v0.15 March 14, 2019
====================================================
no new capabilities, but revised code to work under either python 2 or 3.
Converted all scripts to use optparse for option handling.
For the hq step, the number of processes to use in the calculations is now
an option on the command line, instead of a setting in the config file.
v0.14.1 June 6, 2018
====================================================
mklm.py - Changed from getopt to optparse for command arguments.
hq.py - Fixes to file paths for several files. Was causing an error if user
was not in the working directory when starting hq.
v0.14 Oct 12, 2017
====================================================
Fixed a bug in hsplit.py where if the footprint domain was not the same size
as the inversion domain, the wrong region would be used when extracting the
footprint data out of the grid. This required calling a new routine in
utils.py 'extract_grid'.
v0.13.1 Oct 11, 2017
====================================================
Added a try/except when trying to create a sparse matrix from an H strip file.
If an exception occurs, an error message is printed and program stops.
Fixed bug in hsplit.py. When converting the temporary text files to .npz
files, it now correctly formats an empty .npz file if the text file for a
timestep is missing. This bug caused an error when trying to create a sparse
matrix from an empty .npz file.
v0.13 May 8, 2017
====================================================
Fixed a bug in hq.py. During the loop where it was reading in the h strip
files and computing hq, it was reading the original h strip files, but should
have been reading the hsigma files, which have sigma pre-calculated in them.
Fixed a bug in hsigma.py. The saved hsigma.npz files had the wrong shape,
which was creating errors later when trying to reconstruct the sparse matrix
from the hsigma file.
Modified hq.py and inversion.py slightly to allow importing routines
makeHQBlocks and qht_s into other scripts if desired (mainly for testing).
Added a line to write_sparse() in utils.py to convert insure saved data is a
numpy array, not a matrix. Related to hsigma.npz problem mentioned above.
v0.12 Feb 28, 2017
====================================================
hsplit.py, hsigma.py, hq.py, utils.py, make_z.py
Modifications were made to these files to be able to store the H strip files
as numpy savez files in .npz format in addition to the current binary format.
The savez format is now the default, but the binary format can still be used
with a -b option to the scripts.
Minor changes to apost.py
v0.11 Nov 16, 2016
====================================================
apost.py:
Modified apost.py to fix a bug, and to change logic of how aggregation over
entire time period is done. Before, the complete time period was assumed to
be covered by the individual periods defined by the vshat_dates key in the
configuration file. However, this may not always be true. Also, some of the
variables used for the entire time period also included any extended dates on
the ends, for example 10 days prior to and after the 'actual' starting date.
This resulted in some small error in the total aggregation.
Now a separate total aggregation is done using the 'start_date' and 'end_date'
keys in the configuration file.
Changed description in qsum netcdf files.
utils.py:
Added a check to make sure landmask file exists.
make_z.py:
Added a line to save the sprior data as a numpy file if the given sprior file
was a netcdf file.
inversion.py:
Added a '--resume' option that will skip the calculation of the inverse of
hqht, and the calculation of hqhti*zhsp and the beginning of the inversion.
If --resume is set, then it will read in the saved numpy files hqhti.npy and
hqht_zhsp.npy instead. This gives faster startup in case a previous run of
the inversion was terminated, and it needs to be re-run.
v0.1 Jan 20, 2016
====================================================
First release.