N-LDAS Phone
Conference Minutes
25 January 2002
(REMINDER TO ALL: Email your Powerpoint file of your Jan 02
N-LDAS-related Orlando AMS talk to the core 10 members of the N-LDAS
group. Brian Cosgrove can post to
the
N-LDAS web page.)
A - phone conference Participants:
Ken Mitchell, Dag Lohmann, Paul Houser, Brian Cosgrove, Eric Wood, John Schaake, Qingyun Duan, Alan Robock, Lifeng Luo
B - NEXT FACE-TO-FACE MTG: Feb 26 at Noon at OHD in Silver Spring.
C - NEXT PHONE CONFERENCE: None scheduled at this time.
D - Corrected NASA Retrospective Forcing:
Brian Cosgrove has finished his Version 2 phase of corrections to NASA's 3-year retrospective N-LDAS hourly forcing for the period Sep 96 through Sep 99. Foremost, this phase corrected his error in the hourly temporal partitioning of the daily precipitation forcing, and secondarily improved the fallback treatment for the rare occurrences of out-of-range humidity values from the Eta/EDAS output. All modeling groups can now download this corrected forcing from the NASA site (given below). Then, they can begin the two-part Round #2 (see Section F) of their 3-year retrospective land model runs.
ftp anonymously to hsbserve.gsfc.nasa.gov (note this new address)
cd to LDAS
cd to LDASFORCING.V2
Those groups that downloaded this Version 2 retrospective
forcing before 15 Jan 02 or so (from land.gsfc.nasa.gov), must download a fresh
copy of the Feb 1998 through Dec 1998 data from the above
"hsbserve.gsfc.nasa.gov" server. Data from this 1998 time period now includes a better fix
for the specific humidity problem.
Brian urged us in his related email that "as we make use of this retrospective forcing data, to please carefully check your model output as well as the forcing used as input. The forcing data has been checked over, but because of the complexity of the LDAS forcing generation process, it is always very useful to have more than one pair of eyes checking the data over".
Aside: Brian has also posted retrospective forcing data for the year 2001, which Dag will subsequently use when he resumes his unified scripts for his unified 4-model N-LDAS realtime runs, which he will cycle forward to present time from the final 4-model restart files at the end of the 3-year retrospective runs.
IMPORTANT REMINDER: All four models must save and backup their restart files from the end of the 3-year retrospective runs, and provide them to Dag for the above restart and catch-up of the realtime runs.
E - ADDITIONAL PREPARATION FOR RETROSPECTIVE MODEL RUNS:
1 - (ALL MODELS) Common Soil Texture Map (but not common soil parameters):
Every modeling group will use the 1/8-th degree top-layer soil class map (13 classes) produced by Yun Duan on the common N-LDAS grid. (Contact Yun to acquire the data file, if you have not already done so.) This was derived from aggregation of the 1-km STATSGO soils database of Doug Miller of Penn State. After considerable discussion, the group decided that while the control runs must invoke the above common N-LDAS soil class map, the control runs will NOT invoke/impose/require land model assignment of common soil physical parameters to this common soil class map, because one or more of our four participating land models likely have been tuned to their typical default soil parameter values. For example, Alan Robock described the large negative impacts that simple bucket model simulations in PILPD-2d suffered when, following PILPS directives, non-default or non-typical bucket storage capacities were assigned to the bucket model.
However, the group did agree that follow-on investigations in our N-DAS group will explore,
via more reruns, the impact of invoking common assignment of soil parameters (as well as analogous investigations of common treatments of other factors such as infiltration, canopy resistance, bare soil evaporation, aerodynamic resistance, total soil storage capacity, etc).
Aside: For those groups that want to make use of it, immediately following the phone conference, Dag emailed out documentation of NCEP's assignment of soil parameters in their retrospective and realtime N-LDAS runs, as follows: a) the table that NCEP uses to map the 13 classes in the common N-LDAS soil class map cited above to the 8 classes of the "Zobler" soils classification used in ISLSCP-I, b) the table NCEP uses to map these 8 Zobler classes to 8 of the 11 soil classes given in the widely used Table 3 of Cosby et al. (1984), and c) the tables NCEP uses to assign to each of these 8 classes the Cosby et al (1984, Table 3) values of the four primary soil physical parameters (porosity, slope of retention curve or "b" parameter, saturated hydraulic conductivity, saturated matric potential). Eric Wood stated that the VIC N-LDAS runs use the Cosby parameters also. A noteworthy reminder is that while two models might use the Cosby et al values for the four primary soil parameters just mentioned, they might use different methods to a) derive "secondary" parameters (e.g. upper and lower soil moisture thresholds and exponents in either transpiration or direct evaporation, quartz content for thermal diffusivity calculations, etc), or b) map differently between the 13-class STATSGO texture scheme and 11-class Cosby scheme. (Reminder: non-unitless entries in Table 3 of Cosby et al. (1984) are not in MKS units; the NCEP tables use MKS equivalents)
2 - Before Yun Duan executes the 3-year retrospective of the SAC model, he needs to examine and resolve a likely problem in the potential evaporation forcing he is using. The problem is erroneously low values in the southwest U.S. that was apparent in some of the plots he showed at the Orlando AMS meetings. (This EP problem would have surfaced earlier, if a "dry-run" of the AMS presentation would have been made to the Jan 8 N-LDAS meeting, as per our N-LDAS group policy.)
