Dallas Fort Worth International Airport Traffic Microsimulation Model

Problem: Lower traffic levels resulting from the COVID crisis provided DFW airport with the opportunity to complete maintenance, repairs, and construction of infrastructure. To determine the impacts under construction conditions, vehicular traffic on the airport access roadways needed to be predicted under COVID conditions.

Solution: A form of Machine Learning which uses the general principles of Boosted Decision Tree Regression was used. Data sets: Passenger data pre-COVID and during COVID; TSA throughout dataset; and R values¹ and Inference R value projections from an ensemble of published SEIR² Epidemiology Models.

The “Inference Projections” or the projections resulting from expected state intervention levels (No Action Taken, Social Distancing, or Shelter-in-Place) based on current intervention trends³ were used to model the propensity to travel, which helps forecast the passenger levels in the near term. Load Factors and Connecting Ratios forecasted using Machine Learning models together with airline schedules were used to estimate the originating and terminating passengers.

Resulting vehicles were determined by applying show-up profiles (how long before flight time passengers arrive at airport curbside or how long after the plane arrives do passengers get to the curbside), adjusted for the social distancing measures. These forecasted vehicle volumes were adjusted for changes in mode shares based on observed mode shifts during the pandemic. These volumes were input into microsimulation models to determine if there would be any adverse impacts to the roadways during construction. The forecasts were tested for accuracy by comparing them with daily new actual data, and the model performed within acceptable tolerance limits.

Impact: The forecasted vehicle volumes were found to be 37% lower as a result of the pandemic, and the simulation showed that there will be no adverse impacts to the passengers. This resulted in much simpler traffic control plans that were easier to implement because much lower staffing levels were required and fewer signage/striping changes were needed.