Comparing Machine Learning Using UAVs to Ground Survey Methods to Quantify Milkweed Stem Density and Habitat...

Area of Science:

• Ecology
• Conservation Biology
• Remote Sensing

Background:

• Monarch butterfly populations are declining across North America.
• Conservation efforts are underway, particularly in energy and transportation sectors.
• Effective vegetation management requires accurate monitoring of milkweed (Asclepias spp.), the monarch's larval host.

Purpose of the Study:

• To compare the accuracy and scalability of five different methods for quantifying milkweed stem density and habitat characteristics.
• To identify the most efficient and economical approach for land managers.

Main Methods:

• Five methods were evaluated: Site assessment, Transect plot, Square plot, Large transect (Monarch CCAA methodology), and Machine learning (ML) using UAV imagery.
• Methods included full coverage ground counts, partial ground counts, and aerial imagery analysis.
• Study sites encompassed various land types, including utility rights-of-way and solar arrays.

Main Results:

• Site assessment and Machine learning methods demonstrated the highest consistency in quantifying milkweed stem density.
• Partial ground count methods (Transect plot, Square plot) frequently led to over or underestimation of milkweed populations.
• Estimates of habitat characteristics (woody, broadleaf, grass, bare ground) varied inconsistently among methods and sites.

Conclusions:

• Site assessment and Machine learning are recommended for accurate milkweed population monitoring.
• Land managers should exercise caution with partial ground count methods due to potential inaccuracies.
• Accurate habitat quantification is crucial for effective monarch butterfly conservation strategies.