Year: 2025
Type: article
Abstract: It also emerges that climatic anomalies are increasing at an alarming rate, and the current demand is for advanced prediction systems that can detect anomalies as they occur and respond to them. Incorporating data fusion methods with AI, this study proposes an original Adaptive Atmospheric Fusion Network (AAFN) to assist in enhancing the discovery of climate change and prediction. To make the data on climate more representative, the proposed system will incorporate a Recurrent Temporal Fusion Network (RSTFN) to combine data from multiple sources, including IoT sensors, satellite images, and meteorological stations. To improve feature extraction and the prediction of extreme weather conditions with greater precision, a Dual Attention LSTM Transformer (DAL-T) model is also employed. Lightweight AI models are deployed on edge computing nodes to provide instantaneous anomaly alerts, minimizing latency and enabling quicker decision-making. The system also employs Federated Learning, applying it to collaboratively train the model using decentralized data sources, thereby strengthening the models while preserving the privacy of the data. The application of XAI methods enables an AI-based visualisation dashboard to provide meteorologists and policymakers with transparency on a list of actionable insights, enabling them to respond to climate risks accordingly. It has been experimentally demonstrated that AAFN can identify climate anomalies with an accuracy of up to 98%, surpassing existing models in terms of response time and Accuracy. This novel approach aims to mitigate climate risk faced by organizations through real-time prediction and environmental resilience.
DOI: 10.14419/9svg1s46
Authors: Ragini Kushwaha, Anjali Krushna Kadao, Praveen Kumar Joshi
Institution: Kalinga University
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: diamond
OpenAlex ID: W4415764103
FWCI:
Cites:
Cited by:
Related to: