General information

GEM is Environment Canada and Climate Canada's Global Environmental Multiscale (GEM) weather and seasonal forecast model. Most of the source code is written in Fortran. Just a few basic routines are written in C. The call of the executable is surrounded by pre- and post-processing shell and Python scripts. A hand full of configuration scripts allow the user to set basic parameters.

At the ESCER Centre of UQAM we take a copy of GEM and modify the code and scripts to convert the weather and seasonal forecast model into a climate model. The main differences between GEM and GEM-UQAM are:

The scripts of GEM-UQAM are changed to run simulations by month and to store the output in monthly directories.
GEM-UQAM contains the option to use the land surface model: CLASS (Canadian Land Surface Scheme)
GEM-UQAM contains the option to use the 2D lake model: FLake
In GEM-UQAM all accumulators are averaged and therefore "per second" must be added to accumulator units.

Depending on the version, the official name of the UQAM version of GEM is:

"Developmental version of CRCM5/GEM3, CRCM6/GEM5"
"MRCC5/GEM3, MRCC6/GEM5 en développement"

Model grids

Global and regional grid types

GEM5 supports rotated lat-lon grids:

global uniform
global YinYang (two LAM grids, each covering a little more than half the globe, overlapping like a tennis ball)
regional LAM (Limited Area Model) grids

Enseignement département des sciences de la Terre et l'atmosphère > Introduction to GEM-UQAM > GEM_grids_small.png

LAM grid areas

For LAM grids we distinguish three main regions:

model grid - dynamics: whole grid, including halo/pilot region)
- physics : core grid plus 3 points
core grid - model grid minus halo/pilot region but including blending/merging area/Davies sponge
free grid - inner area

The pilot region or halo contains only the driving interpolated to the model grid. The model need this region to check where the air comes from but the dynamics do not do any calculations in this region. Usually 10 points wide.

In the blending/merging area, also called Davies sponge, the model results get merged with the driving data, giving more weight to the driving data on the outside and more to the model results on the inside, following a cos² curve. This is also the grid specified in the gem_settings.nml. Usually 10 points wide.

The free area contains only the model results. Usually, this is the grid used for the model output.

Enseignement département des sciences de la Terre et l'atmosphère > Introduction to GEM-UQAM > GEM_grid_areas.jpg

Model levels

Vertical model-level structure

Vertical model-level structure showing ζ levels as functions of pressure for values below 200 hPa (where the surface is at sea level) in a typical operational configuration. A sinusoidal mountain extending to 600 hPa in the center of the domain demonstrates the compaction of levels on high terrain. (left) The rectification coefficients are r_min = r_max = 4.5; (right) r_min = 2 and r_max = 100 provide a much faster rectification.

Citation: Monthly Weather Review 142, 3; 10.1175/MWR-D-13-00255.1

Grid points

Grids of GEM version 4 and up are horizontally (Arakawa C-grid) and vertically (Charney–Phillips) staggered.

Enseignement département des sciences de la Terre et l'atmosphère > Introduction to GEM-UQAM > GEM_staggered_grid.jpg

Source code

The source code of GEM consists of the dynamics and the physics plus several libraries.

gemdyn - Dynamics:
                   Takes care of domain, MPI, processor topology, model configuration, domain decomposition, model geometry
                   Main loop over timesteps (in gem_run), input, output, initialization
                   Horizontal advection (also of tracers)
                   Calls the physics
                   Main routine: gemdm.F90
rpnphy - Physics:

gemdyn modelutils rmn rpnphy massv rpncomm vgrid