Memos
Memos sent to the GLASS and Rhône-AGG mailing lists with
information about this project are posted here.
Last Updated
July 24, 2001
Sent out April 17,
2001
The Rhone-Aggregation experiment
We would like to announce the start of the
the Rhone-AGGregation SVAT scheme intercomparison experiment.
It is an initiative within the GEWEX/GLASS(Global Land-Atmosphere
System Study: Polcher et al. 2000) /GSWP (Global Soil Wetness
Project: Dirmeyer et al. 1999) panel of the WCRP. As the name
suggests, one of the main goals of this project is to examine the
impact of the horizontal resolution of the computational grid
(and the associated parameter scaling assumptions used by the
models) on the simulated riverflows. At this time, we would like to
hear from modeling groups who are interested in participating, and
any feedback they might have on the proposed experiment.
A document is now available which summarizes the scientific
objectives for the proposed intercomparison, and it describes
the experimental design, the input data sets, the validation
data and the requested model output variables. The Assistance
for Land-surface Modeling Activities (ALMA)
convention is used for the model input and output data.
The project is modeled after the PILPS-2c (Wood et al. 1998),
PILPS-2e (Lettenmaier and Bowling, 2000) and GSWP intercomparison
projects.
Participant SVAT schemes are to be used to simulate
the mass and energy exchanges at the land-atmosphere interface
for multiple annual cycles at a regional scale using prescribed
atmospheric forcing. River discharge will be calculated at CNRM
(Centre National de Recherches Meteorologiques: National Center
for Atmospheric Research) using outputs from the various SVAT
schemes as input into a river-routing model (similar to PILPS-2c
and 2e). A series of experiments are proposed in which the
spatial resolution is reduced and the input parameters are
aggregated and the atmospheric forcing is averaged.
This project differs from the aforementioned ones in that
the spatial resolution of the input data is much greater
for Rhone-AGG, which will, in connection with the relatively dense
observational river gauge network, facilitate an examination of
the impact of model parameter scaling on the SVAT simulated water
balance components. The project is being coordinated by Aaron Boone,
Florence Habets and Joel Noilhan of CNRM at Meteo-France. Pending
the preliminary results of workshop planned for the end of October,
additional experiments might be added to the baseline Rhone-AGG
experiment proposed in the document.
The Rhone-AGG instructions/outline is available online at
http://www.cnrm.meteo.fr/mc2/projects/rhoneagg/index.html
The web pages will be updated and will correspond with the document
on April 18, 2001.
It can be obtained at the CNRM-Meteo-France ftp site:
ftp cnrm-ftp.meteo.fr
LOGIN: anonymous
PASSWORD: your email address
cd pub-ext/boone/rhoneagg
get the file rhoneagg.ps.gz
It is a GNU-zipped PostScript file which is approximately
1.03 megabytes in size: if you wish to obtain it via email as
an attachment, please send an email to boone@cnrm.meteo.fr.
Figures are in grey-scale, however, certain figures are in color
(as images) on the Rhone-AGG web site.
The input atmospheric forcing, parameter datasets and
calibration basin information are contained on a single
CDrom and will be available to all participant groups
by the end of May.
Sincerely,
Aaron Boone, Florence Habets and Joel Noilhan
Sent out May 4, 2001
This is an update for the Rhone-AGGregation experiment.
The data will be ready for distribution (on a single CD)
during the week of May 14-18. We would like to hear from any
groups/people who are interested in participating (aside from
those who have already responded) by May 12, so that we can
prepare the appropriate number of CDs and send out the data as
quickly as possible.
Sincerely,
Aaron Boone, Florence Habets and Joel Noilhan
Sent out May 22,
2001
Dear Rhone-AGG participants,
We currently are in process of writing the CDs
with the Rhone forcing and parameter files. What we
now need from you is to sign the attached data
convention. Please note that we need the form
with the original signature (i.e. not a photocopy or
a FAX). As it might take some time for us to receive
the form via "old-fashioned" mail, we shall mail out
the CD to each group (sometime this week) *before* we
receive the signed convention. If you do not wish to
sign the convention for any reason, please contact us.
We will also be updating the web page soon with
some information on the 2 calibration basins
(the observed discharge, masks, and contributing area
fractions are included on the CD). As a reminder,
a PostScript document describing the experiment
is available at
http://www.cnrm.meteo.fr/mc2/projects/rhoneagg/index.html
(click on "Document" on the menu panel on the left)
Information about the data layout (on the CD),
file format, and some additional information related
to the experiments will be sent with the CD.
