Predicting Changes in the Catch of Major Fish Species in the Jurisdictional Waters of the Republic of Korea Using GIS and Machine Learning

Abstract

어업 관리는 기후 변화, 환경 악화, 남획으로 인해 심각한 도전에 직면해 있으며, 이는 해양 생태계와 연안 지역 사회의 생계를 위협하고 있다. 이 연구는 대한민국 관할 해역의 주요 어종 어획량 변화를 예측하기 위해 지리정보시스템(GIS)과 기계 학습(ML) 기법, 특히 인공신경망(ANN)과 합성곱신경망(CNN)을 통합하여 지속 가능한 자원 활용과 적응형 정책 전략을 위한 프레임워크를 제시한다. 이 연구는 동해, 황해, 동중국해와 같은 주요 해역의 독특한 해양 조건과 생산적이지만 변동성이 큰 생태계를 중심으로 수행되었다. 해수면 온도(SST), 엽록소-a 농도, 염도, 수심 등 환경 요인을 분석하여 어종 분포와 풍부함에 영향을 미치는 주요 요인을 파악하였으며, 단위 노력 당 어획량(CPUE)을 핵심 지표로 활용하였다. ANN과 CNN 모델은 환경 변수와 CPUE 간의 복잡한 관계를 효과적으로 포착하며, ANN은 시간적 추세를, CNN은 공간 패턴 인식을 우수하게 수행하였다. 특히 CNN은 5월과 10월의 주요 어획 시기에 황해에서 높은 예측 정확성을 보였다. GIS 기반 시각화는 모델 결과를 명확히 제시하며, 생태학적 중요 지역과 어획 구역을 강조하여 데이터 기반 의사결정을 지원한다. 이 연구는 GIS와 ML 기술의 통합이 어업 관리 과제를 해결하는 데 있어 변혁적인 가능성을 제시하며, 해양 자원의 시공간적 역학에 대한 이해를 통해 자원 할당과 지속 가능한 관행을 최적화한다. 또한, 실시간 환경 모니터링과 고해상도 예측 모델은 기후 변화와 인위적 압력에 대한 정책적 대응을 강화하며, 제안된 프레임워크는 생태적 지속 가능성과 경제적 회복력의 균형을 유지하는 데 기여한다. 결론적으로, 본 연구는 GIS와 ML의 통합을 통해 현대 어업 관리에서의 첨단 기술의 중요성을 강조하며, 기후 변화와 남획 문제를 해결하기 위한 혁신적인 접근법을 제공한다.|Climate change, environmental degradation, and overfishing have created unprecedented challenges for fisheries management, causing significant disruptions to marine ecosystems and the livelihoods of coastal communities. In order to forecast changes in the catch of important fish species within the jurisdictional waters of the Republic of Korea, this study investigates the integration of geographic information systems (GIS) and machine learning (ML) techniques, specifically artificial neural networks (ANN) and convolutional neural networks (CNN). The research offers creative frameworks through which to improve the sustainable use of resources, support flexible policy approaches, and bolster the resilience of marine ecosystems by tackling these pressing issues. The study focuses on the East Sea, Yellow Sea, and East China Sea, which are ecologically and economically important for Korea. These waters have special oceanographic features, such as the meeting point of cold and warm currents, which produce highly productive, but unstable, ecosystems. In this study, such environmental drivers as sea surface temperature (SST), salinity, water depth, distance from shore, and chlorophyll-a concentration are examined in order to identify important factors affecting fish distribution and abundance. The catch per unit effort (CPUE) metric is a critical tool for assessing fish stock abundance and fishing efficiency, and allows for detailed spatial and temporal analysis. In order to capture the intricate, non-linear relationships between environmental variables and CPUE, this study uses such ML models as ANN and CNN. ANN excels at modeling temporal trends, whereas CNN outperforms in spatial pattern recognition. The combined application of these models, which provides improved prediction accuracy and reliability across various regions and temporal periods, is a significant accomplishment of this study. For instance, in the Yellow Sea, CNN performed well during important fishing months (e.g., May and October), obtaining a high Spearman correlation with measured CPUE values. The incorporation of GIS further improves the interpretability of the model outputs. GIS-based visualizations provide spatially explicit maps which highlight ecological hotspots, high-efficiency fishing zones, and conservation priority areas. For stakeholders and policymakers, these tools have been extremely helpful in facilitating data-driven decision-making and coordinating fisheries management plans with the Korean Act on Marine Spatial Planning and Management. Furthermore, the seasonal analysis revealed dynamic changes in fish populations, providing useful information for adaptive management and zoning. The results highlight how the combination of GIS and ML technologies can revolutionize how fisheries management is approached. In order to maximize resource allocation and encourage sustainable practices, the findings highlight the sheer importance of comprehending the spatial and temporal dynamics of marine resources. This study helps create flexible strategies that balance economic goals with ecological conservation objectives by using data-driven tools. Furthermore, the study offers practical implications for marine spatial planning (MSP). Incorporating high-resolution predictive models and real-time environmental monitoring systems can help policymakers more accurately predict how anthropogenic pressures and climate variability will affect fish stocks. The adaptive frameworks in this study provide a scalable method for striking a balance between the demands of economic resilience and ecological sustainability. In conclusion, this study emphasizes the importance of advanced technologies in modern fisheries management. Combining GIS and ML improves prediction accuracy while bridging the gap between academic study and real-world policy implementation. By aligning technological advancements with sustainability goals, this study paves the way for novel solutions to the pressing challenges of climate change, environmental variability, and overfishing in the jurisdictional waters of Korea.