The surfaces pack includes a vectorised set of layers that describe the nature of the ground surface in an area.

These layers are vectorised in a fairly straightforward manner from the raster AI Layers,  so to find out about typical error cases and precise definitions, check out the AI Layer Glossary pages below. Each feature (polygon) has a confidence attached that corresponds approximately to the median confidence score of the raster within that feature.  Some features are multipolygons, where neighbouring high confidence features were linked by a shared low confidence area.


Offline Delivery

AI Feature API

MapBrowser Visualisation

(All Generations)

MapBrowser Export*

(Gen 1 & 2 Data)

(tick)(tick) Feature class for each included AI Layer (vector polygons).

(tick)  Lawn Grass

(tick)  Water Body

(tick)  Asphalt

(tick)  Concrete Slab

(tick)  Natural (soft)

(info) Not available in Gen 1 / 2 Data

Characteristics and Recommended Use

There are a wide variety of potential uses for the Surfaces AI Pack:

  • Using the Water Body features to assess flood risk and proximity to water, particularly in conjunction with our DSM and DEM/DTM data.
  • Combining layers (along with AI Pack: Vegetation and AI Pack: Building Footprints) to determine pervious and impervious areas. Calculating a union of pervious layers (such as Lawn Grass, Natural (Soft), and the vegetation layers), and impervious layers (Concrete, Asphalt, and Building) can be used to model stormwater run-off and assess flood risk.
  • Using the  Asphalt layer as a proxy for a road network, to identify roads, changes in roads, and car parks.
  • Using the Concrete Slab layer as a proxy for sidewalks in an area (alongside asphalt roads).

The distinction between asphalt and concrete can become highly challenging, particularly when lighter coloured asphalt is detected, and in the presence of shadow from tall buildings. For some applications, it may be best to consider combining Asphalt and Concrete.

API Notes for Large Objects

The AI Feature API  truncates feature classes in the surfaces pack to the intersection of the feature with the query AOI. This is because a surface feature can be almost unbounded in size, such as a contiguous road network made of asphalt, which clearly cannot be returned in a sensible way in a single API payload.