Update model for robustness in high northern latitudes

[pahbs]

Until resolved, this cloudmasking issue prevents robust analyses of change (eg, in shrub extents across a time series).

Here, this example shows a cloud-free extent from a WV03 scene where the cloudmasking incorrectly masks non-cloud areas.

WV03_20190822_M1BS_104001005054F700 (no cloud mask)

WV03_20190822_M1BS_104001005054F700 (with cloud mask)

[pahbs]

Another example come from the same extent is from: WV03_20170719_M1BS_104001002F7A2E00:

Here, SRlite results (with same contrast stretch as previous post's image).

• black = masked cloud.
• red = edges of cloud that are unmasked
• blue is unmasked cloud shadows.

Weird colors of cloud/shadow-free area are the result of feeding SRLite data that includes cloud shadows.

Finally, an example of a good cloudmask result where clouds are likely close to the ground (b/c no shadows) and SRlite result looks good: WV02_20110727_M1BS_103001000D914900