The 2001 GLASS panel meeting, Toulouse 7-9 November 2001

Nov. 07

[14:00] Jan Polcher

• Confirmation of Agenda.
• Taikan Oki was appointed to take notes of the meeting.

[14:10] PILPS-2e by Bart Nijseen

• The aim of PILPS-2e is to evaluate land-surface Parameterizations in Northern Regions. Preliminary results from the PILPS-2e Model inter-comparison prepared by Laura Bowling, Dennis Lettemaier, and Bart Nijseen were presented.
• PILPS-2e is the first step in a 3 stage process
  • Stage 1: Torne/Kalix Rivers, 58 000 km2 Sweden/Finland
  • Stage 2: (proposed) Mackenzie River, 1 800 000 km2
  • Stage 3: (proposed) Lena River
• 21 models from 11 countries have participated in PILPS-2e.
• The 58 000 km2 basin of the Torne/Kalix rivers was modeled with 218 computational grid cells (0.25 degree grids) in this experiment.
• The meteorological forcing was provided at an hourly time step over the 1979-1998 period. For spin-up the 1979-1998 period was used.
• Only 20 rain gauge stations are available over the basin and thus there is a limit to the accuracy of the data. Precipitation was gridded from SMHI station data. Lapsed for elevation and adjusted for gauge under-catch per SMHI recommendations. Finally it was statistically disaggregated to hourly data. A similar procedure was applied for temperature, Pressure and relative humidity. Clouds were derived from a 1 degree SMHI dataset. Wind speed was taken from NCEP/NCAR reanalysis and finally long-wave radiation was estimated from air temperature and the other variables.
• Original short-wave radiation was underestimated and corrected afterward.
• Moisture and energy balances were calculated to verify the consistency of the data set.
• Using the annual mean latent and sensible heat fluxes the LSMs can be divided into 4 groups.
• Some models have large sublimation fluxes which are compensated by a negative sensible heat flux to maintain net radiation.
• For the mean annual turbulent fluxes the question was raised if an error bar exists for observations.
• Simulated discharges were evaluated at the monthly time scale. They reflect the large differences between models on the simulated sublimation and snow melt. The runoff differences among LSMs is highest in May, when the snow melt season starts.
• Different sensitivities for net radiation have been found. In May it is controlled by albedo while in January is controlled by surface radiative temperature (4.6W/m²/K).
• At high latitudes available net radiation rather than available water is the limiting factor for latent heat.
• largest evaporation rates occur in the summer, but inter-model variations in mean annual runoff are primarily related to the model-predicted winter snow sublimation and melt.
• A participant's view on the value of the ALMA infrastructure was given :
  • PILPS-2e participants were pleased with the ALMA variable definitions.
  • Better documentation of the NetCDF interface on the ALMA web page is needed and example of drivers for LSM should be provided.
  • The ALMA web page should have an example scripts to covert between NetCDF and ASCII for basic input/output.
  • The programs that calculate the energy and water balance errors need to be revised for different computer platforms and generalized.
• Organizer's view of ALMA:
  • For many modeling groups this was their first exposure to NetCDF and thus a lot of help was needed.
  • From a data processing point of view, a more strict respect of the convention for describing coordinate axes and names is needed.
  • Standard format that includes meta-data facilitates data processing and quality checking.
• The question remained on the quality of the forcing data in the region. But the community has little choice as the Torne/Kalix is the region with the highest density of observations for the high latitude climates.
• The group leading PILPS-2e was encouraged to investigate the reasons in the design of LSMs which could explain the difference and attempt more quantitative descriptions of the spread of results.
• The question was raised on the utilization of the hourly output data? It seems that up to now only very little use has made of it.
• There is a need to archive the output datasets (or a subset of it) together with the forcing data for future use and possible re-runs.

[15:38] Rhe-AGG by Aaron Boone

• Aaron Boone presented a quick overview of the project and the workshop which was held in Toulouse early in the week.
• A classification of LSMs in terms of the treatment of surface and subsurface heterogeneity was attempted to understand the results.
• All of the schemes successfully simulated snow in coarse resolution for higher altitude where a lot of snow is present.
• The question was raised on the legitimacy of the comparison of snow depth observations with LSM outputs, even at high resolution.

