Centralized chemistry time step
On this page we provide information about the time at which the chemistry, emissions, photolysis, and drydep operations are performed.
Overview
GEOS-Chem requires that the "chemistry" time step, ΔT(chemistry), must be a multiple of the "transport" time step, ΔT(transport). By "transport" we mean the following dynamical operations:
- Advection
- PBL mixing
- Convection
And by "chemistry", we mean the solving of the chemical mechanism matrix, plus the related operations that are used to compute reaction rates, i.e.
- Emissions
- Dry deposition
- Photolysis
- Chemistry
Historically in GEOS-Chem (prior to v11-01):
- ΔT(transport) was a function of the grid resolution
- = 30 min for 4° x 5°;
- = 15 min for 2° x 2.5°
- = 10min or smaller for nested grid simulations (0.5° resolution or finer)
- ΔT(chemistry) was usually set to 60 minutes, for many grid resolutions.
- This may not have always been the optimal setting.
Our most recent recommendation (cf. Philip et al 2016) is:
- ΔT(chemistry) = 20 min
- ΔT(transport) = 10 min
CAVEAT: Running GEOS-Chem with the recommended timesteps from Philip et al. (2016)] has been shown to increase run times by approximately a factor of 2. For tips on improving run time, please see our Speed up a slow simulation guide on ReadTheDocs.
Timestep Durations
Optimal Configuration
This update was validated with 1-month benchmark simulation v11-01f and 1-year benchmark simulation v11-01f-geosfp-Run0. This version was approved on 16 Apr 2016.
Sajeev Phillip and Randall Martin recommend the following default GEOS-Chem timestep durations for future GC simulations when sufficient CPU resources are available.
Sajeev Philip wrote on 9/25/2015:
Philip et al. (2016) examines the sensitivity of GEOS-Chem simulations to the duration of chemical and transport operators. Based on that study, we recommend chemical timesteps of 20 minutes and transport timesteps of 10 minutes (aka C20T10) for simulations when sufficient CPU resources are available. Fine horizontal resolution should take priority over fine temporal resolution. In some cases, it may be beneficial to use coarser timesteps for initial simulations, and the recommended resolution (C20T10) for final simulations. We encourage specification in publications of the duration of operators due to their effects on simulation accuracy.
On 03 Feb 2016, the GEOS-Chem Steering Committee approved the optimal timestep settings recommended by Sajeev Philip and Randall Martin (listed above) for use with GEOS-Chem simulations. Starting with GEOS-Chem v11-01f, the input.geos.template files in the GEOS-Chem Unit Tester now use these optimal timestep settings. When you generate GEOS-Chem run directories for v11-01f and higher versions, these optimal timesteps will automatically be added to the input.geos files as defaults.
CAVEAT: Running GEOS-Chem with these recommended timesteps has been shown to increase run times by approximately a factor of 2. Users may choose to run with coarser timesteps for faster turnaround with lower accuracy. For more information please see our Speed up a slow simulation guide on ReadTheDocs.
Default in v9-01-02 and Later Versions
Currently, in the GEOS-Chem main.F driver program, "transport" is always done before "chemistry". This is due to historical usage. This order may be changed in the future; however we need to make sure that by doing so we do not break the current functionality. We also must keep in mind that the state of the atmosphere (i.e. advected tracer concentrations in the STT array) is used to initialize the "chemistry" operations since (1) gas-phase tracer concentrations are fed into the chemistry solver (SMVGEAR or KPP), and (2) aerosol optical depths are computed from the aerosol tracer concentrations and fed into the Photolysis mechanism. In addition, GEOS-Chem currently computes the photolysis with the sun angles at the center of the "chemistry" timestep. The state of the atmosphere that is used to initialize the "chemistry" should therefore be close to the center of the "chemistry" timestep.
In GEOS-Chem v9-01-02, the time at which "chemistry" is done is closer to the midpoint of the "chemistry" time step, ΔT(chemistry). This update was tested in the 1-month benchmark simulation v9-01-02q and approved on 18 Oct 2011, and is illustrated in the diagram below.
--Bob Y. 16:16, 20 October 2011 (EDT)
References
- Santillana, M., L. Zhang, D.J. Jacob, R. Yantosca, Estimating numerical errors due to operator splitting in global atmospheric chemistry models: Transport and chemistry, in preparation, 2011.
--Bob Y. 11:05, 21 October 2011 (EDT)