TelFit Code

This code will fit the telluric absorption spectrum in data, using LBLRTM. More details can be found in the documentation and examples, but the installation and run procedure are outlined below. If you use this code, please cite my paper.

Installation

This code originally only ran on python 2, but should now ostensibly run on python 3.

The following packages are required (version numbers are a carryover from the python 2 version):

• matplotlib
• numpy v1.6 or greater
• scipy v0.13 or greater
• astropy v0.2 or greater
• lockfile
• pysynphot v0.7 or greater
• fortranformat
• cython
• requests

The bolded entries are required before installation, so make sure you get them from pip, apt-get/yum, or conda (depending on your OS and python distribution). The setup script will attempt to install the rest if you don't have them, but I suggest doing it yourself just to make sure nothing goes wrong. Once you have the dependencies, then run

git clone https://github.com/freddavies/Telluric-Fitter-py3.git
cd Telluric-Fitter-py3
python setup.py install

to install TelFit. It may take a while, as it needs to build the LBLRTM code and some of its standard input files.

If there are Fortran compile errors

If you are running gfortran version 8 or higher, you will very likely run into some errors when LBLRTM compiles. Fortunately, there is a way out. The TelFit installation places the LBLRTM and LNFL codes in the directory ~/.TelFit/ and attempts to compile them there. The compilation statement for whatever system you are running needs to be modified to include the -std=legacy flag, and then run python setup.py install once more to finish the installation.

WARNING: Any changes you make to the makefiles, as described below, will be overwritten if you simply run python setup.py install again. Comment out line 210 in setup.py to bypass re-un-packing the tar files,

#subprocess.check_call(["tar", "-xzf", '{}{}'.format(TELLURICMODELING, fname), '-C', TELLURICMODELING])

Here are examples for modifying the makefiles. For my system, in the file ~/.TelFit/lblrtm/build/makefile.common, the relevant block is:

osxGNUsgl:
    ${MAKE} -f ${MAKEFILE} all P_TYPE=sgl FC_TYPE=gnu \
    PLTFRM=OS_X \
    FC=gfortran \
    FCFLAG="-frecord-marker=4" \
    UTIL_FILE=util_gfortran.f90

To fix it, I add the std=legacy flag to FCFLAG, like so:

osxGNUsgl:
    ${MAKE} -f ${MAKEFILE} all P_TYPE=sgl FC_TYPE=gnu \
    PLTFRM=OS_X \
    FC=gfortran \
    FCFLAG="-frecord-marker=4 -std=legacy" \
    UTIL_FILE=util_gfortran.f90

The same flag must also be added to FCFLAG in the file ~/.TelFit/lnfl/build/makefile.common:

osxGNUsgl:
    ${MAKE} -f ${MAKEFILE} all P_TYPE=sgl FC_TYPE=gnu \
    PLTFRM=OS_X \
    FC=gfortran \
    FCFLAG="-Wall -frecord-marker=4 -std=legacy" \
    UTIL_FILE=util_gfortran.f

After changing the makefiles, re-run python setup.py install to compile LBLRTM and finish installing TelFit.

Remember: you must comment out the line in setup.py that unpacks the original tar files, or these changes will be overwritten!

Running TelFit

To run TelFit, you should create a script like in the examples. The key parts of the script are the inputs to the TelluricFitter class. You should:

• Initialize fitter: fitter = TelluricFitter()
• Define variables to fit: must provide a dictionary where the key is the name of the variable, and the value is the initial guess value for that variable. Example: fitter.FitVariable({“ch4”: 1.6, “h2o”: 45.0})
• Edit values of constant parameters: similar to FitVariable, but the variables given here will not be fit. Useful for settings things like the telescope pointing angle, temperature, and pressure, which will be very well-known. Example: fitter.AdjustValue({“angle”: 50.6})
• Set bounds on fitted variables (fitter.SetBounds): Give a dictionary where the key is the name of the variable, and the value is a list of size 2 of the form [lower_bound, upper_bound]
• Import data (fitter.ImportData): Copy data as a class variable. Must be given as a DataStructures.xypoint instance
• Perform the fit: (fitter.Fit): Returns a DataStructures.xypoint instance of the model. The x-values in the returned array are the same as the data.