ESTIMATING THE EXPECTED FLIGHT CANCELLATIONS DUE TO WEATHER CONDITIONS AT DALLAS FORT WORTH INTERNATIONAL AIRPORT (DFW)

Abstract

Prevailing weather conditions significantly influence flight cancellations by airlines, airports and the Federal Aviation Administration (FAA). As a result, Dallas Fort Worth International Airport’s (DFW) Operations (OPS) Division is interested in estimating the anticipated departing flight cancellations to efficiently, and safely mitigate the negative effects caused. Accurately estimating the anticipated number of departing flight cancellations well in advance can assist airport operators to plan efficiently, and safely execute flight operations with minimal impact on passenger movement, airport capacity and throughput; consequently minimizing or completely reducing the chances of irregular operations (IROPS).This study focuses on weather related departure flight cancellations. Operations and weather data associated with at least one departure cancellation from 2010 to 2015 are used for mathematical modelling, analysis and validation. The Empirical Bayes (EB) model is used to evaluate the various weather factors that significantly affect flight departure operations. Using Pearson matrix correlation, five weather factors, each with 1047 observations, (6282 entities) are determined and used in the model. The use of EB model accounts for the regression-to-mean (RTM) phenomena not accounted for by the negative binomial (NB) distributions used in the evaluation process. Both qualitative and quantitative statistical validation approaches are used to justify the results found. Actual departure cancellation DFW data from 2016 is used against the estimated results for validation. Results indicate that the number of weather related departure flight cancellations increases when the dew point temperature, the visibility distance and the cloud ceiling height decrease. Results, however, indicate that the number of weather related departure flight cancellations increases with increasing precipitation and cloud cover index. With a strong and positive correlation of R2 = 0.9869 and the means of both the estimated and actual cancellations being statistically different, this study renders the variable selection process and mathematical approach appropriate and acceptable.