========================================================================== Retro experiment list at http://ruc.noaa.gov/internal/RR_runs/RRretrolist.txt To do list: 1) structure based on Subversion 2) partial cycling : done perform cold start: 09Z and 21Z How long to go back? 6 hours, so use GFS 00Z 3h forecast valid at 03Z for 09Z start and GFS 12Z 3h forecast valid at 15Z for 21Z start 3) install system in NCEP time line September 2009 !!! executables data source single runs (using binary) cycling 5) adding and enhancing RR features cloud analysis RUC LSM surface variables cycling surface data analysis ... 6) tuning and bug fixing obs and background error covariance crash on boundary and level 30 ... ========================================================================== ========================================================================== oprRR: account: rtrr xml: /whome/rtrr/RR13/xml bin: /whome/rtrr/RR13/bin exec:/whome/rtrr/RR13/exec code:/whome/rtrr/RR13 working directory: /misc/whome/rtrr/rr_databasedir MODEL: WRF-ARW3.1 GSI: Q1FY09 devRR: account: mhu xml: /whome/mhu/RR13V31/xml bin: /whome/mhu/RR13V31/bin exec: /whome/mhu/RR13V31/exec code: /whome/mhu/RR13V31/trunk working directory: /mnt/lfs0/projects/wrfruc/mhu/RR13V31 MODEL: WRF-ARW3.1 GSI: Q1FY09 ========================================================================== 2009-11-16: Ming Hu, (Mon Nov 16 18:31:44 GMT 2009) For both RR, update WPPP to r107 (MDL2P.f) ========================================================================== 2009-11-09: Ming Hu, For oprRR (start from 2009110916): update to latest version of the cloud analysis and GSI which include: 1) bug fix for cloud analysis when uses METAR 2) put cloud analysis after iteration 3) do mositure supersaturation and negative adjustment 4) dump cloud analysis check file 5) no NASA LaRC data yet (need to change ksh) Before this switch, GSI is global_gsi_CloudR38_NoNASA_NoQadjust ========================================================================== 2009-11-9 Ming HU: 1) for oprRR, use GSI r43 (2009110916) This include all fix and tunning for cloud analysis and runing the cloud analysis after GSI and mositure adjustment 2) for both oprRR and devRR make sure snow trim only work once a day (at 02Z). Now snow trim include for the temperature check and snow/rain/graupel mixing ratio checks. 3) add snowc to full_surface_cycle ========================================================================== 2009-11-3 Ming HU: tunning cloud top pressure in cloud analysis, the last change in devRR is 20:17Z 2009-11-3 ========================================================================== 2009-11-2 Ming HU: tunning the cloud analysis, the last change in devRR is 9:11Z 2009-11-1 this change impact cycle: 2009110109 The last change in oprRR is 2009-10-26 ========================================================================== 2009-10-30 (and 29, 28) Ming HU: For devRR: fixed several bugs in the cloud analysis when uses METAR. These bugs impact cloud base analysis mostly. ========================================================================== 2009-10-25 Ming HU: for oprRR and devRR, cloud analysis in FY09 (start: 2009102521) 1) fix the bug that uses ASL instead of AGL in METAR cloud base. 2) for devRR, using cloud top pressure to suppress the convection and q adjustment and cloud analysis is after optimal iteration. ========================================================================== 2009-10-25 Ming HU: for oprRR, cloud analysis in FY09 (start: 2009102521) 1) fix the wrong values in rh_cld3_p and rh_clear_p. 2) this version include using cloud top pressure to suppress the convection. ========================================================================== 2009-10-23 Ming HU: 1) Use new WPP (revision 106) for devRR (18Z) and for oprRR (22Z) 2) for oprRR, turn of snow trim because of partial cycling. ========================================================================== 2009-10-13 Ming HU: (start from cycle 2009101315) for oprRR: 1) put fix of radar temperature tendency DFI ========================================================================== 2009-10-10 Ming HU: (start from cycle 2009101104) for devRR: 1) put fix of radar temperature tendency DFI 2) fix of ruclsm (r100 