General UrbanSim Block Model Questions¶
Who is the intended audience for the census block model?¶
The census block-level model is a forecasting tool aimed at regional planning agencies. In particular, it targets Metropolitan Planning Organizations (MPOs) that need a zonal land use forecast for their Regional Transportation Plan (RTP), may have limited staff time, and don’t want to undergo a multi-year model development effort in order to have a running UrbanSim model.
What is UrbanSim?¶
UrbanSim is a behavioral simulation of regional real estate markets. It relies on statistical models estimated off of observed data. Individual agents such as households, employers, and real estate developers are simulated over time. See the UrbanSim Overview page for more detailed information.
What spatial scale does the model operate at?¶
This depends on the model template being used. The block model utilizes a simulation engine that operates at the census block-level, the parcel model utilizes a simulation engine that operates at the building and parcel level, and the zone model utilizes a simulation engine that operates at the zonal-level.
How long does a typical UrbanSim model region take to simulate?¶
At the census block level, smaller regions run typically in less than 60 seconds per year, while larger regions of 3-5 million population may take 3-5 minutes per year to simulate. Run times are sensitive to the number and complexity of scenario inputs and the size and complexity of any transportation networks used.
How is model calibration achieved?¶
For the latest census block model: Models are being calibrated to observed growth trends at aggregate geographies (i.e. county and Public Use Microdata Area (PUMA)). Households and units are calibrated using 2013-2018 ACS data and jobs are calibrated using 2010-2017 LEHD data. Location variables corresponding to the calibration geography are introduced into the model specifications, and their coefficients are then iteratively adjusted so as to ensure the simulation matches observed growth trends within some margin. An additional place type based calibration routine is available upon request. For the block model see here for more details.
How is model validation achieved?¶
For the latest census block model: Simulated tract-level change in household growth, employment growth, and residential unit growth between 2010 to 2013 is compared with observed data on tract level changes. Households and units are validated using 2018-2019 ACS data and jobs are validated using 2017-2018 LEHD data. The difference between simulated and observed change is then visually assessed, and metrics are calculated such as root-mean-square error (RMSE) and correlation coefficient.
Can I assert area-specific outcomes in the model? If so, how is this achieved?¶
Yes. Model adjustments can be specified through the Adjustments dialog box. Adjustments specify the number of agents of a certain type (e.g. households with income between $30k and $60k) within a chosen set of geographies. UrbanSim will then assert those adjustments in simulation.
Are development constraints allowed? If so, how is this achieved?¶
In the current census block-level model, employment and residential unit capacities can be specified per census block and will constrain development within the UrbanSim model. With this capability you can set new capacities for blocks to take effect in later years in order to, for example, upzone blocks surrounding a new transit station in the year 2025. See block-level constraints and base year capacity data pages for more information.
How are zoning data applied to blocks?¶
In the census block-level model, uploaded zoning shapefiles are intersected with blocks and their density attributes are applied to those blocks proportionally. This results in a single employment and a single residential unit capacity for each block. The resulting capacity values become a set of constraint records.