DustNet: Skillful Neural Network Predictions of Saharan Dust

Suspended in the atmosphere are millions of tonnes of mineral dust whichinteracts with weather and climate. Accurate representation of mineral dust inweather models is vital, yet remains challenging. Large scale weather modelsuse high power supercomputers and take hours to complete the forecast. Suchcomputational burden allows them to only include monthly climatological meansof mineral dust as input states inhibiting their forecasting accuracy. Here, weintroduce DustNet a simple, accurate and super fast forecasting model for24-hours ahead predictions of aerosol optical depth AOD. DustNet trains in lessthan 8 minutes and creates predictions in 2 seconds on a desktop computer.Created by DustNet predictions outperform the state-of-the-art physics-basedmodel on coarse 1 x 1 degree resolution at 95to ground truth satellite data. Our results show DustNet has a potential forfast and accurate AOD forecasting which could transform our understanding ofdust impacts on weather patterns.