Using Replica in modeling

2 posts / 0 new
Last post
Using Replica in modeling

Hi Everyone,

My name is Josh, and I’m the transportation data analyst at the Tahoe Regional Planning Agency, the MPO and environmental protection agency serving the Lake Tahoe watershed of California and Nevada. TRPA is a small agency, serving a permanent population of about 56,000, and I am the sole modeler on staff. We’re beginning the process of upgrading our travel demand model, and I was wondering if (or how) other agencies were attempting to incorporate data from Replica into their travel demand models, and if so, how successful those efforts have been. The fact that Replica itself is an activity-based model muddies the water here, as unlike other common sources of travel data, the provided product isn’t just raw data and there are issues with using one model’s output as another model’s input. We just started using Replica about 2 months ago and most of our time with the platform has involved validation.

Several of our bordering agencies have started using Replica, and I have heard rumblings of other agencies in the western United States, mostly smaller ones without the funding necessary to conduct full travel surveys, looking at how to incorporate Replica into their travel demand models. In an ideal world, we would be conducting an O-D travel survey this year, but that is likely infeasible for us to conduct accurately due to the resources required. Traveler surveys are quite expensive to conduct and response rates across all types of surveys in recent years have trended toward wealthier and more-educated respondents. Furthermore, intercept surveys are difficult to conduct, with few safe places for us to intercept people along the winding mountain roads entering our jurisdiction.

A major thing that attracted us to Replica was the ability to see the estimated trip patterns and demographics of visitors to our region and those commuting into the region from Reno, Carson City, and the surrounding exurbs. Our model currently uses parameters from a travel survey conducted about 20 years ago, when casinos were the main driver of visitation to the region as opposed to outdoor recreation.  We know there has been a substantial shift away from casino visitation, and the data Replica provides to support this shift is quite insightful and is likely to inform our planning efforts outside of our travel demand model, but Replica is a model, not raw data, and I know that we can’t just use Replica data directly in our model.

We want to be cautious here, and because of that, we aren’t committed to incorporating Replica data into the model, nor do we want to be the first agency to make this jump. But if we have Replica data and other agencies have figured out how to incorporate it in a way that is robust, valid, and insightful, it seems like a logical jump to make given the cost benefits over other alternatives. Being a small agency, the resources we can put into our model are limited.

How are other agencies currently using Replica, particularly when it comes to their travel demand models? Any insights would be greatly appreciated.



Hi Josh,

The question of how to leverage new data sources in travel modeling is both timely and relevant. We would like to draw your attention to two presentations made at this year's TRB Innovations in Travel Analysis and Planning Conference (Indianapolis, June 4-6, 2023) which give examples of automated travel model calibration using big data sources in applications for Ohio DOT and Maricopa Association of Governments.

By applying this methodology it is possible to a) automate calibration of a travel model to new data sources in addition to HHTS and b) understand the outliers between travel model results and third-party data sets which may remain even after the calibration process. While the presentations below illustrate model calibration results using Streetlight and Airsage data, the methodology is portable to other data sources and has been used as well with Teralytics data and transit smartcard data. It is also portable to other model structures (e.g. 4-step, tour-based, ABM).

Here are the TRB links with abstracts and slides.

Weekend Activity-Based Model “…In the absence of sufficient data for weekends from HTS the entire model calibration approach needed to be reoriented towards new sources of information such as big data which is readily available for weekends. The paper includes a detailed description of the innovative ways big data were effectively used for the calibration of disaggregate sub-models of the ABM.” 

Utilizing big data for calibration of a microsimulation ABM “The main question is how big data… could be effectively used for calibration of such disaggregate models as mode choice or tour/trip frequency or car ownership. The paper suggests a new general approach where different types of data including HTS and big data, can be used in one systematic process of travel model calibration… The developed methods are illustrated for an example of ABM developed for the cities of Phoenix, AZ, and Lima, OH."




Daniel Florian

Senior Director, Mobility Simulation

Bentley Systems, Inc.