AI and Big Data for Assessing Carbon Emission in Tourism Areas: A Pilot Study in Phuket City ^†

Abstract

Artificial intelligence (AI) and big data technology were applied to assess carbon emissions in a high-tourism area in this study. In the study site, the Thalang Road in Phuket Old Town, Thailand, visitors and vehicles (including cars, motorcycles, trucks, vans, and Tuktuks) were counted using closed-circuit television (CCTV) footage and classified via the real-time detection transformer (RT-DETR) algorithm. The data were combined with records of electricity usage. From March to October 2024, 20,000 visitors per month visited the site. Electricity was the main source of carbon emissions, averaging 88 ± 11 tCO₂-eq monthly. Transport accounted for 500 ± 14 kg CO₂-eq. The average emission per visitor was calculated as 4.2 ± 0.4 kg CO₂-eq. The results showed how sustainable tourism policies and urban planning strategies need to be developed in Phuket. Based on the results, indirect emissions from the site need to be estimated.

1. Introduction

The tourism industry accounts for about 10% of global gross domestic product (GDP), and significantly contributes to worldwide carbon emissions [1,2]. Between 2009 and 2019, the industry was responsible for 8.8% of global greenhouse gas emissions, equivalent to 5.2 GtCO₂-eq [3]. More than 75% of these emissions were related to transportation, accommodation, and food services [4]. The rapid growth of the tourism industry demands the development of low-carbon technologies [5]. Low-carbon tourism must be promoted urgently by implementing coordinated strategies across industries, including the utilization of big data [6,7]. The emissions reduction in tourism [8,9,10] and the relationship between the industry’s growth and the environmental impacts [11,12] have been extensively examined. The concept of a “touristic ecological footprint” enables a useful framework for assessing the impacts by accounting for energy use, land occupation, and water consumption throughout the tourism process [13,14,15]. Artificial intelligence (AI) and big data are becoming more utilized in tourism research [16,17,18,19] as the technologies enable real-time analysis of travel patterns, resource use, and emissions. The outcomes support more effective, data-driven policy development.

Phuket Island, Thailand, has transitioned from a tin-mining economy to a tourism-dependent destination since the mid-1980s. The rising number of visitors and revenue highlight Thailand’s economic reliance on Phuket’s tourism. While contributing to the nation’s GDP, overtourism and environmental degradation are challenges to be addressed. Through AI-based monitoring of visitors and traffic based on electricity consumption data, we generated a temporally resolved emissions profile for Phuket Old Town. This bottom-up, site-specific methodology offers a new method to estimate urban tourism emissions as a scalable means for destination-level carbon monitoring.

2. Methodology

2.1. Study Site

Figure 1 presents the study site, Thalang Road and Soi Rommanee, in Phuket Old Town. The site comprises souvenir shops, restaurants, and hotels. Figure 2 shows a high influx of tourists in the area.

2.2. Data Collection

Vehicles and tourists were counted using closed-circuit television (CCTV) footage at three locations, as shown in Figure 3. The cameras recorded the footage between 9:00 and 22:00 on Thursdays, Fridays, and Saturdays, one week per month from March to October 2024. The Big Data Institute (BDI) of Thailand processed the footage and trained the real-time detection transformer (RT-DETR) algorithm. The model detected and classified objects into people, cars, motorcycles, trucks, and vans as illustrated in Figure 4. The top left image in the figure shows dense pedestrian activity in a night market, while the remaining images illustrate vehicle detection with confidence scores and entry/exit tracking. These visual outputs demonstrate the application of machine learning in real-time classification and movement analysis in tourism areas.

2.3. Estimation

Electricity usage data were obtained from the Phuket Provincial Electricity Authority to estimate energy consumption in the study site. The total carbon emissions (kg CO₂-eq) incorporate transportation and electricity, as illustrated in Figure 5. For transportation, emissions were calculated by multiplying travel distance (km) by the emission factor (kg CO₂-eq/km). For electricity, emissions were estimated based on the quantity consumed (kWh) and its emission factor (kg CO₂-eq/kWh). For transportation, carbon emissions (kg CO₂-eq) were calculated using the method in Ref. [20].

