Remote Sensing Rhododendron

Using hyperspectral imagery to identify rhododendron.

Dr Gary Llewellyn

Remote Sensing Consultant at 2Excel Geo

 

Gary has a PhD and first class degree (Hons) from the University of Southampton. He has over 20 years of experience in remote sensing and over 12 years in airborne remote sensing.

His expertise is with spatial and spectral aspects of remote sensing terrestrial vegetation, experience in calibration and validation of airborne and satellite data products and has supporting airborne research projects in a wide range of applications and in challenging environments across the world.

 

Rhododendron (Rhododendron ponticum) has been identified as an invasive non-native species in the UK and a potential carrier of plant pathogens. Therefore, in the UK, there is a management focus to identify and map the presence of Rhododendron to facilitate its removal.

This presents a complex series of challenges for remote sensing because Rhododendron is an understorey species and is often obscured by an overstorey. Rhododendron is also very similar other evergreen understorey species such as Cherry Laurel and Holly. The size of Rhododendron cover can also vary greatly from less than a metre across to covering several hectares of woodland.

The use of remote sensing to identify and map rhododendron was investigated as part of a series of projects funded by the Woodland Trust and South Downs National Parks Authority. From this work:

  • Satellite systems have been used to identify the presence of ‘potential rhododendron’, i.e., evergreen understorey vegetation present under a tree canopy that was denuded of leaves but could not discriminate rhododendron from other in-leaf understorey species.
  • Fine spatial resolution hyperspectral data (collected 2Excel) has been used to discriminate rhododendron from other understorey vegetation and hence can specifically identify and map rhododendron coverage.
Figure 1

High-resolution airborne photography. Hover over the image for understorey classification including Rhododendron.

Figure 1

Top: High-resolution airborne photography. Bottom: Understorey classificiation including Rhododendron.

The collection of data in the winter season was necessary to prevent the deciduous overstorey from obscuring the rhododendron understorey. However, this presented two problems: shadow and cloud. Shadow was presented as long shadows cast from the tree stems by the low sun angle and required careful analysis for removal. Obscurement by cloud limited the availability of cloud-free satellite imagery and viable conditions for aerial surveys. Airborne surveys provided a finer spatial resolution than satellite options and allowed greater flexibility in data collection. However, airborne surveys collect smaller areas compared with satellite alternatives. Consideration of both systems provides a dual-scaled solution with satellites proving a course overview of ‘potential rhododendron’ coverage over a region and airborne hyperspectral data allowing a detailed specific assessment over a local area.

In 2021, work was conducted in conjunction with the Woodland Trust and South Downs National Park Authority. This followed on from previous investigations and allowed two sites within the South Downs National Park to be investigated for the presence of rhododendrons. Rhododendrons were successfully mapped using airborne hyperspectral remote sensing and validated by field survey. Classification of the evergreen understory species including Rhododendron, Holly, and Cherry Laurel was performed using field surveys outside of the classification area to test the universality of the technique.

The results demonstrated a >85% accuracy for the specific identification of Rhododendron from Cherry Laurel and Holly. Confusions in identifying rhododendrons were primarily from mixed and obscured signals within the immediate areas proximal to rhododendrons, but these caused a slight underestimate of rhododendron coverage rather than falsely identifying its presence where it was absent.

The algorithms are currently being developed further to account for additional evergreen understorey species (e.g., Shallon), differences in geographic location, land history and land management. The aim is to ensure that these techniques and methodologies can be applied across Great Britain.

Want to find out how Geo can unlock
added potential in your business?

Hall Farm 2
Sywell Aerodrome, Sywell
Northamptonshire, NN6 0BN

Find Us

© 2Excel Geo 2024. All rights reserved.