永續計畫資料庫

The Post-2020 Global Biodiversity Framework mandates that countries protect 30% of terrestrial areas and ensure effective restoration of 30% of ecosystems by 2030, enhancing biodiversity, services, and connectivity. This is particularly challenging in Taiwan and East Asia due to the extensive and varied forest landscapes which complicate traditional assessment methods. Our research focuses on the quantification and prediction of structural diversity in these complex ecosystems, essential for assessing carbon capacity and plant diversity.

We propose using generative AI and deep learning algorithms to merge multiple data types at varying scales and resolutions. By integrating high-resolution satellite imagery, airborne and terrestrial laser scanning, and field biodiversity surveys, we plan to develop a detailed framework to analyze and predict forest structure in Taiwan and East Asia. This method will utilize the extensive coverage of satellite imagery and the precise three-dimensional data from laser scanning. Advanced self-supervised learning models will be employed to learn from field data and effectively integrate multi-scale data. This will allow us to create detailed, spatially explicit predictions of forest structure under different climate scenarios, offering new insights into forest dynamics and biodiversity.

Should the project succeed over the three-year funding period, it will greatly aid in meeting Sustainable Development Goal 15 (SDG-15) and Global Biodiversity Framework (GBF) Target 3. Our approach will furnish precise predictions of forest structure and carbon stocks across the region, supporting the strategic planning of protected areas and restoration initiatives. This will enhance governmental capacity to develop land-use policies that consider both conservation and development needs. The AI methodologies developed could serve as models for similar conservation challenges worldwide, promoting the integration of multi-scale data and advanced AI in forest management. Our findings will equip stakeholders, including governments and non- governmental organizations, with crucial data to craft effective forest conservation strategies amid climatic and anthropogenic pressures. Ultimately, our research will enhance understanding and management of forest ecosystems, contributing significantly to biodiversity preservation and climate change mitigation efforts in Taiwan, Asia, and globally.

全球生物多樣性框架要求各國在 2030 年前保護 30%的區域,並確保 30%的生態系統得到有效恢復。對於台灣和東亞地區而言,由於森林景觀多樣,傳統評估方法難以實施。我們的研究重點是量化和預測這些複雜生態系統的結構多樣性,這對評估碳存量和植物多樣性至關重要。我們利用生成式人工智慧和深度學習算法,將不同尺度和解析度的多種數據類型進行融合。透過整合高解析度衛星圖像、機載和地面雷射掃描以及實地生物多樣性調查,我們計劃開發一個詳細的架構來分析和預測台灣和東亞地區的森林結構。我們先進的自監督學習模型將從數據中學習,並有效整合多尺度數據。這將使我們能夠在不同氣候情境下創建森林結構的詳細預測,為森林動態和生物多樣性提供新的見解。