This repository provide a collection of shell scripts to produce a ensemble analysis (3D+time) of ocean colour observations, using prior statistics from an ensemble ocean ecosystem simulation.
These scripts make use of the SESAM toolbox (https://github.com/brankart/sesam), which requires the EnsDAM (https://github.com/brankart/ensdam) libraries. The installation of these software also requires a FORTRAN-90 compiler and the NetCDF library (with f90 support).
The scripts also make use of the NCO NetCDF operators.
The scripts can be used to perform the following operations (see the README file in the script directory for more details):
- prepare configuration (grid, mask, ...);
- prepare prior ensemble (unconstrained by observations);
- prepare ocean colour observations;
- sample the posterior ensemble (conditioned to observations, using an MCMC sampler);
- diagnose the posterior ensemble (RMS misfit, probabilistic scores, ...).
The scripts use the following datasets:
-
L3 ocean colour products. This corresponds to the tag OCEANCOLOUR_GLO_CHL_L3_REP_OBSERVATIONS_009_085 catalog (https://marine.copernicus.eu/). These data are used as observations to constrain the prior ensemble.
-
A prior ensemble simulation of the ocean ecosystem. These data are used as a prior ensemble (describing the prior probability distribution).
The parameters are specified in the script 'param.ksh', which is sourced in all other scripts so that they all see the same parameters. The parameters include:
- directory settings;
- grid and mask configuration;
- observation parameters (time window, observation error,...);
- multiscale prior ensemble parameters (size, localization scale,...);
- MCMC sampler parameters (sample size, number of iterations, localization,...);
- diagnostic parameters.
The output is an ensemble of possible solutions (in 3D+time), including any combination of variables that are present in the prior ensemble.