# Data and Fortran 90 code for the BGEnKF WRF OSSE study

## Reference

The potential benefits of handling mixture statistics via a bi-Gaussian EnKF: 
Tests with all-sky satellite infrared radiances

Authors: Man-Yau Chan (chanmanyau@gmail.com), Xingchao Chen, Jeffrey L. Anderson

Submitted to the Journal of Advances in Modelling Earth Systems.


## Contact info
If you have questions, free to contact Man-Yau Chan (chanmanyau@gmail.com).


## Directory overview
This directory contains:

1) Experiment data for the three experiments in the manuscript
2) Fortran module containing the PSU-EnKF implementation of the BGEnKF.


## Description of the experiment data

The set up of the experiments are described in the Methods section of the 
referenced manuscript. 

The sub-directories containing the experiments' data are:
1) For the   NoDA experiment: `expt_data/noda`
2) For the   EnKF experiment: `expt_data/enkf`
3) For the BGEnKF experiment: `expt_data/bgenkf`


Within an experiment's sub-directory, the data is further broken into dates. 
E.g., sub-directory `expt_data/noda/201110152300` contains WRF data files at 
23 UTC on 15t October 2011.


The WRF data file names have the following structure: `wrf_enkf_INOUT_d01_MEM`.

The `MEM` indicates the ensemble member identification number. Because we have
50 ensemble members, `MEM` runs from 001 to 050.

The `INOUT` indicates whether the WRF data file is from the prior ensemble 
(`input`) or the posterior ensemble (`output`). E.g., `wrf_enkf_input_d01_023` 
is the data file for the prior ensemble member 23.


## Description of Fortran 90 code
The Fortran 90 code for the BGEnKF can be found in the `fortran_code` directory.
1) `fortran_code/gmm_enkf.f90` is the PSU-EnKF's BGEnKF module.
2) `fortran_code/module_structure.f90` contains the modules referenced by the 
   BGEnKF.
3) `fortran_code/sub_enkf_utils.f90` containing utility subroutines and functions.

Note that the BGEnKF is parallelized using the high-latency strategy proposed by 
Andersonand Collins (2007; "Scalable Implementations of Ensemble Filter Algorithms 
for Data Assimilation").

For details on how the BGEnKF works, see the referenced manuscript's supporting 
information document.