ADAPTATION OF THE MESOSCALE METEOROLOGICAL MODEL MM5 (Penn State University and National Center for Atmospheric Research, USA) at the CRAY-YMP platform (CESCA, Catalunya, Spain)



- INTRODUCTION
 
 

The mesoscale meteorological model MM5 (PSU/NCAR) is a model applicable to limited areas (not on global scale). It has the

possibility of running under hydrostatic and non-hydrostatic modes. The model has been designed by using sigma coordinates to

simulate and predict the mesoscale air circulation.
 
 

The MM5 has been developed by the University of the State of Pensilvania and the National Center for the Atmospheric

Research (NCAR) of Boulder (Colorado). The model is continuing to be improved due to contributions from many useres ariund the

world (Universities, laboratories, etc.)

The Fifth Generation of the Model of Mesoscale NCAR/Penn State (MM5) is the last version of a series of

developments that began in years 70. Since then the model has undergone continuous changes like:

Model MM5 needs certain auxiliary programs for its complete yield. All this joint of programs along with the main modul

MM5 is what constitutes the modelling system MM5.
 
 

The relation between all these programs is in (figure 1, please click here to see the figure). This figure shows to the logical order of

execution of the programs as well as the data flows of such.
 
 

The terrestrial, meteorological and isobaric data, are interpolated horizontally by programs TERRAIN and DATAGRID

respectively, from a mesh of latitude-lengths to a variable domain of high-resolution. This interpolación can be carried

out by means of a projection type Mercator, Lambert or Stereographic.
 
 

The interpolación does not provide the sufficient level of detail, reason why the interpolated data can be improved

(program RAWINS) by means of the inclusion of observations of surface and radiosondages using for this aim the

technique of Cressman.
 
 

Program INTERP carries out the vertical interpolación from pressure levels to the system of coordinates sigma of the

model.

The surfaces sigma near the ground faithfully follow the topography of the terrain, whereas in the high levels but they

tend to come near to the isobaric surfaces. Because as much the resolution as the size of the domain is variable, the

different programs from the model use parametrizaciones for the dimensions, requiring a variable amount of storage

memory.
 
 

Functions of the programs:
 
 

TERRAIN
 
 

It defines the domain of the model and the projection of the map.

It generates the topography and uses of the ground in the cells of the model.
 
 

DATAGRID
 
 

It generates the First-Guess of all the fields of cells of the model from the meteorological data set of entrance.
 
 

RAWINS
 
 

It develops the objective analysis: it combines the First-Guess of the fields of the model with

the radiosondages and the observations in surface.
 
 

INTERP
 
 

It interpolates the data of pressure levels from RAWINS or DATAGRID to the sigma coordinates

of the model.
 
 

GRAPH
 
 

It shows the outputs of each one of the components of model MM5.
 
 

MM5
 
 

It carries out the time integration.
 
 

  -OBJECTIVES:
 
 

The objective of the present Project is to execute and to study the Mesoscale Meteorological Model MM5 at the

Cray-YMP platform of the Supercomputer Center of Catalonia (CESCA) migrating it of its original platform in

the NCAR (National for Center Atmospheric Research) for its possible application to any domains, as

well as to incorporate it in a hierarchy of models that allow to improve the predictions on the quality of the air in

environments from continental to urban scales.
 
 

In order to carry out these objectives, we had to migrate the Fortran code of the model (more than 100,000 lines of code and

near 1000 subroutines) from the NCAR center, with his peculiarities as far as use of his own libraries, compiler and

calls to operating system, to a Cray-YMP machine with 4 processors of the CESCA. It had to know the philosophy

launching of all the components of the MM5 modeling system which are based on scripts C-Shell (use of queues, remote

procedure calls (RPC), configuration of variables, etc.) to be able to modify them and to adapt them to the new machine

Cray-YMP in which we tried to execute the model. The results are goiven in such a way that we have ready all scripts to run

properly the model in opoerational way.

Due to the later closing of the machine Cray-YMP of the CESCA, it was necessary to migrate the model to the IBM SP2

(with 48 nodes), also from CESCA, in which we left an executed code for the main component of the program (MM5).
 
 

Also it has been put into operation, either for CRAY-YMP and IBM-SP2, the possibility of being able to

visualize the results predicted by model MM5 with tool VIS5D (when transforming the binary formats of output of the

MM5 to an format understood by the VIS5D). In addition, a complete work of documentation of the model has been done so that

it will allow to future students and users to carry out changes and new implementations easily and shortly.
 
 

Main changes carried out in the source code:
 
 

 -SIMULATION: CASE EUROPE (NESTED SPAIN)

DESIGN OF THE EXPERIMENT
 
 

The input datas of this simulation come from the IMTA (Mexican Institute of Technology of the Water). I would like to thank the

special contribution from Dr. J.A. Salinas who provided the input data to execute the MM5 main code in our machine.

This case of simulation covers good part with the European continent and in special with Spain. . The data are for 0000, 1200 and 0000

UTM of 29 and 30 of October of 1997.
 
 

(Figure 2, please click here to see the Figure) shows the domain where the simulations was run in this case. The configurations of the

domain for the case of EUROPE study will be described next.
 
 

Main domain (domain 1, approximately Europe):
 
 

Trim in the point 41,0 degrees N and 3,0 degrees W. The size of the grids is of 90 km and the dimensions of domain IX

(direction the North-South) and JX (East-West direction) are respectively of 30*30. It uses domain expanded for the

objective analysis. The expansion in all the faces of the domain is of 360 km.
 
 

Nested domain 1 (domain 2, approximately Spain):
 
 

It has his origin in the mother domain. Domain 2 (figure 3, please, click here to see the Figure) has a size of grids of 30 km and their

dimensions are 25*22.
 
 

REFERRED DATA TO GROUND AND TOPOGRAPHY
 
 

Program TERRAIN must send with two domains according to figures 2,3 to obtain the data referred to the topography

and necessary uses of the ground for DATAGRID.
 
 

OBJECTIVE ANALYSIS
 
 

It is necessary to use the fields of first guess NMC and NAVY SST for DATAGRID. They obtain the first guess of the

fields in 10 levels of standard pressure. These levels of standard pressure are 1000, 850, 700, 500, 400, 300, 250, 200, 150

and 100 mb.
 
 

In RAWINS, 10 new levels of pressure are created: 950, 925, 900, 800, 750, 650, 600, 550, 450 and 350 mb.
 
 
 
 

INTERPOLATION
 
 

Sigma coodinates at 23 vertical levels. These sigma levels are 1, 99, 98, 96, 93, 89, 85, 8, 75, 7, 65, 6, 55, 5, 45, 4, 35, 3, 25,

2, 15, 1, 0,05 and 0.0. It uses INTERP to obtain the file of conditions of contour for the initial domain 1 and fields for

domain 1 and 2 for 3 periods of time.
 
 

SIMULATION OF THE MODEL
 
 

The simulation of the MM5 begins to 0000 UTC of the 29 of October of 1997. Domain 2 simultaneously begins its

simulation that the main domain.The 30 of October finishes the simulation of the model to 0000 UTC.
 
 

The following physical options in the simulation of the MM5 were used:
 
 

 - cloud Equations of Grell with explicit humidity for domain 1 and 2 (including simple ice in the explicit equation).

 - Equation of radiation of Dudhia frequently of radiation of 30 minutes.

 - vertical Diffusion of the humidity in clouds.

 - Coriolis force for nonhydrostatic simulation.

 - Blackdar Pbl.

 - Effect of clouds in the radiation.

 - Without considering the effects of the snow cover.

 - Analysis of embroider in domain 1. Non-Fdda in domain 2.
 
 

COMPUTING TIMES

The simulation with these configuration and input datas, took place in a IBM-SP2 with 48 nodes in the Supercomputer Center of

Catalonia (CESCA). The simulated time was was of 1440 minutes (24 hours). The CPU time required for this simulation was of 30

minutes and 34 seconds.

VISUALIZATION OF RESULTS WITH VIS5D

See Figures 4 , 5, 6 , 7, 8. (PLease, click on the numbers to see the visualization results).
 
 

ACKNOWLEDGMENTS

 We would like to thank the following institutions and persons for his/her collaboration to carry out this project:
 
 

- CESCA: Use of machines Cray-YMP and IBM-SP2.
 
 

- NCAR Wei Wang: Support in all the doubts on any component of the model
 
 

-IMTA (Mexican Institute of Technology of the Water) D. José Antonio Salinas: Input data of

Europe.
 
 

-Steve Emerson: Ordered to give support to the installation of library NETCDF

(support@unidata.ucar.edu).


To see an Spanish version of the PowerPoint presentation, click here