.png){kind=small}
Scale Effects in Remote Sensing of Mangrove Forests
Ms ZHANG Hanwen
( Supervisor: Prof Hongsheng Zhang )
Abstract:
Mangroves are essential ecosystems in tropical and subtropical coastal regions, covering only 0.7% of the tropical forest area but providing crucial ecological and economic benefits. They are also among the most carbon-dense forests, significantly contributing to carbon export to oceans and sediment deposition. Remote sensing has been instrumental in large-scale mangrove monitoring, supporting efforts in distribution mapping, species classification, and functional parameter inversion. However, the influence of scale on mangrove remote sensing remains underexplored, which limits the use of fine-scale data. Spatial scale affects the accuracy of species identification and ecological parameter estimation, yet current studies on this issue are insufficient. This research gap hinders effective monitoring and management. To address this, further exploration of scale effects in mangrove monitoring is necessary, focusing on identifying optimal observation scales and developing methods to manage scale-related challenges. By refining these approaches, we can enhance classification accuracy, reduce data redundancy, and improve the efficiency of mangrove ecosystem management and conservation.
In this study, we will propose novel spatial aggregation methods for scale transition and conversion, combined with high-resolution remote sensing data, and deep learning classification models, to investigate scale effects in mangrove monitoring. The objectives of this study are: (1) to define and quantify the scale effects influencing mangrove distribution, species composition, and biomass estimation; (2) to analyze how these scale effects impact our understanding of mangrove ecological functions, processes, and conservation strategies; and (3) to determine optimal observational and operational scales for effective mangrove monitoring and sustainable management.
Preliminary results from published study demonstrated that (1) Using 0.2 m ultra-high-resolution (UHR) aerial photos and the DeepLabV3+ model, the study achieved an overall mapping accuracy of 92.1%, with up to 53% improvement over existing datasets, effectively delineating complex boundaries and fragmented mangrove patches; (2) the proposed scale effect quantification method revealed that transitioning from 0.2 m to 30 m resolutions could result in an average area underestimation of 5000 m² and up to 25% accuracy degradation; (3) scale sensitivity analysis identified 6 m as the optimal resolution for monitoring fragmented habitats, beyond which accuracy declines significantly, dropping below 82% at 10 m and to 66% at 30 m. This research and findings aim to improve methodologies for scale-aware scientific mangrove ecosystem monitoring, and operational mangrove conservation and management.
Comments