3- Before Dag Lohmann executes the 3-year retrospective of the NOAH model, he will identify the cause of and solution to the very high bias in NOAH soil temperatures that Lifeng Luo identified in the retrospective NOAH runs. This bias really puzzles us at NCEP, because it was not present in the realtime NOAH LDAS runs, nor in NCEP's coupled operational Eta/NOAH runs, nor in the various uncoupled PILPS and GSWP runs of the NOAH model. We expect the problem will turn out to be something external to NOAH model physics, such as an I/O glitch or an inadvertent glitch in model configuration set-up. The positive bias that Lifeng uncovered in the NOAH model "echo" output of the input common solar forcing turned out to be a glitch in the NOAH GriB-table entries for solar insolation. Internally, the NOAH model physically had used correct values of the input solar insolation.
F - THREE-YEAR RETROSPECTIVE LAND MODEL RUNS (2-PARTS):
On their own local institution computing platforms (with exception of SAC model, which will execute on the NCEP platform) and after downloading the above corrected forcing, all the modeling groups will re-execute their land model for the 3-year retrospective period of Oct 96 through Sep 99. This will be done in two sequential parts as follows:
Part 1: A "faster" re-run on the reduced-mask grid of Lifeng Luo (Oklahoma/Kansas region). Each modeling group will ftp their land model output from this reduced grid to Lifeng Luo at Rutgers. Lifeng will perform some early sanity checks/intercomparisons of these "fast" results.
Part 2: The "full-domain" rerun, executing over the full continental land domain of N-LDAS. Each modeling group will ftp their land model output from this full-domain grid to Brian Cosgrove at NASA.
G - ABSTRACT SUBMISSIONS TO THE ANNUAL GAPP MEETING IN MAY
Feb 15 is the deadline for submission of abstracts to the annual GAPP meeting in mid May, this year being held in conjunction with the "Mississippi River Climate and Hydrology Conference in New Orleans. Conference brochures are now available from the GAPP Program Office and a conference web site resides at http://www.ogp.noaa.gov/mpe/gapp/mississippi/call.htm.
The N-LDAS phone conference ended with considerable discussion of the strategy and coordination of submission to this conference of abstracts from the N-LDAS group. We decided to submit around half a dozen abstracts to the conference Special Session on Data Assimilation, which Matt Rodell of NASA/GSFC/HSB is chairing. This special session will address both hydrologic and atmospheric models. Ken Mitchell spoke with Rick Lawford at the AMS meetings about the N-LDAS group desire to have a focus session on the N-LDAS project. Rick responded that he would like such a focus session to be worked out with Matt Rodell as a land assimilation focus "sub-session" within the already advertised Special Session on Data Assimilation. Ken Mitchell relayed this suggestion of Rick to Matt Rodell at the AMS meetings. Ken will further interact and coordinate with Matt on this.
NOTE: During the phone call, the group agreed to submit around 6-7 distinct abstracts on the topics listed below spearheaded by the designated lead authors. Also listed with each topic are co-authors expected to be material contributors. The lead authors indicated below will draft a strawman abstract for their topic to be distributed to the N-LDAS group members by Feb 4 for coordination, input, and revision by the Feb 15 submission deadline.
1 - N-LDAS Project Overview: Lead - Ken Mitchell, Material Contributors - "the gang of 13"
2 - N-LDAS Realtime and Retrospective Forcing: Lead - Brian Cosgrove, Material Contributors - Dag Lohmann, Ken Mitchell, Paul Houser, Dan Tarpley, Rachel Pinker, Wayne Higgins, John Schaake, Jesse Meng, Ying Lin
3 - N-LDAS Validation: Lead - Lifeng Luo; Material Contributors: Alan Robock, Dag Lohmann, Eric Wood, Justin Sheffield, Jesse Meng, Dan Tarpley, Andy Bailey
4 - N-LDAS Energy Budgets: Lead - Eric Wood; Material Contributors: Princeton colleagues, Dag Lohmann, Lifeng Luo, Alan Robock
5 - N-LDAS Water Budgets and Storages: Lead - John Schaake; Material Contributors: Yun Duan, Dag Lohmann, Eric Wood, Lifeng Luo, Alan Robock
6- N-LDAS Streamflow: Lead - Dag Lohmann; Material Contributors: John Schaake, Yun Duan, Eric Wood
7 - 50-Year Retrospective N-LDAS with VIC: Lead - Dennis Lettenmaier, Material Contributors: Eric Wood
Other possibilities:
Dan Tarpley on LDAS skin temperature validation
John Schaake on PRIZM based precipitation analysis
Rachel Pinker and Dan Tarpley on GOES-based solar insolation forcing
Huug van den Dool on 50-year Retrospective N-LDAS with NOAH
Paul Houser on Global LDAS and or HPCC Ultra-res global LDAS
Dan Tarpley on LDAS skin temperature validation
Jared Entin on Kalman filter pilot studies for assimilation