Quebec's homegrown satellite land-cover map holds up well against the global giants, according to a new academic review - it just draws different lines around cities, wetlands, and bare rock.
Researchers systematically compared the COTQ, a 10-meter land-cover product built for Quebec from Sentinel-2 satellite imagery, against three widely used global maps: ESA WorldCover, Esri LandCover, and Google DynamicWorld. The team standardized all four datasets to a common legend, then measured how closely their shapes, object sizes, and spectral signatures lined up across eight Sentinel-2 tiles spanning Quebec's temperate forests, boreal zones, and northern tundra. They backed the statistical comparison with manual photo-interpretation of the areas where the maps disagreed most. The paper is explicitly a benchmarking exercise, not a pitch for a new mapping method.
The headline finding: COTQ's structure and spectral behavior track ESA WorldCover more closely than the other two products, but real disagreements persist around urban areas, wetlands, and rocky or cryptogamic (lichen- and moss-covered) surfaces - exactly the terrain where regional knowledge should beat a one-size-fits-all global model. That is the actual case for provincial mapping efforts: not that they are more accurate everywhere, but that they can be tuned to local classes global products treat as afterthoughts.
Global land-cover maps get reused constantly for climate and planning work, often without much scrutiny of where they quietly diverge. A study that just measures the gap, tile by tile and class by class, is less exciting than a new model but arguably more useful to anyone about to build policy on top of one of these maps.