Climate change has precipitated recurrent extreme events and emerged as an imposing global challenge, exerting profound and far-reaching impacts on both the environment and human existence. The Universal Thermal Climate Index (UTCI), serving as an important approach to human comfort assessment, plays a pivotal role in gauging how the human adapts to meteorological conditions and copes with thermal and cold stress. However, the existing UTCI datasets still grapple with limitations in terms of data availability, hindering their effective application across diverse domains. We have produced the GloUTCI-M, a monthly UTCI dataset boasting global coverage, an extensive time series spanning from March 2000 to October 2022, and a high spatial resolution of 1km. This dataset is the product of a comprehensive approach leveraging multiple data sources and advanced machine learning models. Our findings underscore the superior predictive capabilities of CatBoost in forecasting UTCI (MAE = 0.747°C, RMSE = 0.943°C, R² = 0.994) when compared to machine learning models such as XGBoost and LightGBM. Utilizing GloUTCI-M, the geographical boundaries of cold stress and thermal stress areas on a global scale were effectively delineated. Over the span of 2001 to 2021, the mean annual global UTCI registers at 17.24°C, with a pronounced upward trend. Countries like Russia and Brazil emerge as key contributors to the mean annual global UTCI increase, while countries like China and India exert a more inhibitory influence on this trend. Furthermore, in contrast to existing UTCI datasets, GloUTCI-M excels at portraying UTCI distribution at finer spatial resolutions, augmenting data accuracy. This dataset enhances our capacity to evaluate thermal stress experienced by the human, offering substantial prospects across a wide array of applications. The GloUTCI-M is publicly available at https://doi.org/10.5281/zenodo.8310513 (Yang et al., 2023).

UTCI dataset,human thermal stress,machine learning