add linke turbidity formulas module to atmosphere #278

mikofski · 2016-12-03T12:15:06Z

closes calculate Linke turbidity factor from AOD, precip water and airmass #269
add kasten 1996 pyrheliometric formula for linke turbidity factor
move atmosphere into package, just for fun, why not?
make init with same api as before

Signed-off-by: Mark Mikofski [email protected]

* add kasten 1996 pyrheliometric formula for linke turbidity factor * move atmosphere into package, just for fun, why not? * make __init__ with same api as before Signed-off-by: Mark Mikofski <[email protected]>

* append linke_turb_forms.py to atmosphere.py * rm -rf atmosphere/ Signed-off-by: Mark Mikofski <[email protected]>

* import xrange from six for python 3 compatibility * fix pyrheliometric formula is from 1980

* just use `range` instead of importing `six` * change function to `kasten80_lt` to match `gueymard94_pw` * start adding references

* add lots of references * cite text quoted directly from Ineichen * acknowledge Armel * fix docstrings for numpydocs

mikofski · 2017-01-25T01:41:16Z

I think this is ready. The Travis failure is not related to kasten96_lt()

wholmgren

Thanks @mikofski.

Is there any way to simplify the API? It feels very complicated to me. Also, the various types of potential inputs are not tested.

Please also add the functions to the api.rst file and update whatsnew.

wholmgren · 2017-01-26T19:06:43Z

pvlib/atmosphere.py

+
+    Method can be either ``'Molineaux'`` or ``'Bird-Huldstrom'``. Airmass less
+    than 1 or greater than 6 will return ``NaN``. Precipitable water less than
+    zero or greater than 5[cm] will also return ``NaN``.


It's not clear to me that we need to set these limits. Garbage in, garbage out.

okay, removed filters and put these limits as a .. warning::

wholmgren · 2017-01-26T19:06:50Z

pvlib/atmosphere.py

+    than 1 or greater than 6 will return ``NaN``. Precipitable water less than
+    zero or greater than 5[cm] will also return ``NaN``.
+
+    Based on originam implementation by Armel Oumbe.


wholmgren · 2017-01-26T19:10:32Z

pvlib/atmosphere.py

+        Linke turbidity
+
+    References
+    ----------


there are a handful of typos in the references

1-3 good references might be more useful than 8.

I fixed the references errors. Not sure I agree about the references. It's hard for me to tease out all of these pieces from just one or two references. I suppose we could just use the Ineichen references which include almost all of the other ones, although some are incorrect:

"A new airmass independent ..." (2002) includes: Kasten-1980, Kasten-1996, Linke-1922, Molineaux-1989

"Conversion function between Linke ..." (2008) includes: Berk-1989,1996, Bird-1980 (incorrect reference), Mueller and Mayer

What is your preference?

I'm ok with a longer references section if that's your preference.

wholmgren · 2017-01-26T19:25:36Z

pvlib/atmosphere.py

+    if method == 'Molineaux':
+        # get aod at 700[nm] from alpha for Molineaux (1998)
+        delta_a = angstrom_aod_at_lambda(aod, alpha)
+    else:


probably best to elif method == 'bird-hulstrom' and then else: raise ValueError('invalid method')

wholmgren · 2017-01-26T19:27:18Z

pvlib/atmosphere.py

+        Angstrom :math:`\\alpha` exponent corresponding to wavelength in ``aod``
+    lambda0 : numeric
+        desired wavelength in nanometers, defaults to 700[nm]
+


Need a Returns section

wholmgren · 2017-01-26T19:30:41Z

pvlib/atmosphere.py

+    return lt
+
+
+def angstrom_aod_at_lambda(aod, alpha, lambda0=700.0):


this function needs its own tests

wholmgren · 2017-01-26T20:59:58Z

pvlib/test/test_atmosphere.py

+    atmos_path = os.path.dirname(os.path.abspath(__file__))
+    pvlib_path = os.path.dirname(atmos_path)
+    melbourne_fl = tmy.readtmy3(os.path.join(
+        pvlib_path, 'data', '722040TYA.CSV')


why add a new tmy file instead of using the existing sample file?

no particular reason, I like that Melbourne-FL site, it's close to Cocoa where I have irradiance data. also although no site matches the Linke turbidity data well, the Alaska site has bigger differences, and also bigger differences between broadband approximations relative to Melbourne-FL. Maybe these are not good tests?

