Log of Changes to Pacific Northwest WRF/MM5 Numerical Forecasts
Return to Main MM5 Page
Return to Information Page
9/30/2026
Beginning with the 2026100100 run, all WRF-GFS domains have turned
off the aerosol optical depth and the aerosol angstrom exponent fields
for the season. They will be turned back on with the 2027070100 forecast.
7/16/2026
Beginning with the 2026071600 run, all WRF-GFS domains are using
aerosol optical depth and the aerosol angstrom exponent fields from
NASA's GEOS system.
Three-hourly values of these parameters will be used in WRF to
interact with the radiation scheme in an effort to incorporate the
temperature effects of smoke. See this page for more information.
These aerosol effects are scheduled to be turned back off with the
October 1 run.
NOTE: the extended WRF-GFS will not be using these effects since
GEOS5 model data out to 240 hours is not available soon enough for us
to use with the 00Z cycle and since the 12Z cycle of GEOS5 only
reaches 120 hours.
10/16/2025
Beginning with the 2025101612 run, the WRF-GFS will only produce
3-hourly graphics for both the 36-km and 12-km domains. The 4-km and
1.33-km domains will retain hourly graphics.
1/28/2025
Beginning with the 20250112812 run, the WRF-GFS 4-km domain has been
extended 12 hours to run out to 120 hours. The 36 and 12-km domains
continue to run to 240 hours and the 4/3-km domain to 72 hours. The
expected delay in completion of the runs is around 12 minutes.
11/1/2023
Beginning with the 2023110100 run, the WRF-GFS 4-km domain has been
extended 12 hours to run out to 96 hours. The 36 and 12-km domains
continue to run to 240 hours and the 4/3-km domain to 72 hours. The
expected delay in completion of the runs is around 12 minutes.
5/24/2023
The problem with the smoke and aerosol mods have been corrected and
the mods are turned back on beginning with the 2023052400 run.
5/22/2023
Due to a problem affecting the 4/3-km grids, the smoke and aerosol
mods have been turned off and the 2023051900-2023052200 4/3-km domain
cases were rerun.
5/19/2023
Due to fires in Alberta, Canada, beginning with the 2023051900 run,
all WRF-GFS domains are using aerosol optical depth and the aerosol
angstrom exponent fields from
NASA's GEOS system.
Three-hourly values of these parameters will be used in WRF to
interact with the radiation scheme in an effort to incorporate the
temperature effects of smoke. See this page for more information.
These aerosol effects are scheduled to be turned back off with the 2023100100
run though they may be turned off earlier if smoke conditions over
the Pacific Northwest improve substantially.
NOTE: the extended WRF-GFS will not be using these effects since
GEOS5 model data out to 240 hours is not available soon enough for us
to use with the 00Z cycle and since the 12Z cycle of GEOS5 only
reaches 120 hours.
3/15/2023
The MM5-NAM runs were retired after the 2023031512 run.
3/13/2023
Beginning with the 2023031312 run, the deep stable layer calculations
have been corrected. Areas that showed positive values are now
showing higher percentages due to the addition of a missing layer
in the previous calculation.
12/15/2022
Beginning with the 2022121512 run, the 4/3-km domain has expanded
by roughly 30% to cover all of Idaho.
11/3/2022
Beginning with the 2022110300 run, the extended 36/12-km WRF-GFS
run now goes out to 240 hours instead of 180 hours. This increases
runtime by 20-25 minutes.
4/15/2022
Beginning with the 2022041512 run, all air stagnation advisory
products are now updated with the March 22, 2022 NWS zone shapefile.
3/4/2022
Beginning with the 2022030500 run, all WRF-GFS and most ensemble runs
are once again using the Noah-MP LSM instead of the Noah LSM.
12/29/2021
Beginning with the 2021122900 run, all WRF-GFS and most ensemble runs
are using the Noah LSM instead of the Noah-MP LSM since there are
noticeable problems with the Noah-MP over snow cover.
10/7/2021
Beginning with the 2021100800 run, all WRF-GFS domains are no longer
using aerosol optical depth and the aerosol angstrom exponent fields from
NASA's GEOS system.
These were turned off about 10 days earlier than normal due to
recent problems acquiring the GEOS data in a timely manner.
They are scheduled to be turned back on for the July 1, 2022 run.
7/1/2021
Beginning with the 2021070100 run, all WRF-GFS domains are using
aerosol optical depth and the aerosol angstrom exponent fields from
NASA's GEOS system.
Three-hourly values of these parameters will be used in WRF to
interact with the radiation scheme in an effort to incorporate the
temperature effects of smoke. See this page for more information.
These aerosol effects are scheduled to be turned back off with the 2021100100
run though they may be turned off earlier if smoke conditions over
the Pacific Northwest improve substantially.
3/22/2021
Beginning with the 2021032212 forecast, all WRF-GFS are now using the
new GFS v16. More information on the changes to the GFS can be found
in this document.
8/20/2020
Beginning with the 2020082012 run, all WRF-GFS domains will be using
aerosol optical depth and the aerosol angstrom exponent fields from
NASA's GEOS system.
Three-hourly values of these parameters will be used in WRF to
interact with the radiation scheme in an effort to incorporate the
temperature effects of smoke. See this page for more information.
These aerosol effects are scheduled to be turned back off with the 2020100100
run though they may be turned off earlier if smoke conditions over
the Pacific Northwest improve substantially.
2/18/2020
Beginning with the 2020021900 run, we are now using WRF version 4.1.3
for the WRF-GFS runs. This includes:
. the hybrid vertical coordinate which smooths out model levels
above terrain. Compare vertical levels in this
cross section using with this loop
. using moist theta (instead of dry theta) as the default temperature variable
. using effective radii calculated by the moist physics schemes in the
RRTMG radiation scheme (use_mp_re=1)
More info on WRF release changes can be found here.
1/11/2020
Beginning with the 2020011100 run, switched back to Noah LSM as the
default due to some cold bias problems for Noah-MP over snow in
Western Washington.
7/15/2019
Beginning with the 2019071600 run, the aerosol effects are turned
off. If large wildfires develop in the Pacific Northwest, they will
be turned back on.
7/1/2019
Beginning with the 2019070100 run, all WRF-GFS domains will be using
aerosol optical depth and the aerosol angstrom exponent fields from
NASA's GEOS system.
Three-hourly values of these parameters will be used in WRF to
interact with the radiation scheme in an effort to incorporate the
temperature effects of smoke. See this page for more information.
These aerosol effects are scheduled to be turned back off with the 2019100100
run.
3/28/2019
RAP grids were missing and not used for the 2019032700 and 2019032712
initializations.
1/28/2019
Fixed and updated the meteogram, time-height, and sounding generators.
12/24/2018
Beginning with the 2018122500 forecast, have updated the landuse for
all WRF-GFS and ensemble runs to that derived from MODIS 15-arcsecond
files. Some points around the waters of Western Washington were
manually altered to better reflect the underlying water/land feature.
A comparison of old and new landuse features can be found here.
9/5/2018
The 2018090612, 2018090600 and 2018090412 WRF-GFS runs experienced CFL problems
on multiple attempts due to hurricanes Norman and Olivia traversing
the lower boundary of the 36-km domain. Switching from 1/4-degree to
1-degree GFS input files seems to resolve the problem.
8/15/2018
Beginning with the 2018081600 run, all WRF-GFS domains will be using
aerosol optical depth and the aerosol angstrom exponent fields from
NASA's GEOS system.
Three-hourly values of these parameters will be used in WRF to
interact with the radiation scheme in an effort to incorporate the
temperature effects of smoke. See this page for more information.
5/12/2018
The 2018051212 4/3-km run did not run in real-time due to networking
problems that caused excessive delays in transferring files from our
run computers to the post-processing computer.
4/24/2018
The 2018042412 runs did not use the RAP grids due to problems with the
grids.
3/30/2018
Added new fields and ability to export images as PNG files to the
ensemble plumes page.
3/26/2018
The 2018032612 runs were delayed by NCEP server and CONDUIT problems.
The extended forecast from hours 87-180 did not run.
12/16/2017
The 2017121612 4/3-km run did not run in real-time. A large set of past-case
runs interfered with the model starting properly. The run was
completed on the 17th.