Please let us know if you have any problems
down-loading the attached Word file.
Sincerely
Aaron Boone, Florence Habets and Joel Noilhan
Sent out June 7,
2001
Dear Participants,
We sent out the CD data on June 01, so that most
of you should receive it within the next couple
of days (if you have not already received it).
Please contact us if you do not receive the data
by Friday, June 7.
We would like to mention a few updated items, give
some reminders/highlight already mentioned points, and
to give you more information about the requested
outputs (format). Please note that the portions of this file
related to outputs are also posted on the Rhone-AGG web site
(http://www.cnrm.meteo.fr/mc2/projects/rhoneagg/index.html)
with links (click on "Output" on the menu at the left).
1) Input Data
-------------
Please read the README files included on the
CD data distribution. They give further details
on the content of the various model input files.
2) Calibration Data
-------------------
The calibration basins have been selected, and some
information can be found on the website (click on
the button marked "Calibration" on the menu on the left).
The daily observed discharge for the two basins are
contained on the data distribution for a total of 4 years,
along with grid masks and the contributing area fractions (CAF).
Participants can evaluate their simulated discharge
by comparing their daily total runoff (converted to
m3 s-1 using the provided CAF and the fact that each grid
box has an area of approximately 64 km2).
3) Soil Parameters
------------------
As mentioned in the Rhone-AGG experiment description,
we have provided soil hydrological parameters [wilting
point volumetric water content (VWC), field capacity VWC,
VWC at saturation, matric potential at saturation,
hydraulic conductivity at saturation, and the slope of
the retention curve] using two models: the regression
relationships of Cosby et al. 1984, and those from Noilhan
and Lacarrere (1995) using parameter values from Clapp
and Hornberger (1978). We ask that participants use
one of the following models:
i) Cosby et al. (provided for all Experiments)
ii) Noilhan and Lacarrere (provided for all Experiments)
iii) their own parameters derived from the provided
soil texture (and possibly surface type also in the
case of the wilting point VWC) fields.
Please use i), ii) or iii) consistently for all
experiments. Note that the provided aggregated
soil hydraulic parameters for Experiments 2a,b,c
and 3 have been calculated using the provided
*aggregated soil texture*. Modelers may choose to calculate
the aggregated parameters for Experiments 2a,b,c and 3 by
*aggregating the parameters themselves*. This choice is up
to the modelers: we only ask that you tell us which
aggregation method you use.
4) Output
---------
i)Format
Please write the outputs in netCDF format. A
good source of documentation on the netCDF format can be
found at http://www.unidata.ucar.edu/packages/netcdf/guidef/.
The online users manual is filled with example code blocks.
Also, Jan Polcher has written and assembled some general
software for reading and writing netCDF format
(see http://www.lmd.jussieu.fr/ALMA/alma_bazar.html).
ii) Names and Units
The variable names and units are to follow the ALMA protocol
for Rhone-AGG (see http://www.lmd.jussieu.fr/ALMA/). See Tables
5.1 and 5.2 in the Rhone-AGG description for a list of the
output variable names and units.
iii) Naming convention and content
We would like the participants to create two netCDF files
for each experiment (5 total), and for each year
(3 total). The first file should contain the output data
requested at a 3-hour time step, and the second with the output
data at a daily time step (see Tables 5.1 and 5.2 in the Rhone-AGG
description to see which variables are requested at 3-hour or
daily time steps).
Please use the following naming convention:
"model"_exp"number"_"time step"hr_"year".nc
where
model............acronym for your model. It can be whatever
length you choose, but please limit the
characters to numbers, letters, - and _.
number...........experiment number...either 1, 2a, 2b, 2c or 3
time_step........in hours, either 3 or 24
year.............data are to be output using the same starting and
ending times as the atmospheric forcing less
the first (spin-up) year, i.e.
August 1, 1986 (0300) through July 31, 1989 (2400).
