Hi @WalkerPowell
Just wondering if you included 3D digital surface model to include the rooftops for your analysis? For example taken from Lidar?
You say the pixels are scatted, could you provide a screen grab for reference of your data. I can see variation along any given rooftop for many reasons. Technically a pole, tree , air conditioner (if burned into the dsm) or adjacent buildings could partially shade an area of the roof part of the time and cause some variation during the analysis (also accounting for edges).
With that I assume you are you trying to identify are the ideal areas for all (possible) rooftops in your study area. Correct?
1. If you have the building footprints you could clip the radiation raster. You could use CON or RECLASSIFY tools to help identify higher "areas" to start to identify areas of interest. However you will still have many one cell or scattered pixels.
2. You could calculate the mean insolation for rooftops using ZonalStatistics and your rooftops as zones. This wouldn't provide you with the best contiguous area to place your panel on any particular rooftop but will give you an better way to classify rooftops.
On the scale you are looking at (entire town). Once you identify potential rooftops the location, size and angle of your PV systems starts to become a different question and there are many different ways to deal with optimizing panel placement.
Hope this helps a little. Or let me know if you have any additional questions.
