Jl. Hilliker et Jm. Fritsch, An observations-based statistical system for warm-season hourly probabilistic forecasts of low ceiling at the San Francisco International Airport, J APPL MET, 38(12), 1999, pp. 1692-1705
A computational system that uses statistical equations to forecast hourly p
robabilities of marine stratus burnoff (via ceiling) at the San Francisco i
nternational Airport for 1-6-h lead times is developed. The system is based
entirely upon surface and upper-air observations in the San Francisco Bay
Area as predictors.
A test of the product on a 3-yr independent sample shows a 6%-21% reduction
in the mean square error (mse) compared with persistence "climatology." Th
e amount of improvement is noteworthy, considering that a dearth of reliabl
e observations exists upstream from the airport. Moreover, the inclusion of
important upper-air predictors into the forecast equations can reduce the
mse by 3% when compared with a system of equations derived solely from surf
ace data. Paired-difference tests reveal that the upper-air data provide th
e greatest contribution for valid rimes that are nearest to the data's obse
rvational time.
Ceiling forecasts are compared to Model Output Statistics (MOS) probabilist
ic ceiling forecasts for two different MOS lead times. When forecasts valid
at 1500 UTC are verified, 3-h observation-based forecasts result in a 32%
reduction in mse over MOS forecasts having a 13- to 15-h lead time. When a
more competitive 4- to 6-h MOS lead time is allotted (using an 1800 UTC val
id time), 3-h observation-based forecasts result in an 8% reduction in the
mse over MOS forecasts.
Analysis of the predictive system's performance on the 3-yr independent sam
ple reveals that a broad distribution of probabilistic forecasts is produce
d, in contrast with forecasts made from persistence climatology, which can
offer only a limited probability distribution for each case. Because the pr
obabilistic forecasts are shown to be unbiased, it is expected that similar
systems designed for operational use would guide users toward more prudent
decisions on the implementation or termination of air traffic delay progra
ms.