632 Improving a Pyramidal Image Matching Short-Term Cloud Forecast System by Including Normalized IR Imagery and Surface Observations

Tuesday, 9 January 2018
Exhibit Hall 3 (ACC) (Austin, Texas)
Daniel B. Kirk-Davidoff, UL, Albany, NY; and S. H. Young and J. Zack

Short-term cloud forecasting is crucial to solar energy generation forecasting and has important applications in high-resolution satellite surveillance and in agricultural forecasting. In our solar generation forecasting, we have relied on a Pyramidal Image Matching (PIM) approach that provides good skill up to several hours in advance. However, the technique requires several good sequential cloud images to initialize its implicit estimates of cloud motion. Because we have in the past used only visible imagery, early morning forecasts suffered degraded quality due to low dynamic range in early morning visible imagery at low sun-angle.

Since our clients require frequent delivery of very short lead-time forecasts, at five minute time resolution, our methods must be computationally efficient. In this work, we demonstrate the use of three IR channels to simulate visible cloud imagery during hours before and shortly after dawn. The simulated visible imagery is then fed in to the PIM algorithm. A climatology of clear-sky brightness temperature for the past several days is first developed, supplemented by an elevation-based model of ground temperature where clear days are infrequent. Cold anomalies from this background state are correlated with visible reflectance during daytime calibration periods. We find that linear functions of brightness temperature imagery are able to simulate visible reflectance for mid- and high-level clouds, but do not reliable reproduce fog. To address this problem we are developing a method that combines mesoscale observations of dew point depression and visibility with an IR image analysis algorithm that depends on the warming evident in IR imagery of foggy scenes as the fog develops. We will present results of this work for stations in California and North Carolina, and explore how the additional time and space resolution afforded by GOES-16 impacts the accuracy of the technique.

- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner