[RegCNET] Bug in the interactive lake model
bixq
bixq at ictp.it
Tue Sep 28 09:32:07 CEST 2010
Hi, Michael:
We have been asked to not Cc Regcnet at lists.ictp.it until
we get the final frozen version of RegCM-4.0_lake, so when you
reply (or other RegCM_lakers want to join us), please not Cc
Regcnet at lists.ictp.it.
And RegCM-4.0_lake is put as one branches at the svn site, so you can
svn checkout --username anonymous
https://eforge.escience-lab.org/svn/regcm/branches/regcm4.0_lake
On Mon, 27 Sep 2010, Michael Notaro wrote:
> Xunqiang and Nellie,
>
> Thanks for much for the hard work.
> Azi will perform the necessary tests
> and provide the full code.
I am downloading her code right now.
>
> I do have some questions:
>
> - Does the model now correctly produce snow on
> ice? As I mentioned, the precipitation values get
> lost in the previous model's lake subroutines
> and as a result, no snow on ice is produced.
The precipitation seems being passed correctly, one issue is possibly
related to the time step (dtlake), in my understanding, dtlake
should be always equal to dabat, you know, lake is also calculated
in BATS (also in Zeng), the time step for these three schemes should
be the same.
By default, dtlake=3600., it's not good in my opinion.
>
> - For the ocean fluxes (iocnflx), are you suggesting that
> we not use Zeng and instead use BATS in param.F?
I plan to move calling lake from vecbats into tend, and put it after
zeng, so if lakemod=1, lake output (tg1d, tgb1d) would overwrite the other
two schemes' output.
>
> - We originally were running an experiment in which one of
> the great lakes is replaced by a forest type. How should
> that be done in this new version? Just keep lakemod=1
> and replace the 14s in the land use file with a forest type
> number over that specific lake?
You know, from RegCM3 to RegCM4, there is a easy way to do this kind of
LANDUSE change experiment, after first run of terrain, you'll have
LANDUSE file, then you can edit this ASCII file, rerun terrain (by set
lndfudge=.true.) to perform the LANDUSE change.
If it's a small lake (2~3 points), especially they far away of ocean
points, I usually set the landuse type to the same type as the neighboor
point according to Filippo's suggestion. You know, the inland water (lake)
is treated exactly the same as ocean in BATS, but the SST for lake points
usually is a fake value.
>
> I assume that, with this new version, the lake module will
> be executed over all the major lakes in the world (180 of them)
> as long as i_lake=1, lakemod=1 and the land use category
> is set to 14 in bats, right? But, over a specific lake, such as
> Lake Superior, if the land category is changed to a forest
> type, then the lake module will be skipped and the model will
> indeed treat it as forest, right?
If you don't want to consider the small lakes' effect, you could
edit the LANDUSE file and replace the small lake points to other tupe, and
rerun terrain by set lndfudge=.true., so you'll get an
only_consider_big_lake DOMAIN file.
>
> - Is there a time frame for the parallel version since
> it would be a significant improvement over the slower
> serial?
The parallel mode should have exactly the same results as serial one. For
the parallel efficiency, you know, the lake package just take small
fraction of CPU time, so the parallel efficiency should be roughly the
same as no_lake's one.
Personally, I don't too much about lake scheme. You know, among 180 lakes,
some of them are shallow (1 meter depth), do you think Hostetler lake
scheme still work. Is the newer version of Hostetler lake model available?
>
> Thanks a lot,
> Michael
>
>
>
>
> On Sep 27, 2010, at 2:49 AM, bixq wrote:
>
>>
>> Dear RegCM-lakers:
>>
>> We have just finished the draft version of RegCM-4.0_lake. We'll commit
>> it later after Nellie do some benchmarks.
>>
>> Michael and Azar, if you (or other RegCM-lakers) have time, you're welcome
>> join us, have a test and raise your comments. The package is now at:
>> http://www.ictp.it/~bixq/RegCM-4.0_lake.tar.gz
>>
>> I have made the following modifications in the above package:
>>
>> 1. at Terrain step, about 180 lakes are added in subroutine lakeadj
>> (mod_fudge.f90), the elevation of lakes have been adjustes, and the
>> mean depth of lakes have been written into DOMAIN file;
>>
>> 2. get rid of fort.40, instead, the code reads lake depth from #14
>> record of DOMAIN file; There is a limitation here, if the user
>> want to consider the variation of lake depth for certain big lake,
>> he/she have to write a simple code add the information into DOMAIN
>> file;
>>
>> 3. get rid of the swap files (fort.41 and fort.42), instead, I add several
>> 2D (or 3D) fields and use them as swap, this make things easier;
>>
>> 4. I have just checked the serial run, the code can write LAK output
>> (Nellie will do further check). The issue Michael raised, "the
>> interactive lake module doesn't work" probably due to Zeng's air-sea
>> flux schemes used, and there exists inconsistency among them. We'll fix it.
