炭素貯留における将来のAI応用のためのマルチソースグラフベース地理空間データリソースを支えるデータと補足資料

Unofficial AI-generated summary based on the public title and abstract. Not an official translation.

🔬研究者:GNNや他の機械学習手法を用いた炭素貯留適地評価のための再利用可能なデータセットとベンチマークを提供する。

🏢実務担当者:CCSプロジェクトの初期段階におけるサイトスクリーニングやリスク評価に活用できるデータ基盤を提供する。

🏛政策担当者:CCS適地選定の透明性と再現性を高めるデータ駆動型アプローチの重要性を示唆する。

Overview This data set and supplementary code accompanies the journal article “A Multi-Source Graph-Based Geospatial Data Resource for Future AI Applications in Carbon Storage.” It integrates multiple geospatial data sources into a graph-based resource intended to support research and the development of AI applications related to carbon storage.  The purpose of this work is to introduce a relational, multi-source spatial data resource designed to support the development of future AI-driven framework for CO₂ storage site selection. Carbon storage suitability assessment requires integrating heterogeneous datasets describing geological, environmental, infrastructural, and socio-economic conditions. Traditional spatial datasets are typically stored as independent layers, limiting their ability to represent interactions among systems. This dataset addresses that limitation by organizing spatial information into a graph-structured framework suitable for future machine learning applications, particularly Graph Neural Networks (GNNs). The framework enables AI models to: Learn from interconnected spatial systems rather than isolated variables Detect spatial and functional relationships among candidate storage locations Evaluate site suitability using integrated environmental and infrastructure constraints Produce scalable, reproducible, and flexible optimization metrics for decision support Dataset Description The dataset integrates multiple spatial data sources across Louisiana into a unified relational structure composed of nodes and edges.  Nodes represent spatial entities relevant to CO₂ storage suitability, including geological formations and subsurface characteristics, candidate CO₂ storage locations, industrial emission sources, population density and demographic indicators, critical habitats, and environmental constraints. Each node contains attributes describing physical, environmental, or socio-economic properties used as model inputs. Edges describe relationships and interactions between nodes, such as hydrological connectivity, transportation and pipeline proximity, spatial adjacency and distance relationships, and source-to-sink connectivity. File Descriptions Supplementary Material 1 Describes the metadata and data dictionaries for both the edge list and node attribute databases, including variable definitions, units, and data sources. The SM1 also details the data quality control, assessment, and data usability performed during the pre and post-processing stages of the data integration. Supplementary Material 2 Contains the Python notebooks documenting the full processing workflow used to generate the graph structure. Supplementary Material 3 Includes the finalized edge list and node attribute databases used for graph construction and machine learning applications. Supplementary Material 4 Includes the 37 data layers (vectors and rasters) integrated in the study. Supplementary Material 5 Details the data inputs and outputs, descriptive statistics, and summaries. Supplementary Material 6 Represents an example of a sensitivity analysis conducted to evaluate the influence of selected data construction thresholds.

• Zenodo https://zenodo.org/records/23422722first seen 2026-10-20 04:12:39 · last seen 2026-10-21 04:13:28