Bark Beetle Disturbance Model for Northwest British Columbia Using Landsat Imagery

A 2024 study by Nathan Mital aimed to model historic bark beetle disturbances in northwest British Columbia using spectral signatures. The process used Landsat imagery and Aerial Overview Survey data to train a random forest classifier with 76% accuracy on test data. The model's transfer to the target region was deemed unrealistic, flagging 31% of forestland as disturbed, indicating a need for site-specific data.

Use Cases

• Training a site-specific spectral signature classifier for bark beetle detection based on vegetation indices.
• Assessing the transferability of remote sensing models between different forest regions.
• Analyzing the spatial patterns of potential historic forest disturbances in data-scarce areas.

Strengths

• Model achieved a reported 76% accuracy on the testing data used for spectral signature extraction.
• Methodology is clearly described, involving Landsat imagery and established Aerial Overview Survey datasets.
• Study identifies a concrete limitation (31% unrealistic flagging) which provides a clear direction for future work.

Limitations

• The model failed to transfer effectively, with results deemed unrealistic for the target region.
• Column-level documentation is absent; field semantics must be inferred after download.
• Row count and specific data formats are unknown, which may limit suitability assessment.

Provenance

Source

Borealis Harvested Dataverse, author Nathan Mital.

Collection Method

Model derived from 2024 Landsat imagery and Aerial Overview Survey (AOS) datasets using a random forest classifier.

Time Range

Primary analysis uses 2024 imagery; historic modeling was attempted but not completed.

Freshness

Last updated 2026-05-02 04:11:39; freshness should be verified.

Geography

Northwest British Columbia, Canada, with reference data from forests near this area.