Get Full Government Meeting Transcripts, Videos, & Alerts Forever!
Get email alerts on the Tree Inventory topic
No spam. Unsubscribe anytime.
Commission pilots PlantNet for tree inventory, flags accuracy and training needs for volunteers
Summary
A Tree Advisory Board pilot using PlantNet and EpiCollect5 identified roughly 65% species‑level agreement with the town GIS tree map; commissioners urged further testing, checkpoints with Parks & Rec, and development of a volunteer training/process before scaling.
Jason, representing the Tree Advisory Board, told the Conservation Commission on July 20 that a pilot to test mobile apps for a town tree inventory produced mixed results.
Jason said he tested PlantNet and EpiCollect5 on multiple parcels and exported results to compare against the town’s GIS tree map. PlantNet correctly matched the town’s identifications about 65% of the time in his sample, while EpiCollect5 proved difficult to use. “PlantNet... the percentage that I calculated out of those two was 65% accurate compared to the GIS interactive tree map of Groton,” Jason said, and noted different angles sometimes produced different top identification results.
Commissioners cautioned that 65% species‑level accuracy is not adequate for policy decisions where invasive versus native distinctions matter. One member noted that misidentifying Norway maple (invasive) as sugar maple (native) would materially affect stewardship priorities. Members urged more repeated testing of the same trees, further pilot parcels, and a checkpoint with Parks & Rec before relying on volunteer-collected app data. They also recommended developing a clear volunteer-training protocol and a plan to combine consultant work, volunteer input and app-based methods as appropriate.
Next steps: Jason will run additional sample parcels, compare repeated PlantNet identifications on the same trees, and then checkpoint findings with Parks & Rec to determine whether to scale volunteer collection or pursue consultant-driven inventory work.
