Verifying My Assumptions about Emerging Hot Spot Analysis Behavior

332
0
01-04-2022 09:46 AM
GrantShirts
New Contributor

I've been invited to present an application of the emerging hot spot analysis tool at AASHTO's 2022 GIS for Transportation Symposium. The inputs into the tool were non-overlapping point features derived from a directional roadway network in turn used to create space-time cubes from defined locations. While northbound traffic is spatially near the southbound traffic (e.g. highway) they are not related in terms of congestion. Since I am using a normalized travel time reliability metric, I felt it wouldn't be necessary to create a large complex network from the lines and generate network weights. Instead, I selected a fixed distance of 1 foot with the number of spatial neighbors set at 0. To preserve statistical integrity, I set the neighborhood time step (i.e. weeks) to 8 (recommended number of neighbors for Getis-Ord Gi*) which would be the timestep in question plus the previous 8 weeks for a total of 9 weeks. My intention for setting these parameters was to only include temporal neighbors and avoid any spatial neighbors. Does the community here agree that this would accomplish my intention? Would this be an appropriate application of the emerging hot spot analysis tool? If not, why?

For my remaining question, assume my space-time cube has 52 timesteps. As you step back in time week by week and include the previous 8 temporal neighbors, what happens as you approach the beginning of the time series where there would be less than 8 previous temporal neighbors to include? When that happens, does the emerging hot spot analysis include fewer than 8 temporal neighbors until selecting 0 temporal neighbors for the first chronological time step? If I'm correct about how the tool handles this situation, it would seem that I would need to include 60 timesteps (52 + 😎 in my space-time cube to get more or less a year's worth of proper analysis. In the case of having 60 timesteps, wouldn't it make sense to throw out the first 8 chronological timestep results as less valid due to those having less than 8 neighbors?

Thank You!

ArcGIS Pro 2.8.1

0 Kudos
0 Replies