Vegetation height from L-band SAR backscatter and interferometric temporal coherence measurements

Abstract

This paper presents an algorithm for mapping tree height from L-band spaceborne synthetic aperture radar measurements. The proposed tree height retrieval algorithm integrates dual-polarimetric (HH, HV) backscatter and interferometric coherence time series using a model-based approach. Physics-based models relate radar backscatter and interferometric temporal coherence to forest properties and capture the varying environmental conditions over time. These models are calibrated using lidar measurements, and then jointly inverted to retrieve vegetation height at the pixel level. The algorithm is evaluated using a time series of 6 dual-polarimetric ALOS-2 acquisitions over the Lenoir Landing (AL) site, supported by high-resolution lidar data, weather measurements, and the land cover map from the National Land Cover Database. Over the study area, temporal coherence alone proves reliable for estimating forest height, with varying performance depending on vegetation height and class. Comparison with lidar-derived tree height confirms that integrating multi-temporal interferometric coherence and backscatter significantly improves the estimation performance compared to using backscatter time series alone. In particular, incorporating interferometric coherence extends the dynamic range of the estimated tree heights, overcoming the saturation of backscatter signals for tall vegetation. The algorithm is applicable to repeated radar acquisitions collected within a narrow orbital tube and represents an opportunity to map forest properties with upcoming and future low-frequency radar missions.