LDAS Meeting Minutes
November 14th
2001
NWS/OHD
I. ATTENDEES:
Ken Mitchell, Paul Houser, John Schaake, Alan Robock, Dag Lohmann, Brian Cosgrove, Qingyun Duan, Justin Sheffield. (Representing: NCEP/EMC, NASA/GSFC, NWS/OHD, Princeton U., Rutgers U.; Not represented: NESDIS/ORA, U. Maryland)
II. DATES OF NEXT MEETINGS:
- Given accelerating efforts toward submission of first N-LDAS publication, the group agreed to hold roughly bi-weekly phone conferences (setup by NCEP), in between the quasi-monthly face-to-face meetings. The two phone conference dates of Nov 28 (3:30 PM) and Dec 17 (2:00 PM) were agreed upon.
- The next face-to-face meeting will be held on January 8 at Noon at OHD. With this meeting, OHD will have hosted the location of the last three N-LDAS meetings, owing to security access difficulties at NSAS/GSFC and the difficulty of getting an NCEP conference room for meetings starting at Noon.
III. STATUS OF REALTIME N-LDAS RUNS:
- Given the present emphasis by our group on the 3-year retrospective runs (see next section, IV) to support our first group journal manuscript submission, re-runs of the realtime runs from April 1999 to present have fallen somewhat behind, owing to such factors as quasi-saturation of the computer resources set aside for this project, resulting from the simultaneous execution of "retrospective" and "realtime re-runs".
- All four modeling groups are in the process of re-running the "realtime" runs for the period 16 Apr 1999 to the present, for the purpose of having continuously cycled realtime state variables from all four models initialized 1) on the common start date of 16 April 1999, 2) from initial soil moisture and soil temperature conditions obtained from the NCEP Global Reanalysis II, and 3) utilizing the common N-LDAS forcing that was generated in realtime. Prior to this re-run effort, the various realtime runs of the four models (NOAH, MOSAIC, VIC, SAC) had been initialized from different start dates spanning 1999-2000, and from different sources of initial soil moisture. At the time of this meeting, the status of realtime reruns follows:
-- NOAH realtime reruns are complete up to the present day
-- MOSAIC realtime reruns are complete up to January 2001 (runs continue toward present)
-- VIC realtime reruns are complete up to October 2000(runs continue toward present)
-- SAC realtime reruns are complete up to present day (but may rerun with different EP source)
- Once all four model realtime re-runs are caught up, the group agreed that Dag Lohmann would develop and implement a script that would control the initiation of the daily realtime runs of all four models (e.g. first checking that the daily generation of realtime forcing had completed normally and passed a routine set of QC checks) and the daily ftp of output from all 4 models to the NASA/GSFC N-LDAS output archive.
IV. STATUS OF RETROSPECTIVE N-LDAS RUNS:
- All four modeling groups have finished the 3-year retrospective runs from 01 Oct 1996 to 30 Sep 1999, using the retrospective N-LDAS forcing generated by NASA/GSFC. One goal is to stage all the output of these retrospective runs from all four models to the NASA/GSFC LDAS SGI workstation platform known as "Land". This staging of output has been completed for VIC and MOSAIC, and is well underway for the NOAH and SAC models, which awaited the adding of additional storage media by NASA to their LDAS platform.
-- Corresponding to the remote login access that all LDAS modeling groups have to the realtime LDAS computer platform at NCEP, NASA/GSFC will establish remote login access accounts to the NASA LDAS workstation for all four LDAS modeling groups and Rutgers U., to facilitate the ftp of retrospective and realtime LDAS model output to and from the NSAS LDAS archive. Additionally, via this remote login capability, NASA will make available its three LDAS evaluation tools known as the a) Image Generator, b) GRADS-based data sub-setter, and c) menu-driven download of original LDAS forcing or model output data files.
- As a reminder, the NASA/GSFC LDAS web site is:
http://ldas.gsfc.nasa.gov
(Therein, click the link to "NLDAS")
- To facilitate the LDAS soil moisture validation work by Rutgers University, all four modeling groups are required to ship (ftp) a specified subset of their 1996-1999 retrospective LDAS model output to Rutgers U. For that purpose, Lifeng Luo of Rutgers has provided a script to be used for the subsetting. The required VIC and MOSAIC model outputs have been provided to Rutgers. The required NOAH and SAC model outputs are in the process of being delivered, and this will be completed in the next 1-2 weeks.
