A geospatial skeleton framework for unveiling the 3D structure and dynamics of marine heatwaves from earth observation data

Understanding the multi-dimensional structure and evolution of geospatial phenomena is a fundamental challenge in Earth system science. Marine heatwaves (MHWs), as critical 4D spatio-temporal events in the ocean, exhibit complex vertical heterogeneity and dynamic interactions across latitude, longitude, depth, and time. However, traditional geospatial methods often oversimplify such phenomena into 2D profiles or homogeneous 3D blobs, limiting their ability to capture intrinsic 3D structure and evolutionary patterns. To fill this gap, we propose a novel data-driven framework for constructing 3D skeleton representations of MHWs, enabling fine-grained analysis of their internal structure and spatio-temporal evolution. This framework represents MHW skeletons as graph structure, where nodes denote localized intensity centers and edges quantify morphological correlations of spatial aggregation between subsurface layers. The MHW skeletons are constructed using graph algorithms applied to each snapshot and tracked over time through five evolutionary operators. Applied to the 2014–2016 Northeast Pacific (NEP) and Tropical Pacific in the epipelagic zone (0–200 m), the proposed method reveals distinct vertical structures: ENSO-driven MHW skeletons exhibited greater vertical complexity than coastal MHWs, featuring steep longitudinal slopes (up to 120°). Additionally, two foundational MHW skeleton structures, I-shape and Y-shape, were identified. Notably, dynamic merging and splitting process between subsurface MHWs were captured, highlighting the framework’s capability to uncover hidden interactions in ocean systems. Beyond MHWs, this methodology provides a transferable paradigm for analyzing other 3D geospatial phenomena, enhancing our understanding of their structural complexity and dynamic evolution.