Meeting Minutes
March 1st 2001, NASA Goddard
Discussion
of Forcing Issues
--As soon as Rachel posts data on her website, forcing files will include CAPE, skin temperature and diffuse radiation.
--Presentation on reprocessing GOES SRB data for GCIP
--Introduction of people on reprocessing project.
--Cloud Detection Improvement in GCIP Reprocessing
--GOES replacing GOES will help split window
--UMD group says that some data is missing and that they will be doing some reprocessing.
--Ken advised that for consistence they try to use the Wisconsin data to fill in missing data.
--Discussion about tests in Clouds from AVHRR (CLAVR).
--Agreement that there needs to be some offline discussion about this whole process
--Cloudiness information transforms well into LDAS 1/8 grid but its only monthly data.
--Improved estimates of surface albedos from GOES and improved albedo models discussed.
--Attempting to derive seasonal dependant albedo models for each surface type, as the LDS land classification.(Used AVHRR channel 1 and 2 monthly mean albedos at the TOA from GVI data set)
--Alan points out that differences exist within classes based on when they grow.
--John points out that the EROS data center provides subclasses and that would really be helpful for matching.
--Question is raised as to whether this will matter for the albedos for the purposes of the LDAS project.
--Ken suggests presenting radiation and albedo issues at NESDIS where there are people who specialize in albedo/radiation issues.
--Ken suggests to Rachel not to be tied to the LDAS vegetation classification scheme.
--Rachel states that they also want to improve the optical depth and optical parameter information. Could do this by using AERONET. Monthly climatologies used to improve guess of values.
--GLOBAL LDAS may make use of NASA Pathfinder 1/2 degree (sampled at 30 km)
--Some breaks in areal coverage exist in monthly product
--If multiple satellites are used, you can get differences of +/- 30 W/m2
--Large amounts of binary data stored in Mosaic and Princeton accounts.
--Binary data needs to be deleted within 1 week. Groups should utilize clear directory structure, and output data in GRIB format.
--Data output: 5.6 Mbytes for each hourly output = 50 Gigs a year. So 100 Gigs needed for real-time runs up to present. Only 30 Gigs needed for Sacramento model.
--Disk 108 has 173 Gigs plus some additions
--Use dm in front of commands (dmls dmput dmget etc)
--Data can get migrated to tape system (1-2 month lag) and free up disk space.
--PC raidzone has 225 gigs available. (but the ftp is slower)
--Data on SGI 100 tape system needs to be deleted, and no new data should be stored on that tape system.
--So the question is when to start the rerun. 1996 or later?
--Disk space, runtime and data access issues come into play.
--Paul will move data off of their system and place it here at NASA
--John raises issue of when SNOTEL data starts.
--Ken says given John's point and Eric's push to get science and papers done, we should focus on three years. Use 98 for spin up and then 99 and 00 for analysis.
--Paul would like someone to look at the forcing data to check for errors.
--Alan has soil moisture for 1997-1999, Lifeng will process 2000 data.
--Ken says is there flux data, but 2000 agreement with Oklahoma Mesonet does not include flux data.
--14 soil moisture stations in Nebraska up to present, back 3 years (or so).
--Proceed with 98 spin-up. Keep the 98 output until we start 1999 or even 2000.
--Use realistic approach. Yun says start mid year to avoid snow problems.
--John thinks a paper about this spin up stuff to discuss the issues might be a good first paper. Thus we won't have wasted our time when we find out that we should have done it differently. Only need 6 months to a year of model runs that should be okay. How long to run until we can trust what we are doing?
--Need Eric in this discussion and Dennis too. What about using 75% of porosity b/c field capacity is model dependent. John says do multiple reruns. Some that are the same and some that are model dependent. John says we need a systematic context and thus the basis of a paper.
--Don't need to start in April.... can do any month. John likes April because the soil is the wettest and there is less uncertainty.
--Paul says start from the same initial condition and see how they diverge.
-- Ken says go with EDAS because then it isn't any of the models.
--Start from 12Z, Septemeber 30th for the 1997 water year.
--Need EDAS temperatures and soil moistures to initialize models. (Paul says define it as a standard ALMA variable so we all know what's up).
NASA
--Reprocessing data, CLM, GLDAS
--Jared doing Pilps 2E with CLM.
--Eleanor Burke put in the forward trans in Mosaic model. Running 1978-1993 using ECMWF forcing. SMMR satellite. Model precipitation used, but corrected to monthly means.
NCEP
--Cleaning up code, speed up code, tape migration system.
--Jesse working on radiation forcing. (was Brian aware of some problem? MICAS software problem?)
--Some NOAH streamflow validation shown.
--Basins are on the LDAS website.
--John asks Dag to do this for the other models. Dag says as long as the output is in the grib format he can do it. Be careful about regulated basins.
Office of Hydrology
--Trying to finish routing model with Dag's help.
--Talk to Shue Zen about real time hourly data--M.A.P.s for all of the basins for retrospective (need energy forcing and we used the climatological data).
--Will try to improve the energy forcing for this data. Work with Brian to get the data off of the tape (maybe make an appointment to do such), to get the energy forcing for as long as we can.
Dennis
--Has finished LDAS retrospective forcing back to 1948, but it's hard to get and massive.
Princeton
--Has their final version for the drivers.
--Just about finished reprocessing the binary data from before to get in proper output.
--Will remove the binary stuff in the next week.
--Have LDAS ALMA output.
--Going to go back to when it started to convert output over so then they could look at the streamflow data for that too.
Alan
--Have the Oklahoma data for 2000 and Nebraska data soon.
--Lifeng has obtained some output from the groups.Probably needs to get new output from Mosaic.