A customized framework for regional classification of conifers using automated feature extraction

Area of Science:

• Ecology
• Remote Sensing
• Geospatial Analysis

Background:

• Pinyon and juniper expansion threatens Great Basin sagebrush ecosystems.
• Accurate, high-resolution maps are needed for effective land management and restoration targeting.
• Existing regional remote sensing data lack the necessary spatial resolution and accuracy.

Purpose of the Study:

• To develop and validate an automated feature extraction framework for mapping pinyon and juniper expansion.
• To improve the accuracy and precision of conifer distribution maps at a regional scale.
• To enable targeted restoration efforts for early-stage woodland expansion.

Main Methods:

• Implemented an object-based image analysis framework using Feature Analyst™.
• Utilized 1-m² reference imagery from the National Agricultural Imagery Program.
• Applied supervised learning for automated feature extraction of conifers across Nevada and northeastern California.

Main Results:

• Generated a binary conifer map with an overall accuracy of 86%.
• Demonstrated the framework's capability for automated feature extraction on large datasets with very high spatial resolution imagery.
• Achieved high accuracy and precision in mapping conifer distribution at a regional extent.

Conclusions:

• The developed framework offers a significant improvement in accuracy and precision over existing methods for regional conifer mapping.
• Automated feature extraction is viable for large-scale, high-resolution ecological mapping.
• This approach supports more effective land management decisions for combating invasive tree species.