JULES separates output into one or more output ‘profiles’. Within each profile, all variables selected for output are written to the same file with the same frequency (also referred to as the ‘output period’). The output period can be any multiple of the model timestep, including calendar months or years.
All output is provided on the model grid only. Each output file contains the latitude and longitude of each point to allow the points to be located in a grid if desired (e.g. for visualisation). Output files also contain two time related variables to locate the values in time (this is described in more detail below).
JULES is capable of performing three different types of time-processing - snapshot (instantaneous) values, time averages and time accumulations. Snapshots are instantaneous values produced during the first model timestep of each output period. Time averages and accumulations are calculated over the output period. Each output variable is annotated with a CF convention cell_methods attribute to indicate whether it is a snapshot value (time : point), time average (time : mean) or time accumulation (time : sum).
Each profile can be considered as a separate data stream. By using more than one profile the user can, for example:
All output files will be NetCDF if JULES is compiled with ‘proper’ NetCDF libraries (see Building and running JULES). Otherwise all output will be in columnar ASCII files.
JULES output files contain two time related variables to allow model output to be associated with the correct model time:
For each output period, this variable contains two values - the start and end of the output period. The output period is then the half-open interval given by:
time_bounds(1) < time <= time_bounds(2)
This is the interval that means and accumulations are calculated over.
During each model timestep, JULES captures values for output at the end of the timestep (i.e. after all the science code). This means that in output files, snapshot data at a particular timestep is:
Due to the way the model equations work, this ensures that all output at a given timestep in the output files is consistent.
With the formulation given above, the initial state of the model (i.e. the state at the beginning of the first timestep of a section) is never output (except to dump files). For the majority of users, this will not be an issue. If the initial state is required, it is possible for an output profile to output the initial state for each section of a run (i.e. initial state of each spinup cycle and the main run) to a separate file - see output_initial.
Warning
In initial data files, only snapshot values for state variables will be valid. All other variables specified in the output profile will exist in the file, but their values will be garbage - not necessarily NAN - so use these files with caution.
JULES writes dump files (a snapshot of the current model state) at several points during a run. These can be used to restart the model from that point if desired. The times that dump files are written are:
Each dump is marked with the model date and time that it was produced.