repository) ========================================================================== 2009-10-08 Ming HU: for oprRR: 1) fix of ruclsm (r100 repository) ========================================================================== 2009-10-05 Ming HU: (start from cycle 2009091717) for oprRR: 1) analysis switch to Q1FY09 version of GSI (start: 2009100522) 2) start partial cycle (start: 2009100522) 3) use smoothed terrain from Joe (wait for several hours to work) ========================================================================== 2009-09-17 Ming HU: (start from cycle 2009091717) for oprRR and devRR: change damp_opt to 1 from 3 damp_opt = 1, dampcoef = 0.02, 0.01, 0.01 Ming HU: (start from cycle 2009091717) turn off partial cycle to test GSI Q1FY09 ========================================================================== 2009-09-10 Ming HU: (start from cycle 2009091017) for oprRR : use WRF3.1 r98 according to Tanya: changes include: 1. Call Grell's convective scheme on the whole domain 2. Back off zeroing out microphysics temperature tendencies in the 4-point halo along the boundaries. 3. Revised 2-m diagnostics (new subroutine module_sf_sfcdiag_ruclsm.F) 4. Revised RUC LSM for provide correct coupling to MYJ PBL, and also some cleaning of the code. 5. Updated Registry to add QVG and QCG to the history files. ========================================================================== 2009-09-09 Ming HU: for oprRR and devRR fixed a bug in cloud analysis: GSI uses specific humidity in ges_q cloud analysis uses mixing ratio and then cloud analysis put mixing ratio back to ges_q. this fix effect in late afternoon. ========================================================================== 2009-09-09 Ming HU: for oprRR : use executables from /whome/smirnova/RR-svn/trunk/WRF3.1 Tanya's email: John and I think that it could be an interesting test to call Grell's scheme on all points where Thompson scheme is called as we talked about at the meeting. Also, zeroing out microphysics temp. tendencies in 4 rows and 4 columns along the boundaries didn't prevent from a crash - so we backed off this change. I have recompiled the code with these changes. Could you, please, use executables from my directory /whome/smirnova/RR-svn/trunk/WRF3.1 until the next crash? ========================================================================== 2009-08-27 Ming HU: for oprRR : use WRF3.1 revision 90: 0 temperature tendency along southeast corner start at : 2009082717 ========================================================================== 2009-08-18 Ming HU: for GSI in oprRR (/misc/whome/rtrr/RR13/GSI_RR/trunk2/trunk/gsi_code): 1) radar temperature tendency is 0 (missing is -99999) and possible cause problem in DDFI. to fix the problem, change code in unfill_mass_grid2.f90 to turn off the boundary filter for radar temperature tendency and call this special subroutine in wrwrfmassa.F90 call unfill_mass_grid2t_t(tempa,im,jm,temp1) This fix made at Tue Aug 18 17:52:25 GMT 2009 ========================================================================== 2009-08-05 Ming HU: for oprRR and parallel: 1) turn of fraction seaice /whome/rtrr/static_rr13_boundary/wrf.nl: fractional_seaice = 0 /whome/rtrr/static_rr13_cycle/wrf.nl: fractional_seaice = 0 in snow trim: /tg1/projects/wrfruc/mhu/GSI/comGSI/trunk/lib/sorc/imssnow/update_SNOWICE_netcdf_mass.F90: fractional_seaice=0 ========================================================================== 2009-08-04 Ming HU: for oprRR: 1) set the number of tries for the model run (wrf_arw) in XML scripts equal 1. 2) modified start_em.F which save 500 precious seconds (Aug 4 16:23:50 MDT 2009) ========================================================================== 2009-07-21 Ming HU: change the following two aspects in oprCycle. These changes have been tested in parallel cycle for one day. 1) snow/sea ice trim (r19, r20) change surface parameters in grid point that has sea ice changed ! For sea-ice: ! vegcat(i,j)=24 not used ! ivgtyp(i,j)=24 int IVGTYP ! lu_index(i,j)=24 float LU_INDEX ! landmask(i,j)=1. float LANDMASK ! xland(i,j)=1. float XLAND ! isltyp(i,j)=16. int ISLTYP ! ! For water: ! vegcat(i,j)=16 ! ivgtyp(i,j)=16 ! lu_index(i,j)=16 ! landmask(i,j)=0. ! xland(i,j)=2. ! isltyp(i,j)=14. 