PaperlayerView exact revision on arXiv

Source-linked AI summary

City-scale dark fiber DAS measurements of infrastructure use during the COVID-19 pandemic

Nathaniel J. Lindsey, Siyuan Yuan, Ariel Lellouch, Lucia Gualtieri, Thomas Lecocq, Biondo Biondi

arXiv:2005.04861v1physics.geo-pheess.SPphysics.soc-ph

TL;DR

Real-time measures of infrastructure use during quarantine were important, but mobile-phone reports are incomplete and conventional seismic networks have limited spatial resolution. This paper uses dark-fiber DAS in Palo Alto to detect vehicles from roadbed strain, finding a large Sand Hill Road traffic decline and a strong correlation with mobile-derived activity changes.

  • Problem

    Dense, anonymous city-scale measures of infrastructure use are limited by incomplete mobile-phone sampling and the low spatial resolution or high deployment cost of conventional seismic networks.

    Source: Introduction · pp. 4–5

  • Method

    The study applies template matching to continuous dark-fiber DAS strain recordings beneath Palo Alto to detect vehicle-induced geodetic signals and count traffic.

    Source: Introduction · p. 6, Materials and Methods · p. 12

  • Results

    A 50% decrease in Sand Hill Road DAS vehicle counts correlated strongly with 50–70% decreases in Apple and Google mobile-derived activity data, while hospital-area traffic initially remained sustained.

    Source: Discussion · pp. 9–10, Results · p. 9

  • Takeaways & Limitations

    DAS provides natively anonymized, spatially and temporally granular measurements of infrastructure use across a large urban area.

    Source: Discussion · p. 11

    Vehicle detections are more likely to miss vehicles farther from the fiber, such as traffic on the opposite side of the street.

    Source: Discussion · p. 10

Abstract

from arXiv · show

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.

Introduction

Measuring public-space activity during quarantine mattered for public-health decisions, but mobile-phone reports and conventional seismic networks had important coverage and resolution limits. The paper introduces dark-fiber DAS as a dense, anonymous way to measure infrastructure use and applies it to Palo Alto during the COVID-19 order.

Introduction

  • 60–80% declines in vehicle traffic and in-store purchases, and 20% declines in grocery and pharmacy visits, appeared in Bay Area mobile reports after the shelter-in-place order.Apple and Google began releasing regional Mobility Reports on April 4, 2020.Source: Introduction · p. 3
  • Mobile-phone activity data are relative and incomplete, reflect only some users, and may require combining platform-specific datasets to study a particular city sector.The passage also notes concerns about linking passively generated user data to individuals.Source: Introduction · p. 4
  • Conventional city seismometers provide low-resolution activity maps, while deploying thousands of independent stations is impractical.Previous COVID-19 studies used more than 95 urban seismometers; dense arrays can map infrastructure use but require many costly stations.Source: Introduction · pp. 4–5
  • Dark-fiber DAS uses optical fibers as dense recording arrays, with virtual receivers spaced 1–10 m apart and continuous strain measurements over tens of fiber kilometers.The instrument sends laser pulses and measures returning Rayleigh-scattered light; dark-fiber recordings can capture 10,000–20,000 components.Source: Introduction · pp. 5–6
  • The study uses continuous dark-fiber DAS beneath Palo Alto to compare activity across public sectors during the COVID-19 quarantine order.The Bay Area order began on March 16, three days before California’s statewide order.Source: Introduction · pp. 3, 6

Results

DAS detected a sharp decline in Sand Hill Road vehicle counts after the Bay Area order, while hospital-area traffic was initially more stable. The results also show how vehicle loading and urban background signals vary across DAS channels and frequency bands.

