Global WRF

438 views
Skip to first unread message

Dr. Someshwar Das

unread,
May 4, 2012, 7:43:23 AM5/4/12
to scientists ncmrwf, proj-sc...@ncmrwf.gov.in, project...@ncmrwf.gov.in, india...@googlegroups.com, Thunder Storm, secr...@moes.gov.in, Swati Basu, Swati Basu, swati basu, BN Goswami, Ajit Tyagi, UC Mohanty, shaile...@hotmail.com, Someshwar Das
Dear All,

The WRF model has been run on global domains at 10km, 20km and 100km resolutions on 5, 8 and 16 Nodes of the IBM P6-575 machine at NCMRWF. The CPU time required to run the global WRF model is summarized below.

1. CPU time for Global WRF 100 km resolution, 1 day forecast on 16 Nodes = 2062 S = 0.572 hour
2. CPU time for Global WRF 100 km resolution, 10 days forecast on 16 Nodes = 18757 S = 5.2 hours
3. CPU time for Global WRF 20 km resolution, 1 day forecast on 5 Nodes = 456501 S = 126.8 hours
4. CPU time for Global WRF 10 km resolution, 3 hr forecast on 5 Nodes = 254704 S = 70.75 hours

The last 2 runs have been carried out on dedicated 5 Nodes only in order to avoid hindrance to other users, because these runs take a lot of time. It is estimated that the Global WRF at 10 km resolution will take about 566 hours to make 1 day forecast on dedicated 5 Nodes. Therefore, we can not use this model for operational purpose unless we have a Peta Scale computing power. It may be noted that the WRF model is a highly optimized code developed to run on parallel machines. However, it takes so much time to run, because its physical parameterization schemes including the convection, cloud microphysics, boundary layer and the land surface processes are highly complex compared to other global models.  Sample forecasts made from the above runs can be viewed from the following links.

1. Day-1 forecast from the Global WRF at 100 km resolution: 

2. Day-3 forecast from the Global WRF at 100 km resolution: 

3. Day-5 forecast from the Global WRF at 100 km resolution: 

4. Day-7 forecast from the Global WRF at 100 km resolution: 

5. Day-10 forecast from the Global WRF at 100 km resolution: 

6: Day-1 forecast from the Global WRF at 20 km resolution:

7: Hour-3 forecast from the Global WRF at 10 km resolution: 

I am not trying to make a free advertisement of the model, but the WRF model is also a Unified Model, because it can be run from an eddy scale (a few meters) to the cloud resolving scale (a few kms) to regional scale (a few hundred kms) to the Global scale.  It has provision for 4-D Var Data Assimilation, nesting grids, initializing vortex, movable nest for cyclones, a coupled ocean, and a complex chemistry. Therefore, the model can also be used from nowcasting to short & medium range forecasting, seasonal forecasting and climate scale forecasting. The users have freedom to use any combinations of the schemes from a list of many options for each of the physical processes to suit individual needs. It has the largest user community in the world, many of whom use it for operational weather forecasting in their national meteorological services. The best part is that we are not obliged to pay any royalty for using this model, and yet there is a free community support (thanks to NCAR, the National Centre for Atmospheric Research, USA), a free up-gradation and  a very good scientific documentation, which is updated regularly. The model is used by many research groups and academic institutions/ universities in India, which gives us an advantage to develop it further suitable for the Indian region (I-WRF) by utilizing the potential of the Indian community by working together. 

The main purpose of testing the Global WRF is to explore new avenues for excellence in community development with a scientific spirit. Bricks or Flowers are both welcome as comments.

Cheers, 
**************************************************************************
Dr. Someshwar Das
Scientist-G/ Adviser
National Centre for Medium-Range Weather Forecasting (NCMRWF),
Mausam Kendra, A-50, Sector-62,
NOIDA, PIN-201307, India
.................................................
Phone (O): +91-120-2403928 (Direct), 2403900 (Ext: 277)
FAX ......: +91-120-2403666, 2403611
Email (O): som...@ncmrwf.gov.in
Email (P): some...@yahoo.com, some...@gmail.com
Web : www.ncmrwf.gov.in
**************************************************************************

ajit....@gmail.com

unread,
May 7, 2012, 12:19:46 PM5/7/12
to Prof. D.V.Bhaskar Rao, sto...@googlegroups.com, scien...@ncmrwf.gov.in, proj-sc...@ncmrwf.gov.in, project...@ncmrwf.gov.in, india...@googlegroups.com, Shailesh Nayak, Swati Basu, sw...@ncmrwf.gov.in, swat...@gmail.com, Goswami B N, Mohanty U.C., shaile...@hotmail.com, some...@gmail.com
Dear Prof Bhaskar Rao

Nice to have your mail after a very long time. In fact Indian meso scale modelling community has been missing you.

