Operational concept for observation-based forecasting of clear sky condition

Sky condition forecasting remains a challenging problem for meteorologists. Empirical, observation-based techniques are known to be useful for short-term (0-6 hr) forecasting. In recent years, the ability to store and efficiently process large amounts of data have facilitated the research and development of new, data-driven approaches to predictive learning in the atmospheric sciences. In this study, we substantiate the potential to apply statistical and machine learning schemes to archived time-series of meteorological satellite data along with atmospheric parameters extracted from model data assimilation-based analyses to generate sky condition (cloud-free) forecasts. Based on our findings, we present a new concept for an operational, observation-based cloud prediction system that is relevant to a variety of civil, commercial and military communities ranging from aviation to the energy industry.