Thermal changes along the urban-rural continuums in Southeast Asia

This research has been published in Environmental Research Letters. Please cite it as shown below when using this dataset:

Citation: Zhu, Y., Myint, S., Chen, J., Fan, P., Seto, K., Jain, A. K., Qi, J., & Wang, J. (2025). Thermal changes along the urban-rural continuums in Southeast Asia. Environmental Research Letters. https://doi.org/10.1088/1748-9326/adcad2.

Our study focused on 19 cities across 8 countries: Cambodia, Indonesia, Lao PDR, Malaysia, Myanmar, Philippines, Thailand, and Vietnam. The cities included Phnom Penh, Siem Reap, Jakarta, Surabaya, Denpasar, Vientiane, Pakse, KL, Putrajaya, Yangon, NPT, Quezon City, Iloilo, Taytay, Bangkok, Chiang Rai, HCMC, Hanoi, Cantho. 

To assess the impact of urbanization over the past two decades, we examined the patterns of Land Surface Temperature (LST) changes with Land Use and Land Cover (LULC) changes across 19 cities in SEA along the urban-rural continuums (URCs). These cities include major urban hubs such as Jakarta, Bangkok, and Ho Chi Minh City, as well as medium and smaller cities like Vientiane and Chiang Rai, reflecting a diversity of urban environments and providing a comprehensive sample of SEA’s rapid urban expansion and associated thermal impacts. The boundaries of URCs for each city were defined as an area centered within a 30 km radius from its city center point. This ensured that all cities had the same extent so that the URCs could make the comparison. The city center point was defined as either the center of the central business district or the geometric center of the city’s administrative limits (Estoque et al., 2017). We created 30 ring buffer zones around each city center point at 1 km intervals to analyze LULC and the associated LST change gradient along the URCs. After establishing these buffer zones, we excluded (1) large water bodies or oceans, (2) continuous urban areas beyond the city boundaries—for cities located close to one another, such as Kuala Lumpur and Putrajaya, and Quezon City and Taytay, the city’s URCs were adjusted to exclude overlapping administrative boundaries from neighboring cities, ensuring that the analysis was confined to each city’s specific URCs, and (3) mountains with elevations exceeding 100 m above the mean elevation of each city buffer to minimize the confounding effects of topography on LST variations to ensure that the LST changes we observed were more directly attributable to urban development rather than elevation-related climatic variations. (C. Wang et al., 2016; Z. Wang et al., 2020). These exclusions were implemented to ensure a fair comparison, avoid confusion, and reduce the influence of elevation on the analysis.

For URCs, we retrieved LST and NDVI data for the summer months (June, July, and August) between 2000 and 2022 at a 30-meter resolution based on the Google Earth Engine Platform. This period was chosen due to peak heat-related mortality and morbidity (Hsu et al., 2021; Johnson et al., 2009). A total of 5,805 Landsat 5/7/8/9 scenes were collected to ensure full coverage of all 19 cities. Contaminated pixels (e.g., clouds and cloud shadows) were removed via quality assessment bands and Landsat-7 SLC-off stripes were eliminated. NDVI was calculated from the surface reflectance of the Red and Near-Infrared (NIR) bands. LST retrieval from Landsat data was conducted using the Statistical Mono-Window (SMW) algorithm developed by the Climate Monitoring Satellite Application Facility (CM-SAF) (Duguay-Tetzlaff et al., 2015; Freitas et al., 2013; Sun et al., 2004). Further details are provided in the Supplemental materials.

The final datasets are LST in 2022, LST Sen's Slope (2000-2022), and NDVI Sen's Slope (2000-2022) and are shared.