I'd prefer to define ~2-5 element arrays that span the range of allowed input values, and also define the expected return values. I can't quickly inspect the values that are being tested and expected in this test. Maybe these values span the space well enough?

am = np.array([1, 3, 5]) pw = np.array([0, 2.5, 5]) aod700 = np.array([0, 0.1, 1]) lt_molineaux_expected = np.array('your precalculated values') lt_bird_hulstrom_expected = np.array('your precalculated values') ... np.assert_allclose(lt_molineaux_expected, lt_molineaux) np.assert_allclose(lt_bird_hulstrom_expected, lt_bird_hulstrom)

Then you don't need to read a tmy file or lookup linke turbidity in the test. Seems more simple and more explicit to me. I'm not sure how much I trust the tmy and linke turbidity files anyways.

mikofski · 2017-01-27T01:32:57Z

@wholmgren great feedback on all points, thanks! 👍👍

* remove alpha argument * update docstring to explain that if molineaux method, then aod is shape is only aod at 700[nm] but if bird-hulstrom, then aod should be both aod380 and aod500 * add see also to angstrom_aod_at_lambda * add warning for air mass or pwat out of range * check for bird-hulstrom, raise type error if neither molineaux nor * also simplify angstrom_aod_at_lambda() api, change aod to aod0, change lambda0 to lambda1, and add lambda0, update docstring * update test_atmosphere.test_kasten96_lt with new changes

* separate arguments so their shapes don't change depending on method, ie: before aod could be 1, 2, N, N x 1 or N x 2 depending on whether Molineaux or Bird-Hulstrom were used, now, there are 3 separate args, aod700, aod380 and aod500. * fix misspelling of Hulstrom, no "d", note several sources also make this error :( * add new methods to api.rst * add test_angstrom tests * add raises value error test for bad method * add lower() to compare only lower case of method name

* fix more hulstrom errors, no "d" * fix typos in referecnes

mikofski · 2017-01-31T21:47:21Z

ready for your review:

simplified kasten96_lt API
added tests for angstrom methods
added test for bad method raises ValueError
add new methods to api.rst
update what's new 0.4.4

wholmgren · 2017-01-31T21:53:50Z

Thanks. Maybe tonight. Definitely by the end of the week.

…

On Tue, Jan 31, 2017 at 2:47 PM Mark Mikofski ***@***.***> wrote: ready for your review: - simplified kasten96_lt API - added tests for angstrom methods - added test for bad method raises ValueError - add new methods to api.rst - update what's new 0.4.4 — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#278 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AELiR4eVqvUhj0fooMmIYKi1j0qgUxcFks5rX6vqgaJpZM4LDR0Z> .

wholmgren · 2017-02-01T03:39:24Z

These are great changes, thanks! landscape is complaining about a few things related to the doc strings.

mikofski · 2017-02-01T22:21:57Z

Two of the landscape docstring warnings:

Use r""" if any backslashes in a docstring

are due to math, but I just escaped the backslashes, so they are working fine as is, no need to use a "raw" string

mikofski · 2017-02-01T22:22:00Z

The other landscape errors are due to the DOI external links which are too long. Can we add doi to extlink in conf.py like you did for issues and wiki?

extlinks = {
    'doi': ('https://dx.doi.org/%s', 'doi:'),
}

wholmgren · 2017-02-01T22:28:01Z

Sorry, didn't look carefully enough. All sounds good.