A number of groups in the country are running WRF. We are trying to create mechanism/platform for interaction among WRF groups and promote systematic indigenous WRF development. Dr Someshwar Das is coordinating Ind WRF initiative.

Looking for your views on WRF initiative.

Regards

Ajit Tyagi
Sent from BlackBerry® on Airtel
From: "Prof. D.V.Bhaskar Rao" <dvb...@rediffmail.com>
Date: 7 May 2012 14:23:43 -0000
Subject: Re: [STORMS] Global WRF


I am very happy about the information provided by Dr. Someshwar Das
about the numerical model runs made with Global-WRF at NCMRWF. I deeply
appreciate the efforts made by Dr. Das at NCMRWF and deserve appreciation.
Although NCMRWF may be doing lot of interesting work, rarely they share
their research summaries with scientific community. I congratulate Dr.Das
to break this barrier and wish him success in his modeling efforts.

Sincerely
Bhaskar Rao
(formerly Professor at Andhra University)


On Fri, 04 May 2012 17:13:40 +0530 wrote

> Dear All,
The WRF model has been run on global domains at 10km, 20km and 100km resolutions on 5, 8 and 16 Nodes of the IBM P6-575 machine at NCMRWF. The CPU time required to run the global WRF model is summarized below.
1. CPU time for Global WRF 100 km resolution, 1 day forecast on 16 Nodes = 2062 S =
0.572 hour2.CPU time for Global WRF 100 km resolution, 10 days forecast on 16 Nodes = 18757 S = 5.2 hours3.CPU time for Global WRF 20 km resolution, 1 day forecast on 5 Nodes = 456501 S = 126.8 hours4.CPU time for Global WRF 10 km resolution, 3 hr forecast on 5 Nodes = 254704 S = 70.75 hours
The last 2 runs have been carried out on dedicated 5 Nodes only in order to avoid hindrance to other users, because these runs take a lot of time. It is estimated that theGlobal WRF at 10 km resolution will take about 566 hours to make 1 day forecast on dedicated 5 Nodes. Therefore, we can not use this model for operational purpose unless we have a Peta Scale computing power. It may be noted that the WRF model is a highly optimized code developed to run on parallel machines. However, it takes so much time to run, because its physical parameterization schemes including the convection, cloud microphysics, boundary layer and the land surface processes are highly complex compared to other global models. Sample forecasts made from the above runs can be

viewed from the following links.
1. Day-1 forecast from theGlobal WRF at 100 km resolution: http://www.angelfire.com/sd/somesh/images/raint-100k-d1-16n.gif
2.Day-3 forecast from theGlobal WRF at 100 km resolution: http://www.angelfire.com/sd/somesh/images/raint-100k-d3-16n.gif
3.Day-5 forecast from theGlobal WRF at 100 km resolution: http://www.angelfire.com/sd/somesh/images/raint-100k-d5-16n.gif
4.Day-7 forecast from theGlobal WRF at 100 km resolution: http://www.angelfire.com/sd/somesh/images/raint-100k-d7-16n.gif
5.Day-10 forecast from theGlobal WRF at 100 km resolution: http://www.angelfire.com/sd/somesh/images/raint-100k-d10-16n.gif
6:Day-1 forecast from theGlobal WRF at 20 km resolution: http://www.angelfire.com/sd/somesh/images/raint-20k-d1-5n.gif
7: Hour-3 forecast from theGlobal WRF at 10 km resolution: http://www.angelfire.com/sd/somesh/images/raint-10k-3h-5n.gif
I am not trying to make a free advertisement of the model, but the WRF model is also a Unified Model, because it can be run from an eddy scale (a few meters) to the cloud resolving scale (a few kms) to regional scale (a few hundred kms) to the Global scale. It has provision for 4-D Var Data Assimilation, nesting grids, initializing vortex, movable nest for cyclones, a coupled ocean, and a complex chemistry. Therefore, the model can also be used from nowcasting to short & medium range forecasting, seasonal forecasting and climate scale forecasting.The users have freedom to use any combinations of the schemes from a list of many options for each of the physical processes to suit individual needs.It has the largest user community in the world, many of whom use it for operational weather forecasting
in their national meteorological services. The best part is that we are notobliged to pay anyroyalty for using this model, and yet there is a free community support (thanks to NCAR, the National Centre for Atmospheric Research, USA), a freeup-gradationand a very good scientific documentation, which is updated regularly. The model is used by many research groups and academic institutions/ universities in India, which gives us an advantage to develop it further suitable for the Indian region (I-WRF) by utilizing the potential of the Indian community by working together.

