• a class to fit the module to monostatic measurements has been added (rtfits.py)
• an example-file has been added to show the basic functionality (examples_fitting.py)
• first unittest has been added (test_rtfits.py)

Implement proper setup.py for the package

• setup.py
• PIP distribution

Still not 100% sure if same results like in paper are produced; validate again

• input-datasets that are not sorted in time result in irradic behaviour!
  • add option for "data already sorted"
  • sort dataset by default prior to processing

create .stickler.yml file

stickler-ci configuration

linters:
flake8:
pep8:
files:
ignore: []

use zarr for writing results

• https://pythonspeed.com/articles/mmap-vs-zarr-hdf5/
• https://zarr.readthedocs.io/en/stable/tutorial.html#

add more comprehensive documentation and link to READTHEDOCS

One test that could be implemented is for a complete homogeneous turbid media with an isotropic surface.

The result of the model should then correspond to the zero order solution based on the transmissivity calculated using the Lambert-Beer law.

get rid of the special treatment of omega, NormBRDF and bsf such that they can
be provided similarly to any other V or SRF specifications

The figures for example 01 and example 02 are not the same as those within the Quast & Wagner paper. Seems that there is still something wrong here.

@raphaelquast can you please check if you can reproduce this error?

Examples can be found in the documentation branch.

Automated testing using nosetests and TRAVIS

• add YML file etc.

A documentation on how to manually define phase-functions and BRDF's would be great...

It's currently possible by changing the properties of a given surface- or volume-class element
but I think we should have a clear and save way how to do it ...
(for example by defining a dummy- surface- and volume-class element that is
intended to be defined by the user)

@bulli92 Any ideas on how to properly implement such a possibility?

sympy?

I installed rt1 under conda environment of python 3.10 using the command pip install rt1[full].
Then I tried the example https://github.com/TUW-GEO/rt1/blob/master/doc/examples/example_lin_comb.ipynb. But when I was running

res = pd.DataFrame(dict(zip(('tot','surf','vol','inter'), R.calc())), index=inc)

I got the error like:

runcell(6, '/Users/xushan/research/TUD/test_jackknife/rt1/toy_example.py')
Traceback (most recent call last):

  File "/Users/xushan/opt/anaconda3/envs/py310/lib/python3.10/site-packages/rt1/rt1.py", line 309, in _get_fnevals
    return self.__fnevals

AttributeError: 'RT1' object has no attribute '_RT1__fnevals'


During handling of the above exception, another exception occurred:

Traceback (most recent call last):

  File "/Users/xushan/opt/anaconda3/envs/py310/lib/python3.10/site-packages/rt1/rt1.py", line 275, in _get_fn
    return self.__fn

AttributeError: 'RT1' object has no attribute '_RT1__fn'


During handling of the above exception, another exception occurred:

Traceback (most recent call last):

  File "/Users/xushan/opt/anaconda3/envs/py310/lib/python3.10/site-packages/spyder_kernels/py3compat.py", line 356, in compat_exec
    exec(code, globals, locals)

  File "/Users/xushan/research/TUD/test_jackknife/rt1/toy_example.py", line 71, in <module>
    res = pd.DataFrame(dict(zip(('tot','surf','vol','inter'), R.calc())), index=inc)

  File "/Users/xushan/opt/anaconda3/envs/py310/lib/python3.10/site-packages/rt1/rt1.py", line 690, in calc
    Iint = self.interaction()

  File "/Users/xushan/opt/anaconda3/envs/py310/lib/python3.10/site-packages/rt1/rt1.py", line 1228, in interaction
    Fint1 = self._calc_Fint_1()

  File "/Users/xushan/opt/anaconda3/envs/py310/lib/python3.10/site-packages/rt1/rt1.py", line 1268, in _calc_Fint_1
    fn = np.broadcast_arrays(*self._fnevals(args))

  File "/Users/xushan/opt/anaconda3/envs/py310/lib/python3.10/site-packages/rt1/rt1.py", line 311, in _get_fnevals
    self._set_fnevals(self._fnevals_input)

  File "/Users/xushan/opt/anaconda3/envs/py310/lib/python3.10/site-packages/rt1/rt1.py", line 340, in _set_fnevals
    self.fn,

  File "/Users/xushan/opt/anaconda3/envs/py310/lib/python3.10/site-packages/rt1/rt1.py", line 277, in _get_fn
    self._set_fn(self.fn_input)

  File "/Users/xushan/opt/anaconda3/envs/py310/lib/python3.10/site-packages/rt1/rt1.py", line 292, in _set_fn
    expr_int = self._calc_interaction_expansion()

  File "/Users/xushan/opt/anaconda3/envs/py310/lib/python3.10/site-packages/rt1/rt1.py", line 545, in _calc_interaction_expansion
    volexp = self.V.legexpansion(

  File "/Users/xushan/opt/anaconda3/envs/py310/lib/python3.10/site-packages/rt1/volume.py", line 325, in legexpansion
    assert self.ncoefs > 0

AttributeError: 'LinCombV' object has no attribute 'ncoefs'

So what happened? Could you please help me to solve this issue? Thanks!

fix standard BRDF normalizations such that they are consistent for the isotropic case, i.e.