* add 'doi': ('http://dx/doi.org/%s', 'DOI: ') to extlink in conf.py * change doi links everywhere to use :doi:`doi number` instead of links

* instead of """docstring with :math:`\\alpha` * add Co-Action publishing for original Tellus A DOI before moving to Taylor & Francis

mikofski · 2017-02-02T00:46:37Z

Okay, I think it's ready now. I added doi as an extlink to the conf.py and I change the docstring to use r"""docstring with :math:\alpha""" to make landscape happy. Those are both pretty neat features, so something new to learn. 😺

mikofski · 2017-02-02T00:54:00Z

Okay, that was my last change. Sorry! 😉 I really think it's ready now. Let me know if there are any more comments or suggestions.

mikofski · 2017-02-02T01:04:13Z

For the remaining landscape complaint:

Too many arguments (6/5)

That's the simplified API for the Kasten LT formula. I had some ideas about that but I wasn't what was best. Basically one idea would be to just have one AOD input called aod_bb for broadband AOD. And then provide one more short formulas for Bird-Hulstrom and remove all of the methods checking. That would also remove 2 references.

def bird_hulstrom80_aod_bb(aod380, aod500):
    """Calculate broadband AOD using Bird-Hulstrom approximation"""
    # put Bird-Hulstrom references here
    return 0.27583 * aod380 + 0.35 * aod500


def kasten96_lt(am, pwat, aod_bb):
    """
    Calculate Linke turbidity factor using Kasten pyrheliometric formula.

    .. warning::
        These calculations are only valid for air mass less than 5[atm] and
        precipitable water less than 5[cm].

    Parameters
    ----------
    am : numeric
        airmass, pressure corrected in atmospheres
    pwat : numeric
        precipitable water or total column water vapor in centimeters
    aod_bb : numeric
        broadband AOD or AOD measured at 700[nm] 

    Returns
    -------
        Linke turbidity

    See also
    --------
    bird_hulstrom80_aod_bb
    angstrom_aod_at_lambda
    """
    # remove Bird-Hulstrom references
    delta_cda = -0.101 + 0.235 * am ** (-0.16)  # rayleigh "clean dry atmosphere"
    delta_w = 0.112 * am ** (-0.55) * pwat ** (0.34)  # water vapor
    delta_a = aod_bb
    return -(9.4 + 0.9 * am) * np.log(np.exp(-am * (delta_cda + delta_w + delta_a))) / am

Would you prefer this approach? Let me know and I'll make the change :)

wholmgren · 2017-02-02T04:01:58Z

Nice. I really like the simplicity of your latest proposal. I'm guessing that most users would find it easier to use.

How about this for alternative aod_bb docstring: "Broadband AOD. Note that AOD measured at 700 nm is a good approximation." Up to you. Also, we don't usually use [] around units in pvlib python. I think text reads better with only a space and that seems to be the standard in all of the journals I read (figure labels are a different story).

For consistency with existing functions, we'd want to use airmass_absolute and precipitable_water.

What did you think about simplifying the tests by using a couple of hardcoded values instead of a TMY file?

It's definitely not necessary for this PR, but it would be nice to have a brief section in clearsky.rst (or maybe a new atmosphere.rst) with a few examples and a little bit of discussion. Something for the future.

mikofski · 2017-02-02T04:37:18Z

Sounds good. I'll make the changes and get back to you.

* change kasten96_lt AOD args to just a single arg called aod_bb for broadband AOD * remove "method" arg and if-then-else on method.lower() in kasten96_lt() * add note that according to Molineaux, aod at 700 nm is a good approximation for broadband aod * add new function bird_hulstrom80_aod_bb to atmosphere.py * add it to api.rst * add new test for bird_hulstrom80_aod_bb() to test_atmosphere.py * update test_kasten96_lt() to use aod_bb * use verbose names: - s/am/airmass_absolute/g - s/pwat/precipitable_water/g * fix angstrom aod test to use more realistic values * don't use brackets around units * remove test for raises ValueError in test_kasten

mikofski · 2017-02-03T01:53:56Z

almost there, just cleaning up the data for the test, and removing the TMY file, should be ready soon.

* values span pwat range 0 - 5 cm, and airmass between 1 & 2.5 atm * remove tmy3 file * don't try Bird-Hulstrom, not necessary * make pylint happy, isclose(a, b) a & b should be array or iterable

mikofski · 2017-02-04T09:24:15Z

It's ready. 🚀

I made a separate issue for the docs. 📚 I feel like it may require some iterations, so would rather split into a new PR, OK?

wholmgren

I will merge as soon as the asserts are fixed

wholmgren · 2017-02-04T14:44:36Z

pvlib/test/test_atmosphere.py

+    aod550 = 0.15
+    aod1240 = 0.05
+    alpha = atmosphere.angstrom_alpha(aod550, 550, aod1240, 1240)
+    np.isclose(alpha, [1.3513924317859232])


Need to assert np.allclose or use np.testing.assert_allclose

Weird. np.isclose should work since these are scalars. Pylint told me to change the scalar to an intake and it passes tests fine

changed these back to scalars, although pylint now complains, oh well.

