Resolved From orthographic depth to perspective position?

So I created a shader for dynamic snow/rain coverage.



The final shader is applied in screen space, but in between I do the following:

- Create a grid of the map
- Render each grid cell from top with an orthographic camera, it looks like this



- Now get all visible grid cells in view and render them using the main camera using an unlit shader (to capture depth and the coverage amount):




Now I can use this information when applying the final fullscreen shader.

The problem is: It obviously works only for the same height, meaning the grid-ground has to be even. I could create another grid + cells for each height layer and use transparency-hacks so they can properly overlap etc... But I wonder, is there maybe a way to capture the height or rather depth from the orthographic camera and then use that depth to displace the pixels to their actual "height" when I render all the grid cells in perspective mode?

 

I solved it by rendering each cell on top of the underlying geometry using a decal shader.