Mr Yi-Fei Gu1
1The Swire Institute of Marine Science and School of Biological Sciences, The University of Hong Kong, , Hong Kong, 2Marine Science Center, Northeastern University, , USA
Resource depletion and predation pressure are traditional causes of patch-formation observed in intertidal mussel beds, especially for temperate species. Such processes lead to aggregated patches in the spatial structure of mussel populations that facilitate population survival. On tropical shores, however, thermal stress can also cause mass mortality events altering the spatial structures of mussel beds, and such events have recently been recorded on temperate shores. Under this scenario, mussels are killed and their shells washed away, creating free space onshore and turning previously intact mussel beds into patches, creating fragmented landscapes. To efficiently and accurately assess such mussel patch-formation dynamics I have been applying a novel image sampling framework using computer vision. Structure-from-Motion algorithms are used to create orthophotos of mussel beds from drone images and semantic segmentation trained using Convolutional Neural Networks to help recognize and quantify mussel patches from field images. This image sampling framework has been tested on several exposed rocky shores in Hong Kong, where mussels experience thermal-induced mortalities annually, and will be transferred to assess mussel beds in Boston, MA. The framework will be used to illustrate the temporal process of thermal stress acting on mussel beds by quantifying their patch-formation dynamics on both tropical and temperate shores under changing climates and the possible implications of the loss of these ecosystem engineers.
Presentation Slides – Yi-Fei Gu
