emissions modeling

The escalating energy consumption and carbon emissions of software and AI systems necessitate rigorous measurement methodologies. Method: This paper systematically surveys and evaluates existing energy and carbon measurement approaches, proposing the first unified taxonomy classifying methods into monitoring-, estimation-, and black-box-based categories. It conducts a multidimensional assessment across hardware components (CPU, GPU, RAM) and dual dimensions—energy consumption and carbon emissions—grounded in bibliometric analysis and functional comparison of 87 tools. Contribution/Results: Key gaps are identified, including inadequate GPU dynamic power modeling and insufficient carbon intensity mapping for cloud environments. Three pervasive challenges are revealed: poor reproducibility, high hardware heterogeneity, and ambiguous system boundary definitions across the software lifecycle. The findings provide theoretical foundations and practical pathways for establishing standardized benchmarks and advancing green software engineering.

Carbon emissions from generative AI inference have been systematically underestimated due to ambiguous system boundaries, lack of standardized metrics, and methodological biases in existing assessments. This study employs a scoping review to systematically analyze carbon footprint measurement tools, classification frameworks, and research trends across the full training–inference lifecycle. We propose (i) a standardized measurement protocol, (ii) a dynamic evaluation framework, and (iii) a multidimensional sustainability assessment system—novelly integrating user behavior and system boundary definitions into a unified analytical model. Our analysis uncovers nonlinear relationships between carbon emissions and model scale, prompt complexity, and deployment environment, identifying key leverage points for mitigation. The findings establish a reusable methodological benchmark and practical guidelines for AI environmental governance across technical, social, and operational dimensions.

Existing LLM inference energy and carbon emission assessment frameworks commonly neglect GPU power consumption modeling, leading to inaccurate carbon footprint estimation. This paper proposes the first high-fidelity simulation framework that integrates a fine-grained, utilization-based GPU power model with dynamic grid characteristics, enabling joint quantification of energy consumption and carbon emissions across the full LLM inference pipeline, as well as carbon-aware scheduling analysis. Innovatively, it deeply unifies hardware-level power modeling, LLM inference performance simulation, and energy system co-simulation—enabling, for the first time, interpretable, parameter-level attribution of inference configurations to carbon impact and supporting evaluation of renewable energy integration potential. Experiments show that, under typical deployment scenarios, the framework increases renewable electricity offset rate to 69.2%, providing a verifiable, quantitative tool and evidence-based decision support for low-carbon AI infrastructure design.

This study investigates the net climate impact of AI development on data center energy consumption and greenhouse gas emissions, specifically assessing its dual role in achieving net-zero targets by 2035. Method: We employ an integrated approach combining large-scale data-driven modeling, life-cycle emission assessment, and multi-scenario analysis to quantify both the carbon cost of AI compute growth and its cross-sectoral mitigation potential—spanning energy generation, distribution, and end-use efficiency. Contribution/Results: We introduce a dynamic evolutionary model revealing a nonlinear trajectory: near-term emission increases (pre-2030) driven by computational expansion, followed by substantial net-negative emissions post-2030 via AI-enabled process automation, systemic energy efficiency gains, and intelligent renewable energy integration. Results demonstrate that AI’s long-term abatement benefits systematically outweigh its infrastructure carbon footprint, positioning it as a critical enabling technology for climate governance.

Large language model (LLM) inference incurs substantial energy consumption and carbon emissions, yet existing estimation tools are intrusive, low-accuracy, and require cumbersome input configurations. Method: We propose R-ICE, a lightweight, non-intrusive modeling framework that systematically exploits transferable signals from public LLM benchmark datasets (e.g., OpenLLM, LMSys) for carbon estimation. R-ICE establishes an end-to-end, prompt-level pipeline for fine-grained energy and carbon emission estimation, integrating feature engineering, regression modeling, and dynamic carbon intensity mapping. Contribution/Results: R-ICE enables novel applications such as dynamic LLM routing and carbon accounting, with low runtime overhead, high adaptability, and cross-hardware generalizability. Extensive experiments across diverse models and hardware configurations demonstrate an average estimation error of <12%, significantly outperforming conventional monitoring approaches. R-ICE provides a scalable, infrastructure-ready foundation for green AI evaluation.

Rapid AI advancement has exacerbated challenges in assessing environmental impacts and lacks transparency in corporate carbon footprints. Method: This study develops an enterprise-level AI environmental impact assessment framework covering the full lifecycle—training, inference, hardware manufacturing, and end-of-life—enabling non-LCA experts to quantify carbon footprints. It introduces a novel multidimensional projection model integrating scaling laws, semiconductor energy-efficiency trends, and grid decarbonization pathways, alongside the first “Environmental Return on Emissions” (EROE) metric to align AI development with net-zero goals. Results: Combining streamlined LCA, IPCC electricity scenarios, empirical compute analysis, and generative AI adoption forecasts, we find: (1) large language models consume 4,600× more energy than traditional models; (2) under high-adoption scenarios, global AI electricity demand will increase 24.4× by 2030; and (3) isolated technological optimizations are insufficient to meet climate targets.

This study addresses the challenge of assessing the environmental impact of datacenter task and resource management policies. We propose a simulation-based methodology for quantifying carbon footprint, extending the Batsim simulator to enable fine-grained, dynamic CO₂ emission estimation within the SimGrid framework. Our approach integrates real-time platform power consumption models with time-series regional grid carbon intensity data, computing emissions per task and per node during scheduling simulations. The method is seamlessly embedded into existing simulation workflows, enabling reproducible and scalable evaluation of the carbon efficiency of distributed application scheduling strategies. Our primary contribution is the development of the first open-source, high-fidelity, carbon-aware simulation plugin tightly coupled with scheduling logic—providing a standardized, extensible assessment tool for green computing research.

This study addresses the challenges of identifying software-related carbon emissions and clarifying responsibility boundaries in the ICT sector. We propose the first carbon accountability mapping framework spanning the entire software lifecycle. Leveraging systematic mapping, ICT energy consumption modeling, and multi-stakeholder governance analysis, we trace energy flows and emission sources across hardware-software interactions, and explicitly define climate accountability for developers, operators, and cloud service providers across design, deployment, and operation phases. Our key innovation lies in operationalizing abstract environmental responsibility into a traceable, allocatable role-based model, and developing a standardized Software Carbon Accountability Atlas tool. The framework supports compliance with EU regulatory frameworks—including the Corporate Sustainability Reporting Directive (CSRD) and Corporate Sustainability Due Diligence Directive (CSDD)—and significantly enhances transparency in ICT firms’ carbon disclosures and cross-organizational emission reduction coordination.

This study addresses the lack of standardized methodologies for Scope 3 Category 1 carbon accounting in enterprise AI inference services, where current practices often overestimate emissions by 10–40× due to reliance on coarse-grained industry-average emission factors. The authors propose the first four-tier hierarchical accounting framework aligned with CSRD compliance requirements, which degrades systematically—from a token-level physical energy consumption model to expenditure-driven environmentally extended input-output (EEIO) analysis—based on data availability. Integrating GPU energy benchmarks (ML.ENERGY v3), regional grid carbon intensities (EPA eGRID, Ember), and water efficiency metrics, the framework reveals a previously undocumented carbon-water trade-off: low-carbon regions may entail high water consumption, influencing data center siting decisions. An empirical assessment of a 200-person European firm shows annual AI inference emissions under 1 ton CO₂e, indicating that regulatory challenges stem from methodological gaps rather than emission magnitude, while also uncovering water-carbon synergies overlooked by mainstream ESG tools.

Generative AI (GenAI) poses significant climate risks due to its rapidly growing energy demand and associated carbon emissions, yet it remains inadequately integrated into mainstream climate risk assessment frameworks. Method: We propose G-TRACE—a novel, region-aware, cross-modal carbon accounting framework—that combines empirical data with micro-level simulation to quantify energy consumption and carbon intensity by output modality (text/image/video) and deployment geography. We further introduce a seven-tier AI Sustainability Pyramid model linking digital behavior to actionable governance pathways. Contribution/Results: Empirical analysis reveals that a trend in Studio Ghibli–style image generation consumed 4,309 MWh and emitted 2,068 tons of CO₂, underscoring how decentralized inference amplifies systemic environmental impact. G-TRACE enables granular, auditable carbon accounting for GenAI, advancing both theoretical understanding and practical policy intervention for AI-related climate risk identification, measurement, and mitigation.

This study addresses the current lack of a unified and transparent metric for quantifying and comparing decarbonization rates across diverse integrated assessment model (IAM) climate scenarios. The authors propose a concise, interpretable numerical indicator that, for the first time, enables systematic comparison and ranking of implied decarbonization speeds across 126 IAM scenarios. By integrating statistical analysis, empirical distribution construction, parametric fitting, and bootstrap resampling, the methodology rigorously quantifies uncertainty and validates consistency with representative concentration pathway assumptions. The resulting estimates—providing means, medians, and confidence intervals for decarbonization rates under each scenario—offer a robust and scalable foundation for evaluating climate policy pathways.

Current environmental impact assessments of the AI lifecycle suffer from tool heterogeneity, insufficient coverage of water usage and value-chain stages, and poor cross-study comparability. To address these gaps, this paper proposes a unified operational definition of Green AI and introduces a five-stage lifecycle framework—encompassing hardware manufacturing, model development, training, deployment, and reuse. It explicitly distinguishes Green AI (energy-efficiency–focused) from Sustainable AI (holistically sustainable across environmental, social, and economic dimensions). Methodologically, the framework integrates Life Cycle Assessment (LCA), PDCA-based governance, edge–cloud co-optimized hardware strategies, and a calibrated multi-level metrics system combining estimation and empirical measurement. The resulting methodology enables vendor-agnostic, reproducible quantification of energy consumption, carbon emissions, water use, and embodied impacts. This significantly enhances transparency and comparability, providing researchers, engineers, and policymakers with an evidence-driven, actionable guide for sustainable AI development and deployment.