I see, it's pylint that's the trouble:

In [4]: np.isclose(2, [2]) Out[4]: array([ True], dtype=bool) In [5]: np.isclose(2, 2) Out[5]: True

It's that you have to assert something is true or not. np.isclose only returns an array.

wholmgren · 2017-02-04T14:44:43Z

pvlib/test/test_atmosphere.py

+    alpha = atmosphere.angstrom_alpha(aod550, 550, aod1240, 1240)
+    np.isclose(alpha, [1.3513924317859232])
+    aod700 = atmosphere.angstrom_aod_at_lambda(aod550, 550, alpha)
+    np.isclose(aod700, [0.10828110997681031])


same assert

wholmgren · 2017-02-04T14:44:52Z

pvlib/test/test_atmosphere.py

+    """Test Bird_Hulstrom broadband AOD."""
+    aod380, aod500 = 0.22072480948195175, 0.1614279181106312
+    bird_hulstrom = atmosphere.bird_hulstrom80_aod_bb(aod380, aod500)
+    np.isclose([0.09823143641608373], bird_hulstrom)


same assert

wholmgren · 2017-02-04T14:55:23Z

pvlib/test/test_atmosphere.py

+def test_kasten96_lt():
+    """Test Linke turbidity factor calculated from AOD, Pwat and AM"""
+    timestamps = pd.DatetimeIndex(MELBOURNE_FL[0])
+    timestamps = timestamps.tz_localize('UTC').tz_convert('Etc/GMT+5')


one more that I just noticed... this line errors on the python27-min configuration because timestamps is already tz aware: https://travis-ci.org/pvlib/pvlib-python/jobs/198282511

I don't understand why the python-36 build completed, but we should still fix it.

This was not tz aware on my machine, it gets converted to a naive date time index in utc, that's why I had to localize as UTC then convert to us/Eastern. Is this a pandas or pytz version issue?

this is what I get with pandas

In [21]: pd.DatetimeIndex(MELBOURNE_FL[0]) Out[21]: DatetimeIndex(['1999-01-31 17:00:00', '2000-02-20 20:00:00', '2000-02-22 18:00:00', '2000-02-24 20:00:00', '1995-03-02 19:00:00', '1995-03-11 17:00:00', '1995-03-12 18:00:00', '1995-03-20 16:00:00', '1995-03-20 19:00:00', '1995-03-22 16:00:00', '1995-03-22 19:00:00', '1995-04-07 14:00:00', '1995-04-10 14:00:00', '1995-04-21 14:00:00', '2004-05-01 13:00:00', '2004-05-03 13:00:00', '2004-05-05 18:00:00', '2004-05-16 14:00:00', '2004-05-21 20:00:00', '2004-05-24 16:00:00', '2004-05-31 14:00:00', '2002-06-04 21:00:00', '2002-06-16 22:00:00', '2002-06-17 13:00:00', '2000-07-01 13:00:00', '2000-07-05 12:00:00', '2000-07-06 12:00:00', '2000-07-06 22:00:00', '2000-07-25 20:00:00', '2001-10-14 18:00:00', '2001-10-15 16:00:00', '2003-11-04 19:00:00', '1999-12-20 18:00:00'], dtype='datetime64[ns]', freq=None) In [22]: pd.__version__ Out[22]: u'0.19.1'

the conda list command in the travis builds reports pytz 2016.10 for both builds, so maybe it's pandas. I couldn't find anything about a change in the pandas behavior here, so maybe there's a bug in the newer version. strange.

