Static mask

Motivation

This MR would address issue #1 (closed) . I am attempting to "pre filter" for static sources.

Changes

I put a new function in the FireMain.py that is called during Fire_forward. I also added constants to FireConsts.py.

Review

Before this MR actually merges:

1. I Want to do a test run to make sure it actually does do the filtering I'm expecting
2. Would either @greg or @jsignell glance at my coding style? I'm new to python so I'm not sure if my import statements, function documentation, or putting argument in FireConsts.py is following good practice. Should be a quick review.

requested review from @greg

By Tempest McCabe on 2023-02-15T19:43:53 (imported from GitLab)

added 1 commit

• 65e8eb77 - Prevent function name from being overwritten by bool.

Compare with previous version

By Tempest McCabe on 2023-02-15T20:42:24 (imported from GitLab)

added 1 commit

• f4780a77 - replace gpd.read_file with FEDS custom gpd_read_file function

Compare with previous version

By Tempest McCabe on 2023-02-15T20:58:31 (imported from GitLab)

added 1 commit

• b3c11453 - projection fixes

Compare with previous version

By Tempest McCabe on 2023-02-15T21:02:19 (imported from GitLab)

Test failed. Still included fires that should have been flagged as flares.

By Tempest McCabe on 2023-02-16T14:41:13 (imported from GitLab)

Ok! So my entire approach won't work, because later on the code just takes the bounds of any complex region-file and subsets the observations as a box.

Seems like I'll need edit this sub-setting function instead.

By Tempest McCabe on 2023-02-16T16:55:16 (imported from GitLab)

added 1 commit

• 68f5c4d7 - forcing data-filter to use the complex filtering region, instead of the square.

Compare with previous version

By Tempest McCabe on 2023-02-16T17:20:49 (imported from GitLab)

In this commit, I prevent the algorithm from assuming that a shape is approximately the same as the bounds of the shape. This means that for each timestep, the algorithm has to do point comparisons to a complex shape. This may end up taking longer, even if we track fewer fires.

By Tempest McCabe on 2023-02-16T17:23:34 (imported from GitLab)

added 1 commit

• bd6580aa - make consistent projections.

Compare with previous version

By Tempest McCabe on 2023-02-16T21:40:46 (imported from GitLab)

The code now does the filtering! . The code will filter out static sources in my test case. It's also faster! (At least in this context where most of the fires were fires that were filtered out). Pre filter time: 133 mins. Post filter time: 78 mins.

^The square is the modified region, the blue dots are the real fires, and the red dots are fires that weren't read into the algorithm.

I am now only waiting on style thoughts from @greg or @jsignell.

By Tempest McCabe on 2023-02-16T21:45:37 (imported from GitLab)

Nice work Tess (sorry I didn't get to it sooner).

I don't see anything wrong with the code you've written. Seems pretty succinct and clear to me though not an expert on pandas by any means.

By Gregory on 2023-02-20T15:41:24 (imported from GitLab)

approved this merge request

By Gregory on 2023-02-20T15:39:42 (imported from GitLab)

Thanks Greg :). I'll merge it.

By Tempest McCabe on 2023-02-20T20:26:33 (imported from GitLab)

marked this merge request as ready

By Tempest McCabe on 2023-02-20T20:26:44 (imported from GitLab)

mentioned in commit c1716222

By Tempest McCabe on 2023-02-20T20:26:54 (imported from GitLab)

merged