Add function for accessing MERRA-2

[kandersolar]

• [ ] Closes #xxxx
• I am familiar with the contributing guidelines
• Tests added
• Updates entries in docs/sphinx/source/reference for API changes.
• Adds description and name entries in the appropriate "what's new" file in docs/sphinx/source/whatsnew for all changes. Includes link to the GitHub Issue with :issue:`num` or this Pull Request with :pull:`num`. Includes contributor name and/or GitHub username (link with :ghuser:`user`).
• New code is fully documented. Includes numpydoc compliant docstrings, examples, and comments where necessary.
• Pull request is nearly complete and ready for detailed review.
• Maintainer: Appropriate GitHub Labels (including remote-data) and Milestone are assigned to the Pull Request and linked Issue.

I found a simple way to retrieve MERRA-2 data without needing extra dependencies like xarray and pydap (#1274). It is a little slow, but seems to work well otherwise.

Note, the tests should skip until I add the necessary credentials to the repository's secrets. credentials now added

[AdamRJensen]

Missing a test where the start/end dates have a timezone.

[IoannisSifnaios]

I guess you are maybe missing a "bad request" test? E.g., if username/password do not exist?

Otherwise LGTM! :-)

[adriesse]

Would it be possible to include all these in the variable map? And make that visible in the documentation?

Also, if there is down-welling long-wave, please add it for module temperature modeling.

[kandersolar]

I think the remaining items do not have pvlib names.

Several long-wave variables are available (click to the "Variables" tab): https://disc.gsfc.nasa.gov/datasets/M2T1NXRAD_5.12.4/summary

I guess LWGNT may be the most relevant. Perhaps someone more familiar with this area can confirm?

[williamhobbs]

I would also guess that LWGNT (surface net downward longwave flux) is most relevant. Someone else confirming would still be good.

[AdamRJensen]

The "net" description usually denotes downwelling minus upwelling, but in this case the variable is also denoted as downwelling.

The parameter is described in a NASA appendix here:

LWGNT - Surface net downward longwave flux (W/m2)
The rate of flow of radiant energy reaching the Earth's surface in the thermal infrared spectrum (4-100 µm), expressed in Watts per square meter. LWGNT is a result of atmospheric absorption, emission, and scattering within the entire atmospheric column.

Thus, this is indeed the correct variable of interest.

[AdamRJensen]

I take that statement back. A quick comparison with a BSRN station showed very very different results.

[kandersolar]

I suggest the following mapping:

'LWGNT': 'lwn',
'LWGEM': 'lwu',

We already have ir_down as a parameter in faiman_rad -- how about ir_net and ir_up for these two?

[AdamRJensen]

Longwave irradiance (> 4500 nm) is a subset of infrared (> 780 nm); therefore, in my opinion, calling it infrared is imprecise.

[kandersolar]

Added 'LWGNT': 'lwn', 'LWGEM': 'lwu' in latest commit

[adriesse]

This could be a first for me: I advocate for longer names here! At least lw_down or in more typical pvlib style: longwave_down. The net data is optional.

[williamhobbs]

I think I was confusing "downward" with "net downward". Looks like it's all cleared up, now.

[adriesse]

I found a simple way to retrieve MERRA-2 data without needing extra dependencies like xarray and pydap (#1274). It is a little slow, but seems to work well otherwise.

Is there anything worth describing about this technique? Is it a work-around of some sort?

[AdamRJensen]

LGTM

Do we want to allow inputting pvlib names for the parameters?

[kandersolar]

Do we want to allow inputting pvlib names for the parameters?

I did consider this, but it would require changing the function to iterate over more than one dataset. I am inclined to leave the current (somewhat inconvenient) functionality for now, and if there is demand for a more complex solution then we can add that functionality later.

Is there anything worth describing about this technique? Is it a work-around of some sort?

It's just a simplification of the pydap approach (without actually needing pydap), I suppose. I can't say much about it beyond what the code and comments already document.

[AdamRJensen]

Let's give this PR until the Friday when I suggest we merge it.