Supporting habitat conservation with automated change detection in Google Earth Engine

Area of Science:

• Conservation Science
• Remote Sensing
• Ecology

Background:

• Effective implementation of biodiversity conservation laws is limited by challenges in monitoring habitat degradation.
• Remote sensing data offer potential for quantifying land-cover changes but are underutilized due to a lack of accessible tools.
• Systematic conservation monitoring requires efficient and adaptable methods to track habitat loss.

Purpose of the Study:

• To adapt and evaluate algorithms for automatically detecting land-cover changes using publicly available satellite data.
• To assess the effectiveness of these algorithms in identifying habitat loss relevant to imperiled species.
• To provide a practical solution for improving the monitoring and enforcement of conservation policies.

Main Methods:

• Adapted two algorithms for automated image differencing to detect land-cover changes.
• Utilized free, publicly available satellite data for analysis.
• Validated algorithm performance against ground-truthed data and manually delineated polygons in four case studies of imperiled species' habitats.

Main Results:

• Both algorithms demonstrated high accuracy in discriminating between changed and unchanged land cover (Area Under the Receiver Operating Characteristic Curve >0.90).
• Algorithms effectively delineated habitat loss with low omission and commission rates.
• Algorithm effectiveness varied across different habitat types, with one algorithm showing superior performance in specific cases.

Conclusions:

• Adapted algorithms provide a scalable and efficient method for monitoring habitat loss in biodiversity conservation.
• These tools can help bridge the implementation gap in conservation by improving monitoring and enforcement capabilities.
• A free online tool (conservationist.io/habitatpatrol) is available to facilitate these analyses.