Low-level stratus prediction using binary statistical regression : a progress report using Moffett Field data

Abstract

Various statistical models and techniques were employed to forecast the existence of low-level stratus conditions. They are illustrated for data at a single station (Moffett Field, Sunnyvale, California) using single-station surface meteorological measurements only as explanatory variables. A preliminary exploratory data analysis shows that low (high) dew point depression is associated with the existence (non-existence) of low-level stratus at Moffett Field. Procedures for and results of various methods of fitting logistic models to the data are described. The fitted models were used to forecast stratus op reserved data sets (cross-validation). Results of the crossvalidation are given.