2) change GSI ksh to make sure each cold start have snow/sea ice trim ( old version only partial cycle has snow trim, while full cycle can cause crash in because of bad surface fields from previous forcast) ========================================================================== 2009-06-25 1) Ming HU: put WRFV3.1 into operation RR: Thu Jun 25 11:35:03 MDT 2009 ========================================================================== 2009-06-22 1) Ming HU: turn on radar DDFI at WRF3.1 cycle since 2009062215 ========================================================================== 2009-06-19 1) Ming HU: Add radar DDFI code into WRF3.1 but forgot to turn it on in namelist (Fri Jun 19 18:55:27 GMT 2009) 2) Tanya. change tag 62, 63 are in this time or even on 16, June. ========================================================================== 2009-06-12 1) Tanya: in the main RR cycle and also in 3.1 test cycle: turn on 6th order diffusion in wrf.nl: diff_6th_opt = 2, diff_6th_factor = 0.12, 2) Tanya In both oprRR and 3.1 test cycle: compile WRF and WRF3.1 to include changes with tag 58-61 ========================================================================== 2009-05-19 1) oprRR by Tanya: changed dampcoef to 0.05. It went into 15z bc run, and 16z cycled. ========================================================================== 2009-05-13 1) oprRR by Tanya: Old: HGT_M has no smoothing but HGT_U and HGT_V still have smoothing New: RR with no smoothing for all three topography fields ========================================================================== 2009-05-04 1) oprRR : make the following change to ARW namelist: new values: damp_opt = 3, dampcoef = 0.2, 0.01, 0.01 old ones: damp_opt = 1, dampcoef = 0.02, 0.01, 0.01 The new values work at: 2009050415 ========================================================================== 2009-04-24 1) make the following changes In METGRID.table: ======================================== name=SNOW interp_option=four_pt+average_4pt # masked=water fill_missing=0. flag_in_output=FLAG_SNOW in namelist: &physics usemonalb = .true. 2) oprRR: new surface cycle scheme in cycle. code: /whome/rtrr/RR13/GSI_RR/trunk2/trunk/updatebk schemes (Tanya): When the RR cycle is broken there could be 2 situations: 1 - there IS a previous RR forecast for current time FULL cycling - Cycle ALL surface variable from our list TSLB SMOIS TSK SNOW SNOWH SOILT1 CANWAT 2 - there is NO previous RR forecast for current time, but there is a forecast for the same time of the day but one or two days earlier PARTIAL cycling: a) For the forecast one or two days earlier: Trim snow using NESDIS data - SNOW, SNOWH, SNOWC b) Cycle: SMOIS - all levels of soil moisture SNOW, SNOWH, SNOWC TSLB - levels 3,4,5,6 of soil temperature c) Cycle TSK, SOILT1 and TSLB(1,2) only for snow grid points and keep cold-start TSK and TSLB(1,2) for snow-free points ========================================================================== 2009-04-23 1) oprRR: put snow trim function in cycle. Snow is trimmed at 04z every day code: /lfs0/projects/wrfruc/mhu/GSI/GSI_RR/trunk2/trunk/imssnow ========================================================================== 2009-04-13 1) oprRR: turned on aircraft data. ========================================================================== 2009-03-27 1) oprRR: turned off aircraft data to check the 3 day boundary crash. ========================================================================== 2009-03-25 1) devRR: turned down because of no computer source. ========================================================================== 2009-03-14 1) change qi, qc > 1e-12 to qi, qc > 1e-6 in dfi moisture adjustment . 2) devRR: tune GSI based on March 2008 NAM version: background influence scale and fix files. It seems RR crashes more often. ========================================================================== 2009-03-11 1) add cloud analysis into Q1FY09 veriosn of GSI in devRR ========================================================================== 2009-02-25 1) turn on bias correction for both oprRR and devRR 2) install higher-resolution terrain for both oprRR and devRR 3) start to use Q1FY09 GSI in devRR Now Q1FY09 can run on WJET but the following GSD contributions need to add in: a) Qoption 2 b) boundary filter c) temperature and moisture relation based on virtual temperature d) cloud analysis e) radar