The main purpose of testing the Global WRF is to explore new avenues for excellence in community development with a scientific spirit. Bricks or Flowers are both welcome as comments.
Cheers,**************************************************************************
Dr. Someshwar Das
Scientist-G/ Adviser
National Centre for Medium-Range Weather Forecasting (NCMRWF),
Mausam Kendra, A-50, Sector-62,
NOIDA, PIN-201307, India
.................................................
Phone (O): +91-120-2403928 (Direct), 2403900 (Ext: 277)
FAX ......: +91-120-2403666, 2403611
Email (O): som...@ncmrwf.gov.in
Email (P): some...@yahoo.com, some...@gmail.com
Web : www.ncmrwf.gov.in
**************************************************************************



--
>
Let no thunderstorm escape unnoticed!
>
Did you know, at any given instant more than 2,000 thunderstorms are taking place around the world.
>

>
You received this message because you are subscribed to the Google Groups "storms" group.
>
To post to this group, send email to sto...@googlegroups.com
>
To unsubscribe from this group, send email to storms-un...@googlegroups.com
>
For more options, visit this group at http://groups.google.co.in/group/storms


Prof.D.V. Bhaskar Rao
Visiting Professor
TLGVRC
Jackson State University
JSU Box 18739
Jackson, MS-39217, USA
Tel:(1)+ 601-979-1892
email: venkata...@jsums.edu
Follow Rediff Deal ho jaye! to get exciting offers in your city everyday.

Dr. Someshwar Das

unread,
May 8, 2012, 6:28:04 AM5/8/12
to sto...@googlegroups.com, scien...@ncmrwf.gov.in, proj-sc...@ncmrwf.gov.in, project...@ncmrwf.gov.in, india...@googlegroups.com, secr...@moes.gov.in, swati...@nic.in, sw...@ncmrwf.gov.in, swat...@gmail.com, gos...@tropmet.res.in, ajit....@gmail.com, ucmo...@gmail.com, shaile...@hotmail.com, some...@gmail.com


Thanks Dr. Mitra for mentioning about the MPAS (Model for Prediction Across Scale) being jointly developed by NCAR and LANL. I also discussed about it recently with Dr. Dale Barker (the main person behind developing the 3DVAR/ 4DVAR of WRF, who is now at UKMO). MPAS is basically being developed on Hexagonal grid to avoid polar singularity that requires special filtering. The polar filters do not scale well on massively parallel computers. The model is not  yet released for public. They are still testing and comparing it with the Global WRF. Perhaps it will take many years before they will include all the features available in the WRF (i.e., Data Assimilation, nesting capability, initializing vortex, movable nest for cyclones, various options for an ensemble of physical parameterization schemes, a coupled ocean, and a complex chemistry, etc.).

By definition, the Unified Model is a single model for NWP and Climate prediction,  which can be used across a range of both timescales (nowcasting to centennial) and spatial scales (convective scale to climate system earth modelling). In that sense, the present WRF model is also capable of doing that as shown by a large number of publications available in the literature.

By the way, sharing/ discussing the scientific results among the scientific community is not a self-propaganda.
 
Regards,
Someshwar

From: Ashis K Mitra <ashis....@gmail.com>
Sent: Tuesday, May 8, 2012 8:24 AM

Subject: Re: [STORMS] Global WRF

Dear Prof. Bhaskar Rao,
 
It is not true that NCMRWF does not share its its ongoing works. Our web page is regularly updated.
NCMRWF work progress is reported in major national and international meetings.
We understand that the real progress in research and development is modest and slow, hence abstain from self-propaganda.  
 
Regarding NCAR's moving towards a real unified model, they are working with other partners on a new modelling system
called MPAS - 'Meso-scale Prediction Across Scales'. I was talking to our ex-NCMRWF colleague Dr. Rizvi (at NCAR), soon
MPAS will be available for research community.
 
Regards,
 
Ashis
NCMRWF

Reply all
Reply to author
Forward
0 new messages