>> You know, by default, Zeng is called after vecbats, but lake is in
>> vecbats.
>>
>> 5. The parallal mode requires further work.
>>
>> To Azar Zarrin: please send me the whole version which you used, so we could have a full comparison. I received those codes you sent me 10 days ago, when I compiled them with rest RegCM3, it failed.
>>
>> Thanks and best regards,
>> Xunqiang Bi and Nellie Elguindi
>>
>> On Mon, 20 Sep 2010, bixq wrote:
>>
>>>
>>> Dear all:
>>>
>>> Thank you for finding the bug !
>>>
>>> I am working on the lake model now and hopefully we could find the solution and make it work by next Monday (Sep.27) in RegCM-4.0.
>>>
>>> Regards,
>>> Xunqiang Bi
>>>
>>> On Sun, 19 Sep 2010, Michael Notaro wrote:
>>>
>>>> To the RegCM3 developers,
>>>> It seems that the interactive lake module doesn't work
>>>> correctly in either serial or parallel mode. In both
>>>> cases, the precipitation values become zero in the lake subroutines
>>>> and there is no snow on ice given the lack of precipitation.
>>>> Could you developers of the model please comment on the problems
>>>> of the lake module and how to resolve these issues? To my
>>>> understanding, these issues continue not just in version 3 but
>>>> were not resolved in version 4 either.
>>>> Thanks,
>>>> Michael
>>>> On 09/19/10, Taha Al Hosary <TAlHosary at ncms.ae> wrote:
>>>>> Dear all,
>>>>> When I ran the RegCM model I found that the results of precipitation amount changes in whole by beginning of period. What is the most appropriate start for good predict for the rainfall?
>>>>> Best Regards,
>>>>> Taha
>>>>> -----Original Message-----
>>>>> From: regcnet-bounces at lists.ictp.it [mailto:regcnet-bounces at lists.ictp.it](javascript:main.compose() On Behalf Of Azar Zarrin
>>>>> Sent: Tuesday, September 07, 2010 7:32 PM
>>>>> To: Regcnet at lists.ictp.it
>>>>> Subject: [RegCNET] Bug in the interactive lake model
>>>>> Dear RegCM developers,
>>>>> I'm running RegCM model coupled with Hostetler lake model for a cold
>>>>> time period that ice and snow would be present over the lakes. I'm
>>>>> getting the reasonable values for ice, temperature and evaporation
>>>>> over the lakes, but the snow values are always zero.
>>>>> I've checked the lake.F and ice.F routines. The precip values which I
>>>>> have printed from Ice. F are zero, while there are big amount of
>>>>> precip and hence snow values outputted from the atmospheric model in
>>>>> the SRF field. It seems that the precip isn't being passed into the
>>>>> ice routine from the BATS and atmospheric model.
>>>>> Would you please verify this bug and let me know how to solve this
>>>>> error.
>>>>> Many thanks,
>>>>> Azar
>>>>> ----
>>>>> Azar Zarrin
>>>>> Postdoctoral Research Associate
>>>>> Center for Climatic Research
>>>>> University of Wisconsin-Madison
>>>>> 1225 West Dayton Street, Room 1152
>>>>> Madison, Wisconsin 53706
>>>>> _______________________________________________
>>>>> RegCNET mailing list
>>>>> RegCNET at lists.ictp.it
>>>>> https://lists.ictp.it/mailman/listinfo/regcnet
>>>>> _______________________________________________
>>>>> RegCNET mailing list
>>>>> RegCNET at lists.ictp.it
>>>>> https://lists.ictp.it/mailman/listinfo/regcnet
>>>> _______________________________________________
>>>> RegCNET mailing list
>>>> RegCNET at lists.ictp.it
>>>> https://lists.ictp.it/mailman/listinfo/regcnet
>>>
>>> ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
>>> Dr. Xunqiang Bi email:bixq at ictp.it
>>> Earth System Physics Group
>>> The Abdus Salam ICTP
>>> Strada Costiera, 11
>>> P.O. BOX 586, 34100 Trieste, ITALY
>>> Tel: +39-040-2240302 Fax: +39-040-2240449
>>> ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
>>
>> ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
>> Dr. Xunqiang Bi email:bixq at ictp.it
>> Earth System Physics Group
>> The Abdus Salam ICTP
>> Strada Costiera, 11
>> P.O. BOX 586, 34100 Trieste, ITALY
>> Tel: +39-040-2240302 Fax: +39-040-2240449
>> ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
>
>
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Dr. Xunqiang Bi email:bixq at ictp.it
Earth System Physics Group
The Abdus Salam ICTP
Strada Costiera, 11
P.O. BOX 586, 34100 Trieste, ITALY
Tel: +39-040-2240302 Fax: +39-040-2240449
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
More information about the RegCNET
mailing list