From Data to Insights: Modeling Urban Land Surface Temperature Using Geospatial Analysis and Interpretable Machine Learning

Sensors, Год журнала: 2025, Номер 25(4), С. 1169 - 1169

Опубликована: Фев. 14, 2025

This study introduces an innovative machine learning method to model the spatial variation of land surface temperature (LST) with a focus on urban center Da Nang, Vietnam. Light Gradient Boosting Machine (LightGBM), support vector machine, random forest, and Deep Neural Network are employed establish functional relationships between LST its influencing factors. The approaches trained validated using remote sensing data from 2014, 2019, 2024. Various explanatory variables representing topographical characteristics, as well landscapes, used. Experimental results show that LightGBM outperforms other benchmark methods. In addition, Shapley Additive Explanations utilized clarify impact factors affecting LST. analysis outcomes indicate while importance these changes over time, density greenspace consistently emerge most influential attained R2 values 0.85, 0.92, 0.91 for years 2024, respectively. findings this work can be helpful deeper understanding heat stress dynamics facilitate planning.

Язык: Английский

---

Information-Guided Diffusion Model for Downscaling Land Surface Temperature from SDGSAT-1 Remote Sensing Images

Remote Sensing, Год журнала: 2025, Номер 17(10), С. 1669 - 1669

Опубликована: Май 9, 2025

Land Surface Temperature (LST) is a parameter retrieved through the thermal infrared band of remote sensing satellites, and it crucial in various climate environmental models. Compared to other multispectral bands, bands have lower spatial resolution, which limits their practical applications. Taking Heihe River Basin China as case study, this research focuses on LST data from SDGSAT-1 using three-channel split-window algorithm. In paper, we propose novel approach, Information-Guided Diffusion Model (IGDM), apply downscale image. The results indicate that downscaling accuracy image proposed IGDM model outperforms Linear, Enhanced Deep Super-Resolution Network (EDSR), Convolutional Neural (SRCNN), Discrete Cosine Transform Local Spatial Attention (DCTLSA), Denoising Probabilistic Models (DDPM). Specifically, RMSE reduced by 55.16%, 51.29%, 48.39%, 52.88%, 17.18%. By incorporating auxiliary information, particularly when NDVI NDWI inputs, performance significantly improved. DDPM, decreased 0.666 0.574, MAE dropped 0.517 0.376, PSNR increased 38.55 40.27. Overall, highlight effectiveness information-guided diffusion generating high-resolution data. Additionally, study reveals feature impact different information variations features across regions temperature ranges.

Язык: Английский

---

Spatiotemporal mapping of urban air temperature and UHI under TMY condition: A reference station based machine learning approach

Energy and Buildings, Год журнала: 2025, Номер 343, С. 115923 - 115923

Опубликована: Май 26, 2025

Язык: Английский

---