I am using a path distance function to calculate distances from a set of source cells on a land use raster. Unfortunately, my land use raster and my other inputs have different cell sizes. The land use raster has 300 m pixel sizes and my other inputs have ~90m pixels. I would prefer to avoid resampling. I would also like the entire area of the source pixel to be represented as 0 on the resulting distance raster. I will be pulling out individual pixels from the land use raster to use as source cells, and the easiest way to do this is to convert to points . However, given the size differences in the pixels, I'm not sure that it would be accurate to represent the source cells as points. I was thinking of using polygons instead (which can be expanded to the 300 m size of the pixel), but I'm unsure of how the tool will handle the polygons when calculating distance. If the tool will use the centroid of the polygon rather than the edges, then using a polygon is no improvement over using a point to represent the area. I had also considered using a raster as an input, but, again, I'm not sure how the tool will handle the size differences.
Essentially: Does anyone know how exactly ESRI's distance tools handle different inputs during execution when the source inputs may not match the analysis cell size or may span across multiple pixels in the analysis cell size? How does the program handle resampling and/or source cell definitions in these cases?