About AeroFusion

MW-AeroFusion-Net: Multi-Resolution Wavelet Adaptive Diffusion Graph ConvLSTM Transformer Network for Explainable PM2.5 Forecasting in India.

Explainable air-quality
intelligence for India

The problem

Fine particulate matter (PM2.5) penetrates deep into the lungs and bloodstream. In India, its levels swing with crop-residue burning, winter inversions, traffic and industry — and they travel hundreds of kilometres between cities. Most forecasting tools are black boxes: a number without a reason.

MW-AeroFusion-Net combines multi-resolution wavelets, an adaptive diffusion graph, ConvLSTM, a Transformer encoder and a causal attention gate to forecast PM2.5 1, 24 and 48 hours ahead — and to explain which sources drive each forecast.

SDG 3

Good health

SDG 11

Sustainable cities

SDG 13

Climate action

Team

The people building AeroFusion

Shashank Ravindra Karanam

1BM23CS312

Sohan R

1BM23CS336

Srinidhi B V

1BM23CS339

Srujan K R

1BM23CS340

Mentor

Pallavi C V

Research mentor

Integrity & data provenance

What you are looking at, stated plainly

LiveCurrent and forecast PM2.5 is the Copernicus CAMS model, with Open-Meteo weather, NASA FIRMS fires and NASA GIBS imagery, refreshed every 12 hours by the AeroFusion backend (or on demand with Refresh). CAMS is always labelled as CAMS.
MeasuredHistory and baselines are CPCB monitor measurements from the XKDR India Air Quality archive (558 stations, 2009–2026).
Not trainedMW-AeroFusion-Net has not been trained yet, so no accuracy figures are shown anywhere. Model Lab shows the real training data and its readiness instead.
ScenariosThe What-if simulator applies transparent, documented sensitivities on top of the live CAMS forecast. It is a sensitivity tool, not a model prediction.
ProxiesThere is no free India-wide traffic feed; hour-of-week and festival indicators are used and documented as proxies.