- NASA and Rutgers independently identified a glitch in the retrospective LDAS forcing for the last 4 days of the 3-year retrospective period (27-30 Sep 99). These four days correspond to the first four days following the CRAY C-90 fire at NCEP, during which the traditional Eta/EDAS output was not available from NCEP. Hence, NASA is in the process of regenerating the LDAS forcing for those four days using the NCEP Global Reanalysis I as replacement for the EDAS, and adapting software developed and provided by NCEP to generate N-LDAS forcing fields from NCEP Global Reanalysis.
-- All four modeling groups must re-execute their retrospective runs for the period 27-30 Sep 99, after NASA stages the reprocessed LDAS forcing for this period, and provide the new output to both NASA and Rutgers.
V. JESSE MENG SURFRAD VALIDATION RESULTS CONFIRMING END TO EDAS SOLAR RADIATION
- From the work of Jesse Meng at NCEP (of the NASA-funded G-LDAS group), Ken Mitchell showed monthly, six-station, SURFRAD-based, solar insolation validation of GOES-based LDAS fields, EDAS-based fields in original realtime LDAS forcing files, and EDAS-based fields in corrected retrospective LDAS forcing files. Jesse results from years 1998, 1999, and 2000 (at his web address of http://www.emc.ncep.noaa.gov/mmb/gldas/ldas2.html), show that the 2-3 hour phase error apparent in EDAS-based solar forcing (owing to error in temporal interpolation of 3-hourly EDAS to hourly, and especially apparent for the months of May, Jun, Jul, and Aug 1999) in the 1999 months of the original realtime LDAS forcing files, has been solved in both a) the retrospective LDAS forcing files of Oct 96 to Sep 99, and b) the original realtime LDAS forcing files as of February 2000 (in fact, the bulk of the phase error source is fixed by Sep 1999 in the original realtime LDAS forcing files). We recall that the phase error was not present in the GOES-based solar forcing, which was hourly originally, but we also recall that because of satellite instrument calibration problems, the LDAS solar forcing was EDAS-based and not GOES-based from Jul 99 to Feb 00, hence the LDAS solar forcing for the bulk of 1999 has the phase error. The good news is that by February 2000, both the GOES-based (and hence LDAS) and EDAS-based solar forcing records in the original realtime forcing files is clean, though the EDAS fields continue to show their longstanding and well-known modest positive bias.
VI. LDAS SOIL MOISTURE VALIDATION BY RUTGERS
- Via the web and the overhead computer-driven projector available from OHD at the meeting, Alan Robock showed results from Lifeng Luo's work evaluating LDAS soil moisture and temperature output across the Oklahoma Mesonet domain. So far, Lifeng has produced validation results for VIC and MOSAIC, with NOAH and SAC results coming soon. Lifeng has also evaluated LDAS forcing time series against the local station-measured forcing of the Oklahoma Mesonet (e.g. precipitation and radiation). Alan showed modeled and observed soil moisture time series from several Mesonet sites, and pointed out that the measured soil moisture appears to be surprisingly (and suspiciously) temporally invariant at deeper layers. The VIC and MOSAIC soil moisture time series showed physically consistent responses to the precipitation forcing, with MOSAIC yielding notably drier absolute states of soil moisture.
- The Rutgers LDAS validation web site is:
http://climate.envsci.rutgers.edu/luo/research/LDAS/
and
http://climate.envsci.rutgers.edu/luo/research/LDAS/models.php
- Lifeng Luo will provide OU Mesonet station-based LDAS forcing to all the modeling groups, who will rerun their LDAS models on a reduced land mask for Oklahoma (to be provided by Rutgers), for the purpose of comparing the model responses obtained from the original realtime LDAS forcing (representing a 15-km grid square), and local station-based forcing.
VII. LDAS PRECIPITATION ANALYSIS ALTERNATIVES
- John Schaake showed his recent and expanded results comparing his PRIZM/least-squares-distance-weighting analysis technique with that of the Cressman technique without PRIZM, for the LDAS analysis of daily gage data in Wayne Higgins group (including Wei Shi and Evgeney Yarosh). This is part of the ongoing discussion in the N-LDAS group with Higgin's team about the possible switching of precipitation analysis methods used to produce the realtime daily precipitation forcing for N-LDAS use. John's firmly recommends that the N-LDAS project switch to the PRIZM/least-squares technique, which Higgins team is presently executing (using codes from John Schaake) in realtime mode in parallel with their operational Cressman-based analysis. John will present even more results at the upcoming Fall AGU meeting, after which the N-LDAS group will make a formal decision about whether to switch the precipitation analysis scheme.
