Spatio-Temporal Graph Transformer Network for Climate Change Impact Prediction and Extreme Weather Forecasting
International Journal of Advanced Research in Science Communication and Technology · 26.01.2024 · 797
DOI: 10.48175/ijarsct-15100r ↗Kurz gefasst
Fachartikel in der Zeitschrift International Journal of Advanced Research in Science Communication and Technology, online erschienen am 26.01.2024. Verfasst von Naga Venkatesh Gangabathula. Die Arbeit nennt 14 Literaturangaben.
Abstract (Originalsprache)
Climate change accelerates extreme weather events, threatening global ecosystems and human populations with unprecedented frequency and intensity. Traditional climate models rely on grid-based numerical simulations that struggle with complex spatio-temporal dependencies and fail to capture non-linear interactions across heterogeneous geographical regions. We propose a Spatio-Temporal Graph Transformer Network for Climate Impact Prediction (STGT-CIP) that integrates multi-source climate data using a novel graph-based deep learning architecture. Our framework employs three analytically distinct components: (1) a Multi-Modal Climate Graph Constructor that builds dynamic spatio-temporal graphs from satellite imagery, sensor networks, and reanalysis datasets; (2) a Hierarchical Graph Transformer Encoder that captures long-range dependencies and cross-regional climate interactions through multi-head self-attention mechanisms; and (3) an Extreme Event Prediction Module using temporal attention and uncertainty quantification to forecast extreme weather occurrences with calibrated confidence intervals. Validated on global climate datasets spanning 1980-2024 (N=45,000 geographical nodes), STGT-CIP achieves 94.7% accuracy in temperature anomaly prediction, 92.3% precision in extreme rainfall detection, and reduces forecast error by 34.6% compared to conventional numerical weather prediction models. The framework provides interpretable predictions through attention visualization, highlighting critical climate teleconnections and regional vulnerability patterns, thereby supporting evidence-based climate adaptation and disaster preparedness strategies..
Zitieren
APA
Gangabathula, N. V. (2024). Spatio-Temporal Graph Transformer Network for Climate Change Impact Prediction and Extreme Weather Forecasting. International Journal of Advanced Research in Science Communication and Technology, 797. https://doi.org/10.48175/ijarsct-15100r
Vancouver
Gangabathula NV. Spatio-Temporal Graph Transformer Network for Climate Change Impact Prediction and Extreme Weather Forecasting. International Journal of Advanced Research in Science Communication and Technology. 2024:797. doi:10.48175/ijarsct-15100r
BibTeX
@article{Gangabathula2024,
author = {Gangabathula, Naga Venkatesh},
title = {Spatio-Temporal Graph Transformer Network for Climate Change Impact Prediction and Extreme Weather Forecasting},
journal = {International Journal of Advanced Research in Science Communication and Technology},
year = {2024},
pages = {797},
doi = {10.48175/ijarsct-15100r}
}
EUVIMED