jackie_cupples

Using a Spatial Weights Matrix with Hot Spots Tool

Discussion created by jackie_cupples on Sep 24, 2013
I would like to use the Hot Spots tool in Arc to determine areas for our state transportation agency to install wildlife crossings.  We initially used this tool in two ways:  1) aggregating the data in a fishnet clipped to a 50 meter buffer of the highway; and 2) aggregating the data by highway segments (ranging from 500m ??? 3000m in length).  However, when we contacted someone at Esri, it turns out that the Euclidian distances to nearest neighbor DVCs that are used in calculating the hot spots does not follow the path of the highway.  We were advised that if we generated a spatial weights matrix (.swm) file using the Generate Network Spatial Weights tool and used the output .swm file in the hot spots tool, the Euclidian distances would in essence travel the path of the highway in question and our hot spots results would be more accurate.

However, I???m a little befuddled by the .swm file.  I understand that in running the generate network spatial weights tool, the .swm file calculates the distance from each point (in our case, DVC) to every other point, while traveling the course of the highway.  And, it makes sense the input file for this tool would be the point file with the DVCs (snapped to the highway).  However, when running the hot spots tool, it seems my input can no longer use aggregated data (either by fishnets or uniform lengths of highway segments).  The process returns an error message because the unique IDs of the points used to create the .swm file are lost when they are aggregated.

I have tried using the original DVC point file as the input for the hotspots, but this doesn???t appear to be generating the correct results and the hot spot tool only assigns z-values to the points themselves. We like the visual output generated by the hot spots tool when aggregated data is used because it provides a more continuous understanding of the highway.  I also tried using DVC data intersected with the fishnet around the highway and used the ???join_count??? field as the input field in the hot spots analysis ??? in essence, each point has a join count value which is the total number of points that it shares a fishnet cell with.  But again, this only provides a z-value for the individual DVC point, and not the smoothed results across the entire length of the highway.

Do you have any advice for how I may approach this process?  Thanks for your help.

Outcomes