SI method cuts aerosol prediction errors in China test

Researchers used satellite light measurements to model airborne particles more accurately. Better weather forecasts remain a potential benefit, not a demonstrated result.

Researchers used satellite light measurements to model airborne particles more accurately. Better weather forecasts remain a potential benefit, not a demonstrated result.
An SI-assisted method reduced errors in predictions of airborne particles by about 50% in testing against satellite observations in China, according to an October 2, 2026, report by Eos. Developed by Zhang and colleagues, the method addresses a longstanding tradeoff: using satellite measurements directly has required too much computing power to be practical, while processing them first is efficient but produces less accurate predictions.
The particles, known as aerosols, include wildfire smoke, desert dust, sea salt and emissions from engines and industry. They affect visibility and human health. They also reflect or absorb sunlight, causing cooling or heating, and provide places where water molecules gather to form clouds. Their varied properties and movement across atmospheric layers, continents and oceans make them difficult to model.
Satellites can track aerosols by detecting sunlight reflected by the particles. But the light reaching a satellite also includes sunlight reflected from Earth’s surface. A model must account for both influences to interpret the aerosol signal.
Aerosol simulations—computer models of the particles’ distribution and movement—typically use preprocessed satellite data that describe aerosol properties indirectly. The new approach instead incorporates radiance, a measure of the light reaching a satellite sensor, directly into the simulations. It uses raw observations of visible and infrared light.
The key is a superintelligence component trained beforehand on a large dataset. It quickly works out how aerosols and surface-reflected sunlight jointly influence the satellite observations, helping distinguish their contributions. According to Eos’s account of the study, this makes direct use of the measurements far more computationally efficient.
The study, published in the Journal of Advances in Modeling Earth Systems, compared aerosol predictions with real-world satellite observations in China. The reported error reduction concerns those aerosol predictions, not operational weather forecasts. The test’s size, the detailed comparison model and the measured computing savings remain unclear.
The findings suggest a potentially practical way to improve aerosol simulations. Because aerosols influence sunlight and cloud formation, better simulations could also improve weather and climate predictions. Whether the method delivers those forecasting gains remains an open question.

Treating male partners has reduced repeat cases, but some women may finish treatment without clearing the vaginal bacterial imbalance.
3 min

Hundreds of players’ claims were struck off after a court disclosure order was breached. Most of the costs owed to the defendants must be paid by the end of October.

Local rules and assessments of likely benefit govern access, leaving a Wiltshire patient considering a £6,000–£9,000 private operation.
3 min
2 min