-
Notifications
You must be signed in to change notification settings - Fork 2
Spinning up: steady state and constraining to observations
When spinning up, we have four choices, spinning up until a steady state is reached, spinning up whilst resetting the initial concentrations until steady state is reached, constraining to diurnal measurements or a sequential mixture of all of these.
## There exist two ways of spinning up a model. These are both initiated using the obs argument of the model file, and through the `--obs` and `--spinup` flags.
Both of these rely on the SPINUP variable being set in the initial conditions file.
### --spinup : Classical Spinup
This sets a -ve obs input for the model. This means it will run for SPINUP seconds, and then re-initialise those concentrations as set in the initial conditions.
### --obs : Constraints to a diurnal.
Firstly a model will be constrained to a diurnal as in the observations section.
Following this the spinup variable in the initial conditions shall be set to the number of days + fraction you wish to save.
To stop at midday on the 3rd day you would use `spinup = 3.5`.
If constant values are required for the spinup, please create a constant observation to constrain to.
The simplest way to do this is to supply a negative timestep and a 0 as the observation value (do not use -o) flag.(currently deactivated with start script)
Supply a negative timestep in the ICS and a negative observation number. The latter of which is done through the use of the -s flag when running. Both the spinup variable in the ICS and the -o flags are ignored when this is used. The model runs until the fractional difference is <1e-3 and then proceeds to do abs(INTTIME) timesteps.
First the fitobs package must be run to generate the cubic splines and include.obs files required through the model.
Next the model needs to be recompiled using make.
When running the model we may use the -o flag - The -s flag must NOT be set. The model constrains the runs to the diurnal for the number of whole days defined by the SPINUP variable within the initial conditions and then proceeds until it finishes.
Species can be constrained to observations rather than a single value. This may be used to apply timed emissions, deposition or simply to fit a model to the data until it reaches steady state.
Sometimes a model will not let you constrain NO and/or NO2, meaning it falls over with an integration error. Instead it is better initialising those normally and scaling the total nitrogen oxide species TNOX variable instead. This scales the partitioned NOY groups to