Conference Agenda

Construction is essential for economic development, but places substantial pressure on environmental quality and human health through material- and energy-intensive supply chains. In the U.S., construction spending reached 2.19 trillion USD in 2024, supporting nearly 130 million structures, while buildings and infrastructure account for approximately one-third of national greenhouse gas emissions when operational energy use is included. Continued growth in construction activity intensifies these pressures, contributing to fine particulate matter exposure, smog formation, and other environmental stressors linked to premature mortality and disease.

Recent studies have evaluated construction-related emissions using both bottom-up and top-down approaches. Bottom-up analyses often exclude construction and installation processes, leading to systematic underestimation of total impacts. Top-down studies capture complete supply chains but frequently rely on highly aggregated models with limited sectoral resolution, constraining their ability to represent material specificity and spatial heterogeneity. This study conducts a high-resolution top-down assessment of environmental and public health impacts associated with U.S. construction activity using the U.S. Environmentally Extended Input-Output model, representing more than 400 economic sectors. Construction spending is mapped across detailed residential and nonresidential categories and evaluated under both consumption-based and production-based accounting perspectives. Environmental indicators are quantified using TRACI, and selected midpoint impacts are translated into disability adjusted life years to estimate health burdens. To extend beyond static accounting, this framework is designed to support additional analyses examining how future construction trajectories may influence environmental and health outcomes. Planned extensions include evaluating alternative construction techniques, shifts in residential and nonresidential investment driven by housing shortages, and regional variations in construction intensity and density across the United States. While land use impacts are not directly assessed due to methodological limitations of TRACI, the focus on emissions, energy use, and supply chain structure enables robust comparison across construction types and regions.

Preliminary results indicate that U.S. construction activity generated ~456 million metric tons of carbon dioxide equivalents (MtCO2-e), with non-residential construction contributing 64% (293 MtCO2-e) and residential construction contributing 36% (163 Mt CO2-e). Direct on-site emissions account for 27% of total climate impacts, while upstream supply chains dominate overall burdens (73%). Imported materials also represent a 18% share of embodied emissions, with major contributions from key international trading partners including China, Canada, India, Russia, and Mexico. Beyond climate impacts, construction activity resulted in an estimated annual loss of ~760,000 disability adjusted life years, driven primarily by greenhouse gas emissions and particulate matter formation. Ongoing analysis is refining these estimates and expanding the results to incorporate future construction scenarios and regional differentiation. Overall, this study provides a scalable and extensible framework linking construction expenditures to environmental and human health impacts, supporting evidence-based strategies for construction sector decarbonization and risk reduction.

Reducing scope 3 greenhouse gas (GHG) emissions, which typically account for around 90% of a company’s total emissions, is critical for companies to push towards ambitious net-zero carbon goals. Recently, market-based GHG accounting methods, like book and claim, have been proposed as a solution to curb scope 3 emissions. Indeed, multiple industries are evaluating the feasibility of book and claim frameworks for products that are upstream or downstream a company’s supply chain. This is because companies may have limited physical access to low-carbon products in hard-to-abate industries, such as concrete, chemicals, and textiles. To circumvent this, a book and claim framework would enable a company to purchase an environmental attribute certificate (EAC) of a low-carbon product and claim its carbon emissions without requiring physical traceability.

While proposed book and claim frameworks could be a viable pathway to curb scope 3 emissions, little research has been performed to critically assess if these frameworks are viable and fair in creating an EAC market. Therefore, the first aim of this work is to present a comparative analysis of the existing book and claim frameworks. The second aim is to demonstrate via quantitative case studies how the book and claim frameworks are applied, highlighting differences in corporate GHG emission accounting and underscoring the limitations when applying the frameworks for direct and indirect commodities. Direct commodities, such as ammonia, have a universally agreed upon chemical composition (i.e., product make-up) and thus need fewer guardrails to ensure a fair EAC market. On the other hand, indirect commodities, like concrete and textiles, are typically composed of various proportions of raw ingredients but may have equivalent performance. Lastly, the limitations identified after applying the book and claim frameworks will elucidate recommendations to enable and ensure a fair EAC market.

