A coupled LSTM model for predicting blue carbon and fishery dynamics in tropical coastal wetlands under climate change
気候変動下の熱帯沿岸湿地におけるブルーカーボンと漁業動態を予測する結合LSTMモデル (AI 翻訳)
Yanhua Zhang, Gongguo Wu, Chujun Zou, Di Wu, Pengfei Xie, S Wang, Ping Wang
🤖 gxceed AI 要約
日本語
この研究は、熱帯沿岸湿地におけるブルーカーボンストックと漁業資源の双方向関係を予測する初の結合LSTMモデルを開発した。広東省と海南島の15地点で2018年から2025年までの観測データを用い、高い予測精度（R²=0.93、0.89）を達成。温暖化シナリオではブルーカーボンが7.2%、漁獲量が11.4%減少すると予測した。
English
This study developed the first coupled LSTM framework to predict bidirectional relationships between blue carbon stocks and fishery abundance in tropical coastal wetlands. Using 96 monthly field observations from 15 sites in China (2018-2025), the model achieved high predictive accuracy (R²=0.93 for blue carbon, 0.89 for CPUE). Simulations for 2025-2026 indicate that under a moderate warming scenario, blue carbon stocks could decline by 7.2% with associated fishery CPUE reductions of 11.4%.
Unofficial AI-generated summary based on the public title and abstract. Not an official translation.
📝 gxceed 編集解説 — Why this matters
日本のGX文脈において
日本のブルーカーボン生態系（藻場・干潟等）の評価にも応用可能な手法。LSTMを用いた時系列予測は、日本の沿岸生態系の炭素貯留量や水産資源管理にも展開できる。ただし、データは中国熱帯域に限定されており、温帯域への適用には調整が必要。
In the global GX context
This paper provides a novel AI-based predictive framework for blue carbon and fishery dynamics, relevant to global climate mitigation and adaptation efforts. The coupled modeling approach can inform ecosystem-based management in tropical coastal zones worldwide. For international disclosure frameworks, blue carbon accounting is increasingly recognized under national GHG inventories.
👥 読者別の含意
🔬研究者:A novel LSTM-based approach to model coupled blue carbon-fishery systems, with strong predictive performance and scenario analysis.
🏢実務担当者:For coastal managers and carbon project developers, this model can help predict blue carbon sequestration and fishery impacts under climate scenarios.
🏛政策担当者:The scenario simulations highlight the vulnerability of blue carbon and fisheries to warming, supporting climate adaptation planning.
📄 Abstract（原文）
Coastal wetlands, including mangroves, seagrass beds, and coral reefs, provide critical blue carbon sequestration services and nursery habitats that support fishery resources. However, the bidirectional coupling between blue carbon stocks and fishery abundance remain poorly quantified, particularly under accelerating environmental change in tropical marginal seas. This study presents the first coupled LSTM framework for bidirectional blue carbon-fishery prediction in tropical coastal wetlands. We developed a coupled Long Short-Term Memory (LSTM) neural network model to predict the bidirectional relationship between blue carbon stocks and fishery resource abundance in the coastal ecosystems of Guangdong Province and Hainan Island, China. Field surveys were conducted at 15 representative sites from 2018 to 2025, generating 96 monthly observations of environmental variables (sea surface temperature, salinity, dissolved oxygen, turbidity, nutrients), biological indicators (mangrove above-ground biomass, soil organic carbon, seagrass coverage, coral cover), and fishery metrics (catch per unit effort, species richness, juvenile abundance). The LSTM model achieved superior predictive performance compared to baseline methods, with root mean square error of 0.142 Mg C ha⁻¹ for blue carbon prediction (R² = 0.93) and 0.108 kg h⁻¹ for CPUE prediction (R² = 0.89). Feature importance analysis revealed that mangrove soil organic carbon was the strongest predictor of fishery CPUE (Shapley value = 0.187), while sea surface temperature exerted the greatest influence on blue carbon stock variability. Scenario simulations for 2025-2026 indicate that under a moderate warming scenario (SST increase of 0.8 °C), blue carbon stocks are projected to decline by 7.2% (95% CI: 5.8-8.6%), with associated fishery CPUE reductions of 11.4% (95% CI: 9.2-13.6%). These findings provide a novel predictive framework for ecosystem-based management of tropical coastal wetlands and highlight the vulnerability of coupled social-ecological systems to climate change.
🔗 Provenance — このレコードを発見したソース
- openalex https://doi.org/10.1038/s41598-026-59981-yfirst seen 2026-07-13 06:06:16
🔔 こうした論文の新着を逃したくない方は キーワードアラート に登録（無料・3キーワードまで）。
gxceed は公開メタデータに基づく研究支援データセットです。要約・翻訳・解説は AI 支援で生成されています。 最終的な解釈・検証は利用者が原典資料に基づいて行うことを前提とします。