11/9/2017
The 2017110912 WRF-GFS runs had to use 1-degree GFS files due to
problems at NCEP
4/19/2017
The MM5-NAM forecast could not get past hour 69 due to a problem with
INTERP. The error displayed is:
NHPP vert gradient check 17 9 10.156250E-01-159302. 159302.
nhpp_did_not_converge
3/9/2017
The 2017031000 4/3-km WRF-GFS run had to be run without RAP grids in
order to avoid an error in the water budget for the Noah-MP land
surface model.
2/2/2017
Beginning with the 2017020300 run we have switched to using diff_opt=2
(3D horizontal diffusion) and have turned off 6th-order diffusion
(diff_6th_opt=6).
1/11/2017
The 2017011112 4/3-km WRF-GFS had a CFL error at hour 67:50. A
restart from hour 66 worked fine on the second attempt.
1/3/2017
The 2017010300 and 2017010400 WRF-GFS 4/3-km runs both failed at start
up when the NSIDC snow cover was used for initial conditions. Both
runs were finally completed by dropping that data set.
10/31/2016
Beginning with the 2016110100 run, the WRF-GFS forecasts have been
updated to the nov16 configuration:
. expanded 4/3-km domain south to cover all of Oregon, and
north to cover the Fraser River gap
. the 4/3-km domain is run separately, using 10-minute output from
the 4-km domain is input for the WRF ndown program.
. WRF version 3.7.1 is continued
. physics options remain the same as with the nov15 configuration
For more information, check out the info page.
5/31/2016
The following WRF-GFS runs were initialized without RAP grids due to
problems with the initial conditions.
2016053112
2016053100
2016053012
2016053000
2016052912
2016052900
2016052500
2016052412
2016052312
2016052300
2016052100
2016051912
2016051800
2016051212
5/11/2016
The 2016051112 run used GFS 1/2-degree grids due to numerous missing
1/4-degree files.
Added soil moisture plots.
5/10/2016
Beginning with the 2016051012 run, all WRF-GFS runs are using 1/4-degree GFS
grids instead of the 1/2-degree grids.
4/19/2016
The 2016041912 WRF-GFS runs were run without the RAP initialization fields.
4/8/2016
The 2016040800 WRF-GFS runs were initialized without RAP grids.
4/5/2016
Starting with the 2016040512 run, the 4/3-km domain is being run out
to 72 hours.
3/22/2016
Eliminated the RAP snow fields from our initialization since those are
what caused the numerous 0Z run failures with the Noah-MP land
surface model. Snow initialization come from GFS and the NSIDC.
2/11/2016
Reactivated the sounding, time-height, and meteogram generators for
WRF-GFS runs.
January through March 2016
The following WRF-GFS runs were initialized without RAP grids due to
problems with the initial conditions.
2016011900
2016012900
2016013000
2016013012
2016013100
2016013112
2016020100
2016020200
2016020300
2016020400
2016020500
2016020600
2016020700
2016020800
2016021000
2016021100
2016021400
2016021500
2016021600
2016021800
2016022200
2016022300
2016022600
2016022800
2016022900
2016030100
2016030200
2016030300
2016030900
2016031100
2016031112 - no RAP grids available
2016031200 - no RAP grids available
2016031212 - no RAP grids available
2016031300 - no RAP grids available
2016031400
2016031500
2016031600
2016031700
2016031800
2016031900
2016032000 - no RAP grids available
2016032100
2016032312 - no RAP grids available
2016032400 - no RAP grids available
2016032412 - no RAP grids available
1/23/2016
All 2016012400 WRF-GFS runs were initialized without RAP grids due to
problems with the initial conditions.
1/21/2016
The 4/3-km WRF-GFS run for 2016012100 was initialized without RAP grids due to
problems with the initial conditions.
1/19/2016
All 2016011900 WRF-GFS runs were initialized without RAP grids due to
problems with the initial conditions.
1/13/2016
The 2016011300 WRF-GFS run was initialized without
RAP grids due to problems acquiring the grids.
1/9/2016
The 2016010912 WRF-GFS run was initialized without
RAP grids due to problems acquiring the grids.
1/5/2016
The 2016010512 WRF-GFS run was initialized without
RAP grids due to problems acquiring the grids.
12/31/2015
The 2015123100 and 2015123112 WRF-GFS runs were initialized without
RAP grids due to problems acquiring the grids.
12/16/2015
The 2015121612 WRF-GFS run was initialized without RAP grids due to
problems acquiring the grids.
12/9/2015
The 2015120900 WRF-GFS run was initialized without RAP grids due to
problems acquiring the grids.
12/3/2015
The 2015120312 WRF-GFS forecasts were run without the high-resolution
NSIDC snow data inputs due to a delay in obtaining the data.
11/30/2015
For the 2015113000 and 2015113012 cycles, WRF-GFS runs were all
initialized without RAP grids due to problems acquiring the grids.
11/23/2015
Beginning with the 2015112300 run, the WRF-GFS has switched back to
the Noah-MP LSM. The water budget problems with the Noah-MP and the
high resolution NSIDC snow fields have been diagnosed and fixed.
11/5/2015
Beginning with the 2015110600 run, the WRF-GFS has switched back to
the Noah LSM. There were some water budget problems with the Noah-MP
and the high resolution NSIDC snow fields. Also, we are using
standard MODIS landuse and are not using the lakes until some of our
customers are ready to handle this extra category.
11/3/2015
Beginning with the 2015110312 run, the WRF-GFS has switched to the
nov15 configuration which includes these changes:
* WRF version 3.7.1
* Noah-MP land surface model
* Grell-Freitas ensemble scheme for cumulus convection
* MODIS land-use and lakes
11/1/2015
The 2015103100, 2015103112, 2015110100, 2015110112, and 2015110200
WRF-GFS runs were all initialized without RAP grids due to problems
acquiring the grids.
10/21/2015
The 2015102100 WRF-GFS run experienced a numerical instability at hour
126 for the 36/12-km run. It was rerun from hour 0 without FDDA
and bombed out at hour 121. Ultimately ran it without FDDA and with
1-degree GFS grids as initial and boundary files.
9/3/2015
The 2015090312 WRF-GFS run experienced a numerical instability at hour
168 for the 36/12-km run. It was rerun from hour 0 without FDDA in
order to complete the hours 168 to 180.
8/28/2015
The 2015082812 WRF-GFS run experienced a numerical instability at hour
96 for the 36/12-km run. It was restarted from hour 72 and completed
by simply using fewer processors (4 instead of 156).
8/20/2015
The 2015082000 WRF-GFS runs were initialized without RAP grids.
7/15/2015
The 2015071500 WRF-GFS runs were initialized without RAP grids.
7/8/2015
Beginning with the 2015070800 forecast, the 4/3-km WRF-GFS runs have
been extended from forecast hour 48 to hour 60.
7/2/2015
The 2015070200 WRF-GFS runs were initialized without RAP grids.
6/23/2015
The 2015062300 4/3-km WRF-GFS run failed initially at hour 16.5 due to
a Flerchinger error. It was restarted from hour 12 and it succeeded.
6/21/2015
The 2015062000, 2015062012, and 2015062100 WRF-GFS runs were
initialized without any RAP cloud fields.
There were some post-processing problems for the WRF-GFS 2015062000
run due to a disk mounting problem after a system reboot.
The WRF-GFS 4/3-km domain for 2015062100 only ran to hour 20 in
real-time due to the Flerchinger numerical instability. It had to be
rerun with FDDA turned off.
6/9/2015
The 2015060912 and 2015061000 WRF-GFS runs were initialized without
any RAP cloud fields.
6/3/2015
WRF-GFS extended run for 2015060300 stopped at hour 138 with a
Flerchinger segmentation fault.
5/24/2015
WRF-GFS 2015052412 run did not run in real-time due to problems
with acquiring RAP initialization grids.
5/18/2015
Added deep stable layer graphic to air quality pages for 4-km and
4/3-km domains.
5/15/2015
The 2015051512 WRF-GFS run made it to hour 75 before encountering a
segmentation fault with the Flerchinger error. The model was
restarted without FDDA.
5/12/2015
The 2015051300 runs did not use RAP initialization fields. The
2015051212 run was initialized with 6-hour forecast files from the
RAP 2015050606 run.
5/7/2015
The following WRF-GFS runs have failed at the given forecast hour in
the past several days:
2015050800 - hour 66
2015050712 - hour 54
2015050700 - hour 67
2015050612 - hour 126
2015050600 - hour 91
2015050400 - hour 117
Switching from Noah-MP LSM to the simpler Noah LSM has allowed all of
these runs to complete normally. Therefore, starting with the
2015050800 run, we will use the Noah LSM by default until the Noah-MP
can be fixed.
