Improving cloud nowcasting via incorporation of cloud type

Cloud forecasts on a 0 – 3 hour timescale (i.e. nowcasts) are useful to a variety of disciplines. Astronomy, solar energy production and surface temperature forecasts benefit from good short-term cloud forecasts. The aviation community also benefits from these nowcasts, for instance to determine if less than Visual Flight Rules will apply.

Cloud nowcasting is a formidable problem for numerical weather forecast models. Fast techniques without spin up time are needed to address the cloud nowcast problem. In this research, nowcasting of clouds is studied using a satellite-based technique. First, a cloud mask is determined from the Moderate Resolution Imaging Spectroradiometer (MODIS) instrument onboard the NASA Aqua and Terra spacecraft. The orbits that both Terra and Aqua maintain allow for areas of Earth to be sampled by each within a three-hour window. This allows us to initialize a forecast with the Terra data, and validate it roughly three hours later with the Aqua data.

The effectiveness of a persistence forecast is first examined as a baseline forecast. The nowcast technique is augmented using other meteorological variables such as winds and water vapor content. The improvements in cloud forecast are stratified by cloud type. They hypothesis is that persistence is a more effective forecast for clouds forced by surface topography such as cumulus as compared to higher level clouds that advect with the wind. It will be shown that cloud nowcast performance is a function of cloud type. This knowledge can be used to improve existing cloud nowcast techniques, such as the ADVCLD method used by the U.S. Air Force.