Multi-sensor satellite data fusion and machine learning for Eucalyptus mapping in Meket district, Ethiopia

Area of Science:

• Remote Sensing
• Geospatial Analysis
• Ecology

Background:

• Eucalyptus expansion in Ethiopia's highlands supports the economy and zero-carbon goals but raises ecological concerns.
• Accurate Eucalyptus mapping is vital for management but faces challenges like cloud cover, spectral confusion, and costly commercial imagery.
• Integrated satellite data offers a potential solution for reliable and cost-effective mapping.

Purpose of the Study:

• To develop a reliable and cost-effective method for mapping Eucalyptus trees in Ethiopia's Meket district.
• To assess the effectiveness of fusing Sentinel-1 SAR and Sentinel-2 MSI data for Eucalyptus detection.
• To evaluate the performance of Random Forest, SVM, and CART classifiers for this mapping task.

Main Methods:

• Utilized integrated multi-sensor satellite data: Sentinel-1 Synthetic Aperture Radar (SAR) and Sentinel-2 Multi-Spectral Imagery (MSI).
• Fused eighteen features from spectral bands, radar backscatter, and vegetation indices at the feature level.
• Classified the fused data using Random Forest (RF), Support Vector Machine (SVM), and Classification and Regression Trees (CART).

Main Results:

• Random Forest (RF) achieved the highest performance, with 90% Overall Accuracy (OA) and a kappa coefficient of 0.81.
• Support Vector Machine (SVM) showed comparable performance, with only a 1% difference from RF.
• The study successfully mapped Eucalyptus trees using freely available satellite data.

Conclusions:

• Publicly available Sentinel-1/2 fusion data, combined with appropriate classifiers like RF, provides cost-effective, reliable, and accurate Eucalyptus mapping.
• This approach supports Ethiopia's zero-carbon strategy and aids in sustainable management of Eucalyptus plantations.
• Geospatial technology-based land use planning is crucial for policymakers and planners monitoring Eucalyptus expansion.