[16:40] Continue Rhe-AGG

• The panel suggested that a project overview in multi-author paper on Rhe-AGG (BAMS or EOS are candidate journals).
• The ALMA data exchange format proved useful in this experiment as well and allowed for an efficient quality check on the submitted model output. This data set should be archived as well.
• An important issue for the Rhe-AGG forcing and validation data of discharge, for instance, is to obtain permission to redistribute it.
• This discussion on the distribution of data related to past, present and future experiments raised the issue of agreement from model authors. GLASS needs to inform them about this intention and ask for their agreement. We also need to reconstruct the data of past experiments performed under PILPS.

[17:05] PILPS-C1 by Nicolas Viovy

• The aim, and the innovation this experiment will bring is to compare land surface schemes with in-situ data for both water and carbon fluxes .
• The PILPS-C1 experiment does not only target LSM representing biophysical processes but also BGC models with crude representation of biophysical processes, and models coupling biophysics and biochemistry. This project is an opportunity to compare different kinds of models so we would like to open the project to groups coming from different communities.
• This project aims at a number scientific questions:
  • What is the ability of the models to correctly reproduce both biophysical and biogeochemical processes?
  • Do models which include carbon out-perform those which do not?
  • are the models able to reproduce the sink of carbon observed on the site?
• The chosen site is Loobos in the Netherlands (one of the Euroflux sites).
• Three kinds of simulations will be proposed to the participants :
  • A free simulations where only meteorological and soil characteristics are provided to the participants.
  • A constrained simulation where models are forced by ancillary data which describe the vegetation (LAI, albedo, roughness, ...).
  • A calibrated simulation where the possibility is given to calibrate models using observed fluxes.
• How to organize this experiment ? Two options were discussed :
  1. Give all the data at the same time with the expectation that groups will not use the ancillary and fluxes to calibrate models for the free simulation.
  2. Give data in sequence but this solution is clearly more complex to organize.
• This experiment brings a new spin-up problem: carbon pools. Especially the soil carbon needs hundreds of years to equilibrate unlike the water reservoirs. In Loobos (like most other stations) the soil carbon is not in equilibrium and an annual net sink is observed. Two solutions are proposed and were discussed for this experiment :
  1. Perform an equilibrium simulation using the first year of data until equilibrium is reached. The model will then be compared to the observed CO2 net flux-mean sink. This would suppress any natural and observed long term trends.
  2. Perform a realistic 100 years simulation with forcing data observed at a nearby meteorological station. This long data set will not be of very good quality but probably good enough to simulate the long term evolution of the vegetation.
  The recommendation after discussion is to perform the 100 year simulation.
• The participants will be asked to use the results of their 100 years run to initialize their carbon pools for the shorter simulations.
• The proposed analysis for PILPS-2e will be :
  • monthly means
  • daily means
  • daily amplitude and phase of diurnal cycle
  • analysis by weather types
  • analysis of correlation and determination coefficient.
• Proposed time line
  • November-December 2001: experimental details and forcing data issued
  • March 2002: submission of results for "free simulation"
  • May 2002: submission of results for constrained simulations
  • July 2002: workshop for analysis of preliminary results

Future PILPS experiments.

• There is a huge lack of PILPS experiments in semi-arid and tropical climates and this should be the priority for the next experiment.
• During the discussion a number of sites were suggested but a more detailed analysis of the feasibility for each of them needs to be done :
  1. The San-Pedro river basin in Arizona.
  2. The HAPEX-Sahel data from Niger.
  3. The OASIS observations in Australia.

Nov. 08

[9:10] Some general information on the context of GLASS was given by Jan Polcher.