E_transport = d∑Fi,

(1)

where d is the travel distance (km). In this study, d was set at 0.8 km, the combined length of Thalang Road (600 m) and Soi Rommanee (200 m) (Figure 1). Fi is the emission factor (kg CO₂-eq/km) for each vehicle type i.

Table 1. Emission factors for different vehicle types [21].

Vehicles	Emission Factor (kg CO2-eq/km)
Trucks and vans	0.29
Cars	0.18
Motorcycles and Tuktuks	0.05

lists the Fi values used in this study, based on the emission factors of vehicles in Thailand reported by Nilrit and Sampanpanish [21].

3. Results

Figure 6 presents the distribution of transport modes used by visitors at the study site. Motorcycles dominate, accounting for 50.6% of all vehicles, followed by cars (34.2%), trucks (8.9%), vans (5.0%), and Tuktuks (1.3%). Figure 7a shows the estimated transportation-related emissions from March to October 2024. Total emissions remain steady at approximately 600 kg CO₂-eq per month. Cars contributed to the largest emissions, followed by motorcycles and trucks. The emission of Tuktuks was minimal. Figure 7b illustrates emissions from electricity consumption. Emissions exceeded 90 tCO₂-eq per month from April to June, peaking in May at over 100 tCO₂-eq, linked to a consumption of 236 MWh. This peak probably resulted from an increase in the use of fans and air conditioning during summer. Emissions dropped to around 80 tCO₂-eq in June and continued to decrease slightly throughout the rainy season. Figure 7c presents the monthly number of visitors. March records the highest number at 24,000. From April to August, visitor numbers remained around 21,000, then decreased to approximately 19,000 in September and October. Figure 7d shows carbon emissions per visitor. A peak occurs in May at around 5.0 kg CO₂-eq per visitor, aligning with the increase in electricity-related emissions. After May, the values gradually decreased, reaching 4.0 kg CO₂-eq in August, with a slight increase observed in October.

4. Discussion

The study results demonstrate that transportation and electricity use are key contributors to direct carbon emissions in a tourism-intensive urban area. Electricity use emerged as the largest emission source, averaging 88 ± 11 tCO₂-eq per month. The monthly emission was recorded as the highest in May, exceeding 100 tCO₂-eq due to increased cooling demand during the peak of summer. The hotel and other accommodation sectors contribute a high proportion of the total energy consumption, primarily related to air-conditioning systems, illumination, and water heating [22]. These trends highlight the importance of energy-efficient technologies and strategies in hospitality [23].

Visitor numbers ranged from 24,000 in March to 19,000 in September–October, with a stable number of approximately 21,000 visitors from April to August. Despite the monthly variation, carbon emissions per visitor peaked in May at 5.0 kg CO₂-eq, reflecting higher electricity consumption. The steady decline after May indicated that seasonal factors and energy demand patterns influenced per-visitor emissions. In the transportation at the study site, motorcycles were the most common mode. However, cars contributed the most to emissions due to their larger emission factor. Transport-related emissions remained consistent throughout the observable period: a monthly average was 600 kg CO₂-eq. The emissions reduction in transportation required interventions such as electric vehicle promotion, improved public transit, and enhanced walkability [24,25]. Encouraging alternative modes of travel such as rail or ridesharing can have a substantial impact on the carbon cost of tourism [26].

The results of this study guide policymakers and urban planners working for low-carbon tourism. However, aligning tourism development with climate policy goals remains a major challenge, as the tourism industry prioritizes economic gains over environmental limits. It is important that tourism strategies based on climate awareness be mainstreamed into overarching development planning to achieve sustainability [27].

5. Conclusions

We assessed direct carbon emissions in a high-tourism area by analyzing transport modes, electricity consumption, and visitor activity from March to October 2024 in Phuket, Thailand. Transport-related emissions remained steady at approximately 600 kg CO₂-eq per month, with cars contributing the largest share despite motorcycles being the most used mode. Electricity consumption accounted for monthly emissions of 88 ± 11 tCO₂-eq. The result highlights the importance of the “touristic ecological footprint” and sustainable tourism policies. As a priority, we must promote low-emission transport, enhance energy efficiency in tourism-related infrastructure, and manage peak electricity demand. In future research, it is necessary to incorporate indirect emissions—particularly from food, waste, and water—to enable more accurate carbon emissions assessments.