Marine snow morphology illuminates the evolution of phytoplankton blooms and determines their subsequent vertical export

Marine snow, composed of detritus, organic matter, and inorganic materials, is a key component of the biological carbon pump. Its morphology (size, shape, porosity) and composition significantly influence aggregation, disaggregation, export rate, and biological interactions. Previous studies often simplified marine snow characteristics, hindering a comprehensive understanding of its role in the carbon cycle. The increasing availability of in situ imaging technologies, such as the Underwater Vision Profiler (UVP), allows for detailed morphological analysis of marine snow, necessitating objective classification methods. This research proposes a new method to categorize marine snow into ecologically relevant morphotypes using UVP data from the Arctic marginal ice zones (Baffin Bay and Fram Strait). The study hypothesizes that the method can reveal how marine snow concentrations and vertical distributions change during phytoplankton bloom phases and how this links to carbon export. This will be accomplished by comparing morphotype-resolved patterns to classical analyses of total concentration, biovolume, and size distribution. The potential of this classification method applied to underwater imaging techniques for understanding ocean carbon cycling is discussed.

Existing literature highlights the importance of marine snow in the biological carbon pump, emphasizing the impact of its size and composition on carbon export. However, studies often oversimplify the complex morphology of marine snow, assuming simple geometries like spheres or defined fractal dimensions. This simplification stems from the challenges in studying marine snow's fragile nature and complex characteristics. While optical methods have improved the ability to detect and enumerate marine snow in situ, the establishment of standardized morphological categories remains a challenge. Although optical systems provide vast image datasets, the discrimination of ecologically relevant marine snow groups remains poor. Therefore, an objective and universal classification method is urgently needed to advance our understanding of the biological carbon pump and processes like microplastic deposition.

The study employed data from two Arctic regions: Baffin Bay and Fram Strait. In Baffin Bay, data were collected along transects following sea-ice retreat using the CCGS Amundsen. Fram Strait data were collected from RV Polarstern. The Underwater Vision Profiler (UVP 5hd) collected images, providing morphological properties for over a million marine snow particles. These properties included size (area, perimeter), shade intensity (mean/median gray level), shape (symmetry, elongation), and structure (homogeneity/heterogeneity). Data processing involved removing outliers and applying logarithmic transformations to reduce skewness. Principal Component Analysis (PCA) reduced the dimensionality of the data, creating a morphospace. K-means clustering was then used to categorize marine snow particles into five morphotypes: dark, elongated, flake, fluffy, and agglomerated. The number of clusters was determined by considering the interpretability and ecological relevance, aiming for a balance between detail and clarity. The robustness of the classification was validated by analyzing different datasets independently (Baffin Bay, Fram Strait, and both combined) and comparing cluster memberships using Inter-Rater Reliability Analyses (Cohen's Kappa). The origin of each morphotype was inferred using existing knowledge of bloom dynamics and concurrent data on oceanographic variables (ice retreat, phytoplankton composition via HPLC and microscopy). The study also used classical metrics like total concentration, total biovolume, size spectrum slopes, and vertical attenuation to compare with the morphotype-based analysis. Sinking rates were roughly estimated for some morphotypes by tracking the maximum concentration peak through time. Finally, a Shannon-Wiener diversity index measured morphological diversity of marine snow during the different ice phases.

The PCA analysis revealed a clear separation of marine snow into five distinct morphotypes based on their morphological properties. The study showed distinct spatiotemporal distribution patterns for each morphotype throughout the phytoplankton blooms. In the Baffin Bay, the dynamics were analyzed using Open Water Days (OWD) to align stations relative to ice conditions. Dark compact marine snow showed the lowest vertical attenuation and high concentrations in deeper waters, potentially representing older, fragmented, or densified objects. Elongated marine snow, often associated with diatoms, was concentrated in the upper layers. Flake and fluffy marine snow were abundant in ice-free waters, linked to the bloom of Phaeocystis. Large agglomerated forms appeared later, mostly in subsurface waters. The Shannon-Wiener diversity index highlighted increased diversity in the upper layers before ice break-up and later in deeper waters. In the Fram Strait, the patterns were similar but with distinct variations in the dominance of morphotypes at different bloom phases. The sinking rates for the dark and elongated morphotypes were estimated, and the dark morphotype showed a significantly faster sinking speed, highlighting the importance of factors beyond size in determining particle settling. The vertical attenuation differed between morphotypes and the size-based approach, again emphasizing the value of this new classification. The Arctic bloom succession (diatom to flagellate dominance) correlated with changes in marine snow morphology and composition. This allowed for an extended conceptual scheme of Arctic phytoplankton bloom phenology including marine snow developmental phases, which shows that the first, ice-covered phase was characterized by low concentrations of mostly dark, compact morphotypes in Baffin Bay and elongated forms in Fram Strait. Ice break-up resulted in increased concentration and diversity, while the late bloom phase saw abundant large agglomerated forms.

The study's findings demonstrate that classifying marine snow into distinct morphotypes provides a more nuanced understanding of phytoplankton bloom dynamics and carbon export than size-based approaches. The differing sinking speeds and vertical attenuations of different morphotypes highlight the importance of considering both size and morphological properties to estimate carbon flux. The changes in marine snow composition during different bloom phases reveal how plankton communities and their physiological processes affect the structure and dynamics of the biological carbon pump. The proposed classification method offers a significant advancement in our capacity to study marine snow and is highly valuable given the increasing quantity of in situ imaging data. The consistency of results between the Baffin Bay and Fram Strait datasets supports the method's broader applicability. The observed differences in marine snow composition may also be influenced by distinct water mass regimes. Future research could extend the morphotype catalog, incorporating data from other ecosystems and seasons. Developing algorithms to calculate sinking speed based on image-derived properties (porosity, density) would further enhance flux estimations.

This research introduces a novel, objective classification method for marine snow into ecologically meaningful morphotypes using in situ imaging data. The method effectively captures the complex dynamics of marine snow during different phases of Arctic phytoplankton blooms, improving our understanding of carbon export and its relationship to plankton community structure and processes. The approach is robust, reproducible, and shows potential for wider application. Future work should focus on expanding the morphotype catalog, developing improved algorithms for flux estimation that incorporate morphological information, and applying this methodology across various ecosystems and seasons. This method represents a significant advancement for understanding the biological carbon pump.

The study's focus on Arctic phytoplankton blooms at the ice edge may limit the generalizability of the classification system to other ecosystems or seasons. While the robustness of the five-category classification was tested, exploring the impact of varying the number of clusters could further refine the approach. The inability to directly determine the composition of individual marine snow particles limits detailed interpretations of the origin of certain morphotypes. Also, estimations of sinking speeds were approximate, based on peak deepening. Finally, the time-for-space approach, which substitutes time series with spatial gradients, could introduce uncertainty in the interpretation, especially in regions with complex hydrography.