# What other tools can complete this exercise besides ArcGIS Spatial Analyst?

The objective in the exercise is to determine how many public computers are available per person as a result of spatial distribution of libraries and people.

Part 1: Grid of points

Using Brooklyn, NY boundary, create a grid of points spaced at 250’. Please report the number of grid points resulting.

Part 2: “Smooth” population on a uniform grid of points

Using Brooklyn, NY boundary, census block population data (US Census 2000) and the resulting grid of points from Part 1, construct a “smoothed” population density for Brooklyn. Using Spatial Analyst create a population raster from census blocks. Note the default density and make sure it corresponds to the 250’ grid point. Apply the population raster to the vector grid making sure that the resulting total population is equivalent to original data. Please show results and document steps in detail.

Part 3 : Determining Library computers per population

Using Brooklyn , NY boundary, uniform population grid resulting from Part 2 and the attached table, determine the interpolated density of publicly available computers from Brooklyn’s Public Library System per population (using the uniform grid of points population). Please show results and document steps in detail.

• The spreadsheet link does not work. – johanvdw Jun 19 '11 at 16:38
• @johan It doesn't really matter, because the question doesn't ask for our help solving this (homework) problem: it asks for references to software that can be used to solve the problem. The tasks are (1-2) create a population density grid from polygon data (i.e., convert from vector to raster), (3) create a density grid from a point table, and (3) divide one grid by another. – whuber Jun 19 '11 at 16:43
• In that case, I find the words used very confusing. 'grid of points' instead of grid. Many programs are able to do this. I will just mention SAGA gis saga-gis.org – johanvdw Jun 19 '11 at 16:49
• Thanks for the quick replies. How about ET GeoWizards or QGIS? Is the spread sheet link working now? – user3396 Jun 19 '11 at 17:14

Quantum GIS together with GRASS will do for sure, as it will for sure with many other Open Source packages. Sorry, no spare time to this exercise for you with QGIS/GRASS (for free).

You can do this with Python.

In pseudo code, this is what I would do:

Load both the shapefiles and the table.

Create a uniform grid of points at 250 feet spacing based on a bounding box of the NY boundary. This should probably be clipped to the boundary.

For each population shapefile feature, snap it's centroid to nearest grid point and add the population to the grid point.

You need to geocode the public library addresses, so that you have an X,Y,nComputers table.

Next use inverse distance weighting interpolation to interpolate the number of computers onto the grid points.

Finally, you can divide the number of computers by the population of the each point and you have an answer.

• what tool would you use to perform "For each population shapefile feature, snap it's centroid to nearest grid point and add the population to the grid point." Seems like there should be something in regular ArcToolbox yes? – user3396 Jun 20 '11 at 3:58
• IDW is not an appropriate approach to creating a density grid: it is designed for a different task altogether (interpolation, not density estimation) and will produce incorrect answers. – whuber Jun 20 '11 at 14:19

This is easy to perform in eg saga gis. 1) Convert the borders to a grid (grid/gridding, shapes to grid) 2) Convert the population shapefile to a grid in the same grid system. If the shapefile contains population and not population density, divide the population by the area first 3) Create voronoi polygons from the points of available computers*. Divide the number of computers by the area of the polygons, multiply by the cell size; convert to a grid 4) divide both grids

In fact there is no need to create grids. You could just intersect both polygon files. You have population density in one shapefile, and computer density in the second one.

*: which is in fact a poor approximation of reality