Authors:
K. Srinarayani, Jaffarhameed Abdulkatherjilani, Kumaran Subramaniam, Lee Do-Hyun, Chen Po-Yu, Basant Kumar

Addresses:
Department of Computer Science in Artificial Intelligence and Machine Learning, SRM Institute of Science and Technology, Ramapuram, Chennai, Tamil Nadu, India. Department of Software Engineering, LinkedIn, San Jose, California, United States of America. Department of Information Technology, Jurong Tech Solutions, West Region, Singapore. Department of Information Technology, Hanwha Manufacturing Group, Incheon, South Korea. Department of Information Technology, Xinyi Holdings, Xinyi, Taipei, Taiwan. Department of Mathematics and Computer Science, Modern College of Business and Science, Muscat, Sultanate of Oman.

Abstract:

Climate simulation is a key tool for understanding atmospheric process dynamics and forecasting long-term climate behaviour. However, long-term simulations of high-resolution climate models are computationally costly, forcing scientists to use lower-resolution models that, in most cases, introduce systematic biases. These imprecisions compound, reducing the reliability of long-term climate forecasts, especially when used to examine extreme events such as heatwaves, heavy precipitation, and storms. This work also proposes a probabilistic machine learning framework to improve the statistical precision of long-time climate simulations using short-time, high-fidelity training data. The technique statistically learns the discrepancies between low-resolution climate simulations and accurate reference trajectories. It corrects these discrepancies non-intrusively, without altering the underlying numerical solver. The framework uses a probabilistic neural network architecture to model the distribution of correction terms, reflecting uncertainty in climate predictions. The proposed solution can improve long-term weather statistics and stabilize the initial simulation system.

Keywords: Climate Simulation; Probabilistic Machine Learning; Long-Term Weather Statistics; Probabilistic Neural Network; Initial Simulation System; Neural Network Architecture.

Received on: 28/06/2025, Revised on: 01/09/2025, Accepted on: 30/10/2025, Published on: 07/09/2026

DOI: 10.69888/FTSESS.2026.000787

FMDB Transactions on Sustainable Environmental Sciences, 2026 Vol. 3 No. 3, Pages: 153-164

• Views : 19
• Downloads : 8

Download PDF