Generate city center point shapefile based on satellite luminosity data

Discussion created by sottin on Jan 30, 2013
Hello everyone

I am facing some problems solving the following task:

I'd like to identify the city center points for each polygon by using light intensity, based on (1) a raster file, displaying luminosity at night (values from 0 to 63) at a 1 sqm /pixel resolution and (2) a polygon shape file, consisting of administrative boundaries (say, e.g. counties or districts).

My idea for a solution method is as follows: First I'd want to split the raster based on the individual polygons extent to generate a raster file corresponding to each shape file.

Next I'd like to identify the pixels having the highest raster values, and - given it is only one of these in each new created raster (which is unlikely) - I'd want to create a point.
As there usually is more than one pixel with the highest raster value of 63 (or lower, the maximum value depends on the individual raster) I then want to create "moving" buffers around the points having the highest raster value, in order to identify which pixel actually fits the description of city centers. By that I mean, that as long as to buffers exhibit the same sum of raster values, the buffer radius should be increased repeatedly to finally find out, which single raster cell actually is the center of the most illumined area within each of those polygons, and then append to it to point shape file created in the step before.

On top of all this I'd like to automate the process using ModelBuilder.

Thus my questions are as follows:
(1) How to split one raster into many based on polygons? ("Split Raster" seems to to offer this option)
(2) Which tools to use for finding the city center within each of those polygons?
(3) how to automate this conditional statement using ModelBuilder?

Thank you very much in advance for your tips and comments!