Manual land cover classification is subjective and in the natural environment, features change over time, for example scrub becomes trees. OS carries out a flying programme each year, capturing each area of Great Britain at least once every 3 years. The aerial imagery derived from this capture programme can be impacted by seasonal variances, i.e. the time of year when the area was flown over, or even the time of day. Earlier in the year vegetated features may be captured with leaf-off, long shadows may impact the feature’s classification, or during a dry summer an area of marsh may look like rough grassland.
Additionally, the automated machine learning classification to derive the percentage cover value uses aerial imagery as a top-down view so some features on the ground may have their true nature obscured by overhanging trees.
The automated classification of some feature types is of lower confidence and therefore the percentage cover value for these features may be of lower quality, i.e. ‘Saltmarsh’, ‘Heath’, ‘Rough Grassland’.