key: cord-0462058-n4594s8u
authors: Lindsey, Nathaniel J.; Yuan, Siyuan; Lellouch, Ariel; Gualtieri, Lucia; Lecocq, Thomas; Biondi, Biondo
title: City-scale dark fiber DAS measurements of infrastructure use during the COVID-19 pandemic
date: 2020-05-11
journal: nan
DOI: nan
sha: 329eddefd0bc6965ea76be5b16e0b355c5e6be6a
doc_id: 462058
cord_uid: n4594s8u

Throughout the recent COVID-19 pandemic when government officials around the world ordered citizens to quarantine inside their homes, real-time measurements about the use of roads, hospitals, grocery stores, and other public infrastructure became vital to accurately forecast viral infection rates and inform future government decisions. Although mobile phone locations provide some information about community-level activity, dense distributed geophysical sensing of ground motions across a city are more complete and also natively anonymous. In this paper, we demonstrate how fiber-optic Distributed Acoustic Sensing (DAS) connected to a telecommunication cable beneath Palo Alto, CA captured seismic and geodetic signals produced by vehicles during the COVID-19 pandemic outbreak and subsequent quarantine. We utilize DAS strain measurements of roadbed deformation caused by local cars and trucks in an automatic template matching detection algorithm to count the number of vehicles traveling per day over a two-month period around the timing of the San Francisco Bay Area shelter-in-place order. Using a segment of the optical fiber near a major grocery store on Sand Hill Road we find a 50% decrease in vehicle count immediately following the order, but data from near Stanford Hospital showed a far more subtle change due to on-going hospital activities. We compare the information derived from DAS measurements to other quarantine response metrics and find a strong correlation with the relative changes reported by Google and Apple using mobile phone data.

During the recent global coronavirus disease pandemic that began at the end of 2019 , human movement in public spaces became a public health issue as government officials advised "social distancing" and other behavioral interventions to mitigate the spread of the novel coronavirus (Ferguson et al., 2020; Tian et al., 2020) . In California, nearly 40 million individuals were ordered to stay indoors by state officials beginning 19-Mar-2020, except as needed to maintain critical infrastructure operations, conduct essential services, and obtain food or personal exercise. In the San Francisco Bay Area, the order came from city and county officials three days earlier on 16-Mar-2020.

During public health emergencies of this magnitude, measuring human activity and how public spaces are being used becomes critical for many reasons throughout all phase of the pandemic. Monitoring activity levels during early stages of a pandemic can provide scope of contact tracing and containment efforts. During quarantine, quantifying public activity feeds-back the required information to public health and government officials to make informed decisions such as if the population is properly following quarantine orders (Wesolowski et al., 2012 , Tizzoni et al., 2014 , Oliver et al., 2020 . Historically, this information has been captured in the months or years following the crisis through in-person surveys, but in the past decade public and private mobile phone data has become a tool to understand where, when and how phases of a particular crisis are unfolding in real time. During the COVID-19 crisis, Apple and Google released Mobility Reports beginning April 4, which aggregated mobile phone location services data from individual users into regional statistics about public infrastructure usage (Apple, 2020; Google, 2020). In the San Francisco Bay Area, these reports signaled that vehicle traffic and in-store purchases dropped by 60 -80% in the days following the shelter-in-place order, while grocery and pharmacy visits decreased by 20%. Although valuable, the utility of mobile phone data is limited in many different ways. For example, the data analyzed are only collected from the subset of people who have access This preprint was submitted to Science Advances on 10-May-2020.

Manuscript Template Page 4 of 21 to large data plans and opt-in to use location services. This bias in data sampling likely biases the reported infrastructure usage statistics according to socioeconomic class, age, and also by region.

It also means that the Mobility Reports are relative and incomplete because the statistics are reported as percentage change from background and hence the absolute number of people using a public area is not documented. Second, because mobile phone data are segmented by platform, understanding shelter order impacts in particular sectors of a city, such as a park or a major arterial freeway, requires concatenating multiple, potentially heterogeneous databases. Third, while pseudo-anonymization and aggregation responsibly delivered a public good in this case, users are likely to still be wary of the causal linkage established by monitoring their passively generated mobile phone data.

