Case studies

Using multispectral imagery to monitor urban forests


Share | 03/06/2020

Drone flying over Denver, Colorado
UAS flights over urban terrain require precise flight-planning to comply with FAA restrictions, limitations and safety considerations. Ground Observers and effective communications are critical given the height of the houses and trees that often obscure direct line of sight to the UAS.

Urban parks, tree-lined streets, gardens, greenways, and nature preserves all serve as the lungs of metro areas, filtering air and water, helping control stormwater, providing shade, and helping to conserve energy.

With over 80% of North Americans living in urban areas, these ‘lungs’ are resources worth protecting.

However, like other forests and crops, they are at risk from invasive pests and diseases. In North America, one of the main threats is the Emerald ash borer (Agrilus planipennis), a green beetle native to the forests of Asia that attacks ash trees exclusively. Its larvae feed on the inner bark of a tree for one to two years before emerging, affecting the tree’s ability to transport water and nutrients to its canopy. Since neither the symptoms, nor the beetles are visible in the early stages of infestation, it is very difficult to identify and save affected trees.

The Emerald ash borer first appeared in Michigan in 2002, and since it has spread to 35 of the 50 US states. It has the potential to destroy 90–95% of all ash trees in North America, which account for about 25% of trees in urban areas.

The most effective way to treat an infestation and prevent the spread is to detect the symptoms before they are visible, which is possible with drone-based multispectral imagery. However, the use of drones in urban environments presents some unique challenges. The main roadblock being that according to FAA regulations, drones cannot be flown over private property without the consent of the owner.

Despite the challenges involved in flying over an urban area, Two Colorado-based companies, Spectrabotics and Arbor Drone, joined forces and set out to help protect urban forests from this major threat. For their work, the team chose the MicaSense series RedEdge sensor due to its reliability and spectral resolution.

The first challenge was to find a way to fly over the urban environment in Denver, complying with Federal Aviation Administration regulations and at the same time covering all areas planted with ash trees. Thankfully, much of Denver is set out in a grid pattern and Spectrabotics was able to plan their flight path by aligning with the right of way alongside the road. This allowed them to capture imagery of the trees without violating FAA rules. While, Spectrabotics’ workflow was effective, a lot of observers were needed to cover such a large area.

When analyzing the imagery, the team also had difficulties finding an appropriate tool. The commonly used vegetation index NDVI saturates at a high Leaf Area Index (LAI). LAI is, generally, a measure of the leaf surface-area contained within each pixel (each pixel is roughly 8 centimeters square for the image set). This saturation hides many of the subtle data features needed to identify EAB impacts on Ash trees. So the team went back to the lab to develop their own algorithms that focus specifically on matching the right bands to identify specific features of interest related to EAB.

Emerald Ash Borer (EAB) analysis displayed in Spectralink
Emerald Ash Borer (EAB) analysis displayed in Spectralink. Data collected from a neighborhood in Denver, CO.
Mid, early, and late states of EAB infestations.
RGB image of ash trees
(Above) Imagery shows sample Green Ash trees in various stages of early-onset EAB infestation. Typical visual observations for EAB infestations include yellowing and thinning leaves in the upper portions of the canopy — the yellowing indicates low chlorophyll concentrations in the leaves. In the image, the orange colors identify areas of EAB infestation. Healthy trees have a more uniform coloration and consistency, particularly in the Infrared spectrum. (Below) An RGB image of the same three trees shows very little variance in tree health. The use of additional spectral bands, in this case, the five bands from RedEdge, are required to provide insight into infestation levels.
EAB analytics layers
EAB analytics layer combines multiple spectral bands focused on health and chlorophyll content. The EAB analysis allows analysts to quickly assess and quantify the extent of EAB infestations in complex urban terrain.
Flightpath over Denver, Colorado
Denver, CO. Flightpath set up to avoid flying over private property. Drone passes are restricted to the right of ways alongside the road, which limits image overlap to roughly 10% with adjacent images.

The multispectral imagery obtained with the MicaSense series RedEdge, and the subsequent analysis, allowed for the early detection of EAB, which represents a major advancement in the fight against this threat. The team was able to identify infested trees that foresters could then assess on the ground and determine if a trunk injection or other form of chemical application was required, or if it’s necessary to cut off certain branches of the tree.

Dan Staley from Arbor Drone
Dan Staley is Principal of Arbor Drone, a green infrastructure and remote sensing company specializing in woody plants. A licensed remote pilot, he studied Environmental Horticulture and Urban Forestry at the University of California, Davis; and urban planning with an Urban Ecology minor at the University of Washington. Arbor Drone and Spectrabotics joined forces to develop and continue to scale the EAB analysis.

Newsletter sign up

Sign up to receive updates directly to your inbox.