• The questions was raised yesterday on the link between GLASS and SNOWMIP and especially the overlap of SNOWMIP and Rhe-AGG experiments. SNOWMIP was launched by WGNE before GLASS was set-up. Thus there is no formal link but interactions and communication exists.
• A short report was given on the WGNE/GMPP meeting which took place the week before the GLASS science panel meeting. One of the main points is that at next years WGNE/GMPP a effort should be made to present to the forecasting community how GLASS can help them. In particular which GLASS projects they should follow closely.
• Within GEWEX and GMPP there is a need to increase the interaction between the various projects. For instance there is the question from the GCSS on the table of how cloud models should deal with land surface processes. Can we recommend one land-surface scheme or any other procedure ?
• We need to raise the awareness that GLASS needs to have access to an infrastructure which is able to collect GCM runs and redistribute the data if we want to coordinate GCM experiments within the global coupled action. An example for such an infrastructure is AMIP. The need for such an activity will be raised at the GEWEX SSG.

[9:30] A short revisit of the PILPS-C1 discussion :

• The 100 year run for the spin-up of the carbon pools seems to be the best solution.
• In this experiment an opportunity should be given for any model, even without vegetation dynamic, to participate, i.e. it should include DVGMs and SVATs.
• Announce this experiment in GEWEX and IGBP news letters, and to the US/European carbon communities.
• The schedule should start in January. 6 month are required after the data is released for those who are not accustomed to ALMA convention.
• This leads to a workshop in middle of May.

[10:00] Overview of PILPS.

• There is no prominent project in the waiting and we should use the opportunity to push for a semi-arid experiment.
• The Murray-Darling river basin in Australian will probably become a GEWEX CSE. This opens the possibility to do a Rhe-AGG type experiment in the area after 2003. The idea could be submitted to the Bureau of Meteorology.

[10:35] GSWP 1.5 presentation by Taikan Oki.

• The correction of catching ratio of raingauges for the precipitation forcing of GSWP-1.5 was presented.
• Some applications of GSWP1 product were presented. For instance the soil moisture was used as boundary condition for GCM experiments.
• Interesting applications of GSWP1 results for global water resources assessment were presented.
• It was shown that projects are underway to assess the human impact, such as irrigation, in land-surface scheme for an application in GCMs.

[11:40] GSWP-2 by Paul Dirmeyer.

• An outline of the GSWP-2 plan was presented and it's relation to the evolution of the ISLSCP-II was discussed.
• GSWP-2 will present a challenge for the data management. To show possible solutions a short introduction to DODS servers and their potential was given.
• If all goes as planned with ISLSCP-II, the GSWP-2 could start at the end of 2002.

[12:25] Discussion of both GSWP experiments.

• Bill Rossow has proposed to provide upward long wave radiation at the surface derived from satellite observations. This data could be used to validate the land-surface schemes at the global scale.
• GSWP1.5 has to be seen as a preparation for GSWP-2. It should allow us to put in place the infrastructure to handle all the data which is needed for GSWP-2.
• The DODS technology should be testes with GSWP-1.5 so that it is ready for use in GSWP-2. The same is true for the validation with remote sensing data (upward long wave radiation at the surface).
• It is suggested that the models which produced carbon fluxes in PILPS-C1 should do the same during the GSWP-1.5. This would enable to compare the fluxes at the global scale.
• The questions was raised on the scientific objective of GSWP2. It is proposed that a workshop will be organized in Autumn 2002, in conjunction with the next GLASS panel meeting, to decide on these objectives. A few ideas were already suggested :
  1. The inter-annual variability of surface fluxes.
  2. Sensitivity tests of LSMs using 2 kinds of vegetation maps and 3 kinds of forcing data. These will be available in the ISLSCP-II data base.
• It is proposed that the output of the GSWP-2 experiment be available for analysis by anybody who is interested.
• The data of GSWP-2 should be distributed widely through mirror DODS servers in a few places.

[14:10] Jan Polcher killing the time !

• The coupled local experiments were discussed. It is clear that the priorities of this action need to be determined first and this could best be done with a workshop. An ideal moment would be to hold it together in March with the GABLE meeting.
• We quickly returned to the science plan for GSWP2. The other proposals which came up were :
  • general impact of land surface on climate.
  • inter-annual variation of soil moisture and fluxes.
  • transferability of soil moisture from one model to another.
• waiting for Randy Koster on the phone ...

[14:50] Tele-conference with Randy Koster.

