Radio waves don't travel in straight lines over terrain. A transmitter on one side of a ridgeline is effectively invisible to a receiver on the other side, regardless of how close they are on a flat map. Today I taught the platform to understand terrain.
I built a terrain profiling module that ingests high-resolution elevation data — one-arc-second SRTM tiles from the USGS, providing a height measurement roughly every 30 meters. For any transmitter-to-receiver path, the module samples the elevation profile at 30-meter intervals, computes Fresnel zone clearance at each point, and calculates signal attenuation from terrain obstruction using the ITU-R P.526 knife-edge diffraction model. The result replaces a simplistic flat-earth propagation model with one that accounts for mountains, ridgelines, and valleys.
A 27-test validation suite confirmed the math: my test corridor between Catoctin Mountain (492 meters elevation) and downtown Frederick (90 meters elevation) showed 153 dB of path loss over 27 kilometers, compared to 105 dB for a flat 1-kilometer path. The difference is dramatic and explains why signal strength alone, without terrain correction, produces wildly inaccurate range estimates in hilly terrain.
I also completed the talkgroup identification system. The platform now recognizes 459 talkgroups across three jurisdictions, automatically resolving each transmission to a human-readable agency name and category — police, fire, EMS, or other. When a pin drops on the map, it tells you not just where the transmission originated, but who was transmitting.
And then it happened. Driving down Route 355 in Urbana, Maryland, the app dropped its first terrain-corrected pin for a county school board transmission. The pin landed within a quarter mile of the correct location, at 91% confidence. The first moment where the physics, the math, and the code all aligned in the real world.