• Isotropic := NormBRDF / pi
• HenyeyGreenstein(t=0) = Isotropic = NormBRDF /pi
  - currently we have: HenyeyGreenstein(t=0) = N / (4pi)
• CosineLobe(i=0) = Isotropic = NormBRDF/pi
  - currently we have: CosineLobe(i=0) =NormBRDF

-> adapt tests to fit to the new normalizations

• tests on dev-branch are currently failing for python2 (simple syntax problems)
• new versions of numpy, pandas etc will also drop support from 2020 on...

As this is a rather new project, we should ensure that it supports python 3.x from the beginning before releasing it.

Should take minor effort to do this, but would need to build up some expertise here.

Change also travis.yml to perform tests for different python versions

• some more tests for simple analytic solutions would be nice
• use rtfits.performfit() in tests of rtfits module
• change to cse_symengine_sympy backend in tests

use coveralls to check extent of testing

• currently only performfit() and generate_dyn_dict() use pandas features -> optional dependency for rtfits module

Currently the output of the model is the top of canopy (or volume) specific intensity.

For many applications, it is also desirable to have additional information about e.g

• total canopy transmission
• total canopy absorption

These are implicitly simulated, but are not an output of the model at the moment. Is it possible to have these additional parameters as an output using the current technical ?

• load a dump-file
• generate all plots
• call performfit()
• etc..

One could think of using a deterministic model like e.g. the AIEM or others and fit RT1 coefficients to that model.

One could also think of using existing model of dielectric shapes (e..g for vegetation layers) to determine the phase function behavior and obtain coefficients required for RT1.

Not really thought through, just as an idea.

• test unsorted groupings
  • test interpolation with unsorted groupings
• test load() function
• test start-value assignments via dataset
• test dynamics assignment via integers
• test general_functions.scale and others

Correct treatment of array-processing for tau, omega and R (the hemispher. reflect.)

• this is crucial in order to have fast series-processing (e.g. time-series)
• caching makes it possible to calculate series in approx. same time as single values!
• simultaneous least_squares fitting of multiple datasets possible

within the code, the use of theta_0 , theta_i and mu_0
is not fully consistent, evident from definitions such as:

• theta_i = np.arccos(mu_0)
• def thetap(self, thetai, thetas, phii, phis, a=[-1.,1.,1.])
  which is actually defined to represent thetap with zenith-angle arguments (theta_0, theta_s)

-> this needs to be corrected in order to be able to write a consistent documentation

A possibility for choosing delta-distributions as phase-functions and/or BRDF's would
be nice. This could be achieved by directly including the exact solutions which can easily be gained for delta-like scattering-distributions.

@bulli92 can you have a look at this?

http://rt1.readthedocs.io/en/documentation/annex_technical.html

since thetap and theteBRDF only differ due to the initialization of the
a-parameter in the volume- or surface-class, they are actually a duplicate
of each other, and we can combine them to a single function scat_angle( ... )

• tests need to be adapted if functions are changed
• whenever scat_angle() is called within the definition of a phase-function, one
  must provide self.a as argument since the default value of the a-parameter of
  scat_angle( ... ) will change with respect to the class in which it is used.
  i.e. within _set_function always use:
  self.scat_angle(theta_i,theta_s,phi_i,phi_s, self.a)

The error happens when I ran the code:

    import matplotlib.pyplot as plt
    fig = plt.figure(figsize=(25, 7))
    ax1 = fig.add_subplot(111, projection='polar')
    ax2 = fig.add_subplot(121, projection='polar')
    
    # plot BRDF and phase function
    plot1 = SRF.polarplot(inc = list(np.linspace(0,85,5)), multip = 1.5, legpos = (0.,0.5), polarax=ax2,
                         label='Surface scattering phase function', legend=True)
    plot2 = V.polarplot(inc = list(np.linspace(0,120,5)) ,multip = 1.5, legpos = (0.0,0.5), polarax=ax1,
                         label='Volume scattering phase function', legend=True)
    
    fig.tight_layout()

in the example: https://github.com/TUW-GEO/rt1/blob/master/doc/examples/example_lin_comb.ipynb.
The error information is:

Traceback (most recent call last):

  File "/Users/xushan/opt/anaconda3/envs/py310/lib/python3.10/site-packages/spyder_kernels/py3compat.py", line 356, in compat_exec
    exec(code, globals, locals)

  File "/Users/xushan/research/TUD/vodca/demo_rt1.py", line 120, in <module>
    plot1 = SRF.polarplot(inc = list(np.linspace(0,85,5)), multip = 1.5, legpos = (0.,0.5), polarax=ax1,

  File "/Users/xushan/opt/anaconda3/envs/py310/lib/python3.10/site-packages/rt1/rtplots.py", line 263, in polarplot
    return fig

UnboundLocalError: local variable 'fig' referenced before assignment

It seems that the fig is not explicitly assigned in the function rtplots.py if polarax is not None.

the example-notebooks have not been updated for a long time...

• the monofit function that is used in there has been depreciated...
  (it should be replaced with performfit)
• update structure to new configuration
• use already implemented plotfunctions
• ...

implement a proper logging mechanism
(instead of warnings and prints)

there's currently no warning if a parameter is intended to be fitted but a corresponding key is provided in fixed_dict.
(in this case, the fixed_dict values will be used and the parameter will not be fitted!

Make code faster by e.g. using CYTHON for critical parts

currently fn-coefficients are only correctly evaluated
if a single incidence-angle (and not an array) is provided...

-> this should be easily fixable

lambdify is now pickleable in sympy v1.3 -> remove "cse_seng_sp_newlambdify" since it is no longer needed