Exploratory Data Analysis of a North American Whole Building Life Cycle Assessment datasets

• Applied systematic EDA to the largest detailed and harmonized WBLCA dataset in NA. • Gained insights into low-carbon building design and decarbonization from real data. • Demonstrated that EDA framework adequately addresses data challenges. • Elucidated key nuances to enhance the interpretation and decision-making. • Provided key insights to support future advanced analysis and predictive modeling. The building construction sector is a major contributor to global greenhouse gas emissions, with embodied carbon—the emissions from material production, use, and disposal—forming a significant portion. While whole building life cycle assessment (WBLCA) quantifies such impacts, detailed statistical analysis for interpretation of WBLCA data remains underutilized. This study addresses the gap by introducing a systematic exploratory data analysis (EDA) framework applied to a harmonized and high-resolution dataset of 244 real-world North American buildings. The core steps in the EDA framework include distinguishing attributes, univariate analysis, bivariate analysis, and feature engineering. These steps address the different data challenges, such as high dimensionality, mixed attribute types, missing values, outliers, and multivariate relationships through the evaluation of robust selections of statistical methods in EDA. Bivariate analysis (e.g., mutual information, one-way ANOVA, post-hoc analysis, and two-way ANOVA) and feature engineering methods quantified the correlation and differences in pairwise and global relationships. Embodied carbon intensity weakly correlated with most meta-features, while materials and building use emerged as the most influential factors. Given that the dataset is uniquely comprehensive and disaggregated, a more nuanced and detailed understanding of environmental impact patterns and relationships is possible. This research demonstrates that following the systematic EDA framework provides a well-defined process to address the database challenges. Its integration into WBLCA dataset evaluation enhances interpretation and supports informed decision-making for low-carbon building design and decarbonization strategies that a conventional simplified analysis would miss. Moreover, it helps to build the insights necessary for advanced analysis and predictive modeling.