Recently there was a discussion among several AGI engineers concerning the sensitivity of our Urban Propagation (UProp) Extension
results to the quality of building geometry. UProp provides very fast, site-specific path loss predictions for communication links in urban environments and is powered by AGI business partner Remcom’s
Wireless InSite® Real-Time propagation algorithms. To help address these questions, I did a bit of investigating. This is by no means an exhaustive study but I think the preliminary results provide some helpful information.
First, I set up two relatively simple scenarios showing STK/Communications
RIP (Received Isotropic Power) coverage at 2m altitude around some building geometry. The first scenario shows the difference between a highly detailed model and a simplified polygon model. The second scenario evaluates the difference between an original shapefile and one with building heights modified by a random distribution up to 25%.
First Scenario - Detailed vs. Simplified:
You can see from the picture that a significant number of polygons and complexity have been removed from the model. To evaluate the difference in results, I exported the Value By Grid Point report to Excel and started comparing. My FOM (Figure of Merit) is RIP (dBW). Out of a total of 3,400 grid points, the following general results were:
Maximum % Difference: 91%
Average % Difference: 2.7%
So, out of 3,400 grid points about 290 points differ by more than 10% (about 8.5 %).
Second Scenario - Randomly Deviated Building Heights:
This picture may not show as many obvious differences, but I assure you, there are some significant alterations (I created a simple script to modify the building height randomly by up to 25%). Again, the FOM Metric is RIP (dBW) with a total of 26,245 grid points.
Maximum % Difference: 102%
Average % Difference: 0.53%
So, out of 26,245 grid points only about 130 points differ by more than 10% (about .5 %).
It may be a little early to draw any serious conclusions here, but at the onset I think it's pretty obvious that the quality of the building data is not the driving factor for quality in the results. I've shown that the vast majority of grid points tend to agree within 1%, even with significant differences in the shapefile (in height anyway). If anyone is interested in learning more, please feel free to contact me at firstname.lastname@example.org.