Satellite-Driven Machine Learning Approaches for Rainfall Forecasting in Support of Smart Irrigation: The Case of Ankara and Samsun
Fırat Üniversitesi Mühendislik Bilimleri Dergisi · 29.03.2026 · 38 (1) · 213-230
DOI: 10.35234/fumbd.1790288 ↗Kurz gefasst
Fachartikel in der Zeitschrift Fırat Üniversitesi Mühendislik Bilimleri Dergisi, online erschienen am 29.03.2026. Verfasst von Esma Nur Çakir und 3 weiteren Autor:innen. Die Arbeit nennt 51 Literaturangaben.
Abstract (Originalsprache)
Accurate rainfall forecasting is essential for agriculture, water management, and disaster preparedness, and its importance is further amplified by increasing climate variability. Traditional approaches often fail to adequately capture the nonlinear and highly variable nature of rainfall, particularly at the local scale. This study investigates the effectiveness of machine learning techniques for daily rainfall prediction in Ankara and Samsun, Türkiye, to support smart irrigation systems. Rainfall time series derived from the PERSIANN satellite dataset (2020–2024) were enriched with lagged values, rolling statistics, dry-spell indices, and seasonal encodings. Multiple models, including Linear Regression, Random Forest, Support Vector Regression (SVR), Gradient Boosting, and XGBoost, were comparatively evaluated. The findings indicate that model performance varies depending on both the selected algorithm and climatic conditions. SVR achieved the lowest mean absolute error but exhibited instability under extreme conditions, whereas XGBoost provided more balanced performance in both magnitude and occurrence prediction. No single model consistently outperformed others, highlighting the necessity of ensemble and hybrid approaches. This study presents a systematic city-scale evaluation of ML-based rainfall forecasting in Türkiye, integrating satellite-derived data with application-oriented modeling. However, challenges remain in modeling rare events, improving interpretability, and ensuring computational efficiency for real-time applications.
Zitieren
APA
Çakir, E. N., Demirok, İ. B., Koçdeviren, O., & Cömert, Z. (2026). Satellite-Driven Machine Learning Approaches for Rainfall Forecasting in Support of Smart Irrigation: The Case of Ankara and Samsun. Fırat Üniversitesi Mühendislik Bilimleri Dergisi, 38(1), 213-230. https://doi.org/10.35234/fumbd.1790288
Vancouver
Çakir EN, Demirok İB, Koçdeviren O, Cömert Z. Satellite-Driven Machine Learning Approaches for Rainfall Forecasting in Support of Smart Irrigation: The Case of Ankara and Samsun. Fırat Üniversitesi Mühendislik Bilimleri Dergisi. 2026;38(1):213-230. doi:10.35234/fumbd.1790288
BibTeX
@article{akir2026,
author = {Çakir, Esma Nur and Demirok, İbrahim Buğra and Koçdeviren, Onur and Cömert, Zafer},
title = {Satellite-Driven Machine Learning Approaches for Rainfall Forecasting in Support of Smart Irrigation: The Case of Ankara and Samsun},
journal = {Fırat Üniversitesi Mühendislik Bilimleri Dergisi},
year = {2026},
volume = {38},
number = {1},
pages = {213--230},
doi = {10.35234/fumbd.1790288}
}
EUVIMED