5/5/2015
The 2015050600 WRF-GFS runs encountered a segmentation fault with the
36/12-km extended run and failed at hour 91. Rerun attempts also failed.
5/4/2015
The 2015050400 WRF-GFS runs encountered a segmentation fault with the
36/12-km extended run and failed at hour 117. Rerun attempts the
next day did work.
4/30/2015
The 2015043012 and 2015050100 WRF-GFS runs encountered a water budget
problem with the Noah-MP scheme that prevented them from working.
All runs used the Noah LSM instead.
4/22/2015
The 2015042300 WRF-GFS 36/12-km extended run did not process boundary
conditions beyond forecast hour 111.
4/21/2015
The 2015042100 WRF-GFS 36/12/4-km domain developed bad surface
temperatures using the Noah-MP LSM. The model needed to be restarted
from the beginning with the Noah LSM.
4/20/2015
The 2015042012 WRF-GFS 36/12-km domain extended run was only able to
reach hour 147.
4/13/2015
The 2015041300 WRF-GFS 36/12-km domain extended run was only able to
reach hour 114.
The 2015041100 WRF-GFS 36/12-km domain extended run was only able to
reach hour 138.
4/9/2015
The 2015041000 WRF-GFS runs were initialized without RAP grids after
a 1:40 hour delay due to the lack of RAP grids and the installation
of a new infiniband switch.
4/6/2015
The 2015040612 WRF-GFS runs were initialized without RAP grids after
a 1 hour delay.
3/6/2015
The 2015030700 WRF-GFS runs were initialized without RAP grids after
a 3 hour delay.
3/1/2015
The 2015030112 WRF-GFS runs were initializaed with RAP grids from
2015030109 and the model was delayed by 5 hours.
2/19/2015
Beginning with the 2015021912 run, switched from using the 1.0-degree
to the 0.5-degree GFS grids by default for initialization and
boundary conditions.
2/18/2015
The 2015021812 and 2015021900 WRF-GFS runs were run without the RAP
initialization fields. They did use the Noah-MP LSM, however.
2/17/2015
Both, the 2015021712 and 2015021800 WRF-GFS runs were run with the
Noah LSM instead of the Noah-MP LSM due to a water budget problem in
the Noah-MP LSM. This appears to be due to our RAP initialization.
1/28/2015
Beginning with the 2015012900 case we have switched to the new jan15
configuration which has the following highlights:
- using WRF version 3.6.1
- using the Noah-MP land surface model
- initialization moisture fields using the Rapid Update Cycle (RAP)
grids when available. Model graphics will add "INIT: RAP+GFS"
when these grids are used. In case the RAP is unavailable (which
could happen up to 10% of the time), the GFS and NAM 221 grids
will be used for initialization, as before. Boundary conditions
for forecast hours 3 and beyond will continue to be from the GFS,
no matter which initialization combination is used.
8/2/2014
Lightning caused 4 of the machines in our cluster to go down at 11:41
AM and this resulted in failure of the 4/3-km run for the 2014080212
forecast.
7/30/2014
The 4/3-km 2014073012 run crashed at forecast hour 9 due to a
numerical instability and had to be restarted with a shorter time
step (from a factor of 6 to a factor of 5 times the horizontal
spacing).
6/27/2014
Starting with the 2014062800 run, the ventilation index using the PBL
depth has been corrected. All prior products used a layer shallower
than the true PBL with shallower layers over higher terrain features.
Also starting with the 2014062800 run, an experimental agricultural
burn index product has been added to the air quality page for the
4-km and 4/3-km domains.
6/23/2014
The 2014062300 1.33-km run failed to complete in real-time and stopped
at hour 16. It was finished offline on June 26.
5/8/2014
Starting with the 2014050800 run, the time-averaged air quality
products are now produced during the run, not at the end. This also
improved the depth used for the 500-meter ventilation product.
Before, the bottom five layers were chosen which was less than 300
meters. Now, all layers below 500 meters are used, which is usually
the bottom 8 layers.
3/24/2014
Due to infiniband switch problems, no 4/3-km domain was run in
real-time for the 2014032312 and 2014032400 runs. (n21 offline)
11/14/2013
With the 2013110900 run, we have modified our sep13
configuration to include the following:
- began using MODIS land-use for the real-time. Note, for debugging
purposes, the 2013111300 and 2013111312 runs used the USGS land-use files.
11/8/2013
Ran the 4/3-km domain with MODIS land-use to test its viability in the
real-time system. Anticipate changing to this for all domains
between the 2013110900 run and 2013111200 run.
9/16/2013
With the 2013091612 run, we have switched to our new sep13
configuration which has the following highlights:
- returned to using the Noah Land Surface Model
- bagan using the RRTMG (newer Rapid Radiative Transfer Model)
longwave and shortwave schemes.
- WRF version 3.5
9/6/2013
A power spike due to lightning and a subsequent problem with our
Infiniband switch meant that the 4/3-km runs for 2013090512 and
2013090600 did not start and the 4-km domain ran very slowly (at
roughly 1/3 the normal speed). It is believed that a bad port on
the IB switch was the cause.
7/8-7/13/2013
A problem with our Infiniband switch meant that the 2013070812 and
2013070900 WRF-GFS runs did not run in real-time and that the 4/3-km
domain did not run at all for either of those dates or 2013071112
nor 2013071312.
4/13/2013
The 2013041400 run did not run in real-time due to a major power
outage at the University of Washington. The 4/3-km run was not run
at all, but the the 36,12, and 4-km domains were submitted later the
next day to be completed on April 15.
4/12/2013
An attempt was made with the 2013041200 and 2013041212 forecasts to
switch to the RRTMG longwave and shortwave radiation schemes and the
RUC Land Surface Model. Unfortunately, the 4-km and 4/3-km runs for
2013041212 both crashed by forecast hour 6 without any indication of
the failure. We are returning to the Dudia+RRTM radiation and
thermal diffusion surface models and postponing this change until
further testing can be accomplished.
12/20/2012
Beginning with the 2012122100 forecast, we are now using the
simplified Arakawa-Schubert cumulus scheme instead of the old
Kain-Fritsch. Our studies show this
to be a superior scheme that eliminates the non-physical precipitation
bands of Kain-Fritsch.
10/10/2012
Beginning with the 2012101100 case we have switched to the new oct12
configuration which has the following highlights:
- making modifications in our vertical levels to suppress wave
reflection off the model top
- using WRF version 3.4.1 with an unmodified YSU PBL scheme
The following changes to our model configuration contribute to the
first item, the suppression of waves off the model top:
* using the newer implicit gravity-wave damping layer (damp_opt=3)
* expanding the depth of the damping layer from 5 km to 7 km
* changing the vertical levels to more gradually increase in depth
from the surface to a more nearly constant 1-km depth in the upper
troposphere
* raising our model top from 100 hPa to 50 hPa
For the PBL scheme change has these features:
* the newer YSU scheme gives similar wind speed and direction
scores to our u-star modifications for the 3.1.1 YSU scheme in
our summer and winter 2-month tests
* a slight degradation (more pronounced cold bias) in the new YSU
scheme
Research on the PBL scheme will continue in hopes of furthering
improving winds and temperatures.
3/13/2012
A 2.7-hour power outage at UW from 15:12 to 17:52 caused some delays and
made products unavailable.
2/7/2012
Updated relative humidity, veritcal velocity, and surface wind speed
graphics to better serve people with color blindness.
1/26/2012
Recent updates to the web pages:
- merged 1 1/3 km domain into the main WRF-GFS page
- added 1 1/3 km subdomain wind products that include high resolution
roads, counties, and islands maps
- added soundings, timeheights, and meteograms for 1 1/3 km domain
- added all air quality products for 1 1/3 km domain
- removed the experimental title from the extended 36/12-km runs
- added additional, model output snow products
12/21/2011
Switched to RIP4.6 for graphics production.
10/17/2011
Beginning with the 2011101712 run, we switched to the new oct11
configuration which includes an expansion of the 12-km, 4-km, and
4/3-km domains.
9/9/2011
Added the air quality page for the high resolution 4/3-km domain.
8/29/2011
Beginning with the 2011083000 forecast, we're now using the old
Kain-Fritsch cumulus scheme for all WRF runs. The newer scheme with
shallow convection produces unrealistic, numerical precipitation
bands in the southern portion of the East Pacific High. See
this for an examination of this
problem.
6/30/2011
Beginning with the 2011063012 forecast, the high-resolution 4/3-km
domain has extended eastward into Idaho. The terrain is also no
longer smoothed since we are using epssm = 0.2 for this domain.
5/13/2011
Beginning with the 2011051400 forecast, we're now plotting the lowest
sigma level temperature instead of T2 as the 2-m temperature. Over
snow, the T2 field is too cold in the spring, as can be seen in
this loop.
3/1/2011
Beginning with the 2011030200 forecast, we now use NSIDC
high-resolution (roughly 1-km) snow fields for model initialization
over the United States and parts of the Canadian Rockies. Outside
this region, we continue to use NAM 221 grids where available and GFS
outside that region. NSIDC snow analyses are updated once daily and
the most recent one will be used for each forecast. A case
comparison showing the effects of the new snow fields is available here.
1/13/2011
Beginning with the 2011011312 forecast, we made the following changes:
- extended 36, 12, and 4-km runs to hour 84 from 72 hours
- starting 4-km run at the same time as the 36+12 run
- expanded the 4/3-km domain to cover all of Washington state
- extended the 4/3-km domain from 36 to 48 hours
- added a 12Z 4/3-km domain run
- all nests are run with their parent nests, no ndown is used. This
improves precipitation near the borders.
11/17/2010
Beginning with the 2010111800 forecast, we are now producing a wind
gust product which is a factor of 1.5 times the 10-m wind speed.
10/6/2010
Beginning with the 2010100612 run, we have changed the WRF-GFS runs
for the 36, 12, and 4-km domains to a new u-star modification. The
2010041412 change has been dropped in favor of a u-star modification
in the YSU PBL scheme which affects only the momentum. The details
of this are as follows:
- applied only over land
- a terrain-variance based u-star factor ranging linearly from 1.175
to 1.575 for terrain variances of 0-200m^2 is applied in the momentum
calculation of the YSU PBL scheme
- for terrain variances greater than 200 m^2, a u-star factor of
1.575 is applied
The effect of this change will be to decrease the high wind speed
bias of WRF 3.1.1. This is known as case wrfv311ustar71 in the
experiments detailed here.
4/19/2010
Beginning with the 2010042000 run, the high-resolution 4/3-km WRF-GFS
runs will have standard WRF 3.1.1 model code restored. The u-star
modification for this domain only will be turned off. The change
from using Noah LSM to the thermal diffusion scheme will remain.
4/14/2010
Beginning with the 2010041412 run, the WRF-GFS runs have changes as
follows:
1. modified friction velocity (u-star) by increasing by a factor of
1.5 over land
2. switched from Noah LSM to thermal diffusion
1. Surface wind speeds over land have had a noticeable high bias in
speed and geostrophic bias in direction. Experiments have proved
that this modification to u-star improves these biases.
2. The Noah LSM in the WRF 3.1.1 and 3.2 codes has a strong cold
bias over the elevated terrain of the Intermountain West. Daytime
high temperatures and surface temperatures are often well below
observed. Turning off the Noah LSM and switching to the simpler
5-layer thermal diffusion scheme (as was used in our MM5 runs)
improves the surface and 2-m temperatures greatly. This change will,
however, introduce about a 1°F higher dewpoint temperature bias.
12/30/2009
Beginning with the 2009123100 run, the WRF-GFS has switched to
version 3.1.1. This dec09 configuration is identical to the
previous, apr08, except for the change in model version.
The 3.1.1 code now includes a double-moment rain scheme in the
Thompson microphysics.
The test 4/3-km domain has moved to the 0Z cycle.
11/7/2009
Began testing a 4/3-km domain over western Washington for 12Z cycles.
5/20/2009
Two recent cases failed with OFF TOP cfl error: 2009051800 and
2009052000. Both required a rerun with grid_fdda turned off.
2/10/2009
Added meteograms for the extended range forecasts with the 2009021000
run.
1/22/2009
With the 2009012212 case, we have once again been forced to use the
Dudhia shortwave and RRTM longwave radiation schemes because the CAM
schemes do not allow the forecast to run past hour 2.
1/9/2009
With 2009010912 case started running WRF version 3.0.1.1. Modifications
enable restart capability so that extended WRF-GFS runs will
now work. Version 3.0.1.1 fixes the erroneously high wind speeds
over the mountains during our stronger wind events. See for example:
Dec 21 2008 12-km and
Dec 21 2008 4-km loops.
Problems with CAM radiation schemes for 3.0.1.1 forced a
switch to the Dudhia shortwave and RRTM longwave radiation schemes
for just the 2009010912 run. After that, CAM was fixed. Adaptive
time steps have been turned on as well.
11/20/2008
Bug found in the Urban Canopy Model that was causing temperatures to
go far too cold over urban areas. Forecast hour 45 of the 2008111812
run was the worst. For more info, see
this. Beginning with 2008112100 run, ucmcall is now set to 0.
Switched MM5-NAM from MPICH2 to OpenMPI due to continued problems with
the computer locking up. Note: the WRF-GFS and MM5-GFS have/had been
using OpenMPI for over a year.
10/30/2008
Ran 2008103100 and 2008103112 cases with WRF version 3.0.1.1.
Unfortunately, the extended WRF-GFS runs could not be run with this
version due a bug in the restart portion of the code. Returned to
using the 2.2.1 code until that bug can be fixed.
NOTE: both 2008103100 and 2008103112 cases were later re-run in
version 2.2.1 for archive purposes.
8/27/2008
Moved Extended WRF-GFS to new machines resulting in output being
available 3 hours earlier (11:20 am/pm PDT vs 2:20 pm/am PDT).
Added more graphics for 12-km extended run as well.
8/8/2008
Added text output to the standard soundings on the Web page.
7/3/2008
Activated numtiles=4 switch (as suggested at the recent WRF Workshop)
for the 4-km domain with the 2008070312 run to save 20 minutes of
run time.
4/15/2008
WRF replaces MM5 as operational model
Beginning with the 2008041500 run we switched from MM5 version 3-7-3
to WRF-ARW version 2.2.1 as the operational forecast model.
Highlights of this upgrade are:
* expanded 12-km and 4-km domains
* backported the YSU PBL scheme from version 3.0 into 2.2.1
3/3/2008
Beginning with the 2008030312 run, the WRF-GFS is now running a 4-km
forecast at both 0Z (old) and 12Z (new). The 4-km run also now
extends to 72 hours.
1/3/2008
Beginning with the 2008010412 run, the WRF-GFS is using a 216-second
time step (6*dx) for the 36-km run instead of a 180-second time
step. Tests have shown this to be stable, but daily runs will need
to be performed to give it a full test. Should CFL conditions arise
and the model become unstable, we will return to the more costly
180-second time step.
9/28/2007
Beginning with the 2007092800 run, the WRF-GFS is run to 72 hours for
the 36/12-km domains.
7/31/2007
Beginning with the 2007073100 run, the WRF-GFS is using the
positive-definite moisture scheme.
4/8/2007
Beginning with the 2007040800 run, the WRF-GFS is now running a 4-km
forecast for the 0Z run.
8/16/2006
Beginning with the 2006081600 run, the MM5-GFS is using the Thompson
microphysics scheme of 20 June 2006. The CCM2 radiation scheme has
default settings for cloud ice effective radius of 35 micrometers and
cloud liquid effective radius of 10 micrometers. A change in
getdat.F is made for the CCM2 that includes the radiative properties
of snow along with cloud ice.
CIMR(N,NLL)=(QI3D(NLIM,JSLC,NLL)+QNI3D(NLIM,JSLC,NLL)) 09JUN06.1
+ *1000. 05DEC01.1355
Beginning with the 2006081512 run, the WRF-GFS is using the Thompson
microphysics scheme of 11 August 2006.
7/18/2006
Beginning with the 2006071812 run, the MM5-NAM (formerly MM5-ETA) is
using MM5 version 3-6-3 (old was 3-5-4).
6/8/2006
Beginning with the 2006060900 run we switched from using the Thompson
microphysics scheme to the Reisner2 scheme. We have noted a high
temperature bias and have discovered that precipitating clouds allow
too much solar radiation to reach the surface in the Thompson
scheme. We will continue to investigate this and hope to be able to
use the scheme when this issue has been resolved.
5/15/2006
Beginning with the 2006051512 run we switched from MM5 version 3-6-3
to 3-7-3 using the latest Greg Thompson microphysics scheme. No
changes in domain configuration have been made, so end-users should
not need to make any modifications. Highlights of this upgrade are:
* Changed from old Reisner2 microphysics which produced too much
ice at high levels to the new Greg Thompson scheme.
* Changed LANDUSE.TBL to match the one used in WRF.
This new LANDUSE.TBL reduces the albedos for some categories slightly,
which will result in slightly warmer daytime temperatures.
* Changed MRF planetary boundary layer routines:
1) Changed the computation of RLAM to be a function of delta-z.
2) Instead of imposed parabolic mixing coeff profile in BL, use
the max of that AND the value computed from local K-theory
(the latter will increase mixing in cloud-topped BLs).
* Bug fixes to Noah LSM scheme -- note, we do not yet use this
scheme, but intend to run tests to see if this scheme produces
better results than the current 5-layer soil model.
* Preprocessing has upgraded from NCAR 3-6 to 3-7 versions.
Beginning with the 2006051512 run, we implemented the 13 May 2006
changes to the WRF version of the Thompson microphysics scheme.
Results should now be consistent with the MM5 version.
5/9/2006
Updated WRF runs to version 2.1.2 with latest Thompson microphysics
modifications. Previous runs had abnormally high precipitation in
areas where frozen hydrometeors reached the ground.
4/21/2006
Had to update the PREGRID Vtable file for the OTIS SST fields since
the level code changed from 1 to 0 with the 2006042012 file. Now
using '*' instead of '1'.
4/1/2006
Updated WRF runs to version 2.1.2 with Thompson microphysics and added
several PBL output variables to the output files. First run with
these changes is the 2006040100 run.
3/21/2006
Made sounding generator available online to all users.
2/6/2006
Beginning with the 2006020600 run, the MM5-ETA graphics have been
aligned with the MM5-GFS. NOTE: the MM5-ETA is still running on a
slightly smaller 12-km domain (the aug02 domain) and is still not
running the same physics packages.
4/5/2005
Beginning with the 2005040600 run, the MM5-GFS will be upgraded from
the Intel FORTRAN compiler version 7.0 to version 8.1. The overall
speed-up is significant:
Intel 7.0 Intel 8.1
Compiler Compiler
Domains Run Time Run Time Savings
------- ---------- ----------- --------
d1d2 1:15:49 1:03:21 0:12:28 (16.5%)
d3 3:51:15 2:57:27 0:53:48 (23.3%)
Total 1:06:16 (21.6%)
2/4/2005
Added WRF-GFS runs to 48 hours with limited graphics.
1/12/2005
Added run-independent links to all single-image display pages.
1/6/2005
Beginning with the 2005010612 forecast, the snow cover initialization
is turned back on, thanks to our code modifications to the MM5 that
allow IFSNOW=1 to work with IMOIAV=2.
11/2/2004
Added additional Puget Sound area soundings, timeheights, and
meteograms to the MM5-GFS page by giving the user a choice of two map
options.
10/29/2004
Added Mt. St. Helens trajectories.
10/8/2004
Updated the look of the MM5-GFS, MM5-ETA, and MM5-GFS Air Quality
pages to include groupings by color, vertical level, and/or product
type.
8/19/2004
Beginning with the 2004081900 forecast, the MM5-GFS is using the
Bucket Soil Moisture Model and is using the 20-km RUC grids to
initialize the soil moisture. For a comparison of the 2-m
temperatures see this.
8/10/2004
Updated the main home page for our MM5 to better reflect the whole
project and to improve overall look.
8/4/2004
Beginning with the 2004080500 forecast, the 4-km run of the MM5-GFS
has the Kain-Fritsch cumulus parameterization turned on. Recent work
has shown that this is a good idea and that it will help rectify the
discrepancy between convection in the 12-km domain and that in the
4-km domain. For a comparison of runs made with and without the K-F
scheme turned on for the 4-km domain, see this
link. The addition of this cumulus scheme to the 4-km domain
will increase runtime by only 3 minutes.
6/3/2004
Beginning with the 2004060400 forecast the CCM2 effective ice radius
has been changed from 40 microns to 35.
6/3/2004
Beginning with the 2004060212 forecast, rerun on the 3rd, we are now
running MM5-GFS with the following changes:
* The MPI-implementation of the CCM2 modification -- where the
effective ice radius of radcsw.F is changed from 14.6 microns to
40 microns -- has been corrected. All previous forecasts from
2004051800 through 2004060200 used the 40 micron value on
only 2 of the 36 processors and the 14.6 standard value on all
other processors.
This loop shows the dramatic difference in the
amount of radiation reaching the surface due to the presence
of thin, high clouds.
5/19/2004
Beginning with the 2004051900 forecast, the MM5-GFS GEMPAK files are
now calculated at cross points, not dot points. Comparisons with
MMOUT files are much easier.
5/17/2004
Beginning with the 2004051800 forecast we are now running MM5-GFS with
the following changes:
* Radiation scheme is now CCM2 with a slight modification -- the
effective ice radius of radcsw.F is changed from 14.6 microns to
40 microns.
* Reisner 2 is now used without skipping, since this has been shown
to adversely affect temperature forecasts beyond 24 hours.
3/19/2004
We are now placing the MM5 trajectories into
ftp://ftp.atmos.washington.edu/mm5rt/trajectory
They remain there for 4 days before being purged.
There are trajectory files for both d2 and d3. They are 12 hour
backward trajectories from forecast hour 18 back to forecast hour 6.
They are released at sigma level .98 (at hour 18) which corresponds to
a height above ground of about 125 meters. The trajectories have numeric
IDs starting with 1. The riptraj.titles file gives the numeric
ID in column one along with lat, lon, elev, and name of each trajectory
starting location. Elev is the actual point elevation of each site,
not the MM5 grid box elevation, which of course, is what the trajectories
use.
2/9/2004
Beginning with the 2004020912 run we turned off the old aug02 run and
made the dec03 setup the official, real-time run. The key aspects of
this change are as follows:
* Expanded 4-km domain now covers all of Idaho.
4 km terrain and land use.
* Slightly expanded 12-km domain.
12 km terrain and land use.
* MM5 version 3-6-2
* Using Intel Fortran Compiler version 7.1 with compiler flags
FCFLAGS = -O2 -ip -axW -tpp7 -rcd
* Continuing to use MPICH version 1.2.4
since the latest version 1.2.5.2 does not maintain the process
group of the executables launched by mpirun. Maintaining process
groups is essential for our job_control system.
* utilizing the dual-Xeon Beowolf cluster known as sage
* Reisner 2 microphysics on all domains (old was simple ice) this
has a significant computational impact -- 47% slow-down on the
4-km domain
* RRTM longwave radiation scheme (old was FRAD=2 cloud radiation
scheme)
12/17/2003
Beginning with the 2003121700 run we:
* turned on Reisner 2 microphysics for the dec03 (a.k.a testrt)
* turned on RRTM longwave radiation scheme for the dec03 (a.k.a testrt)
12/15/2003
Beginning with the 2003121512 run, we began running a new,
expanded MM5-GFS on the sage, dual-Xeon Beowolf cluster in a test
mode, but on a regular basis. This is known as the dec03
configuration (currently also known as the testrt run). Its key
components are as follows:
* Expanded 4-km domain now covers all of Idaho.
4 km terrain and land use.
* Slightly expanded 12-km domain.
12 km terrain and land use.
* MM5 version 3-6-2
* The old, aug02 configuration is still being run and placed in the
default locations for the MM5-GFS run.
* The new, dec03/testrt run is accessible over the Web at http://www.atmos.washington.edu/mm5rt/testrt.html
* Using Intel Fortran Compiler version 7.1 with compiler flags
FCFLAGS = -O2 -ip -axW -tpp7 -rcd
* Continuing to use MPICH version 1.2.4
since the latest version 1.2.5.2 does not maintain the process
group of the executables launched by mpirun. Maintaining process
groups is essential for our job_control system.
* Experimenting with Reisner 2 and RRTM schemes.
8/27/2003
Beginning with the 2003082712 run, the NetCDF files sent to WRH are
chopped down to fewer fields and levels.
8/14/2003
Beginning with 2003081500 run, TAPFRQ, BUFFRQ, INCTAP are now set as
follows for the 36/12-km runs:
TAPFRQ = 60.,
BUFFRQ = 60.,
INCTAP = 1,1,1,1,1,1,1,1,1,1,
which is a change from the old
TAPFRQ = 15.,
BUFFRQ = 15.,
INCTAP = 4,1,1,1,1,1,1,1,1,1,
Turns out that the 12-km grids do not match up bit-for-bit after
the first hour with this change -- something MMM is looking into.