VIII. EDAS SOIL MOISTURE FIELDS AS A BACKDROP FOR LDAS
- From the NCEP GAPP/GCIP Core Project Web site, Ken Mitchell presented realtime soil moisture fields from the NCEP operational coupled land/atmosphere Eta/EDAS mesoscale model suite (includes the NOAH LSM as the land component), which has been cycling its soil moisture state variables continuously since June 1999. On 24 July 2001 (this past summer), NCEP implemented the operational assimilation of hourly Stage-IV 4-km precipitation into the EDAS, as well as several major upgrades to NOAH LSM physics. Given the operational assimilation of precipitation, the EDAS soil moisture states will provide a competitive "control" case against which the LDAS soil moisture states from the NOAH LSM can be compared.
-- end of month Eta/EDAS land states and monthly land-surface water budget fields are posted shortly after the end of each month at the NCEP web site of
http://www.emc.ncep.noaa.gov/mmb/gcp/h2o/index.html
with the overall NCEP GAPP/GCIP Core Project Web site (which includes link to above) at
http://www.emc.ncep.noaa.gov/mmb/gcip.html
-- Ken Mitchell showed EDAS soil moisture results for the end of October, and these results reflected the dry conditions being experienced along the eastern seaboard over the past two months.
IX. PROPOSED OUTLINE FOR FIRST N-LDAS GROUP JOURNAL MANUSCRIPT
- The group discussed at some length the technical depth and target journal that the group's first journal publication should embrace. Following considerable discussion, it was decided that the first publication should be an overview publication submitted to BAMS, following along the lines of recent BAMS publications of the GSWP and PILPS projects.
- Qingyun Duan is an editor of BAMS and will contact the BAMS Chief Editors about the prospect of their accepting such a submission, given that the format and content of BAMS is reportedly undergoing some changes.
- The group discussed and arrived at the following broad strawman outline for the envisioned manuscript, and section writing assignments (subject to revision at upcoming meetings):
(9 sections, but we may combine Sections 2 and 3 into one, and Sections 4-8 into one results section with subsections)
1.0 Introduction (Ken Mitchell)
- purpose, objectives, and approach of the project
- cast the different purposes of realtime and retrospective executions
2.0 LDAS Configuration and Forcing
(Ken Mitchell lead, Dag Lohmann, Brian Cosgrove, Lifeng Luo, and Jesse Meng contributors)
- domain, grid, land mask, terrain field, other surface fields (soil type, veg type, LAI, greenness)
-realtime forcing sources and processing to LDAS grid
-- figures: both time series and spatial map examples
-- retrospective forcing sources and processing to LDAS grid
-- figures: both time series and spatial map examples
-- validation of solar forcing against SURFRAD (Jesse Meng)
-- comparison of precip forcing with OU Mesonet obs (Lifeng Luo)
3.0 Summary of participating LSMs (Ken Mitchell lead, contributions from each group)
-- include a table of major LSM characteristics of NOAH, MOSAIC, VIC, SAC
4.0 Monthly Surface flux and Water Balance intercomparisons
(Brian Cosgrove, Qingyun Duan, Dag Lohmann lead, others)
-- intercompare spatial maps of monthly averages among the four models
--- latent heat flux, sensible heat flux, ground flux, net radiation
--- rainfall, snowfall, runoff, snowmelt, soil moisture storage and change
-- validate hourly time series at flux observing stations (e.g. ARM, Bonneville)
5.0 Soil moisture validation (Lifeng Luo and Rutgers)
-- over OU Mesonet
-- over Illinois ?
6.0 Skin temperature validation (Justin Sheffield, Dag Lohmann, Andy Bailey, Dan Tarpley)
-- LDAS simulated skin temperature versus GOES-retrievals
7.0 Snow cover intercomparison and validation (Justin Sheffield, others)
8.0 Streamflow validation (Dag Lohmann, others)
9.0 Summary and future efforts (Ken Mitchell, all)