Machine Learning–Based Prediction of Temperature and Precipitation Using Historical Meteorological Time Series Data

Zoljoodi, Mojtaba; Souri Laki, Mohammad; Alipour, Maryam Mary

doi:10.22034/ijcoe.2026.606852.1290

Machine Learning–Based Prediction of Temperature and Precipitation Using Historical Meteorological Time Series Data

Articles in Press

Document Type : Original Research Article

Authors

Mojtaba Zoljoodi ¹

Mohammad Souri Laki ²

Maryam Mary Alipour ³

¹ Faculty member of Research Institute of Meteorology and Atomespherice Science (RIMAS)

² Kharazi Highway, Pezhohesh

³ Research Institute of Meteorology and Atomespherice Science (RIMAS)

10.22034/ijcoe.2026.606852.1290

Abstract

Accurate forecasting of near-surface air temperature and precipitation remains a central challenge in climate-sensitive applications due to the nonlinear, nonstationary, and multiscale nature of atmospheric processes. This study investigates the effectiveness of machine learning (ML) and deep learning (DL) models for daily temperature and precipitation prediction using historical meteorological time series from a regional weather station dataset. Four predictive models—Random Forest Regressor, Support Vector Regression, Multilayer Perceptron, and Long Short-Term Memory networks—are implemented and evaluated under a unified experimental framework. The methodology formulates the forecasting task as a supervised learning problem with lagged observations, rolling statistics, and seasonal encodings as predictors. Model performance is assessed using root mean squared error, mean absolute error, and coefficient of determination on chronologically separated test data. Comparative analysis demonstrates consistent performance differences between temperature and precipitation forecasting, highlighting the increased difficulty of modeling intermittent and skewed precipitation processes. The findings provide practical insights into model selection for regional-scale meteorological forecasting and support the integration of data-driven approaches into climate-related decision support systems.

Keywords

Machine learning

Time series forecasting

Temperature prediction

Precipitation prediction

LSTM

Subjects