Evaluating Sentinel-1 Capability in Classifying Dieback in Chestnut and Oak Forests

This letter analyzes the contribution of the Sentinel-1 (S1) satellites, which provide C-band synthetic aperture radar (SAR) data, to the monitoring of forest dieback. Multispectral satellites (typically Landsat 8 or Sentinel-2, S2) have been found to be effective in detecting early signs of dieback, while little work has been done with SAR data despite its sensitivity to canopy structure and water content and its ability to pass through clouds. Our analysis is conducted on two study sites in France, where the dieback of chestnut and oak plots have been labeled. Classifications have been conducted to measure the ability of S1 data to identify plot in dieback. Our results show that S1 time series are not very sensitive to forest dieback. Using a single S2 image leads to more powerful classification results than using 1 year of S1 data. While S1 data may not be suitable for stand-alone forest dieback detection, it could be interesting to use it for other forest monitoring applications that should not be affected by dieback (species identification and clear-cut detection).