Advancing non-invasive tree root detection by creating a training data set of GPR tree root signatures

2019, Andrew Millward, PhD, Co-PIs Mr. Justin Miron, Mr. Nikesh Bhagat, Ryerson University

To ensure the protection of existing trees, especially in expanding and densifying urban centers, urban forestry practitioners must have accurate, rapid, cost-effective and ideally—non-invasive—tools to identify and map tree roots under the ground. Ground Penetrating Radar (GPR), while not a new tool for this purpose, remains the most promising of approaches to realize these practitioner goals, especially for protection of heritage and culturally significant trees. Our current research focusses on improving the accuracy of identifying tree roots in GPR data, early results of which will be published in the forthcoming Morton Arboretum’s Landscape Below Ground IV conference proceedings. Further advancement of GPR for tree root detection and mapping requires improvement to software, which is possible by creating a library of root scenarios that can be continually enhanced using machine learning. With Tree Fund support, we propose to develop a predictive model that successfully links material properties of tree roots with a corresponding library of GPR signal properties; this will augment GPR accuracy for root detection across a variety of soil conditions.

Year: 2019

Funding Duration: 2 years

Grant Program: Jack Kimmel International Grant

Grant Title: Advancing non-invasive tree root detection by creating a training data set of GPR tree root signatures

Researcher: Dr. Andrew Milward

Key words:

Peer Reviewed Publications from Grant:

General Audience/Trade Publications:

Professional Presentations:

For more information on this project, contact the researcher via TREE Fund at treefund@treefund.org.