As an example, the output file containing the 3-hour variables
from the ISBA SVAT scheme for experiment 2b from August 1, 1987 -
July 31, 1988 would be named
ISBA_exp2b_3hr_8788.nc
The total number of files from each group should then be
2 (3 and 24 time steps) * 5 (number of experiments) * 3 (total years
to report) = 30
Please use the Linux/UNIX compress or gzip (GNU zip) to
compress the output files. For those using PCs, WinZip is also
acceptable.
iv) Array structure
Please write the outputs using the same array
format as the provided input atmospheric data: i.e. 2-D arrays
in space and time for which space is to be treated as a
1-D array (1471 points for Exp1, 20 points for Exp2a, etc.)
using the provided masks. The length
of the time dimension should be defined using "NF_UNLIMITED".
Time should be reported in the same manner as in the inputs:
using the number of time steps and the number of seconds
since Aug. 1, at 0000 LST [eg. for the 3hr output, the first
reported time step (count) is 1, and the first time at 0300
is 10800 seconds]. The daily averages and totals correspond to
the values from the first model time step after midnight of the
previous day to midnight of the current day.
v) Software
As time goes on, we hope to make some relatively simple
software packages/code, which have been developed specificly
for Rhone-AGG, available via the web. Currently,
there is a Pvwave program available for reading data from
the CD, and plotting fields on a rectangular grid
(to obtain the code, click on the button marked
"Input" on the web site menu at the left). This program
is located on the ALMA site.
Participants should check the ALMA software bazaar site
periodically. Multi-purpose software is currently available on the
ALMA site (checking model outputs, read/write in netCDF, etc.).
vi) Errors/Problems
As in previous PILPS exercises and the GSWP, we will
notify participants as soon as possible if we detect
questionable results. The deadline for submission of
results is currently Sept. 1, 2001.
5) Experiments
--------------
The Exp1 run will serve as a model's baseline or control
run. The output total runoff will be used as inputs
into the distributed hydrological model at CNRM. The
resulting baseline river discharge will be compared to
the observed values at various gauging stations within the
basin. The 10 largest sub-basins will be used (see Fig. 2.11
of the Rhone-AGG description), along with other smaller
basins. Grid box average snow depth will also be compared
to the observed snow depths.
The SVAT scheme results (atmospheric fluxes, snowpack
variables, soil moisture, surface state variables...)
from Exp1 will be averaged to the Exp2ac, Exp2b and Exp3 grids,
and the resulting quantities will be compared to the results
from Exps 2a, b, c and Exp 3. Exp2a and c will be used
to examine the impact of parameter aggregation and grid
resolution on the aforementioned SVAT-simulated variables.
Exp 2b will be used to examine the impact of using dominant
parameters only (as opposed to effective parameters).
Exp3 results will be used to examine the impact of aggregation
on treating a basin as the computational "area",
as opposed to an imposed rectangular grid which does not
necessarily respect the "natural" boundaries of the basin.
The simulated discharge will be compared to the corresponding
baseline result and the observed values.
6) Additional Information
-------------------------
We would like the participants to send us a
concise text summarizing the following by the
end of August (we will use this to help us
interpret the results):
i) The most recent/important model reference(s)
for their model
ii) A summary of what model parameters were calibrated
and how/if they were extrapolated over the domain
iii) What type of sub-grid parameterizations are used by
their model (what best describes their model):
a) effective parameters (and which parameters)
b) tile/mosaic approaches (structure: i.e.
multiple columns or many surfaces over a single soil,
etc.)
c) integration using probability distributions
(gamma, normal, etc...and for which variables...)
d) some combination of the above
iv) What sub-grid parameterizations they use
a) sub-grid runoff scheme
b) sub-grid precipitation interception
c) sub-grid saturated fraction for evaporation
c) fractional snow cover area parameterization
and how it influences the overall surface energy budget
d) other?
A model might use only a), or all of the above, etc.
v) which soil parameter model and which aggregation method
they used (see item number 3 above)
vi) If the aggregated surface/vegetation parameters
on the data distribution are not used in favor of
different aggregation methods, please explain what
method was used and for which parameters.
vii) Surface energy budget structure
a) single soil/snow/vegetation composite
surface energy budget
b) distinct canopy and soil with composite snow
c) distinct soil, canopy and snow
d) other?
7) Questions/Problems
---------------------
We will post questions (anonymously) and answers on the FAQs
page of the Rhone-AGG web site (as was done in PILPS-2e). We will
make every effort to update this site every day (that we receive
email) by 1200 GMT.
Good luck!
