Parallel work

There are literally scores of different land use and land use-transport modeling systems reported in the literature, and a few that we know of that aren't highlighted in journals. Many (most?) have not been used to inform public policy in a formal sense. The projects described here are only those that the TLUMIP team are familiar with, have found innovative and inspiring, and that are used in practice.

LUSDR

One model that we're very high on is LUSDR (Land Use Scenario DeveloperR), developed by Brian Gregor at the Oregon DOT. LUSDR is a stochastic microsimulation of household and business location to support land use and transportation planning. Brian developed it from scratch to be a sketch planning land use model. It incorporates a lot of current ideas in land use modeling, including a highly graphical user interface. The package is written in the R statistical language (see below). Several resources are available for download:

• Model documentation
• A PowerPoint presentation (in Adobe Acrobat format) of an overview of the model and application in Medford, Oregon
• A paper, "Land Use Scenario DevelopeR: Practical land use modeling using a stochastic microsimulation framework," that was presented at the 2006 TRB annual meeting and appeared in Transportation Research Record 2003 (2007)

Further information about LUSDR can be obtained by contacting Brian (see the Contacts page).

UrbanSim and PECAS

The TLUMIP work has helped launch two other land use modeling systems: UrbanSim and PECAS. The first generation TLUMIP work focused on conducting a proof-of-concept using the TRANUS package (links below). Paul Waddell, then part of the TLUMIP team, also started on the development of an entirely new dynamic disequilibrium land use model. A prototype of this first generation of UrbanSim was tested in Eugene-Springfield. Since then Paul has extensively overhauled and enhanced the package at the University of Washington.

The PECAS package is evolving from the work on the second generation models, based upon work by Doug Hunt and John Abraham of the University of Calgary. PECAS has evolved from the PI (production allocation and activity interaction) module, which allocates households and businesses to small geographic areas within the study area. A webpage for PECAS is not in place yet, but further information can be obtained from Doug or John (see the Contacts page).

Other projects

The TRANUS package was used to implement the first generation statewide TLUMIP model, and continues in occasional use. TRANUS is an aggregate equilibrium fully integrated land use-transport model, and has many similarities to MEPLAN. The TRANUS website has a number of documents, and the software is freely downloadable (executable format). The final report on the implementation of the TRANUS model in Oregon describes in detail the development, calibration, and application of the model.

The work of the TLUMIP team has been highly influenced by several others models. Two EU research projects in particular, PROPOLIS and SPARTACUS, have provided a number of ideas and comparisons against which to measure progress. The PROPOLIS (Planning and Research of Policies for Land Use and Transport for Increasing Urban Sustainability) project was funded under the EC Fifth RTD Framework Programme, and completed in 2004. Further information about project, including downloadable final report and brochure, are available from the PROPOLIS website. The SPARTACUS (System for Planning and Research in Towns and Cities for Urban Sustainability) project was an earlier EC research project (1996-98) that sought to evaluate different strategies for sustainable development. It was based on MEPLAN, but included many innovative extensions (such as grid-based noise and emissions analysis). Unfortunately, not much remains accessible about this project, although a brief summary is available. Michael Wegener's presentation notes from the Second Symposium provides insight into both projects.

Three other modeling systems, although not land use-transport models, were quite influential in the design of the second generation TLUMIP models. ALBATROSS (A Learning Based Transportation Oriented Simulation System), developed by Theo Arentze and Harry Timmermans, was presented at the Second Symposium. They have also developed AMADEUS, an agent-based activity-based travel demand model. They published a book in 2000 about ALBATROSS that is unfortuately out of circulation, but an amazing number of journal articles and papers about both articles are available on Theo's website. If you only have time to read one thing about it, find this article:

Hensher and Tu's TRESIS model is a transportation, land use, and environmental strategy impact simulator for urban areas. It was developed at ITS at the University of Sydney, and is an elegant and internally consistent model formulation that allows it tackle problems as diverse as greenhouse gas emissions and economic analyses, as well as traditional multimodal transport scenarios. However, the land use modeling component is not as sophisticated as those in mainstream land use-transport models, although further progress in that vein is contemplated. The definitive paper on the model is:

An earlier version of the paper can be found on the ITS website.

Finally, the influence of TRANSIMS on many projects in addition to this one can hardly be overstated. Originally developed at Los Alamos National Laboratory, it advanced a true microsimulation modeling environment for transportation systems. Work continues on the model, which is now an open source project. The FHWA maintains the current project website, with links to documentation and the TRANSIMS Open Source site.

Software tools

The Java programming language has been used to develop the first generation UrbanSim and second generation TLUMIP models. The latter are based on a cluster computing architecture. XML-RPC is the backbone of the distributed applications in the TLUMIP models. The model currently runs on eight dual-core dual-processor computers at the Oregon DOT and development team offices. Python is also used as a scripting language, for summarizing many of the model outputs, and prototyping new functionality.

The R statistical language is used extensively by the project team to visualize, summarize, and communicate the model outputs. Like the other software tools used by the project team, R is a freely available open source package. Brian Gregor and Ben Stabler have put together a great introduction to the package, R for Data Analysis and Modeling, that is available for download.