(Vector) masks

[philip-monk]

Hello again,

Would it be possible to create a Jupyter Notebook that demonstrated the use of masks?

A bit of context may help clarify the intent:

I am using Landsat 8 (and eventually Sentinel 2) data to evaluate the thermal effects of renewable energy (on land surface temperature in and around site locations).

The methodology I have inherited and am currently using is very manual.

1. Vectors are hand-drawn for the area(s) of panels, and for the boundary of the site itself.
2. Vectors are drawn around areas that may distort LST calculations - topography, hard surfaces, water, urban areas, disturbed ground et al..
3. The boundary of the site is buffered into 100m strips going outwards for 2km.
4. Distorting areas are masked from these buffers
5. Masked buffers are used to interrogate Landsat 8 values to calculate LST
6. (ANOVA is used to determine any variation in LST pre/post site construction).

At present, I am tangled in a mess of geometries at stage 4 - see https://code.earthengine.google.com/?scriptPath=users%2Fprmonk%2FLST_CI%3AOutline_and_buffers_Stateline. I have overlapping geometries in a geometry collection - some described clockwise, and some anti-clockwise. Naively, I assumed I would be able to just draw some shapes, bundle them together, and subtract them from the buffers. It seems to be a lot more complicated than this!

I know that there is a built-in method, ee.image.mask, and various API's you provide that are related, including ee_to_geojson, ee_to_shp and vector_to_geojson, which may be part of the solution in terms of managing geometries in this kind of task.

I can see the utility in being able to mask an image and control where processing/analysis/sampling/interactions occur in a way that is derived from either an imported geometry and/or a user-input geometry.

I wonder if you could create a notebook to unpick this for the next person (and hopefully help me work out where I am going wrong).

Thanks again for such an amazing resource.

Best wishes,

Philip

[philip-monk]

I imagine it's best not to get yourself in a muddle of geometries in the first place, so obviously demonstrating the best approach is most useful. I wonder, though, if there is also scope to show how to get yourself out of a pickle if you find yourself, as I do, trying to understand geometry behaviour that makes no sense and for which documentation is self-referential (and therefore of very limited use).

[giswqs]

You can use Map.user_rois to return drawn shapes as a FeatureCollection and then use it to do analysis with other FeatureCollection.

[philip-monk]

Thanks for the suggestion.

The only mention of this I can find is https://geemap.org/workshops/GeoPython_2021/#use-drawing-tools but I can't find it in the API list for geemap (or any other module) so I'm not sure how to implement it. Also, how would it be different to the Earth Engine IDE vector tools (Python/JS aside)?

[giswqs]

Map.user_rois is just an attribute of the map object. You won't find it in the API Reference which only lists methods and functions. If you want to see how geemap records drawn shapes, see https://github.com/giswqs/geemap/blob/master/geemap/geemap.py#L484