temperature tendency ========================================================================== 2009-02-13 1) oprRR: According to Dezso Devenyi Ming, please add the following line in read_prepbufr.f90 if(kx>179.and.kx<190) toe=toe*r0_5 !DEDE 12 Feb 2009 Here is the code environment Temperature if(tob) then ppb=obsdat(1,k) call errormod(pqm,tqm,levs,plevs,errout,k,presl,dpres,nsig) toe=obserr(3,k)*errout if(kx>179.and.kx<190) toe=toe*r0_5 !DEDE 12 Feb 2009 qtflg=tvflg(k) if (inflate_error) toe=toe*r1_2 if(ppb < r100)toe=toe*r1_2 cdata_all(1,iobsout)=toe ! temperature error cdata_all(2,iobsout)=dlon ! grid relative longitude cdata_all(3,iobsout)=dlat ! grid relative latitude cdata_all(4,iobsout)=dlnpob ! ln(pressure in cb) ========================================================================== 2009-02-12 1) Build a new RR cycle based on RR repository: The new cycle is on lfs0, /mnt/lfs0/projects/wrfruc/mhu/RR13 2) Now we name the RR cycle on rtrr as oprRR: which is on /mnt/lfs0/projects/rtrr/RR13 name the RR cycle on mhu account as devRR: which is on /mnt/lfs0/projects/wrfruc/mhu/RR13 3) devRR is identical to the oprRR NOW ========================================================================== ========================================================================== 2009-02-09 1) 0 h forecast on our web is now analysis grids instead of DFI since 2009020919 ========================================================================== 2009-01-30 1) RR1h in lfs0: start to use radar temperature tendency in DFI 2) update WRF SVN ========================================================================== 2009-01-28 1) RR1h in lfs0: start from cycle : 2009012823 * use modified cloud analysis package (Russ comments and radar temperature) * save surface variables. Here is a list of variables saved: TSLB:description = "SOIL TEMPERATURE" ; SMOIS:description = "SOIL MOISTURE" ; TSK:description = "SURFACE SKIN TEMPERATURE" ; SNOW:description = "SNOW WATER EQUIVALENT" ; SNOWH:description = "PHYSICAL SNOW DEPTH" ; RHOSN:description = " SNOW DENSITY" ; 2) update GSI SVN ========================================================================== 2009-01-26 1) RR1h in tg1: use modified cloud analysis package ========================================================================== 2009-01-02 1) RR1h in lfs0: use exec in trunk to do the cycle since 2009010918 (ARW) also use Tanya post-process: /whome/wrfruc/WPPV3_ceil/exec/wrfpost.exe ========================================================================== 2009-01-02 1) RR1h in lfs0:get rid of observation in southeast corner (100X100 grid) to check possible reason for RR crash RR cycle in lfs0 cold start from 2009010217 2) RR1h in tg1: turn off w_damping and cold start cycle at 2009010222 ========================================================================== 2008-12-23 1) get rid of observation in southeast corner (30X30 grid)to check possible reason for RR crash RR cycle in lfs0 cold start from 2008122319 ========================================================================== 2008-12-19 1) according to Tanya: cold-start RR uses option 2 for SW radiation rather than Dudhia option 1 I put it into RR to include effects of cumulus scheme on radiation. SO, RR switched to option 2 for SW today: For tg1 cycles: new option starts at 2008121918 THE run time at 121918z is 3096s, while run time at 16z is 3104 and 15z is 2998s For lfs0 cycles: new option starts at 2008121920 THE run time at 121920z is 2786s, while run time at 18z is 2894 and 19z is 2879s ========================================================================== 2008-12-15 1) according to Tanya: make changes to postprocess for visibility: You can use the new directory /whome/wrfruc/WPPV3_vis or just copy CALVIS.f and MDLFLD.f from /whome/wrfruc/WPPV3_vis/sorc/wrfpost. Changes were made for both cycles tg1: /whome/mhu/WPPV3_test -- with copygb (copied from /whome/wrfruc/WPPV3_vis) lfs0: /whome/rtrr/RR13/WPPV3 Changes should be seen SINCE cycle 2008121522 ========================================================================== 2008-11-21 1) make changes to qcmod.f90 to fix bug: Wanshu: The bug is in errormod. The bug gives unrealistically large oberrors to the raobs, aircraft and profiler data in the upper layers. The main fix is changing from "max" to "min" 2) Use new nam_errtable.rd3v 3) make TKE_MYJ change to both RR cycles 4) fix a bug in sfcextract 5) add GSI analysis to post-process ========================================================================== 2008-09-23 1) Use filter in boundary area to reduce the impact of analysis increment along boundary This is the reason of RR crash. After filter boundary, RR already successfully continue running 4 days The filter is in unfill_mass_grid2.f90 2) Start to use new cloud analysis code send to NCEP. The new GSI is in: /misc/whome/mhu/GSI_RR/trunk2/trunk/ ========================================================================== 2008-09-15 The following changes made to the RR recently 1) SST uses GRIB2 data 2) read PREPBUFR and BUFR from public directory, which is stable and a little earlier /public/data/grids/rr/para/prepbufr /public/data/grids/rr/para/dump 3) Start to use hjet 4) Some change in post process and ncl to make Alaksa and CONUS plots ========================================================================== 2008-08-08 Use new post process and ncl. Starts from 2008080820 ========================================================================== 2008-08-07 Use new terrain that was smoothed by Joe in south boundary. The run started at 2008080723 ========================================================================== 2008-08-03: Extend relax boundary zone for ARW forecast from 5 to 10 grid points ========================================================================== Special features of GSI and ARW (2008-07-11): GSI: March 2008 NAM version with the following changes: 1) fixed lots of IO problems related to the use of ARW 2) moisture analysis (use qoption = 2 right now) background error variance for qoption =1 (compute_derived.f90 from NCEP) turned off relation between temperature and moisture (tv_to_tsen.f90) qoption=2 works for ARW 3) parallel cloud analysis and cloud observation IO 4) turn off soil moisture and skin temperature updating. 5) normal wall time for GSI is 720 s with 64 processors 6) updating SST at 01Z ARW: Version 3: 0) DFI is used (option 3, 40 min backward and 40 min forward) 1) use quilt to reduce IO time (8 processors) 2) combined wrf_input and wrfout 3) have boundary condition problems for dfi in last three cycles. right now use extra boundary files to bypass the problem. 4) normal wall time for ARW is 3200 s with 120 processors 5) use short DFI integration time step time_step_dfi = 40, 6) microphysics step: mphdat = 1 ========================================================================== RR setups in lfs0(2008-11-20): Workflow: /whome/rtrr/RR13/xml WRFRR_v3_BOUNDARY.xml : outside cycle (6 hours) WRFRR_v3_CYCLE_1H.xml : 1-h RR cycle Scripts: /whome/rtrr/RR13/bin Code: /whome/rtrr/RR13/ ========================================================================== RR setups (2008-07-11): Currently, Rapid Refresh 1-h cycles is running on WJET and tg1 disk Workflow: /tg1/projects/wrfruc/mhu/xml_RR13 WRFRR_v3_BOUNDARY.xml : outside cycle (6 hours) WRFRR_v3_CYCLE_1H.xml : 1-h RR cycle Scripts: /tg1/projects/wrfruc/mhu/bin_RR13 Code: /whome/mhu/ WPSV301 WPPV3 WRFV3x GSI_RR/trunk2/trunk Outside cycle runs the following tasks every 6 hours to generate boundary condition and cold start background: ungrib_GFS ungrib_SST metgrid_arw real_arw3 real_arw2 real_arw wrf_arw RR13_clean 1-h RR cycle runs the following tasks every hour: lightning_gsi mosaic_gsi ssrc_arw gsi_arw wrf_arw post ... bucket ... sfcextract ... ncl ... crontab used in the cycles: MAILTO=mhu@ou.edu 0,5,10,15,20,25,30,35,40,45,50,55 * * * * /tg1/projects/wrfruc/mhu/xml_RR13/run_WRFRR_BOUNDARY > /tg1/projects/wrfruc/mhu/xml_RR13/croninf 2>&1 0,5,10,15,20,22,25,28,30,33,35,40,42,45,47,50,52,54,56,58 * * * * /tg1/projects/wrfruc/mhu/xml_RR13/run_WRFRR_CYCLE_1H > /tg1/projects/wrfruc/mhu/xml_RR13/croninf2 2>&1 run_WRFRR_BOUNDARY /whome/harrop/workflowmgr/workflowmgr.rb --xml=/tg1/projects/wrfruc/mhu/xml_RR13/WRFRR_v3_ BOUNDARY.xml --store=/tg1/projects/wrfruc/mhu/xml_RR13/WRFRR_BOUNDARY_V3.store run_WRFRR_CYCLE_1H /usr/local/esrl/bin/workflowmgr.rb --xml=/tg1/projects/wrfruc/mhu/xml_RR13/WRFRR_v3_CYCLE_ 1H.xml --store=/tg1/projects/wrfruc/mhu/xml_RR13/WRFRR_CYCLE_1H_V3.store ==========================================================================