남극 은암치(Pleuragramma antarctica)의 음향 식별 및 머신러닝 기반 분포 예측

Abstract

This study presents a comprehensive investigation into the acoustic scattering characteristics and distribution predictions of Antarctic silverfish (Pleuragramma antarctica) inhabiting the Ross Sea. As a key forage species in the Antarctic ecosystem, P. antarctica serves as essential prey for top predators such as penguins, seals, and baleen whales. Its unique life cycle, which involves spending its entire lifespan in the water column, renders it ecologically significant. Due to these physiological and ecological traits, P. antarctica is recognized as an indicator species for the conservation of Antarctic marine resources and ecosystem health, necessitating quantitative assessments of its distribution and population dynamics. This study is structured around three major components: (1) analysis of acoustic target strength (TS) using theoretical scattering models, (2) evaluation of spatiotemporal distribution characteristics using multi-frequency acoustic data, and (3) construction of machine learning models utilizing environmental variables as predictors. The primary objective of this research is to estimate the frequency-specific target strength of P. antarctica based on body length and tilt angle using two theoretical models: the Distorted Wave Born Approximation (DWBA) and the Kirchhoff-Ray Mode (KRM) model. Specimens were classified into adults and juveniles based on a threshold length of 80 mm, with different density (g) and sound speed (h) ratios applied accordingly. Acoustic responses at 38, 70, 120, and 200 kHz were simulated, and TS-length regression formulas were derived. The DWBA model considers the body as an elastic continuous cylinder, while the KRM model incorporates structural details to calculate backscatter and diffraction effects with greater precision. The results indicated that juveniles exhibited stronger responses at higher frequencies, while adults were more responsive at lower frequencies. TS variation also differed with tilt angle. Both models demonstrated that TS increases with length, providing essential data for future biomass estimations. Secondly, acoustic survey data collected between 2018 and 2024 across the Ross Sea and polynya regions were analyzed to quantify the horizontal and vertical distribution of P. antarctica. Adults and juveniles exhibited clear depth-based separation: adults were predominantly found at depths of 100–200 m, while juveniles concentrated in the upper 0–50 m layer. Horizontal distribution patterns varied by season and region, with particularly high NASC values repeatedly observed in specific areas such as TNBP and RSP, suggesting that these regions serve as important habitats. Frequency difference (ΔMVBS) values were employed to distinguish silverfish from krill, and model-derived frequency differences closely aligned with field observations, validating the species identification method. Lastly, a machine learning model based on the Random Forest algorithm was developed by integrating environmental variables such as temperature and salinity with acoustic backscatter data (NASC). Input features included both average and depth-specific values of temperature and salinity, with separate models trained for adults and juveniles. The adult model demonstrated strong predictive performance (R² = 0.77, RMSE = 2.06), while the juvenile model exhibited relatively lower accuracy (R² = 0.39). Variable importance analysis revealed that average temperature and salinity at depth had the greatest influence on adult distribution, whereas surface temperature was the most influential factor for juveniles. Extrapolation using the model further indicated that P. antarctica is highly sensitive to specific environmental conditions, with anticipated distribution shifts in response to future oceanographic changes. This study effectively integrates acoustic scattering theory, long-term acoustic survey data, and machine learning-based prediction models to elucidate the distributional dynamics and environmental sensitivity of Antarctic silverfish. The findings provide valuable baseline data for anticipating structural changes in the Antarctic ecosystem under climate change and offer scientific support for the designation of Marine Protected Areas (MPAs). Furthermore, the results underscore the potential for advanced acoustic classification and prediction of weakly scattering species such as P. antarctica, contributing to future polar fisheries management and ecosystem monitoring efforts.