​Multi-sensor change detection for within-year capture and labelling of forest disturbance

Picture
Figure 1. False Colour Infrared from 2017 of the study area encompassing the Elephant Hill fire and forest harvest in British Columbia, Canada. The highlighted areas indicate the four disturbance outcomes relevant for change capture. 1) Winter Harvest 2) Summer Harvest 3) Fire (early season) and 4) Fire (late season).
Picture
Figure 8. Early Fire region. A) False Colour Infrared image in 2016 B) False Colour Infrared image late in 2017 C) Change attributed to type D) Disturbed Forest vs. Undisturbed.
Picture
Figure 9. Late Fire region. A) False Colour Infrared image in 2016 B) False Colour Infrared image late in 2017 C) Change attributed to type D) Disturbed Forest vs. Undisturbed.
Background

Forests are constantly shaped by disturbances such as wildfire and timber harvest, and knowing when and how these changes occur is essential for effective management, reporting, and scientific understanding. Satellite imagery has long provided a valuable historical record of forest change, but many existing approaches summarize conditions only once per year. As a result, disturbances that happen late in the season or near the end of a time series can be missed or delayed in official records. With the growing availability of frequent satellite observations, there is a clear need for methods that can detect forest change more quickly while remaining reliable over large areas.

Study area
The study area is located near Kamloops, British Columbia, Canada (Fig. 1). It encompasses diverse land-use and land-cover types, including forest, grassland, regrowing forest,  human settlement, rocky outcroppings, agriculture, and wetlands. While most of the area is undisturbed in any given year, stand-replacing disturbances can occur in typical years.

Approach

The lab developed a rapid forest change detection framework that combines information from multiple satellite data streams. We used imagery from two widely available satellite systems to observe forests repeatedly throughout the growing season, rather than relying on a single annual snapshot. Our approach first identifies signs of major forest disturbance by tracking consistent drops in vegetation condition over time. We then combine repeated provisional assessments using a probabilistic method that weighs new evidence as it arrives. This allows us to continuously update maps of forest change and produce a reliable end-of-season summary that reflects both timing and disturbance type.

Key Findings

​Our results demonstrate that rapid, multi-sensor forest monitoring is both feasible and accurate:
  • High overall accuracy: We mapped forest disturbance with an overall accuracy of approximately 95%, validated using independent reference data.
  • Reliable identification of disturbance type and timing: We successfully distinguished wildfire from forest harvest and further identified whether harvesting occurred in winter or during the growing season.
  • Reduced detection delay: By combining multiple satellite sources, we captured disturbances occurring late in the season that would normally be missed or delayed by annual mapping approaches.
  • Meaningful integration with long-term monitoring: Our rapid-response results complemented existing long-term trend analyses, improving estimates for the most recent years of satellite records.
​
Impact
​
This work shows how near–real time satellite data can be transformed into timely, management-relevant information. By reducing delays in detecting forest change, the lab’s approach supports more responsive forest management, improved disturbance reporting, and more accurate environmental assessments. The framework is lightweight, transparent, and adaptable, making it well suited for operational monitoring programs and future expansion as new satellite data sources become available.

Resources

Published Paper : Cardille, J. A., Perez, E., Crowley, M. A., Wulder, M. A., White, J. C., & Hermosilla, T. (2022). Multi-sensor change detection for within-year capture and labelling of forest disturbance. Remote Sensing of Environment, 268, 112741. DOI: https://doi.org/10.1016/j.rse.2021.112741