Model Design
The BKRCast model is a localized implementation of the SoundCast activity-based (AB) model. The AB model employs DaySim to generate demand for resident travel throughout the region. DaySim includes a set of model components that forecasts residents' activity generation, destination, mode, and time-of-day choices; DaySim also includes additional models such as household vehicle availability. DaySim's forecasts of resident demand are combined with forecasts of auxiliary demand from trucks, nonresidents, special generators, and group quarters; the BKRCast model then assigns combined demand to roadway and transit networks to produce estimates of network performance. The model system executes iteratively with feedback to achieve stable, equilibrated results. The figure below illustrates the overall model system flow.
In preparation for a run, the BKRCast model first copy all files from base_inputs folder to a number of other folders, if run_copy_input_files is set to True. It performs accessibility calculations and generates a buffered parcel file for DaySim. The employment data (raw parcel file) used in accessibility calculations is copied from a secure location at the start of a model run and deleted afterwards. The model system then sets up various directories and inputs, which includes creating Emme bank and project folders, and copying DaySim code and large inputs to appropriate locations. Once the setup is in place, the model system imports highway and transit networks are to Emme database for network models. The model system also runs a simple bike model to estimate bike volumes on the network. The model system then assigns input DaySim seed trips to the network and generates an initial set of impedance for demand models to use.
The model system executes iteratively after validating demand model inputs. First, the model system runs only work and school location choice models in DaySim to build stable work and school location shadow prices. After three iterations of the shadow price run, demand estimates are produced. There are two primary types of demand components in the model system: DaySim resident demand and auxiliary demand (trucks, nonresidents, special generators, and group quarters). DaySim predicts the daily activity patterns of all regional residents when they travel within the region. The auxiliary models predict other components of the overall travel demand, such as truck demand, external demand, special generators including airport demand, and group quarters demand. The model system combines the DaySim -generated and auxiliary-models demand, assigns it to highway and transit networks, and generates revised estimates of network impedance. These revised impedance are then fed back into DaySim and the auxiliary models to produce new demand estimates. The model system uses successive averaging to achieve equilibrated results.
After three feedback loops or user-defined model convergence, whichever is reached first, the model system produces detailed reports containing summaries of model outputs. In the end before completing the run, the intermediate files are organized/cleaned-up.
The chart below shall depicts how modules are related in terms of file inputs and outputs.
