Machine Learning based Geospatial Sensitivity Analysis Tool (ML-GeoSAT) for Gross Exposure to Climate-Related Risks

基于机器学习的地理空间敏感性分析工具 (ML-GeoSAT)，用于应对气候相关风险的总体暴露

• 批准号: 10031472
• 金额: $ 6.36万
• 依托单位: ANGLESTACK LIMITED
• 依托单位国家: 英国
• 项目类别: Small Business Research Initiative
• 财政年份: 2022
• 资助国家: 英国
• 起止时间: 2022 至 无数据
• 项目状态: 已结题
• 来源: https://gtr.ukri.org/projects?ref=10031472
• 关键词: Machine; Learning; based; Geospatial; Sensitivity

ML-GeoSAT is a unique tool that will allow underwriters to measure the direct and indirect impacts of physical and transition risks on the entire value chain of a counterparty to determine exposure or vulnerability risks/levels. Particularly, ML-GeoSAT is suited for direct counterparty investments like extractive and mining companies (involved in direct extraction processes like cobalt and other mineral resources mining) and indirect counterparties like battery and electric vehicles manufacturers with transcontinental supply chain networks. For the direct counterparties, underwriters are interested in understanding how weather patterns and local climate-related policies can influence or inflict damage on these investments with or without mitigation (gross or net losses) while for the indirect counterparties, underwriters are interested in understanding how physical and transition risks on supply chain infrastructure can affect a counterparty's ability to service debts or meet other financial obligations.ML-GeoSAT is a disruptive approach to evaluating risks, especially for counterparties as it adopts a Natural Capital Analysis model in risk evaluation. ML-GeoSAT posits nature as capital stock and accordingly assesses how natural degradation negatively impacts a financial institution. This takes the form of a portfolio-level assessment that helps financial institutions identify their natural capital asset dependencies. It is usually carried out in four steps, identifying: (1) relevant geographies, sectors, borrowers and/or assets; (2) relevant natural capital assets (e.g., minerals, water, rubber etc.) that their directly or indirectly tied to counterparties; (3) potential natural and transition-related disruptions that could occur; and (4) geographies, sectors, borrowers and/or assets most at risk and its overall impact on the underwriters' portfolio risk.ML-GeoSAT will seek to achieve the following: (1) dynamically discretise counterparties portfolios by creating a logistics-based pathway of inputs from source to final products; (2) utilise geospatial data and advanced image recognition in evolving a physical risk framework (for real-time and predictive risk assessments) for each material hub; (3) leverage advanced natural language processing techniques in evolving a real-time knowledge-based framework (for digitising real-time climate policies and codifying a transition-risk assessment protocol) for each material-hub (region-specific); (4) integrating physical and transition risk assessments for each material hub into a comprehensive climate-risk assessment tool (CRAT) for real-time sensitivity analysis; (5) providing an intuitive visualisation platform with innovative financial metrics to convey real-time and future-based impairment for varying scenarios.

ML-GeoSAT是一个独特的工具，将允许承保人衡量实体和过渡风险对交易对手整个价值链的直接和间接影响，以确定暴露或脆弱性风险/水平。特别是，ML-GeoSAT适用于直接交易对手投资，如采掘和采矿公司（参与钴和其他矿产资源开采等直接开采过程）以及间接交易对手，如拥有洲际供应链网络的电池和电动汽车制造商。对于直接交易对手，承销商有兴趣了解天气模式和当地气候相关政策如何影响或损害这些投资，无论是否缓解（毛损失或净损失），而对于间接交易对手，承销商有兴趣了解供应链基础设施的物理和过渡风险如何影响交易对手偿还债务或满足其他财务要求的能力。ML-GeoSAT是一种颠覆性的风险评估方法，特别是对于交易对手，因为它在风险评估中采用了自然资本分析模型。ML-GeoSAT将自然设定为资本存量，并据此评估自然退化如何对金融机构产生负面影响。这采取投资组合层面评估的形式，帮助金融机构确定其自然资本资产依赖性。评估通常分四个步骤进行，确定：（1）相关的地理位置、部门、借款人和/或资产;（2）相关的自然资本资产（例如，矿物质、水、橡胶等）（3）可能发生的潜在自然和过渡相关中断;以及（4）风险最大的地理位置、行业、借款人和/或资产及其对承销商投资组合风险的整体影响。ML-GeoSAT将寻求实现以下目标：（1）通过创建从源头到最终产品的基于物流的输入路径，动态离散化交易对手投资组合;（2）利用地理空间数据和先进的图像识别技术，制定实体风险框架（3）利用先进的自然语言处理技术，逐步建立一个实时知识框架（用于数字化实时气候政策和编纂过渡风险评估协议）（特定于区域）;（4）将每个材料中心的物理和过渡风险评估整合到综合气候风险评估工具（CRAT）中，以进行实时敏感性分析;（5）提供直观的可视化平台，并提供创新的财务指标，以传达不同情景下的实时和基于未来的减值。