no change in newest pandas or pytz

In [3]: pd.__version__ Out[3]: u'0.19.2' In [4]: pd.DatetimeIndex(MELBOURNE_FL[0]) Out[4]: DatetimeIndex(['1999-01-31 17:00:00', '2000-02-20 20:00:00', '2000-02-22 18:00:00', '2000-02-24 20:00:00', '1995-03-02 19:00:00', '1995-03-11 17:00:00', '1995-03-12 18:00:00', '1995-03-20 16:00:00', '1995-03-20 19:00:00', '1995-03-22 16:00:00', '1995-03-22 19:00:00', '1995-04-07 14:00:00', '1995-04-10 14:00:00', '1995-04-21 14:00:00', '2004-05-01 13:00:00', '2004-05-03 13:00:00', '2004-05-05 18:00:00', '2004-05-16 14:00:00', '2004-05-21 20:00:00', '2004-05-24 16:00:00', '2004-05-31 14:00:00', '2002-06-04 21:00:00', '2002-06-16 22:00:00', '2002-06-17 13:00:00', '2000-07-01 13:00:00', '2000-07-05 12:00:00', '2000-07-06 12:00:00', '2000-07-06 22:00:00', '2000-07-25 20:00:00', '2001-10-14 18:00:00', '2001-10-15 16:00:00', '2003-11-04 19:00:00', '1999-12-20 18:00:00'], dtype='datetime64[ns]', freq=None) In [6]: pytz.__version__ Out[6]: '2016.10'

I get the same thing on pandas 0.19.2. But also this...

In [11]: pd.Timestamp(MELBOURNE_FL[0][0]) Out[11]: Timestamp('1999-01-31 12:00:00-0500', tz='pytz.FixedOffset(-300)')

see pandas-dev/pandas#11736

mikofski · 2017-02-04T15:30:52Z

Anyway, I didn't like this test and want to change it so this gives me the chance.

* fix assert np.isclose() should be scalar only * super simplify test_kasten96_lt() to ranges of inputs and expected outputs, don't need to compare to actual data, just test that method is working nominally

mikofski · 2017-02-04T16:13:37Z

let's see how this goes. sorry for all the iterations. ☹️

wholmgren · 2017-02-04T16:20:58Z

pvlib/test/test_atmosphere.py



 def test_angstrom_aod():
    """Test Angstrom turbidity model functions."""
    aod550 = 0.15
    aod1240 = 0.05
    alpha = atmosphere.angstrom_alpha(aod550, 550, aod1240, 1240)
-    np.isclose(alpha, [1.3513924317859232])
+    np.isclose(alpha, 1.3513924317859232)


sorry for not being clear. this line and others like it should read assert np.isclose(alpha, 1.3513924317859232)

wholmgren · 2017-02-04T16:26:43Z

no worries. it would help if the "so called judge" (the test suite) actually worked reliably.

mikofski · 2017-02-04T16:37:02Z

Landscape to the rescue - cleaned up unused import (clearsky). Travis reports at least one Python 2 and one Python 3 tests each pass, the others stall or error out on unrelated issues.

wholmgren · 2017-02-06T15:19:30Z

@mikofski In case my comment was buried in the exchange on Saturday morning... the functions that use np.isclose need assert statements in order for them to actually test something: #278 (comment)

I'll also close #101 when this is merged.

* remove old melbourne-fl test data

mikofski · 2017-02-06T20:06:13Z

Oops! Sorry about that! Hopefully good to go.

wholmgren · 2017-02-06T22:36:00Z

Great, thanks Mark!

add linke turbidity formulas module

6e9aaf0

* add kasten 1996 pyrheliometric formula for linke turbidity factor * move atmosphere into package, just for fun, why not? * make __init__ with same api as before Signed-off-by: Mark Mikofski <[email protected]>

wholmgren added this to the 0.4.4 milestone Dec 28, 2016

mikofski mentioned this pull request Jan 15, 2017

advice on modeling at high altitudes? #285

Closed

mikofski added 5 commits January 23, 2017 15:16

ignore venv, sublime

4a2dde7

revert api changes from 6e9aaf0

c492804

* append linke_turb_forms.py to atmosphere.py * rm -rf atmosphere/ Signed-off-by: Mark Mikofski <[email protected]>

move demo to tests

f0b5246

* import xrange from six for python 3 compatibility * fix pyrheliometric formula is from 1980