An alternative, fully anonymous, absolute signal of city-scale human activity is everpresent in the low-level seismic wavefield produced by humans, often called the anthropogenic seismic background. For over two decades, seismologists have referred to these 0.5 -50 Hz vibrations as "anthropogenic noise" because they commonly obscure earthquakes and other natural Earth signals (Meremonte et al., 1996) . This "noise" is actually comprised of seismic waves excited by a multifarious number of moving sources, including vehicles, trucks and trains, buildings and bridges excited by the wind, and generators, pumps and other motorized systems.

Instrumenting an urban area with a few precise inertial seismometers has documented the strong spatial and temporal variabilities associated with the anthropogenic seismic wavefield, from high ground motion amplitudes during the daytime to low ground motion amplitudes during night, weekend, and holiday times; ground motions have even been documented from crowds during sporting matches and between sets of a rock concert (Vidale 2011; Diaz et al., 2017) .

Recently, during the COVID-19 pandemic, recordings from over 95 urban seismometers were studied in a worldwide effort to measure how the quarantine affected anthropogenic seismic these waves also undergo severe anelastic attenuation, which is exponential with frequency.

Therefore, conventional seismometers deployed in cities can only produce low-resolution maps of general human activity levels, biased towards sources close to the stations. Densifying the recording array using thousands of seismometers with sensor separations of 100 m can measure general urban infrastructure usage patterns across a city such as greater noise near commute infrastructure for a period of several months (Inbal et al., 2015) . However, the cost to deploy and maintain thousands of independent stations each with its own sensor, data-logger and power supply throughout a city is impractical for even a brief experiment, and inconceivable over years. Poisson's ratio ( ) using the Boussinesq approximation for a loading point source:

In (1) Methods for a description of the algorithm). Figure 3 shows the number of vehicles detected per This preprint was submitted to Science Advances on 10-May-2020. there is a single-day increase in traffic suggestive of people driving to the store to get supplies.

One DAS experiment provides highly resolved statistics about public infrastructure utilization across many large sectors of a city through time. According to our results from the Stanford DAS-2 experiment, the COVID-19 quarantine order from local and state government officials resulted in a major decrease in commuter traffic along Sand Hill Road but sustained use of critical infrastructure such as the road near Stanford Hospital. The baseload of vehicle traffic on Sand Hill during the quarantine was likely related to essential business traffic like the major grocery store and pharmacies. These changes may seem obvious as Sand Hill is a commuter This preprint was submitted to Science Advances on 10-May-2020.

Manuscript Template Page 10 of 21 corridor connecting the I-280 freeway and CA-82 highway to businesses and workplaces that closed during quarantine while Stanford Hospital remained busy, but accurately quantifying these changes in real-time is a challenge.

A commonly reported metric during the quarantine was derived from mobile phone users.

We find a strong correlation between the 50% decrease in vehicles counted using DAS on Sand

Hill Road and the 50 -70% decrease in mobile phone-derived activity data reported by Apple (for Santa Clara County driving data) and by Google (for San Francisco Bay Area retail and recreation data; see Figure 3 ). The Mobility Report data could be biased downwards from the actual number of vehicles on the street for many reasons. For instance, they exclusively rely on data from a select pool of users, and thus will not include users making trips without data services, public transit vehicles, or police/fire/emergency vehicles, which were all still using Sand Hill Road during quarantine. Furthermore, the Mobility Reports aggregate different regions, which may be less representative of Sand Hill Road.

Over 450,000 automatic vehicle detections were made using DAS on Sand Hill Road during the 8-week experiment. We found that the quasi-static or geodetic strain response to individual vehicle loading/unloading presented clear evidence of a proximate vehicle (Figure 2d ).

