================================================================ FLEXINVERT SYNTHETIC DATA PREPARATION ================================================================ Description: Prepares synthetic input data for testing FLEXINVERT Notes: 1) Output from a forward run of FLEXINVERT (run_mode = 0) are needed before running synthetic data test. 2) For CO2 only NEE fluxes are perturbed. 3) Fluxes are returned on the same domain as the flexpart grid_time files (if nested is true this is the nested domain). Requires that the input fluxes cover the same or greater domain than the flexpart files. 4) If the inversion domain (covered by state vector) is smaller than the flexpart domain, only the fluxes inside the inversion domain are perturbed. Synthetic Data: 1) Prior fluxes: prepared by randomly perturbing prior fluxes as defined in SETTINGS_[spec]_files using the uncertainty statistics given in the prior error covariance matrix B The original prior fluxes (not perturbed) are considered as the true fluxes. 2) Observations: prepared by forward simulation of FLEXINVERT using the original (true) prior fluxes. The observations are randomly perturbed using the uncertainty statistics of the observation error covariance matrix R. Usage: 1) Compile with gfortran using: make 2) Edit SETTINGS_files for: i) path_output: the following files from the forward run must be in this directory: - monitor.txt - nbox_xy.txt - hloc_box.txt - evals.txt - evecs.txt - cort.txt The synthetic observations and perturbed prior fluxes are written to this directory. ii) path_prior: the prior fluxes are read from this directory. Use the same SETTINGS_config as for the forward run. 3) Edit the bash script: job_prep_syndata.sh 4) Run the bash script: ./job_prep_syndata.sh