AI tool
Detect land-cover change anywhere in Nepal
Choose a place and two dates. The U-Net maps both on the same 5.46 m grid and shows where the land changed, how much of each class was gained or lost, and the largest transitions, such as farmland to built-up land.
Selected: Kathmandu core (Durbar Square → Thamel) (27.7090, 85.3110)
Reading the result
Compare like with like: two Sentinel-2 cloudless years are the fairest pair, because they share the same 10 m sensor and processing, and each year is colour-normalised before classification. Mixing sub-metre Esri imagery with a 10 m mosaic exaggerates change. National and district trends with confidence intervals are on the land-cover pages.