Remote Sensing
Change detection
The family of methods that compare imagery across dates to identify where land cover or condition has altered.
Change detection can be as simple as differencing two indices, or as involved as fitting seasonal models and flagging departures. The choice governs how much real change is found and how much noise is reported as change.
Design matters more than algorithm. Matching season, sensor and processing across dates prevents apparent change that is really a difference in observation conditions.
Frequently asked questions
What causes false change?
Cloud, haze, sun angle, sensor differences, seasonality and geometric misregistration.
Is change detection enough to quantify emissions?
No. Area change must be combined with carbon stock factors for the vegetation involved.