Travel demand forecasting inherently involves uncertainties that result in risks of under or over-estimating travel demand. Some of these are due to inherent uncertainties in future conditions such as population, employment and economic growth. Others are due to uncertainties in model parameters or in the structure of the models that are used for forecasting. In most major travel forecasting applications, sensitivity analyses are conducted to determine how these uncertainties affect the forecasts. However, these conventional sensitivity analyses are not formally structured in ways that determine the full risk profile associated with model forecasts. In particular, they typically vary only one or two of these uncertain variables at a time and so do not identify the interactions among these variables. In addition, they typically do not associate probability distributions with the uncertainties and thus cannot estimate the probability density associated with the end forecasts.
More advanced approaches have been used to estimate revenue risks associated with tolled highway facilities and some of these methods could be usefully applied more generally to other travel demand forecasting applications. Generally, these approaches make inferences about the probability distributions around the uncertain model inputs and parameters, determine model response to those uncertainties based on a set of sensitivity runs and then use Monte Carlo simulation to fully define the cumulative probability distribution for the model forecasts. In addition, we have very recently extended that approach using response surface methods to create a closer multivariate approximation of the travel demand model’s sensitivity using model runs that are constructed according to a statistically-efficient design of experiments. This advanced approach was used to evaluate traffic and revenue risks for two recent managed lanes projects in Florida.
This paper will describe the approaches that are commonly used to identify risk associated with urban travel demand forecasts, describe the approaches that are currently being used to identify the demand/revenue risks associated with toll facilities, detail the recent advances that could be more widely applied to provide more robust estimates of the risks associated with travel demand forecasts and describe the application of these advanced methods to recent major planning studies.