Reach-back vs. edge processing

My article Edge ISR: Revolutionizing Intelligence in Contested Environments (Over the Horizon Journal, 2025) argues that when the electromagnetic spectrum is contested, sending raw sensor data back to rear-area analysts (reach-back) stops working, and more of the processing has to happen at the edge, close to the sensor. This toy model shows why. All numbers are notional.

Megabits per collection (for example a few minutes of full-motion video).
Megabits per second back to the rear.
Share of the link you cannot use.
Seconds for on-board or forward processing to turn raw data into detections.
Reach-back
send everything, analyse in the rear
Edge
analyse forward, send detections
Faster

How the numbers are worked out

Reach-back time = raw data ÷ usable link + 30 s of rear analysis. Edge time = edge processing + detections (1% of the raw data) ÷ usable link. Usable link = capacity × (1 − share lost). Every value is an illustrative assumption, not a measurement of any real system.

The article's point is in the shape of the curve: as the link degrades, reach-back time grows without limit, while edge processing barely changes, because it no longer depends on moving the raw data.