Overview of CanESM

History and application

The Canadian Earth System Model is a global model developed to simulate historical climate change and variability, to make centennial-scale projections of future climate, and to produce initialized seasonal and decadal predictions. CanESM has a pedigree extending back over 40 years, having built on many preceeding model versions. CanESM, its predecessors and components have been widely applied for various phases of the Coupled Model Intercomparison Project (CMIP), the Chemistry Climate Modelling Initiative (CCMI), for operational seasonal prediction at Environment Canada, and many other scientific applications.

Well over 200,000 years of climate simulation have been conducted with CanESM, and the results have appeared in hundreds of peer reviewed journal publications, Intergovernmental Panel on Climate Change (IPCC) Assessment Reports, World Meterological Association Ozone Assessements, and Artic Monitoring and Assessment Program Reports, amongst others.

Selected CanESM output for official CanESM simulations is publicly available on the ESGF, and on the CCCma website.

Model description

A complete description of CanESM5 and its components is given in Swart et al. (2019) <https://gmd.copernicus.org/articles/12/4823/2019/>. A brief summary is provided here:

Component

Description

CanAM

Canadian Atmosphere model, a global, 3D general circulation model

CLASSIC

The Canadian Land Surface Scheme and Canadian Terrerestrial Ecosystem Model (embedded in CanAM)

CanNEMO

NEMO ocean model modified for CanESM, including ocean biogeochemistry

CanCPL

The Canadian Coupler, which connects CanAM and CanNEMO

Scientific documentation

The peer reviewed literature provides the definitive source of documentation for the scientific aspects of CanESM. Some scientific documentation also appears inline in the model routines. Key publications and scientific user guides are referenced below.

Overview of basic operations

At a high level, CanESM must be configured, compiled and run, and then outputs must be post-processed. Each step is treated briefly below, and more completely in the remainder of the guide.

Configuration

Configuration is achieved via cpp directives at compile time, and via various namelists at runtime. A set of configuration utility scripting is normally used to handle creating consistent model setups, including specifying cpp and namelist settings, configuring consistent component timing, specifying the correct inputs and outputs, and controlling the runtime and post-processing sequencing.

Compilation

The code for CanAM (including CLASS/CTEM), CanCPL and CanNEMO are each compiled into a separate executable. Hence any complete CanESM run uses three executables. Ocean only runs use only the NEMO executable, while atmopshere only runs (AMIP) rely on CanAM and CanCPL. Compilation is nominally achieved via standard Makefiles, with some additional utility scripting.

Runtime

The model requires various inputs at runtime, including configuration namelists, and various forcing or data files. Each component requires unique inputs, which depend on the experiment and configuration. Each model component also has its own restart files, which are recorded at least once at the end of a block of simulation. The coupled model is currently always started from an existing restart.

The three executables communicate over MPI during a simulation. Hence, runs are launched in so called Multiple Program Multiple Data Mode. Each component controls its own timing. Information is passed between CanAM and CanNEMO, via CanCPL, at a specified coupling interval. During the course of a run, each component will write out data into raw history files, as well as various log files. CanAM input and output occurs in CCCma format, whereas CanNEMO and CanCPL use NetCDF formatted files. Logs are all plain text format.

Long periods of simulation are generally run in chunks. The optimal chunk size depends on the wallclock limits and stablity of the machine being used. A set of sequencing infrastructure is typically used to orchestrate the running of many chunks to form a complete simulation, including various file handling, and potentially the launching of diagnostics which post-process the model output.

Post-processing

Various post-processing of raw model history files is required, nominally joining tiles into global files; producing appropriate time averages, spatial transforms, and derived quantities. The final stage of post-processing is the creation of CMIP-compliant NetCDF data. An extensive set of diagnostic tools exists to achieve this post processing, whose complexity and expense rivals that of the model itself.

CanESM key references

The following comprise scientific references that document the formulation and characteristics of CanESM5 and its model components. More technical references can be found in the Component Documentation section.

Coupled Model and CanESM5 updates

CanAM

CLASS / CTEM

CanNEMO, CMOC and CanOE * Christian et al. (preprint)