Dr. Dave McKay Profile

Dr. Dave McKay

at Univ of Edinburgh

SPIE Involvement:

Author

Publications (1)

Proceedings Article | 12 October 2021 Presentation

Mapping fragmented woodlands in cloudy landscapes using Sentinel satellite data in NW Scotland

David Milodowski, Alex Merrington, Nicholas Ross, Dave McKay, Paula McGregor, Mathew Williams

Proceedings Volume 11888, 118880B (2021) https://doi.org/10.1117/12.2601009

KEYWORDS: Satellites, Associative arrays, Spatial resolution, Remote sensing, Backscatter, Vegetation, Satellite imaging, Radar, Modulation, LIDAR

Read Abstract +

Expansion of forests and woodlands are key elements of global strategies to capture carbon from the atmosphere and therefore mitigate climate change. Fundamental to the successful planning and management of woodland conservation and restoration is the ability to map accurately the spatial extent and character of woodland ecosystems. Satellite remote sensing increasingly provides a powerful tool to facilitate these monitoring efforts at scale. However, woodland cover in Scotland is highly fragmented, with marked differences in structure between the remnant native woodlands and more common commercial plantation systems. In addition, the landscape is topographically complex and frequently shrouded in cloud. These factors pose challenges for satellite remote sensing. Therefore, the capacity to characterise fragmented woodland cover accurately in such landscapes remains uncertain. In this contribution, we assess the extent to which trees can be mapped at 10m spatial resolution using a combination of openly available Sentinel 1 C-band radar backscatter data and Sentinel 2 multi-spectral imagery in fragmented woodland landscapes in NW Scotland. We assess the predictive accuracy of the resultant classifiers using a spatially rigorous site-level cross-validation across six sites of varied woodland cover, and explore the role of canopy characteristics and topography in modulating the accuracy with which trees are detected. We find that the accuracy with which we can detect trees is strongly dependent on the stand structure. Trees are mapped more accurately in dense woodland (≥50% tree cover) than more open woodlands (≥20% tree cover), and especially compared to isolated trees. Accuracy also varied with topography, with highest accuracies in flat terrain and reduced accuracies on steeper slopes. These results demonstrate clear potential for integrating Sentinel satellite monitoring systems within woodland management frameworks, while highlighting the importance of reporting context-dependent accuracy statistics with remotely sensed maps of tree or forest cover.

View contact details

UPDATE YOUR PROFILE

Is this your profile? Update it now.

Sign into your SPIE.org account

Don’t have a profile and want one?

Create an account on SPIE.org

Keywords/Phrases

Search In:

Publication Years