A model-data fusion technique to retrieve three-dimensional cloud distribution

Multiple cloud layers complicate the analysis of cloud bases and cloud thicknesses needed to support a variety of aerospace missions ranging from low-level airdrops to overhead imagery collection. While satellite imagery can be used to derive locations and altitudes of the highest cloud layer in a column, it does not yield the bases and thicknesses of optically thick clouds, nor does it provide information about cloud layers beneath overcast areas. Numerical weather prediction models such as MM5 and the Global Forecast System (GFS) provide three-dimensional forecasts of meteorological fields such as winds, temperature, relative humidity, and cloud water, but cloud forecasts derived from such models can be notoriously inaccurate. To improve three-dimensional cloud analyses for aerospace applications, a cloud-data fusion (CDF) algorithm first developed at the Air Force Research Laboratory has been extended. The CDF technique can now be used to derive the cloud fraction in each of 24 pressure layers over a common model-satellite horizontal domain. The algorithm uses a satellite-derived cloud analysis to determine which pixels are cloudy and model fields to distribute the clouds in the vertical. The user can choose to employ either fuzzy logic or a diagnostic relationship (e.g. Sundquist et al. 1989) to derive fractional cloudiness in each layer. It is planned to apply the CDF technique to GFS fields and the Cloud Depiction and Forecast System-II (CDFS-II) of the Air Force Weather Agency twice per day during the upcoming Northern Hemisphere summer and fall seasons. Initial verification results will be presented.