FYI, I want to be more transparent about how I generated the altitude dataset. I'm trying to consolidate the generation into one python script so is easy for you guys to run it. I will probably release it as a Github Gist soon.

Also, I want to do a better job quantifying the accuracy of the generated map. I will try to find a good source of ground truth for altitudes and release some error stats.

Here is the Gist I promised that allows you guys to reproduce the altitude map.
https://gist.github.com/nicomt/d2c5f08a8ee3500550be42d8dbee6c1d

While compiling the script, I realized that the lower altitude bound in my first commit was lower than the lowest place on earth (The Death Sea). In this version, I fixed the lower bound to -450m, which removes those bad values and allows me to reduce the step 28m instead of 35m.

Also, I compared the results with a few samples from the Google Elevation API, and here is what I got.

It looks like the error is pretty bad in Greenland and Asia but relatively okay elsewhere.
Here are the actual results of my test.

Let me know if there is something else I can do to help.

Thanks so much Nicolas, Two very minor nits that aren't blockers and can be improvements in future PR's, I don't want to hold up merging this asap.

1. in test_lookup_altitude() a fixture might work better instead of looping over test_locations
2. in location.lookup_altitude() it would be nice if we could adopt pathlib.Path and the / operator for all new code and retire osp.join(). EG:

   import pathlib
   pvlib_path = pathlib.Path(__file__).parent
   filepath = pvlib_path / 'data' / 'Altitude.h5'

@mikofski I will fix those in the next PR. Thanks