage in this 36-km loop valid at 00 UTC 22 July 2009.
• Conclusion: The new radiation scheme gives mixed results. There is no clear reason to switch to it.

The Effects of Nudging

• Turning nudging on at a lower level, the 5th model level instead of the 10th (approximately 200m instead of 620m), gives a minor increase in low clouds, as shown for the 2008091800 case in thses loops:
  • 36-km domain hour 24, hour 36,
  • 12-km domain hour 24, hour 36.
• Nudging at a lower level and increasing the nudging coefficients for winds and temperatures by a factor of 3 and moisture by a factor of 300 (to the NCAR defaults) has a greater effect on low clouds, as shown for the 2008091800 case in thses loops:
  • 36-km domain hour 24, hour 36,
  • 12-km domain hour 24, hour 36.
• Nudging at a lower level, using the increased nudging coefficients, and turning nudging on in the 12-km domain increases low clouds still further but eliminates some mesoscale detail as shown for the 2008091800 case in thses loops:
  • 12-km domain hour 24, hour 36, hour 42.
  • The satellite images indicate that the extent of these clouds may be overdone.

Proposed Changes to Real-Time WRF-GFS

• Change from WRF version 3.0.1.1 to 3.1. Tests show improved numerical stability.
• Pending favorable verification statistics for 2 months and 2 seasons, enhanced nudging on the 36-km domain should be implemented.
• Continue to use YSU PBL scheme if ACM2 scheme is determined to be producing too low PBL heights.
• Continue with RRTM longwave and Dudhia shortwave radiation schemes.
• To improve cloud-water initialization, a cycling approach can be used in combination with the DFI. This approach would require running the 36-km domain separately from the 12-km domain and would change our timestep from 216 seconds, 6-delta x, to 225 seconds (an approach used by Brian Ancell). This method has not been extensively tested, but it shows promise.