Marta González, February 24th, 2022

In this Rebalance Talk, Andreu Ulied spoke with professor Marta González – known for her work in transport modelling and an expert in developing computer models – on models based on actual current data known as “Nowcasting”, this new generation of models focus on what people actually do, not on people values or preferences in order to identify and manage the demand in urban infrastructures in relation to energy and mobility.

The evolution of technology over the past decades has given rise to ubiquitous mobile computing. There are 5.31 billion unique mobile phone users in the world today. Even in developing countries, penetration rates are of 83% and growing fast. Mobile phone devices and the applications that run on them passively record information on location with high spatial and temporal resolution. Cellular antennas, wifi access points, and GPS receivers are used to measure the geographic position of users to within a few hundred meters or less.

While the collection, storage, and analysis of mobile data presents very real and important privacy concerns, it also offers an unprecedented opportunity for researchers to quantify human behavior at large-scale. With billions of data points captured on millions of users each day, new research into mobility science has begun to augment and sometimes replace sparse, traditional data sources, helping to answer old questions and raise new.

The approach Dr. González uses to mode mobility is to understand the underlying patterns of individuals using new high resolution data. It ties together the inference of home and work activity locations from data, with the modeling of flexible activities (e.g., other) in space and time. The temporal choices are captured by three parameters: the weekly home-based tour number, the dwell rate, and an exploration rate. Individual parameters are extracted from the data to capture how an individual deviates from the circadian rhythm of the population, and generate the wide spectrum of empirically observed mobility behaviors.

The spatial choices of visited locations are modeled by a rank-based exploration and preferential return mechanism. Meaning that each location’s choice can be labeled either as a return to previous locations or as an exploration (a visit to a location never visited before). For the returns, the probability to select a location is proportional to the number of visits it already has, also known as the “rich get richer effect”. This effect generates a Zipf’s law in the frequency of visits, which is a ubiquitous observation in human mobility data.

Her research team have uncovered `mobility motifs’ by examining abstract trip chains over the course of a day. A daily mobility motif is defined a set of locations and a particular order that a person visits them over the course of a day. These motifs constitute directed networks where nodes are locations and edges are trips from one location to another. For example, the motif of an individual whose only trips in a day are to and from work will consist of two nodes with a two directed edges (one in both directions). Counting motifs in mobility data from mobile phones and traditional travel surveys, we find on average individuals visit 3 different places in a given day. We then construct all possible daily motifs for a given number of locations (n) and compute the frequencies that those motifs appear in human mobility data.

Shockingly, while there exist over 1 million ways for a user to travel between n=6 or fewer locations, 90% of people use one of just 17 motifs and nearly a quarter follow the simple two location commute motif. Her team have found the same results both in travel survey data, mobile phones and their model which reproduces empirical results.

To measure each individual’s hinterland, or area of influence, she borrowed a quantity from polymer physics known as the “radius of gyration”. For each individual, it measures the average distance to her mean location (e.g. her center of mass). In essence, the radius of gyration is a measurement of the characteristic distance an individual travels during an observation period. When analyzing data from cities of diverse latitudes and GDPs, they share the distribution of radius of gyration of their population. It is approximated by a truncated power-law. Interestingly, even when analyzing mobile phone records from Saudi Arabia, where women do not have access to drivers licenses, the distributions of the radius of gyration of men and women do not present significant differences.

González’ modeling framework is based on revealed data, not in stated preferences. She develops mathematical models predicting people behavior without any assumption on people values or wishes. These collection of universal patterns in space and time is what we called individual mobility patterns.

She combines mobile phone data with other sources of information and much richer possibilities for discovery emerge. Her team have a study, that uncovers various spending and mobility habits combined. It uses digital traces left by credit cards and mobile phone data in Mexico City. By grouping consumers by their similarity in purchase sequences, we detect five groups or “lifestyles”. Purchase habits are not only related to socio-demographic characteristics such as age, gender, income and mobility – they are also related to the places people visit, and the people they call. Each of these groups has a core purchase, which is the most frequent of their spending activities. The groups were labeled as: “home makers”, “dinners”, “high tech users”, “youth” and “commuters”.

In the years to come Marta is combining these approaches to uncover pattern of humans and natural systems interactions. To mitigate the current trends of our global
environmental footprint, we need to better understand the sources of greenhouse gas emissions and develop more science-based policies. There’s a disconnect between “ubiquitous computing” and the best practices proposed by those studying environmental sciences, urban planning and public policy. Improving this connection is her current research goal.

Marta C. González is Associate Professor both in City and Regional Planning and Civil and Environmental Engineering at the University of California, Berkeley. She also holds a Physics Research faculty position in the Energy Technology Area (ETA) at the Lawrence Berkeley National Laboratory (Berkeley Lab). With the support of several cities, companies, and foundations, her research team develops computer models to analyse digital traces of information mediated by devices. They process this information to manage the demand in urban infrastructures in relation to energy and mobility. 

Learn more about Marta González