A weighted autoregressive model to improve mobile emissions estimates for locations with spatial dependence

For air quality analysis, travel demand model output must be converted to the fine-grained spatial and temporal resolution required by the photochemical models. In particular, network link volume estimates by modeling period (i.e., AM, PM, off-peak) must be converted to hourly profiles. New methods for accomplishing this have not taken spatial autocorrelation into account. This research proposes a weighted autoregressive model for estimating hourly profiles when geographic ordering (i.e., links on a network) is present. An iterative maximum likelihood estimation procedure is developed and used to estimate allocation factors, which disaggregate period-based travel demand model assignments into hourly profiles, for San Diego. Three different forms of the weight matrix, W1 (binary), W2 (reciprocal distance), and W3 (combination of W1 and W2), are tested in the analyis. The results indicate that autocorrelation from geographic ordering can be eliminated with the proposed technique.