Forecasting Climate Extremes-A Data Driven Approach

Volume 19, Issue 2,  2025

Download

Author(s):

Rana Muhammad Ebrahim Usman Institute of Technology, Karachi, Pakistan, 21B-022-CS@alumni.uit.edu

Muhammad Taha Pasha Usman Institute of Technology, Karachi, Pakistan, 21B-039-CS@alumni.uit.edu

Hanzallah Ahmed Khan Usman Institute of Technology, Karachi, Pakistan, 21B-148-CS@students.uit.edu

Muhammad Faraz Khan Usman Institute of Technology, Karachi, Pakistan, 21B-053-CS@alumni.uit.edu

Muhammad Umer Usman Institute of Technology, Karachi, Pakistan, 21B-120-CS@students.uit.edu

Muhammad Wasim Usman Institute of Technology, Karachi, Pakistan, mwasim@uit.edu

Lubaid Ahmed* Usman Institute of Technology, Karachi, Pakistan, lahmed@uit.edu

Abstract Climate extremes such as heavy rainfall, droughts, and heatwaves pose an increased risk to human life, agriculture, buildings, and nature. Currently Pakistan relies on traditional statistical models that are built on limited station-level data, while machine-learning-based forecasts remain underutilized. The objective of this paper is to introduce a provincial-based approach for accurate and precise short-term extreme weather forecasts adapted to the unique climatic conditions of Pakistan. The research uses the Extreme Gradient Boosting algorithm along with 25 years of historical weather data of Pakistan retrieved from Numerical Weather Prediction models, specifically, GFS—Global Forecast System and ICON—Icosahedral Nonhydrostatic Model, in improving the prediction of heatwaves, droughts, and heavy rainfall events. The models showed excellent performance in identifying heatwaves, classifying droughts, and identifying rainfall severity. The results show the potential of refining data from physics-based weather models with machine learning models to significantly improve forecasts of climate extremes, filling a vital gap in Pakistan's weather prediction landscape. This approach would prove beneficial to emergency management agencies in disaster preparedness and response as well as to the general public to make better decisions.
Keywords Climate extremes prediction, Machine Learning, XGBoost, Pakistan Climate, Heatwave prediction, Drought classification, Rainfall prediction.
Year 2025
Volume 19
Issue 2
Type Research paper, manuscript, article
Journal Name Journal of Information & Communication Technology
Publisher Name ILMA University
Jel Classification -
DOI -
ISSN no (E, Electronic) 2075-7239
ISSN no (P, Print) 2415-0169
Country Pakistan
City Karachi
Institution Type University
Journal Type Open Access
Manuscript Processing Blind Peer Reviewed
Format PDF
Paper Link https://jict.ilmauniversity.edu.pk/journal/jict/19.2/3.pdf
Page 66-72