UrbanSim Parcel-level Starter Model

This repository provides a starting place for parcel level UrbanSim models, and provides two working example models, with public data and configurations included in the repo itself. This version is a replacement for both urbansim_defaults and sanfran_urbansim, and works with the new UDST developer model. Works with Python 2.7 and 3.5.

The examples provided here are discussed in detail in the documentation for the new developer model.

Requirements

• urbansim
• orca
• pandana
• osmnet
• developer

Optional

A good and strongly recommended practice is the use of an isolated Python environment where to run every project, install development and pre-release versions of packages and avoid conflicts with other Python projects running on the same machine.

• Python virtual environment (virtualenv) To install: pip install virtualenv

Installation steps

We recommend running UrbanSim models in a virtual environment through conda or virtualenv. Conda virtual environment setup can follow along with steps in the Travis build file (.travis.yml).

git clone https://github.com/urbansim/urbansim_parcels
cd urbansim_parcels
python setup.py install

The San Francisco model is a simple example:

cd urbansim_parcels/sf_example
python simulate.py

The San Diego example has a more fully-featured set of models. We've included a small subset of the San Diego data and network in this repo. The full data can be downloaded at this link.

Run the subset model:

cd urbansim_parcels/sd_example
python Simulation.py

Run the full model:

cd urbansim_parcels/sd_example/data
curl -O https://dl.dropboxusercontent.com/u/69619688/sandiego_data_12042015.zip
unzip sandiego_data_12042015.zip
# replace the "store" and "net_store" variables with the correct names in settings.yaml
cd ..
python Simulation.py

Running the simulation scripts off the bat relies on pre-estimated configurations; you can also run the estimation workflows using estimate.py or Estimation.py in each example. Note: Estimation on San Diego region should be done using the full dataset, rather than the subset.