ADVANCING WOODY COVER CHARACTERIZATION IN DRYLAND ECOSYSTEMS USING LIDAR AND OPTICAL TIME-SERIES DATASETS

Abstract

Dryland woody cover plays an important role in terrestrial ecosystems while undergoing transformation due to the changing climate and human impact. However, characterizing the extent and dynamics of dryland woody cover is challenging given the great variations in dryland woody vegetation life forms, structure, and spatial distribution. Clear thresholds defining the structural limits of woody cover classes are needed to map and validate woody cover extent, revealing map reliability and usability. The Land Cover Meta Language (LCML), which has succeeded the United Nations Food and Agriculture Organization's (FAO) Land Cover Classification System (LCCS), provides a detailed conceptual and operational framework for applying classifiers in building legends for mapping land cover. The physiognomic-structural traits - height and cover, have been used by a number of global classification schemes to define the extent of woody cover. However, in the existing studies on dryland woody cover mapping, the definitions of “tree”, “forest” or “woody cover” are often vague or lacking in terms of definitive physiognomic structural limits. In addition, inconsistencies frequently exist between the map class definition and the reference data used for guiding the training data collection. To fill the gap, increasingly available remotely sensed data source brings the chance to characterize and extrapolate the woody cover structure parameters suitable for implementation of consistent structural thresholds at large scale. This study is focused on promoting a systematic definition framework to advance dryland woody cover characterization, using multi-source remote sensing data. The implementation was prototyped in Senegal, which features a high variation in height and cover compositions of dryland woody vegetation. First, we reviewed the discrepancies in the definitions adopted by the existing dryland maps and examined the resulting disagreements in the derived map products, using unoccupied aerial vehicle (UAV) lidar data. The result showed that existing continental-scale maps have low accuracies and spatial agreement. Then, we proposed and implemented a definition framework based on height and cover, for consistent woody cover mapping. We applied a regression tree model and combined field-based UAV optical and lidar data, with spaceborne HLS and Sentinel-1 SAR data, to create a set of 30 m wall-to-wall woody vegetation cover maps under different height thresholds. The map validation yields R-square 0.67 to 0.86, demonstrating the capability of woody cover characterization under various physiognomic structural thresholds. Finally, built upon the mapping capability and the framework, we extended the mapping of woody cover over temporal domain to monitor the cover dynamics of the dryland ecosystems. We found that between 2016 to 2023, 1.45 (±0.32) Mha of woody cover were lost in Senegal, and 0.98 (±0.28) Mha was gained, as a result of both natural regenerations/disturbances and human interventions. The area estimations were within decent uncertainty level, proving the capability and validity of the framework in woody cover monitoring. The study emphasizes the need, as well as the capability to relate remote sensing-based observations and derived products with the explicit structure of woody cover. Through this study, we call for applying consistent physiognomic structure-based definition framework, in future woody cover characterization efforts.