Because this strain falls off like 1/r 2 , vehicles passing far from the fiber such as on the opposite side of the street have a greater likelihood of being missed, as can be seen in Figure 2b where Northbound vehicles are weaker than Southbound ones (fiber is located on the Southbound side).

While the template matching algorithm performs better than a simple threshold detection algorithm, more sophisticated methods that utilize machine learning or the distributed nature of DAS data are required to refine these results (Martin et al., 2018) .

Spatially-localized roadbed deformations from vehicles passing in the immediate vicinity of the fiber provide a more accurate detail about traffic patterns than is available to point sensor seismic measurements. This is because one urban seismometer, or more commonly a short This preprint was submitted to Science Advances on 10-May-2020.

Manuscript Template Page 11 of 21 geophone, only records high frequency anthropogenic surface waves. Such energy could be produced by vehicles traveling anywhere within a radius of several hundred meters. In Figure 4 , we show that the DAS data from Sand Hill Road documents a relationship between the number of vehicles and the total surface wave strain energy recorded in two different high frequency bands In times of crisis, access to DAS-derived information about public infrastructure utilization is valuable to public health and government officials because it provides city-wide measurements which are natively anonymized and aggregated with the requisite level of spatial and temporal granularity to make decisions. DAS retrieves data about how infrastructure is being used across a large urban area with an easily deployable instrument. power of the seismic trace. We used a 0.01-s duration STA window and a 10-s duration LTA window, and automatically determined the detection threshold such that about 200 templates were detected per day. These settings were validated against human picks for short segments and it was determined that the STA/LTA algorithm accurately detected the higher amplitude vehicle signals.

Using the median of a large population of STA/LTA detections effectively avoided corruption by earthquake signals. We then computed the full cross-correlation between the template and the daily time series, labeling windows with cross-correlation value above 0.7 as vehicles. 

An investigation of transmission control measures during the first 50 days of the COVID-19 epidemic in China

Quantifying the impact of human mobility on malaria

On the use of human mobility proxies for modeling epidemics

Mobile phone data for informing public health actions across the COVID-19 pandemic life cycle

Urban seismology-Northridge aftershocks recorded by multi-scale arrays of portable digital seismographs

Seattle "12th man earthquake" goes viral

Urban seismology: On the origin of earth vibrations within a city

COVID-19 societal response captured by seismic noise in China and Italy

Github repository

Imaging widespread seismicity at midlower crustal depths beneath Long Beach, CA, with a dense seismic array: Evidence for a depth-dependent earthquake size distribution

Rayleigh scattering based distributed sensing system for structural monitoring

Contributed Review: Distributed optical fibre dynamic strain sensing

Continuous subsurface monitoring by passive seismic with distributed acoustic sensors-The "Stanford Array" experiment

Dynamic strain determination using fibre-optic cables allows imaging of seismological and structural features

Distributed acoustic sensing using dark fiber for near-surface characterization and broadband seismic event detection

First field trial of sensing vehicle speed, density, and road conditions by using fiber carrying high speed data

Distributed acoustic sensing for seismic monitoring of the near surface: A traffic-noise interferometry case study

Passive imaging and characterization of the subsurface with distributed acoustic sensing (Doctoral dissertation

Urban Seismic Site Characterization by Fiber-Optic Seismology

Urban Near-surface Seismic Monitoring using Distributed Acoustic Sensing

Fiber-optic network observations of earthquake wavefields

Monitoring of remote seismic events in metropolitan area fibers using distributed acoustic sensing (DAS) and spatio-temporal signal processing

Potency-magnitude scaling relations for southern California earthquakes with 1.0< M L< 7.0

Finite element model of ballasted railway with infinite boundaries considering effects of moving train loads and Rayleigh waves

Train traffic as a powerful noise source for monitoring active faults with seismic interferometry

ObsPy: A bridge for seismology into the scientific Python ecosystem

This preprint was submitted to Science Advances on 10-May-2020.

Manuscript Template Page 17 of 21