The 15-minute files are obsolete.
7/29/2003
Updated mm52ncdf.exe code that generates Netcdf files for NWS to use
2-m temperatures and mixing ratios, and 10-m winds for the surface
fields.
6/11/2003
For just the 2003061112 forecast, we were unable to obtain ETA 221 grids
for initialization. As such, MM5-ETA and MM5-GFS used surface
temperatures, soil moisture, soil temperatures (the deep soil
reservoir temperature), and snow cover from the ETA 104 and GFS
1-degree grids respectively.
3/14, 3/17/2003
Moved many ventilation, air-quality, pbl height graphics to a new Web
page. Turned back on hourly graphics generation for 12-km domain
starting with 2003031712 run.
2/12/2003
Fixed a bug in GEMPAK files that corrupted all precipitation amounts
since 2003020312 switchover. NOTE: this does not affect the RIP
graphics displayed on Web or standard verification files.
Fixed a bug that was not incorporating the short 1-minute time step
values for the 4-km domain. This bug meant that fields such as
10-meter wind and 2-m temperatures were 0 at hour 6, the initial hour.
2/4/2003
We have scaled back on graphics generation -- previously we were
producing 5952 plots per forecast run! As of the 2003020500 run, we
are only producing graphics for the 12-km domain every 3 hours of
forecast time.
2/3/2003
Beginning with the 2003020312 forecast, we have made two key changes to
the MM5-GFS forecast.
1) the outer, 36-km domain, is now nudged toward the GFS solution
every 3 hours using the following criteria:
* 3-hourly 3D analysis nudging is applied (no surface analysis nudging)
* 3D (not surface) wind field nudged at 1.E-4
* 3D (not surface) temperature field nudged at 1.E-4
* 3D (not surface) mixing ratio field nudged at 1.E-4
* Boundary layer nudging is on for wind but not temperature or
mixing ratio.
2) the 36-km and 12-km domains are split off from the 4-km
domain and run first.
3) the 4-km domain begins at hour 6 instead of hour 0 and uses
1-hour 12-km domain files for initial and boundary conditions
as processed by NESTDOWN.
10/5/2002
Beginning with 2002100500 forecast, now running MM5 on Linux cluster
using a local, non-RAID IDE drive. Files are moved to RAID array
during the run, but this allows for an approximate 10 minute earlier
completion of the model run.
9/30/2002
Due to demands on the web server at U. of W., forced to change the
policy for storing old forecast data on the Web from 21 days to 14 days.
9/12/2002
Beginning with the 2002091200 forecast, the real-time is now beginning
the 4 km domain at hour 0 and running it simultaneously with the
outer grids. This change was made because a 36/12 km simulation
presents the following logistical problems: 1) we have are not receiving
the GFS grids (for boundary conditions) fast enough to keep up, and
2) we cannot postprocess (create graphics) fast enough to keep up --
we had to throttle the 12-km graphics. Overall runtimes for the two
methods are as follows:
36 and 12 (IDEALIZED) 1:25
36 and 12 (realized) 2:25 (model must wait for grids to arrive)
4 km alone 2:25
36/12 run plus a 4 km run 4:50
36/12/4 simultaneous 4:25
9/9/2002
Beginning with the 12Z forecast on Monday 9 September 2002,
2002090912, we began utilizing the new Athlon Beowolf cluster for
forecasting. Changes are:
* Expanded 4-km grid farther into Idaho.
4 km terrain and land use.
* Extended length of runs:
: 4 km domain forecasts from hours 9 to 48 (old was 12 to 36).
: 36/12 km domains out to 72 hours (up from 60).
* Initialization of most fields from NCEP's GFS (formerly known as
the AVN) at 1-degree resolution. The 36-km and 12-km domains
continue to be initialized with surface temperature, soil
temperature, soil moisture, and snow cover using the higher
resolution ETA 221 grids (roughly 40-km resolution).
* Continue a 36/12 km domain run using only ETA 221 grids for
initialization and ETA 104 grids for boundary conditions. This
MM5-ETA is run on the old Sun E6500 and is identical to the
previous real-time with the exception of the discontinuation of the
cloud water initialization. A separate Web page has been
established for this here.
* Switch to version 3.5 of the MM5 with MPP instead of OpenMP for
the GFS run
* discontinuation of cloud-water initialization
More details can be found here.
8/30/2002
8/29/2002
Reverted back to jun01 MM5 setup, rerunning 2002082912 to 2002083012
cases in old configuration. New configuration will continue as
a test until issues related to grid projection can be resolved with
all dependent software.
With 2002082912 run, BEGAN TESTING A NEW MODELING SYSTEM with:
* 108-km grid and expanded 4-km grid.
All domains
4 km terrain and land use.
* Extended length of runs:
: 4 km domain forecasts from hours 9 to 48 (old was 12 to 36).
: 108/36/12 km domains out to 72 hours (up from 60).
* Initialization of most fields from NCEP's GFS (formerly known as
the AVN) at 1-degree resolution. The 36-km and 12-km domains
continue to be initialized with surface temperature, soil
temperature, soil moisture, and snow cover using the higher
resolution ETA 221 grids (roughly 40-km resolution).
* Continue a 36/12 km domain run using only ETA 221 grids as a
backup or "Quick Look" run.
* switch to version 3.5 of the MM5 with MPP instead of OpenMP for
the GFS run
* discontinuation of cloud-water initialization
4/9/2002
Beginning with 2002040900, began running a test ensemble member,
avn-lkb, with a modified surface scheme for MRF initialized with
AVN.
Beginning with 2002041000, OUTPSIM, OUTPSIH, OUTBR, OUTWSCALE,
OUTHGAMT, OUTHGAMQ, and MOL now have correct values for 36-km
domain. From 2002031112 through 2002040912 the 36-km domain had
12-km domain values inserted. Click here
to see the comparisons of the OUTBR (Bulk Richardson number).
4/1/2002
With the 2002040100 run, we now initialize MM5 snow cover with the
water-equivalent snow depth field obtained from the ETA 221 grids.
The old source was the USAF snow cover data obtained from
ftp.ncep.noaa.gov in directory /pub/gcp/sfcflds/oper/live/. Serious
problems were evident with these data for several days and example
plots and further information is available here.
3/11/2002
With 2002031112 run, added the following fields to MM5 output files
with no effects on results.
Name Alt. Name Units Description
OUTPSIM PSI_M DIMENSIONLESS STABILITY PARAMETER FOR MOMENTUM
OUTPSIH PSI_H DIMENSIONLESS STABILITY PARAMETER FOR HEAT
OUTBR BR DIMENSIONLESS BULK RICHARDSON NUMBER
OUTWSCALE W_SCALE M**2 / S**2 MIXED-LAYER VELOCITY SCALE
OUTHGAMT GAMMA_T S**(-1) COUNTER-GRADIENT TERM FOR HEAT
OUTHGAMQ GAMMA_Q M**(-1) COUNTER-GRADIENT TERM FOR MOISTURE
MOL T_STAR Kelvin MIXED-LAYER TEMPERATURE SCALE
MAXWSP10 M/S MAXIMUM 10-meter windspeed
The last occurrence of the
maximum single timestep 10-m
wind speeds occurring in the
hour prior to the reported
hour.
MAXWDIR10 DEGREES Corresponding grid-relative
wind direction for MAXWSP10.
MAXT2 Kelvin Maximum single timestep 2-m
temperature occurring in the
last hour.
MINT2 Kelvin Minimum single timestep 2-m
temperature occurring in the
last hour.
1/24/2002
With 2002012412 forecasts, began using Workshop 7 f90 and Guide 4.0
under Solaris -- previously using Workshop 6 (non-updated) f77 and
Guide 3.8. Comparisons show that these results differ SLIGHTLY (with
radiation fields having the largest differences near convective
regions). Compared to results obtained under OSF1 f90, the
discrepancies may have decreased, but the overall pattern of
differences appears the same. Click here for a loop of
difference plots, here for a loop of raw fields to compare, and
here for a description of the plots.