The synthesis of the existing book and claim frameworks suggests that EAC eligibility requirements differ between direct and indirect commodities. For example, indirect commodities might have starkly different material compositions yet still satisfy a similar performance requirement (e.g., achieving a specific compressive strength class for a concrete mixture). Because of the nuances inherent in indirect commodities, one key recommendation is to require additional guardrails, such as industry-specific matching rules, to ensure a fair EAC market. These additional guardrails go beyond industry-specific requirements defined in product category rules. These guardrails would similarly influence how the GHG emissions are accounted for (e.g., direct substitution method). Overall, this work formulates recommendations and demonstrates their need to establish a fair and reputable EAC market for hard-to-abate sectors.

Recent calls have been made for fair/equitable tools and frameworks to be integrated throughout the research and design life cycle —from conception to implementation—with an emphasis on reducing inequity in artificial intelligence (AI) and machine learning (ML) applications. Simply stating that equity should be integrated throughout, however, leaves much to be desired as sustainability researchers, practitioners, and decision-makers attempt to employ equitable practices. In this forum piece, we use a critical review approach to explain how socioecological inequities emerge in ML applications across their life cycle stages by leveraging the food system. We exemplify the use of a comprehensive questionnaire to delineate unfair ML bias across data bias, algorithmic bias, and selection and deployment bias categories. Finally, we provide consolidated guidance and tailored strategies to help address AI/ML unfair bias and inequity in sustainability science applications. Specifically, the guidance and tools help to address sensitivity, reliability, and uncertainty challenges. There is also discussion on how bias and inequity in AI/ML affect other sustainability science research and design domains, besides the food system—such as living labs and circularity. We conclude with an explanation of the future directions sustainability research should take to address unfair bias and inequity in AI/ML. Last, we call for systemic equity to be embedded throughout sustainability applications to fundamentally understand domain-specific socioecological inequities, identify potential unfairness in ML, and select mitigation strategies in a manner that translates across different research domains.

The cooling requirements for data centers can significantly increase electricity and water use, a concern that is intensifying as data centers are increasingly installed in water-stressed regions across the United States (US). The findings from current studies evaluating the environmental impact of individual data centers cannot be generalized to other datacenters as the local conditions (cooling technology and specifications, local weather, and electricity grid mixes) vary. Consequently, there is no generalizable approach which quantifies the electricity, water and greenhouse gas (GHG) impact of data centers across an increased geographical scale (e.g., county) and simultaneously accounts for variations in local conditions.

To address this knowledge gap, this study develops the first county-specific and geographically scalable assessment of the water, electricity, and GHG impact of cooling data centers. This research uses an artificial intelligence (AI) and physics-based framework to quantify the said environmental impacts for data centers in Arizona (AZ). The AI model is trained on and applied to publicly available satellite imagery to classify the AZ data centers as water-cooled (i.e., cooling tower is present on the rooftop) or air-cooled (cooling-tower absent). The AI model extracts the dimensions (e.g., fan radius) and geometry of the cooling tower (if present). The cooling tower make and model is determined by matching the AI-determined dimensions with those of 2500 commercially manufactured cooling towers (compiled from manufacturer websites).

For the 10 AZ data centers with cooling towers (250 MW), the physics-based model quantifies the environmental footprint by accounting for 28 parameters covering the cooling tower make and model, data center operations, local weather, chiller performance, and grid water and GHG-intensity. For the remaining 35 AZ data centers without cooling towers (1320 MW), the physics-based model similarly accounts for 22 parameters and quantifies the environmental impacts by leveraging the parametric equations in the Department of Energy’s EnergyPlus program.

AZ’s data centers annually consume 73 million m³ of water (4% of AZ households) and 19 billion kWh of electricity (49% of AZ residential electricity demand) and emits 4.7 million tonnes CO₂e yr⁻¹ (11% of AZ household emissions). No single cooling technology universally minimizes water-use across AZ data centers. Air cooling has a lower water impact than water cooling in 14 AZ counties, whereas water cooling has a lower water impact in Coconino County. Coconino County’s electricity is 60% hydropower, which results in high evaporative losses at the upstream hydroelectric powerplant. Therefore, in Coconino County, the upstream water penalty from generating electricity for electricity-intensive air cooling is significantly higher than the water penalty from using less electricity-intensive water cooling (i.e., cooling towers). Therefore, water cooling is preferable to air cooling of data centers in Coconino County.

Based on these findings, this research defines principle to (i) extend the AI and physics-based framework to assess the location-specific environmental footprint of cooling data centers across various US counties; (ii) optimize the cooling portfolio of data centers to minimize environmental impacts; and (iii) determine thresholds for water and GHG-intensity of electricity at which cooling towers are preferable to air coolers.