Sincerely,
Aaron Boone, Florence Habets and Joel Noilhan
Sent out June 12,
2001
Dear Participants,
The purpose of this email is to identify
a correction with respect to one of the time attributes of
the atmospheric forcing files (rhoneagg/Atm_dat/Exp*/*). This
correction does not necessitate rewriting the data, but you
need to be aware of it nonetheless (when reporting results).
A ncdump of rhone_agg_exp*_8586.nc yields the attributes
for time:
float time(tstep) ;
time:units = "seconds since 1985-08-01 00:00:00" ;
time:calendar = "gregorian" ;
time:title = "Time" ;
time:long_name = "Time axis" ;
time:time_origin = " 1985-AUG-01 00:00:00" ;
int timestep(tstep) ;
timestep:units = "timesteps since 1985-08-01 00:00:00" ;
timestep:title = "Time steps" ;
timestep:tstep_sec = 10800.f ;
timestep:long_name = "Time step axis" ;
timestep:time_origin = " 1985-AUG-01 00:00:00" ;
The time axis (the variables time and timestep) and
all of the attributes are correct except
for the *staring time*. The first time step does not
correspond to 3am LST, but to *midnight LST* so that
the time:units and timestep:units attributes (starting
time) for ALL of the atmospheric forcing files should
be:
time:units = "seconds since 1985-07-31 21:00:00" ;
timestep:units = "timesteps since 1985-07-31 21:00:00" ;
So, the first time for reporting results should
be year-08-01 00:00:00, not year-08-01 03:00:00. The last
time step, therefore, corresponds to 07-31 21:00:00.
This correction does not impact the other attributes
or variables. It is important to know for reporting the
requested outputs at the correct times
(and output runoff for calibration
if needed/performed), and for those schemes which
explicitly calculate a solar zenith angle (for surface
or snow albedo, etc.). I am sorry for any inconvenience.
Sincerely,
Aaron
Sent out July 10,
2001
Dear Participants,
At this time we would like to let everyone know
some details about the Rhone-AGG workshop.
The workshop will be held at the CIC (International
Conference Center) on the Meteo-France campus in
Toulouse, France. It will run from Monday, November 5,
through mid-day Wednesday, November 7. We will provide
information about meals and lodging in mid-August.
Tourism information (at least as a starting point)
can be found at
http://www.tourisme.fr/us/annu/index.htm
(the region is the Midi-Pyrenees, and the
department is the Haute-Garonne)
We will update the "Workshop" page on the
Rhone-AGG web site in the upcoming weeks.
Another note: I will be out of the office
from July 16 though July 23, so please contact
Florence if you have any urgent questions during
that time.
Sincerely,
Aaron Boone, Florence Habets and Joel Noilhan
Sent out July 17,
2001
Dear Participants,
I would like to inform everyone who is *calibrating*
their model using the observed discharge of a correction
to make to the conversion from simulated (SVAT) runoff
to total discharge from each basin (m^3/s). If you are
*not* calibrating your model, then you can ignore this email.
* * * * * *
In order to compare the simulated total runoff to
the observed discharge, one computes the simulated
discharge for the 2 calibration basins as:
Discharge_sim = (1/rho_w)*Area_sim*Runoff_sim
where
- Runoff_sim = simulated runoff (kg m^2/s)
- rho_w = density of liq water
- Discharge_sim = simulated discharge (m^3/s)
- CAF = contributing area fraction (0 to 1)
- Area_sim = sum(CAF(:))*area_max_gridbox
(simulated area)
- area_max_gridbox = 64.e6 m^2
But the true drainage area or
"observed area" for each basin is actually slightly
smaller than that calculated using the CAF. The Area_obs
(observed area) is given for each basin on the CD in the
corresponding README file, while the
simulated area is given on the Rhone-AGG Calibration web page
(which is readily computed using the CAF from the CD).
The simulated and observed areas are given in the following
table:
Name ID Area_sim (km^2) Area_obs (km^2)
Ognon at Pesmes 1 2129 2040
Ain at Vouglans 2 1251 1120
Therefore, the simulated discharge calculated by each group
must be calculated using the corresponding observed area.
For example, using the above relation, the "true" simulated
discharge (comparable to the observed discharge given on the CD)
should be calculated as
Discharge_sim' = Discharge_sim*Area_obs/Area_sim
The above results in about a 4% change in the simulated discharge
for the Ognon, and a 10% change for the Ain.
|