POST
|
I am trying to create a traffic density raster (AADT) for wildlife habitat modeling with a kernel distribution. I'm starting with road layers for different states with traffic volumes in the attribute table as AADT. For higher trafficked roads (interstates, etc), I'd like them to be weighted more heavily and have a larger search radius for the kernel density curve away from the road. For instance, I'd like highways to have a 2km search radius but also retain their initial value at the actual road, so it seems multiplying the output raster by 2 gives a value close to the original value. For more local roads, I'd like to use a search radius of 1km, which then does not need to be multiplied. I have several questions pertaining to this method: Does anyone have the exact formula that ArcGIS uses for the kernel line density function in Spatial Analyst? Is there a way to manipulate the search radius and weighting of the kernel based on a separate attribute of the line (for instance, different road classes)? Or is the best solution to just run separate kernel density functions on each class of a line shapefile using a different search radius, multiply them by a factor, and then add them together? I've included an example output that used a 3km search radius across the state of NC to give an idea of what I'd like for an output.
... View more
06-05-2012
07:39 AM
|
0
|
0
|
1595
|
POST
|
Conceptually, I understand that kernel density (for my example) takes the value of a line, smooths it out and distributes it to the given search radius using Silverman's equation. The value of the input line equals the value under the curve of the kernel, but I'm wondering how I could manipulate the spread so that the output raster gives the value of the line for the pixels located at the line. If the line value is 1, for instance, how would I get a kernel density output raster whose value is 1 for pixels at the line? I am using traffic AADT values and would like to stratify my search radii for given road classes (larger search radii for highways, smaller for local roads) to show an increased effect of highways on the adjacent landscape than smaller rural roads. If I increase the search radius to 3000 m, the pixel value at the source is close to but slightly greater than 1/3 of the input value at the line. Is the best way to get an output raster whose values equal the input shapefile simply to multiply it by the search radius (using 3000 m search radius, multiply raster by 3?) Again, this gives me a value slightly larger than the initial value. If I leave the output rasters as is, the increased search radius gives smaller values that are stretched farther out, which is not what I want. Also, for the purposes of potential publication, is this a defensible approach, or is there a better way to go about this?
... View more
05-29-2012
12:47 PM
|
0
|
0
|
374
|
Online Status |
Offline
|
Date Last Visited |
11-11-2020
02:23 AM
|