1/11/2002
With 2002011100 forecasts, implemented rewriting of initial forecast hour
PBL-scheme and radiation-scheme 2-d fields with values taken as near
to 1 minute into the simulation as the timestep allows. Fields
affected are:
PBL HGT, REGIME, SHFLUX, LHFLUX, UST, SWDOWN, LWDOWN, SWOUT, LWOUT,
T2, Q2, U10, and V10.
This is intended to show reasonable fields even at hour 0.
Adjustments do occur rapidly and these values appear reasonable.
Also with 2002011100, began writing out mmout_d[123].pblfile files
containing just the near 1-minute PBL-scheme fields of:
PBL HGT, REGIME, SHFLUX, LHFLUX, UST, T2, Q2, U10, and V10.
1/3/2002
Added probability of frozen precipitation, 1-hour snow accumulation,
and freezing level graphics for 12 and 4 km domains.
9/19/2001
With 2001091900 run, added additional graphical products such as
thickness, max CAPE, 300 mb isotachs, and 700 mb temperature.
Also added cloud fields to the 12km domain.
8/3/2001
With 2001080400 run, made default runs on tahoma use a 27-processor set.
8/1/2001
With 2001080200 run, added one more CPU tahoma, bringing us to 30.
Locking 29 of 30 (2:57,4:15 and 2:56,4:08) did not improve runtimes
over 27 of 28 (typically 2:53 and 4:05).
7/30/2001
With 2001073100 run, added one more CPU tahoma, bringing us to 29, and
replaced memory resulting in full 16-way memory interleaving. No
change in model runtimes evident.
7/27/2001
With 2001072800 run, replaced a bad memory module, SIMM, on tahoma.
Now at mostly 4-way and 8-way memory interleaving. Still waiting for
SUN to replace bad board. No change in model runtimes evident.
7/24/2001
With 2001072500 run, added 2 more CPUs to tahoma. Now at 28 of 30,
still waiting for SUN to replace bad board. 36/12 km now runs in 2:53
and 4 km domain runs in 4:08. Complete model runs are finished by
2:18 am/pm.
7/23/2001
With 2001072400 run, added 2 extra CPUs to tahoma. SUN sent 3 boards
with 2 CPUs each and only one board was working. 36/12 km runs went
from 3:08 to 3:00, 4 km domain went from 4:55 to 4:20.
7/9/2001
With 2001070912 run, incorporated the following changes to the
modeling system:
* expanded 12 km domain to cover Idaho and western Montana
(click here for a description of the domain boundaries and
TERRAIN namelist)
* added 5 new layers in the vertical, mostly in the boundary layer,
bringing us to the following full sigma levels:
1.00, 0.995*, 0.99, 0.985*, 0.98, 0.97, 0.96, 0.95*,
0.94, 0.93*, 0.92, 0.91*, 0.90,
0.88, 0.86, 0.83, 0.80, 0.77, 0.74, 0.71,
0.68, 0.64, 0.60, 0.56, 0.52, 0.48, 0.44, 0.40,
0.36, 0.32, 0.28, 0.24, 0.20, 0.16, 0.12, 0.08, 0.04, 0.
* = new
* converted from MM5 version 2.12 to MM5 version 3.4 (at a cost in
speed of approximately 10%)
* using all MM5 version 3.x preprocessing (TERRAIN, REGRID, INTERPF,
and NESTDOWN)
* using RIP 3-0 for most graphics
* complete rewrite of real-time scripts
12/21/2000
Correctly implemented deep soil reservoir temperatures that were intended
to begin 7/31/2000. First forecast using these is 2000122200.
* Deep soil (30 cm) reservoir temperature is now initialized with the
10-40 cm values given in the ETA 221 initial conditions. Should an ETA
221 forecast not be available for a given date, the most recent data
will be used.
11/28/2000
Noted that forecasts from 2000111300 to 2000112800 erroneously used SST
data from 2000111112 -- due to change in gdinfo in GEMPAK 5.6 upgrade.
11/23/2000
Returned to proper KMP_LIBRARY environment variable setting, thus improving
model run times by approximately 26 minutes for the 36/12 km run and by
15 minutes for the 4 km run starting with the 2000112300 simulation.
11/21 and 11/29/2000
Added Washington state ventilation index graphics effective with 2000112200
run for 4-km domain, and 2000113000 run for 12-km domain.
10/12/2000
Added 1-hour precip graphics to 36 km domain and 24-hour precip at hour
60 to 36 and 12 km domains with the 2000101200 run.
8/18/2000
Added Fire Weather Index, Surface Wind Speed, and Surface RH graphics
with the 2000081812 run. A brief description of the fire weather index
can be found at: http://met.rfl.psw.fs.fed.us/fwi/forecast_general.html.
Our MM5 calculation of the fire weather index uses 10 m winds and
2 m temperatures and relative humidities.
7/31/2000
Began running 4 km domain over the all of Washington and Oregon with the
2000073112 run.
This continues as MM5 v2.12 with the following additions/fixes:
* As mentioned, 4 km domain covers Washington and Oregon.
* We are once again forecasting the 4 km domain from forecast hours 12 to 36.
* 36 and 12-km domains are now forecasting out to hour 60.
* 36-km domain has been expanded south and west to near Hawaii in SW corner.
* No u-star sub-grid scale drag modifications are included.
* Deep soil (30 cm) reservoir temperature was to be initialized with the
10-40 cm values given in the ETA 221 initial conditions; HOWEVER, due to
a programming error, this was not properly implemented until Dec 22, 2000.
All deep soil temperatures are initialized with average values from the
INTERP job which generally just the average between the 0 and 3 hr
forecast.
* New graphics have been developed.
6/15/2000
After noticing extremely low wind speeds over largely sloping terrain,
at Mt. Rainier locations, we have removed the Ustar wind-speed drag
modifications from our MM5 starting with the 2000061600 run.
6/1/2000
Upgrade to real-time computers.
4/10/2000
New 4 km graphics were put into place. These include pblht and vent.
This involved changes to create_graphics.pl and create_gifs.pl.
Changes were also made to the web scripts.
2/25/2000
Air temperatures plotted in meteograms have been changed from being
the temperature at the lowest sigma level (about 40 m) to an
estimated "2 meter" temperature which is computed as:
T = ( ground surface temperatue + 40 m temperature ) * .5
1/4 to 1/24/2000
MM5 ensembles begun.
The 36/12 km domains are forecast with the AVN, NGM, NOGAPS, and
Canada's GEM model for the 00Z cycle (see this link for more information).
Full suite of ensembles made public 1/24/200.
12/21/1999
Bug fixes for latest, dec99, modeling system:
* Several fields in the mifc character array of the mmout files were seriously
garbled. Terrain variance file reading moved back to location in code prior
to CALL ADDALL fixed this. Bad headers for 1999121300 through 1999122012.
* Soil temperature data initialization using 40-day running mean was inoperative
for forecasts 1999121300 through 1999122112. Turned back on with 1999122200.
12/13/1999
Began running 4 km domain over the whole state with the 1999121300 run.
This continues as MM5 v2.12 with the following additions/fixes:
* We are now forecasting the 4 km domain from forecast hours 6 to 36.
* A bug that was introduced in the may99 version has been fixed. The
bug was incorrectly reading the terrain variance file in calculating
the u-star mods for domain 3 only.
* We are now using the version 2-12 TERRAIN program and are now using
the standard landuse.tbl and the 24 land-use categories of the
USGS datasets.
* The following are the values used in the TERRAIN namelist:
PHIC = 48.7 ; CENTRAL LATITUDE (minus for southern hemesphere)
XLONC = -138.8 ; CENTRAL LONGITUDE (minus for western hemesphere)
NESTIX= 101,100,124, 76, 1, ; DOMAIN IX
NESTJX= 137,103,172, 91, 1, ; DOMAIN JX
DIS = 36.,12., 4., 1., 1., ; GRID SIZE(KM)
MTHRD = 1, 1, 2, 3, 1, ; MOTHER DOMAIN ID
NESTI = 1, 31, 42, 39, 1, ; LOWER LEFT I OF NEST
NESTJ = 1, 85, 31, 18, 1, ; LOWNER LEFT J OF NEST
10/12/1999
Corrected a bug in our version of interp that was causing step-like features
in the boundary and initial fields for one-way nests, i.e., domain 3.
The effect of this bug on the model has not yet been determined.
10/5/1999
Meteograms for 12 km domain sounding stations added to all Web pages.
Meteogram generation starts when all 48 hours have been completed.
Process is currently separate from MM5 real-time scripts.
