Examples

Useful Tip:

Every registered user can access a free public example of an AutoMap run via the monitoring system in order to showcase the potential output of the process.

Scales

The most notable difference between satellites is the spatial resolution. A coarse resolution (big pixels) results in less detailed results but is much easier to handle for large areas. Detailed results (small pixels) help to better understand small-scale processes. Another thing to consider is the revisit-time. MODIS revisits locations every day, Landsat approximately every 16 days, Sentinel-2 every 2-3 days, depending on the proximity to the equator.

Hover over the image part you want to expand.
  • Sentinel
  • MODIS
  • Landsat
  • OSM

Classifications

Using just a few manually collected sample points in four basic classes (Urban, Forest, Land, Water), the optimized classifier is capable of producing very accurate results.

Hover over the image part you want to expand.
  • Sentinel
  • MODIS
  • Landsat
  • OSM

A classification can only be as good as its training. We provide a default input dataset based on MODIS Land Cover Classification (MCD12Q1) - Type 1 at 500 m pixel resolution, globally. This dataset differentiates 17 classes, but has many limitations. Applied to higher-scaled satellite results, the idea is to spatially refine the input, which drastically limits the possible accuracy. For this default option, a Cohen's Kappa only around 0.3 can be expected.

Hover over the image part you want to expand.
  • Sentinel
  • MODIS
  • Landsat
  • OSM