A Comparison of Methods for Anticipating and Understanding Uncertainty of Outputs from Transportation and Land Use Models

Uncertainty is important to appreciate in the outputs of complex and dynamic urban systems. This study demonstrates three methods for uncertainty propagation in transportation and land use models: Local Sensitivity Analysis with Interaction (LSAI), Monte Carlo (MC) methods, and Bayesian Melding (BM) method. Two case study settings are used to illustrate how these methods work, allowing for inter-method comparisons. LSAI can provide the sign of change implied by the changes in exogenous variables, the relative importance of the change in exogenous variables, and the decomposition of the change of output into the individual and interaction changes in exogenous variables. The computing time for multiple model runs is determined by the number of (groups of) exogenous variables. However, that approach obtains only point estimates, while MC and BM can deliver the entire distribution of each output. However, the MC and BM meethods require an understanding of the uncertainty of all model inputs and parameters. MC methods can be used to solve problems with probability structures, or non-probabilistic problems. It is straightforward to obtain a model’s output distributions via random sampling from the distributions of inputs and parameters. However, MC methods require a large number of samples (and thus full model runs), especially for more accurate results. Both LSAI and MC only solve the deterministic problems, while BM can solve both deterministic and stochastic problems, explicitly. BM can estimate the posterior distribution of outputs from prior probability distributions and likelihoods of inputs and parameters. However, BM can be extremely expensive in terms of computing time, since it requires several hundred runs of the model. Moreover, all outputs must be known at an intermediate point (in time, typically), to allow foror intermediate validation.