ChangeLog

version 1.6
	- Added erythemal_weight option to shortwave optimization (see
	optimize_lut_sw.sh) to add extra weight to surface ultraviolet
	spectral fluxes improving the accuracy of UV index for the
	"window", "fine" and "vfine" band structures
	- Added capability to use "training1" line-by-line dataset of 50
	profiles, new quasi-idealized profiles that sample parameter space
	better, via ADDITIONAL_TRAINING_CODE in test/config.h. The
	generated files then have a "t" suffix (possibly in addition to
	the "b" suffix indicating that both the evaluation1 and
	evaluation2 datasets are used in training). The additional
	training1 dataset does not necessarily improve the quality of the
	CKD models generated.

version 1.5 (July 2023)
	- Added hybrid-logarithmic-transmission-3 averaging method
	- optimize_lut exits with a non-zero error code if minimizer
	returns anomalous status (e.g. NaNs or dodgy bounds)
	- If absorption coefficient min=0 and max>0, then ensure minimizer
	bounds in log space are at least 1.0 apart
	- test/run_[lw|sw]_lbl_evaluation.sh can now do the radiative
	transfer to provide all the training data needed in the longwave
	and shortwave

version 1.4 (November 2022)
	- Save solar spectral irradiance (corresponding to
	gpoint_fraction) in output file

version 1.3 (October 2022)
	- Added capability to create-LUT to use transmission averaging
	over 2, 3 and 10 layers instead of just 1.
	- Scripts use directories on ECMWF's Bologna HPC

version 1.2 (May 2022)
	- Added "fine" and "window" shortwave band structures for
	reference calculations including high resolution 290-340 nm for UV
	index calculations
	- create_look_up_table stores the min and max molar absorption
	coefficients, which can be used by optimize_lut to set hard bounds
	on the optimized coefficients and also to set the prior errors

version 1.1 (January 2022)
	- Made use of Adept L-BFGS minimizer the default
	- Removed lbfgs library

version 1.0 (January 2022)
	- Improved optimization of relative-linear representations (CH4
	and N2O), included optimization of CFC11
	- LW scripts produce models with g-point counts divisible by 4
	- Fixed bug in find_g_points shortwave where SSI was used
	without having been reordered
	- Introduced use of sub-bands for the near-infrared
	- Added capability to use Adept's minimizers in optimize_lut.cpp,
	requiring Adept version >= 2.1

version 0.8 (June 2021)
	- Added copyright statements to source files