Spatiotemporal Analysis of Urban Surface Cover Structure in Ho Chi Minh City from 2015 to 2025: A Big Data and Machine Learning Approach

Tóm tắt

Land-use structure transformation in megacities such as Ho Chi Minh City (HCMC) not
only reflects rapid economic growth but also constitutes a fundamental driver of geohazards,
particularly land subsidence caused by increasing static and dynamic loads. To quantitatively
assess this process, the study developed an automated monitoring framework on the Google Earth
Engine (GEE) platform, integrating the Random Forest algorithm to process multi-temporal
satellite imagery from Landsat 8/9 and Sentinel-2 over 11 years (2015–2025). Accuracy
assessment results indicate robust classification performance, with Kappa coefficients ranging
from 0.85 to 0.96 and Overall Accuracy between 88.1% and 97.4%. The findings reveal a clear
expansion of built-up impervious surfaces, increasing from 5,500.45 ha in 2015 to 6,395.12 ha in
2025. The study successfully captured the spatiotemporal dynamics of five major land-cover
classes, highlighting the pronounced growth of “built-up impervious surfaces” and the complex
fluctuations of “bare land,” which reflect different construction preparation stages. Statistical
analysis shows a strong spatial correlation between impervious surface expansion and areas
identified as subsidence-prone. The resulting dataset provides reliable input data for geotechnical
models, enabling clearer differentiation between static structural loads and dynamic traffic loads
in ground deformation prediction.