fix docstring

17f3b9f

* just use `range` instead of importing `six` * change function to `kasten80_lt` to match `gueymard94_pw` * start adding references

test kaste96_lt against linke turbidity using tmy3 at 700nm

33f6374

* add lots of references * cite text quoted directly from Ineichen * acknowledge Armel * fix docstrings for numpydocs

wholmgren reviewed Jan 26, 2017

View reviewed changes

mikofski added 6 commits January 30, 2017 15:15

removing filters from kasten pyrhelometric formula

086b640

add angstrom_alpha() method in atmosphere to calculate alpha

02d1fe5

Merge branch 'master' into linke_turb_forms

645b2fb

add new linke and angstrom functions to api.rst

ec3521c

* fix more hulstrom errors, no "d" * fix typos in referecnes

mikofski added 2 commits February 1, 2017 16:33

add doi to sphinx docs as extlink

1964e20

* add 'doi': ('http://dx/doi.org/%s', 'DOI: ') to extlink in conf.py * change doi links everywhere to use :doi:`doi number` instead of links

make landscape happy and use r"""docstring with :math:\alpha"""

6c3f380

* instead of """docstring with :math:`\\alpha` * add Co-Action publishing for original Tellus A DOI before moving to Taylor & Francis

reformat long lines to make landscape happy :)

fd382bf

mikofski added 2 commits February 4, 2017 01:06

use hard coded test values instead of TMY3

8f667c2

* values span pwat range 0 - 5 cm, and airmass between 1 & 2.5 atm * remove tmy3 file * don't try Bird-Hulstrom, not necessary * make pylint happy, isclose(a, b) a & b should be array or iterable

remove unused imports

e64345e

mikofski mentioned this pull request Feb 4, 2017

add discussion and examples of Angstrom and Kasten formulas to docs #302

Closed

wholmgren reviewed Feb 4, 2017

View reviewed changes

fix test kasten

74524c1

* fix assert np.isclose() should be scalar only * super simplify test_kasten96_lt() to ranges of inputs and expected outputs, don't need to compare to actual data, just test that method is working nominally

wholmgren reviewed Feb 4, 2017

View reviewed changes

remove clearsky import from test_atmosphere

00166d0

mikofski added 2 commits February 6, 2017 11:20

add missing assert statements for tests

59368c8

update hulstrom expected value

13bb4cd

* remove old melbourne-fl test data

wholmgren merged commit 9a120dd into pvlib:master Feb 6, 2017

wholmgren mentioned this pull request Feb 6, 2017

find new turbidity data #101

Closed

mikofski deleted the linke_turb_forms branch February 7, 2017 00:09

mikofski mentioned this pull request Jun 30, 2017

use generic broadband AOD in Bird model instead of only Bird-Hulstrom approximation #341

Open

		return lt


		def angstrom_aod_at_lambda(aod, alpha, lambda0=700.0):

add linke turbidity formulas module to atmosphere #278

add linke turbidity formulas module to atmosphere #278

Conversation

mikofski commented Dec 3, 2016

mikofski commented Jan 25, 2017

wholmgren left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

mikofski commented Jan 27, 2017

mikofski commented Jan 31, 2017

wholmgren commented Jan 31, 2017 via email

wholmgren commented Feb 1, 2017

mikofski commented Feb 1, 2017

mikofski commented Feb 1, 2017

wholmgren commented Feb 1, 2017

mikofski commented Feb 2, 2017

mikofski commented Feb 2, 2017

mikofski commented Feb 2, 2017

wholmgren commented Feb 2, 2017

mikofski commented Feb 2, 2017

mikofski commented Feb 3, 2017

mikofski commented Feb 4, 2017

wholmgren left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

mikofski Feb 4, 2017 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

mikofski commented Feb 4, 2017

mikofski commented Feb 4, 2017

Choose a reason for hiding this comment

wholmgren commented Feb 4, 2017

mikofski commented Feb 4, 2017

wholmgren commented Feb 6, 2017

mikofski commented Feb 6, 2017

wholmgren commented Feb 6, 2017

mikofski Feb 4, 2017 •

edited

Loading