An Automated Low Cloud Prediction System

An Automated Low Cloud Prediction System

At the Air Force Geophysics Laboratory (AFGL) Weather Test Facility (WTF) at Otis AFB, MA, a network of cloud base height, visibility, and wind measuring instruments were used to explore techniques for the short range prediction of low cloud ceiling. AFGL developed this system in response to the USAF Air Weather Service's requirements to modernize its basic weather support capabilities. This system allowed AFGL to evaluate the ability of statistical forecasting techniques to provide decision assistance significantly improved over the decision assistance currently provided by climatology and persistence. The approach relies upon the use of a hierarchical clustering algorithm to transform the raw cloud base height data into an automated low cloud observation. Four prediction techniques (Regression Estimation of Event Probabilities, Equivalent Markov, climatology, and persistence) yielding probability estimates of low cloud ceiling were evaluated and comparisons made to determine their respective accuracy and reliability. In addition, thresholding techniques were used to convert probability forecasts (unit bias, maximum probability, iterative, and persistence). Analysis of the data collected at the AFGL WTF demonstrates the accuracy and reliability of the automated low cloud prediction system. Regression estimation of event probabilities provided accurate, reliable, high resolution probability forecasts with results superior to climatology, persistence, and Equivalent Markov.

Scientific and Technical Aerospace Reports

An Automated Low Cloud Prediction System

At the Air Force Geophysics Laboratory (AFGL) Weather Test Facility (WTF) at Otis AFB, MA, a network of cloud base height, visibility, and wind measuring instruments were used to explore techniques for the short range prediction of low cloud ceiling. AFGL developed this system in response to the USAF Air Weather Service's requirements to modernize its basic weather support capabilities. This system allowed AFGL to evaluate the ability of statistical forecasting techniques to provide decision assistance significantly improved over the decision assistance currently provided by climatology and persistence. The approach relies upon the use of a hierarchical clustering algorithm to transform the raw cloud base height data into an automated low cloud observation. Four prediction techniques (Regression Estimation of Event Probabilities, Equivalent Markov, climatology, and persistence) yielding probability estimates of low cloud ceiling were evaluated and comparisons made to determine their respective accuracy and reliability. In addition, thresholding techniques were used to convert probability forecasts (unit bias, maximum probability, iterative, and persistence). Analysis of the data collected at the AFGL WTF demonstrates the accuracy and reliability of the automated low cloud prediction system. Regression estimation of event probabilities provided accurate, reliable, high resolution probability forecasts with results superior to climatology, persistence, and Equivalent Markov.