5. Output

Aronnax produces output in two formats. The full fields are saved in Fortran unformatted files, that are compact and efficient, but not particularly user friendly. Layerwise statistics are saved into one csv file per variable.

5.1. Reading the data

Aronnax includes a number of helper functions for dealing with the unformatted Fortran output.

The simplest of these is interpret_raw_file which loads a single output of a single variable into a numpy array.

aronnax.interpret_raw_file(name, nx, ny, layers)

Read an output file dumped by the Aronnax core.

Each such file contains one array, whose size depends on what, exactly, is in it, and on the resolution of the simulation. Hence, the parameters nx, ny, and layers, as well as the file naming convetion, suffice to interpret the content (assuming it was generated on the same system).

Aronnax also ships with a function for lazily loading multiple timestamps of a single variable. This function uses xarray and dask to create labeled n-dimensional arrays.

aronnax.open_mfdataarray(files, grid)

Open a number of output files into an xarray dataarray. All files must contain the same variable.

Uses dask.delayed to lazily load data as required.

Parameters:

• files (list) – a list of file names to load
• grid – a grid object created by aronnax.Grid

Returns:

xarray.DataArray of the desired variable