Evaluation of short duration nowcasting algorithms for optimizing projected runway visual range (RVR) conditions

This paper examines the attributes of simple algorithms to determine the potential of utilizing Runway Visual Range values (RVR) calculated from previously measured extinction coefficients to project RVR over different future short-term time intervals. The focus is on examining RVR conditions during the two-minute future interval considered to be most applicable to aircraft on final approach. This time period is considered representative of the time it takes for a modern aircraft to reach the touchdown point on the runway from the outer marker which ranges from round 4- to 7-nm from the runway threshold. Algorithms for forecasting RVR for time intervals ranging from 1-7 min are also desired for use in coasting visibility sensor (VS) readings in the event of a VS outage. This capability would conceivably ensure safe operations when equipment loss of service occurs during aircraft final approach or planned takeoff.

VS data, covering many types of natural weather conditions that result in decreased RVR, are used for the study. The algorithms consist of backward averaging over intervals ranging from 2-5 min and using linear regression over 3-min, 5-min and 9-min time intervals. The results are compared relative to simple data coasting by comparing concurrency statistics of predicted versus actual RVR. The results are given for all the events considered as well as for subsets of events such as snow, fog and by the occurrence of the highest RVR Category of the event. They show that the best predictor depends on the characteristics of the event. Tradeoffs in the frequency of missed alerts and false alerts with differing prediction basis times are also examined. Means of ensuring safe operations are also explored.