Diagnostic time step
A problem with the calculation of diagnostics in GEOS-Chem was spotted by Jintai Lin. This page explains what was the problem, what was done to correct it and then what impact is to be seen in the results. You can find a full explanation and some output comparisons in this pdf
Time step splitting in GEOS-Chem
In GEOS-Chem, we use the Strang operator splitting. In such a splitting, we do not use the same time step for all processes. Some processes have a time step twice the time step of other processes. Typically, in GEOS-Chem, chemistry, emissions, dry deposition time steps (60 min in 4x5 resolution) are twice advection, convection, wet deposition, PBL time steps (30 min in 4x5 resolution).
So, in 4x5 resolution, GEOS-Chem would run with a time step of 30 min and have the transport (advection, convection, wet deposition and PBL) processes done at each time step. But chemsitry (chemistry, emissions and dry deposition) processes are only done every other time steps.
Because of the time step splitting, at the end of the first time step in GEOS-Chem, tracers concentrations are changed by 30 min of transport and 60 min of chemistry. This state is not physically measurable. At the end of the second time step, tracers concentrations are changed by 60 min of transport and 60 min of chemistry. Only, the concentrations at the end of the second time step should be taken into account for the diagnostics.
Before version 8-02-03 of GEOS-Chem, diagnostic calculations for tracers concentrations were done at the end of each time step, instead of the end of each other time step. This was changed in version 8-02-03.
The problem only concerns concentrations that are affected by processes with different time steps. The calculation of the following diagnostics have been changed:
- ND33 : Column tracer
- ND35 : Tracer at 500mb
- ND45 : Tracer concentrations with time window average.
- ND47 : 24h tracer concentrations
- ND48 : Station timeseries
- ND49 : Instantaneous timeseries
- ND50 : 24h average timeseries
- ND51 : Satellite timeseries
Note: ND31, ND33 and ND68 calculations are changed but without consequences on the results because they output data from the met fields or time-independent data.
The changes in the output tracer concentrations are dependent on the tracers and the time averaging. Typically, more the results are average, less changes are to be expected.
Hourly output present high levels of changes, they can easily reach 10% or more. Simulations that output hourly data may have to be re-done using version 8-02-03 of GEOS-Chem.
Monthly output are almost insensitive. The changes are smaller than errors on measurements. Monthly outputs from previous GEOS-Chem versions can be kept.
--Ccarouge 17:11, 13 October 2009 (EDT)