9/29/1999
The modeling system has been updated to more seamlessly handle switching
to alternate models when the ETA 104 does not arrive on time. The following
scripts have been updated:
run_mm5rt.pl
preprocess.pl
waitpreprocess.pl
bdycom.pl and bdycomd3.pl
rm.old.data.files.pl
9/16/1999
When running gridder and using snow files, note that the format
has changed in the GEMPAK snow files as of 1999091600. Instead of
being at glevel = 0, gvcord = pres (@0%pres), the snow and ice fields
are now at glevel = 0, gvcord = none (@0%none). This is a simple
change to make in the gridder namelist or in preprocess.pl:
snowcoord = 'none',
The real-time preprocess.pl has been changed.
Why the change? It is because the newly installed GEMPAK
(NAWIPS-5.4.12) has a slightly different
$GEMTBL/grid/vcrdgrib1.tbl file for use by nagrib.
Same change goes for SST as for snow. Use
sstcoord = 'none',
in gridder's namelist.
OTIS grib file names have also recently changed, but the GEMPAK file
names are consistent.
6/23/1999
Beginning with run 1999062312, model executable is compiled with Guide 3.7.
6/12/1999
Beginning with run 1999061200, corrected cloud-water initialization
has been turned on.
6/8/1999
Beginning with run 1999060812, switched to MM5 version 2.12 from our
modified version of 2.7. See NCAR's change log
for more information on the differences between 2.7 and 2.12.
5/28/1999
Beginning with 1999052812, model is initialized with 0-hour ETA 221
32-km gridded data, as available. We continue to interpolate from
ETA 104 80-km grid for boundary conditions at later forecast hours.
5/26/1999
Initialization of cloud mixing ratio fields for 36 and 12 km domains, using
Eta model gridded cloud fields begins with model run initialized 1999052800.
This data was included in the following model runs:
1999052800
1999052812
1999060300
1999060312
Cloud initializatoin has since been discontinued pending further testing.
5/20/1999
Improved calculation of soil substrate temperature array, using a combination
of 40-day and 2-day running mean air temps. This first implemented for model
run initialized 1999052100 UTC.
Previous model simulations used forecast future temperature for this purpose.
5/18/1999
Major changes in MM5 real-time modeling system operations implemented with
the 1999051812 forecast:
* 36 and 12 km domains run to 48 hours before 4 km starts
this results in the following time changes:
Domains Forecast Approximate Finish Time Time Change
1,2 24-hr 9:10 am/pm -1 hr 50 min (earlier)
1,2 36-hr 10:00 am/pm -3 hrs (earlier)
1,2 48-hr 10:45 am/pm -3 hrs (earlier)
3 12-hr 10:45 am/pm +1 hr 45 min (later)
3 24-hr 12:20 am/pm +1 hr 20 min (later)
3 36-hr 1:10 am/pm +20 min (later)
All complete 1:10 am/pm -30 min (earlier)
* 12-km domain has been expanded from 82x82 to 88x88
* can now start up with input cloud water fields
* can use ETA 221 32-km grids for initialization of hour, this is better than
the 80-km ETA 104 grids
* can now start up with improved soil temperatures
* preprocessing for all domains is done incrementally on the fly,
rather than all at once
* preprocessing has been moved to other computers
* output files all use 4-digit years
Note, we continue to use the same physics packages and MM5 version 2.7, though
we have upgraded to Guide 3.6 and Fortran 5.0.
Known bugs:
4/22/1999
Fixed a major bug in postprocessing script that was introduced in
March 1998 configuration of the real-time.
Result: model now finishes at least 75 minutes earlier!
2/22/1999
Changed postprocessing to include new "ripdp" version of rip. Also added
Ernie's changes which create improved labelling and color bars.
10/26/1998
Fixed problems with the domain 3 (4km) vertical cross section plots. Some
parameters were being plotted on a cross section shifted about 10 km from
where it should have been located.
8/10 and 8/17/1998
Re-formatted the alternate format pages (current and past date
versions) to make them narrower. Lines now wrap, and the table now
has a border drawn where possible.
7/23/98
Fixed the alternate format page for non-current dates so that all available
graphics are now accessible. No changes to current page or standard page
needed.
7/22/98
Added mods to zero out initial accum rainfall in initnest.F. Changes take
effect beginning with model run 98072200. No significant changes expected
in results or physics.
5/22/98
Changed from winter to summer color fill scheme for 925 and 850 mb
temperature plots in RIP.
5/22/98
Beginning with model run 98052300, we have turned off snowcover in
latitudes south of 55N. In November, the snowcover will apply to
latitudes south of 50N if elevations exceed 1219 meters. In January,
February, and March, snowcover will apply wherever the USAF snowcover
has interpolated snowcover onto the domain.
5/8/98
Beginning with model run 98050900, we have modified the 5-layer soil model
to correct for soil temperatures below snow and to correct surface
temperatures over snow. Prior results in warm weather had very cold
surface temperatures (order of 250 K) and very hot upper-level soil
temperatures (order of 340 K).
3/31/98
Beginning with model run 98040100, we have modified the preprocessing
programs to calculate RH with respect to water, at all levels.
3/18/98 (later in the day)
Beginning with model run 98031900, we have switched to using 3 hourly
boundary conditions from the ETA model, now being supplied by NCEP.
Previously, this data was only available at 6 hour intervals.
3/18/98
Several events have taken place during the past two days.
98031612: mm5 crashed
98031700: mm5 crashed
98031712: switched to an older version of mm5, which does NOT contain
ustar "form-drag" modifications and other bug fixes, while in the
process of diagnosing the problems with the mm5. This run was
successful.
98031800: NCEP hardware problem, ETA model output unavailable.
Beginning with model run 98031812, we have returned to mm5v2.7, including
the ustar modifications. This real-time mm5 system is now configured with
an expanded outer domain to extend further west, north and south. We are
also ingesting snow cover data from the U.S. Air Force, as well as new,
1/4 degree (higher resolution than before) SST data from the U.S. Navy's
OTIS ocean model.
The mm5 problems were traced to a bug in the mm5's "MRF" pbl scheme, and
this problem has been corrected.
2/4/98
Replaced the u-star modifications to the pbl scheme, since they have
been cleared of any wrongdoing, at least for the present.
So, u-star mods were missing from 98012900 until 98020412.
1/28/98
Removed the ustar mods from the mrf pbl scheme, which had been added
on 1/20/98, due to apparent problems which affected some runs.
The runs which failed to integrate through the entire 48 hour forecast
are: 98012700, 98012712 and 98012812. It is not yet clear, but possible
that all runs from 98012000 through 98012812 contain compromised
output.
1/20 - 1/21/98
Switched to version 2.7 of MM5 with ustar mods. Expect 5-10% speed up
of model.
Added numerous sounding locations and 700 mb vertical velocity graphics to
the home pages.
The ustar mods are a simple scheme which increases "form drag" (otherwise
known as friction) on the low level winds, depending on the roughness of
the terrain. The variance in terrain height is calculated from a high
resolution dataset, and then read in by the mm5. The mm5 adds a term to
the frictional component of the total wind field which is proportional to
the total wind speed and terrain variance, so that in rough terrain,
wind speed is reduced, while in other areas wind remains unaffected.
1/12/98
Added 1-hour and 24-hour (at 36 and 48 hours) precipitation graphics to
home pages.
12/18 - 12/19/97
Added hourly output to 12 km and 4 km domains after forecast hour 12.
Updated home pages to retrieve new hourly graphics and to auto-refresh
every 10 minutes instead of every 5 minutes.
12/1/97
Added map showing locations of vertical cross sections.
11/12/97
Added links to several alternate format pages that provide a faster
interface to the graphics than the standard, form request method.
11/9/97
Added 4 km domain vertical cross sections.
10/24/97
Updated MM5 home page (http://www.atmos.washington.edu/data/mm5.cgi)
to reflect the latest graphics--(specifically added links to 36km
domain cloud water fields and to 500 mb heights/vorticity plots for
all domains).
Added a link to this log file.
10/16/97
Added links to access statistics and revised info page.
Made the home page automatically refresh every 5 minutes (so users
always get the latest status message on the current forecast).
10/13/97
MM5 runs initiated once the 12-hour ETA forecast is available, rather
than waiting for all 48 hours of ETA forecasts. This is made
possible by the restarting of the model at forecast hour 12 to begin
the 4 km domain.
9/29/97
4 km domain added.
Back to MM5
Home Page