Google Street View

as a source of retrospective ground truth.

by Dr Chloe Barnes
07/06/2018

During the tree species campaign in August 2016, high resolution true-colour and hyperspectral imagery was acquired for a mixed deciduous woodland in Suffolk, South East England. Tree species surveys were conducted in the field in conjunction with the airborne collect in order to fulfil the objective of the campaign.

The presence of ash dieback ‘Chalara’ (Hymenoscyphus fraxineus) was noted at the site by forest managers at the time of collection but was not recorded in any detail during ground sampling. Google Street View imagery for the roads surrounding the forested area was acquired in September 2016, one month following the airborne acquisition.

Individual trees situated in the hedgerow were assessed from various roadside angles to make an assessment of dieback. In most instances where foliage was still present, ash (Fraxinus spp.) was visually discriminated from other species along the roadside using leaf shape.

Ash tree with evidence of dieback (Google, 2016)
Ash tree with evidence of dieback (Google, 2016)

To capture the visual observations from Google Street View, the high resolution (6 cm) true-colour imagery acquired by was used to manually generated polygons corresponding to the observed tree crown. The virtual survey results were added to each of these tree crown polygons as attributes. A dataset including 29 individual ash trees with canopy dieback percentages from 0 to 90% was generated for the study area. These polygons were subsequently utilised to extract spectral reflectance and narrow-band vegetation indices [add vegetation indices link] from the hyperspectral imagery for each of the individual ash trees surveyed.

Ash tree crowns identified on the true-colour imagery
Ash tree crowns identified on the true-colour imagery

The analysis of the hyperspectral imagery for these ash trees informed the research and capability development conducted by in relation to the applications of hyperspectral remote sensing for tree disease assessment.