Enhancements to a Three-Dimensional Cloud Analysis Scheme

We discuss an enhanced version of the ARPS Data Analysis System (ADAS), part of the Advanced Regional Prediction System (ARPS) developed by the University of Oklahoma. The ADAS has two major components: (1) a Bratseth objective analysis scheme that evaluates potential temperature, pressure, horizontal winds, and specific humidity; and (2) a three-dimensional cloud analysis algorithm, which is based on the Local Analysis and Prediction System developed by the NOAA Forecast Systems Laboratory. ADAS builds analyses using conventional weather observations (METARs, rawinsondes, and wind profilers) as well as alternative data sources (Doppler radar reflectivity and radial winds, satellite IR and visible imagery, METAR cloud reports). These analyses can be used to initialize a numerical weather prediction model (ARPS) or for other applications.

Several improvements to the ADAS have been devised. First, our Enhanced ADAS uses a 4-km CONUS cloud analysis from the TASC Cloud Mask Generator (CMG). The CMG ingests GOES multi-spectral imagery and METARs, and applies a series of single- and multi-spectral tests to detect clouds. This provides high quality information about the distribution of clouds among the lines of site between GOES satellites and the Earth's surface; however, it does not indicate cloud altitude, thickness, or fraction.

Second, the Enhanced ADAS assimilates 10-km cloud top pressure and effective cloud amounts derived from the GOES sounder via the "CO₂ slicing" technique. This results in additional detection of high, thin cirrus clouds often missed in IR imagery. These data also provide crucial information on the vertical location and coverage of the observed clouds.

Finally, the Enhanced ADAS can assimilate cloud drift winds and H₂0 winds prepared by the University of Wisconsin/Cooperative Institute for Meteorological Satellite Studies (CIMSS). The program uses a quality control algorithm based on procedures from the Canadian Meteorological Centre and the European Center for Medium Range Forecasting. The algorithm rejects light winds, applies threshold checks to CIMSS-provided RFF and QI quality control values, and rejects winds with low elevation angles. This can provide fairly high resolution upper-air wind data to the analysis.

Comparisons of the original and Enhanced ADAS show dramatic differences in 3D clouds when the CMG and CO₂ data are used. There is a significant reduction of false detections in the horizontal, elimination of some spurious high clouds, and better detection of thin cirrus. Some significant differences to the wind analyses are also noted, particularly over the oceans.

Work is underway to use these analyses to initialize a statistical Cloud Scene Simulation Model as well as the ARPS numerical weather prediction model. Results will be presented at the conference.