• The first presentation was about the Poor Man's land data assimilation experiments.
• The second presentation was the inter-comparison study which was conducted with 4 models on the land-atmosphere feedback. This work is currently being published in J. of Hydrometeorolgy.
• The question arose quickly if the Poor Man's LDAS should be propose to AMIP as an experimental subproject. GLASS on it's own would have difficulties to manage the data associated with such an experiment.
• The poor man's LDAS will be circulated among GEWEX/WGNE/AMIP in order to evaluate the interest and to find help to organize this experiment.
• For the 2nd experiment a 2-3 page description of the project is required and should be distributed widely within the community. Groups are expected to start submitting their results in the middle of the next year. Only a limited number of variables will be collected in order to make the project manageable.
• In a final talk Randy Koster examined the standard PILPS approach to inter-comparisons. It seems that it does not investigate some important aspect of land surface models.
• The relations between E/R_n and soil moisture or R/P and soil moisture as they are implemented in land-surface scheme have been found to affect strongly the annual energy and water budget and the seasonal memory of soil moisture. They also impact the requirements for soil moisture initialization for seasonal precipitation forecasts.
• It is proposed that a PILPS experiment be planed which not only examines the reproduction of evaporation and runoff fluxes but also the sensitivity of these fluxes to soil moisture variations. The analysis would center on a comparison of this sensitivity (the fitted slopes) to the sensitivity seen in the observational data.
• The requirements a site would need to fulfill for such an experiment are first the usual PILPS measurements and secondly an adequate measurement of both evaporation and soil moisture. One possible study site could be the ARM-CART site in Oklahoma. The PILPS-C1 site might also allow for such a study.

[16:40] Paul House came on the phone and discussed with those present the local-coupled action.

• In this action a change in the coordinator will occur. Louis Bastidas will replace Hoshin Gupta.
• The LDASs looked very much like the GSWP experiment at first. But this is shifting as more and more data assimilation starts to be done in the LDASs.
• The local coupled action will concentrate on the interactions between the atmosphere and the surface with a special attention to the implications for data assimilation.
• It is proposed to organize a workshop at ECMWF in spring 2002 to raise the awareness of the problems linked to the atmosphere/surface interactions and try to propose some comparisons of methods for forcing land-surface models which go beyond what is done today in PILPS or GSWP.
• GABLES (The atmospheric planetary boundary layer project within GMPP) held it's first meeting in August 2001.
  • The major motivation for this project is to revive the research on the PBL which had a declining activity after the middle of the 1980s.
  • David Randall initiated the project and it will be coordinated by Bert Holtslag.
  • The first focus of GABLS will be to study the stable PBL.
  • More than 20 groups participated enthusiastically at the August meeting and it is expected that GABLS will be a very active project.

[17:00] General discussion of the local/coupled action.

• The question was raised whether a similar experiment to Randy's 2nd proposal could be performed with meso-scale models and how that could contribute to the understanding of the local coupling between the surface and the atmosphere. Roni Avissar and Paul Dirmeyer offered to explore the idea.
• The GCSS panel submitted to GLASS a request regarding their handling of surface processes. For the moment they impose to their cloud models estimates or measured surface fluxes and thus they freeze any possible interaction between the atmosphere and surface. They are thus asking what kind of standard surface model they should couple to their cloud models. It should be the same for all models as they wish to ensure a consistency between all models which will be compared.
• The forcing of LSMs in cloud models is quite different from what happens in GCMs. For instance the time step is 3--10 seconds and thus the fast variations of fluxes can more easily be represented.
• The simplest proposal, and possible a good candidate, could be the bucket model with an interception reservoir added.
• One solution would be to propose a simplified version of CASM. But it would need to be fitted out with the PILPS4c interface.
• The best solution would probably be the improved bucket Chris Milly has developed. A series of papers has been submitted by Chris Milly which describes the value of each addition which was done to the original bucket model. It is thus a very well documented simple LSM.
• There is no perfect LSM and thus no best solution for cloud models. But what GCSS is looking for is a minimal LSM which represents the major feedbacks in a physically reasonable way.
• VIC would also be a good candidate as it performs very well for scaling through various resolutions.
• The point was raised that it could be ALMA's role to maintain a standard simple LSM. This scheme would be useful to a number of GLASS activities as it could be considered as the minimum requirement for a LSM and other models could be evaluated by their added value.

Nov 09

[9:10] Revisit and continuation of the GMPP/GCSS request

• It was decided that Chris Milly would be asked if he accepts that his model is considered as the "minimal reference" model.
• This activity of a "minimal reference" model should be part of ALMA activities.
• Such a "minimal reference" model should be outfitted with the PILPS4C coupler and an infrastructure which makes this model easy to use. The PILPS4C interface is getting common in Europe, USA, and Australia and ALMA should start coordinate these developments.

[9:40] CLIVAR monsoon panel has sent a request for GLASS to help them in determining the role surface processes play in climate variability.

• As the experiments proposed by Randy Koster will address this issue we propose that he presents with Paul Dirmeyer the GLASS activity at the CLIVAR monsoon panel. The next meeting of that panel will be next summer in Texas.
• It appeared that the global coupled action will address these issues and there is no need today to plan for anything more.
• Some actions on the validation of LSM in semi-arid conditions could also help the CLIVAR monsoon issues. Thus the planed next PILPS experiment should be linked up with the CLIVAR request.
• Would it be possible to do a PILPS experiment with the GAME-tropic dataset ? This would be probably the best option to evaluate the ability of our LSM to represent surface processes in monsoonal climates.

[10:00] The C4MIP project was briefly discussed.

• The GLASS panel needs to pay attention that a contact be established between PILPS-C1 and C4MIP. PILPS-C1 can be viewed as the point validation for the LSM used in C4MIP.
• Peter Cox or somebody from the Hadley Centre should be on the GLASS panel to represent the C4MIP project and ensure that the contact is maintained.

[10:40] The ALMA action.

• There is a need to update the data convention after the experience gained in PILSP-2e and Rhe-AGG. Comments from Aaron Boone will be integrated. Help needs to be provided for data preparation for those organizing experiments. This could be done by providing examples. Some of the documents need to be corrected.
• The list of output variables needs to be expanded so that it include the requirements of the PILPS-C1 experiment.
  • Jan Polcher will take care of updating the convention.
  • The names used for the axis by the groups sending in data should be standardized. The meta-data convention ALMA uses caters for that but there is a need to make sure that the participants are aware of this and use it.
  • There is a need for a simplified description of the data exchange convention. A simple "How To" could be useful.
• ALMA needs to keep an archive of previous experiments.
  • All PILPS experiments should be archived in the ALMA format. This will require some digging to find some of the data sets. It is essential to have the forcing data, the verification data is important and it would be very nice to have the output of all participating models.
  • There is a need to include in the forcing data files a description of the reference level at which wind, humidity and temperature have been measured. This would allow model to adapt to this automatically.
  • The ISLSCP-I data is archived at IIS/UT (Tokyo). It is archived in the standard ALMA format and a number of problems have been identified or corrected :
    • long wave correction
    • land/sea mask in Canada should be corrected.
    The full dataset should be ready before Spring, 2002 for GSWP1.5 as well as for GSWP2.
  • For the PILPS 2e and Rhe-AGG experiments the panel needs to ask permission for archiving and distributing the results from participants.
  • DODS servers should be operated at COLA and IIS/UT for GSWP2.
    • This should be started with PILPS2e forcing data.
    • The forcing data server should be ready in Spring 2002.
  • The output variables from GSWP-1.5 amd GSWP-2 should be consistent with the ALMA data exchange format.
• This activity is essential in the preparation for the GSWP2 experiment.

[11:20] Discussion of the implementation Plan.

• The executive summary of the implementation should be published in BAMS or EOS.
• First the last version of the implementation plan (November 2000) needs to be verified and updated.
• How will the executive summary be prepared ? Andy Pitman will edit the implementation plan for such an article.
• The improvements of ALMA with the introduction of DODS needs to be described in the implementation plan.
• For those GLASS panel members who need to have the support of their home institution to ensure their participation in meetings, a letter from David Carson (WCRP director) can be arranged.
• Paul Houser and Bart van den Hurk will organize the workshop of the local coupled action in spring 2002.

[12:05] Review of the action items (see below)

Finish at 12:30.

Actions decided during the 2001 GLASS panel meeting