Calculating population in a polygon from a raster

[AlexD1]

Hi all

I am trying to calculate the population within a polygon (SA2 in Australia) from a population grid with a 1 km2 resolution. I don't want the grid value to be double counted when the raster intersects more than one polygon, which I believe excludes zonal statistics. I understand the assumptions with population distribution within the raster cells (eg evenly distributed) but I am having trouble even getting to a point I can do the calculation to get the SA2 populations.

There are a couple of methods I have tried, but none seem to be working as I expected. 

Firstly, turning the raster into points (vector), then summing the points in each polygon. When I use the raster to point tool, not all raster cells convert to points.

Secondly, when I convert the raster to a polygon (gridded vector) to do an area weighted calculation, a lot of the boundaries dissolve so rather than having a 1x1 grid of polygons I have a map where some polygons are 1x1 and others are completely random shapes. Where boundaries have dissolved, they only get the base value of the raster (e.g. if 2 raster cells were each 2 and combined into one polygon, the polygon has a value of 2 rather than 4).

Open to other ideas too if anyone has them!

Thanks

Alex

[DanPatterson]

When I use the raster to point tool, not all raster cells convert to points.

Raster to Point (Conversion)—ArcGIS Pro | Documentation

did you follow the instructions to ensure that the extent parameter is set to avoid resampling and perhaps shifting?

If you have a visual of where cells were not converted to points, that would be useful.

for Zonal statistics

How zonal statistics tools work—ArcGIS Pro | Documentation

did you follow the discussion and comments about cell size, snap raster and cell alignment for zonal statistics?  It may be the cause of some of your issues

[JonathanDelura]

For Zonal Statistics, the value of a raster cell is only caclulated for a zone if it's center is within that zone, so as long as the SA2 polygons don't overlap, there shouldn't be any double counting.