Measuring vegetation loss from archives of drone orthophotos: permanent-loss rule and QGIS validation

Technical note (not peer reviewed). Low-cost, reproducible method to measure year-by-year vegetation loss from existing archives of RGB drone orthophotos taken with different drones, cameras, dates and georeferencing. Orthophotos are co-registered to a reference flight on fixed structures (SIFT + MAGSAC), colours are harmonised by histogram matching, vegetation is detected with the Excess Green index and a single Otsu threshold, and only permanent loss is counted (vegetation that does not come back in any later flight), so that seasonal browning of grass is not taken as loss. Results are delivered as a QGIS project and validated with stratified random points reviewed by photo-interpretation. In an anonymised case study (ten flights over six years) the method reached 81 % user's accuracy for the loss class, 98 % for the no-loss class and 90 % overall. The chain is in operational use by Vibokair with reliable, validated results. The general maps are synthetic; the only real map is an authorised woodland sector. English and Spanish versions included.