I'm currently working with Geotiff mask data (simple 1 layer files that serves just to mark some area of interest). I've found few solutions to serve this data on frontend but I'm wondering if there is any known "good practice".

There are 3 solutions so far I'm considering

1. Prerender tiles, and then serve them as static image files stored as blob.
2. Store cloud optimized geotiff and then send range http request to get required tile
3. Serve raw data and render it via JS components (eg. plotly heatmap)

So far second approach was in my opinion best way to go, but what if i need to do some operations with geotiffs in future, like subtract one mask from another or append those masks ?

For example if we consider masks subtraction two problems emerges here:

• It could easily generate a lot of temporary blob result files
• Generating COG from result will take some time on backend

On the other hand extracting data from geotiff, doing some numeric operations on it and serving raw number to frontend to be rendered with JS library is tempting in this scenario, but I'm afraid about performance for larger masks. Right now i'm thinking about dynamic interpolation of data in frontend to decrease array size before rendering. Has anyone got any experience with similar situations and could share thoughts about it?