![[2602.19654] NEXUS : A compact neural architecture for high-resolution spatiotemporal air quality forecasting in Delhi Nationa Capital Region](https://arxiv.org/static/browse/0.3.4/images/arxiv-logo-fb.png){kind=logo}
[2602.19654] NEXUS : A compact neural architecture for high-resolution spatiotemporal air quality forecasting in Delhi Nationa Capital Region
Summary
The paper presents NEXUS, a compact neural architecture designed for high-resolution air quality forecasting in Delhi NCR, achieving impressive predictive performance with minimal parameters.
Why It Matters
Air pollution is a critical issue in megacities like Delhi, impacting public health. NEXUS offers a novel approach to accurately forecast air quality, which can inform policy and public health decisions. Its efficiency and effectiveness make it a valuable tool for real-time monitoring.
Key Takeaways
- NEXUS architecture forecasts CO, NO, and SO2 with R² values over 0.9.
- It uses only 18,748 parameters, significantly fewer than competing models.
- The model identifies key meteorological factors affecting air quality.
- It integrates advanced techniques like patch embedding and adaptive fusion.
- NEXUS enables real-time deployment for air quality monitoring systems.
Computer Science > Machine Learning arXiv:2602.19654 (cs) [Submitted on 23 Feb 2026] Title:NEXUS : A compact neural architecture for high-resolution spatiotemporal air quality forecasting in Delhi Nationa Capital Region Authors:Rampunit Kumar, Aditya Maheshwari View a PDF of the paper titled NEXUS : A compact neural architecture for high-resolution spatiotemporal air quality forecasting in Delhi Nationa Capital Region, by Rampunit Kumar and Aditya Maheshwari View PDF HTML (experimental) Abstract:Urban air pollution in megacities poses critical public health challenges, particularly in Delhi National Capital Region (NCR) where severe degradation affects millions. We present NEXUS (Neural Extraction and Unified Spatiotemporal) architecture for forecasting carbon monoxide, nitrogen oxide, and sulfur dioxide. Working with four years (2018--2021) of atmospheric data across sixteen spatial grids, NEXUS achieves R$^2$ exceeding 0.94 for CO, 0.91 for NO, and 0.95 for SO$_2$ using merely 18,748 parameters -- substantially fewer than SCINet (35,552), Autoformer (68,704), and FEDformer (298,080). The architecture integrates patch embedding, low-rank projections, and adaptive fusion mechanisms to decode complex atmospheric chemistry patterns. Our investigation uncovers distinct diurnal rhythms and pronounced seasonal variations, with winter months experiencing severe pollution episodes driven by temperature inversions and agricultural biomass burning. Analysis identifies critical mete...