Page 88 - Read Online

Yu et al. Carbon Footprints 2025, 4, 17 https://dx.doi.org/10.20517/cf.2025.12 Page 5 of 23

Table 1. A sample of the top 10 terms with high scores from the NLP model
Keywords Score
Estimate road traffic 0.0052
Accurately estimate road 0.0076
Global greenhouse gas 0.0101
Road traffic 0.0116
Formulate effective emission 0.0221
Estimate road 0.0332
Carbon dioxide 0.0397
Greenhouse gas 0.0436
Emission reduction policies 0.0451
Effective emission reduction 0.0453

NLP: Natural language processing.

Figure 2. Final refined search query.

account for both vehicle types and energy sources.

Actual operating conditions
Vehicle operating modes (e.g., acceleration, cruising, deceleration, idling) significantly affect emissions.
Among these, acceleration typically produces the highest carbon emission rates, while idling results in
extremely high emission factors due to zero speed but ongoing fuel consumption . Traffic conditions such
[22]
as average speed, congestion level, and vehicle density also impact emissions. Intersections characterized by
frequent stops and starts are often identified as localized high-emission zones [23,24] .

Other factors
Additional environmental and infrastructural variables also influence CO emissions. Road gradient can
[25]
2
alter engine load and fuel use, while meteorological factors such as ambient temperature and altitude affect
fuel efficiency and combustion processes, thereby influencing emission levels [26-28] .

TRADITIONAL EMISSION MODELS
Traffic emission models are generally categorized into two broad types: traditional models grounded in
mathematical or physical principles, which are widely adopted by governments and research institutions;
and data-driven models, which have recently gained popularity due to their adaptability to specific data
conditions and application needs.

In this review, traditional models are further classified into average speed, traffic situation, and modal
models, following the framework proposed by Smit . Alternative classification schemes in the literature
[29]
include categorizing models as macroscopic, mesoscopic, or microscopic based on their application scope,
or as white-box, gray-box, and black-box models depending on their level of interpretability . However,
[30]
there is currently no universally accepted taxonomy, and even the widely used macro-meso-micro
[31]
classification remains controversial .