Results

  • The Palo Alto array recorded more than 14.5 TB of continuous DAS strain data from March 1 through May 1, 2020.Daily power spectral density was computed for channels near Stanford Hospital and Sand Hill Road.Source: Results · p. 6
  • Power spectra varied across channels: Sand Hill amplitudes were 20–30 dB higher, most channels showed reductions up to 7 dB, and hospital channels lacked systematic peak change.Some Upper Sand Hill channels also showed minor, narrow frequency peaks.Source: Results · p. 7
  • High-frequency DAS data tracked vehicle motion 100–200 m from the road, including vehicles on Sand Hill traveling at about 15 m/s.The observed seismic background transitioned sharply at X=3.18 km where the fiber turned away from Sand Hill toward Stanford Hospital.Source: Results · p. 7
  • Filtering at 0.1–1 Hz revealed localized wavelets with peak amplitudes around 5–10 nanostrain that migrated at the road’s speed limit.Source: Results · p. 7
  • Vehicle loading produces a localized geodetic strain response: displacement depends on vehicle mass and geometry, and the strain signal falls off rapidly with distance.The modeled vehicle signal has energy around 0.1–1 Hz; the passage notes that four-wheel vehicle response also depends on wheel separation or vehicle area.Source: Results · p. 8
  • A template-matching algorithm detected and characterized Palo Alto traffic changes during COVID-19 quarantine; Sand Hill Road counts fell from about 11,000 to 5,500 vehicles per day, a 40–60% decrease versus the first-week median (1–7 March 2020).Counts were reported for morning commute hours, 4:00 am–12:00 pm.Source: Results · pp. 8–9
  • About 11,000 vehicles per day on Sand Hill Road fell to around 5,500 after the March 16 order, a 40–60% decrease from the first-week background; hospital-area counts initially held near 2,000 per day.Sand Hill traffic remained reduced through quarantine; hospital-area counts began declining around April 6 to about 80% of pre-quarantine levels.Source: Results · p. 9

Discussion

The quarantine coincided with sharply reduced commuter traffic on Sand Hill Road but continued use near Stanford Hospital. DAS provides localized vehicle counts that complement mobile-phone activity reports and can characterize infrastructure use across city sectors.

Discussion

  • DAS measured a 50% decrease in Sand Hill Road vehicles, while activity near Stanford Hospital remained comparatively sustained.The Sand Hill traffic baseline during quarantine was likely related to essential businesses such as the grocery store and pharmacies.Source: Discussion · pp. 9–10
  • DAS vehicle counts on Sand Hill correlated strongly with Apple’s 50–70% decrease in driving activity and Google’s 50–70% decrease in retail and recreation activity.The mobile reports cover broader regions and omit some road users, so their values may not represent Sand Hill Road directly.Source: Discussion · p. 10
  • Roadbed deformations from vehicles near the fiber provide more spatially localized traffic detail than point-sensor seismic measurements.The strain response falls off as 1/r^2, making vehicles farther from the fiber more likely to be missed; the authors note that more sophisticated detection methods could refine results.Source: Discussion · p. 10
  • DAS vehicle counts relate to surface-wave strain energy in both the 4–10 Hz and 15–30 Hz bands, helping calibrate the vehicle component of seismic background.The template-matching algorithm made over 450,000 automatic vehicle detections during the eight-week experiment.Source: Discussion · pp. 10–11
  • DAS provides city-scale infrastructure-use measurements that are anonymized and spatially and temporally resolved for public-health and government decision-making.The paper describes the instrument as easily deployable.Source: Discussion · p. 11

Materials and Methods

The Stanford DAS-2 experiment recorded strain along a telecommunications fiber and processed the data to characterize vehicle-related signals. Vehicle counts were obtained by matching channel- and day-specific templates to continuous recordings.

Materials and Methods

  • The Stanford DAS-2 instrument recorded optical phase shift at 250 samples per second across 1,250 channels, with 8.16 m channel spacing and a 20 m gauge length.About 350 channels covered the Sand Hill Road section between near Stanford Hospital and SLAC.Source: Materials and Methods · p. 11
  • Data processing converted optical phase to nanostrain, removed trends, means, and common-mode noise, and applied bandpass filters when isolating signals.The optical-phase conversion used a multiplication factor of 10.77097; PSD strain values are expressed in nanostrain^2/Hz or dB/Hz, and RMS strain was computed from 30-minute power spectra.Source: Materials and Methods · pp. 11–12
  • The modeled vehicle-loading signal assumed a fiber depth of 2 m, a lateral offset of 10 m, vehicle speeds of 10–15 m/s, and vehicle masses of 1,000–3,000 kg.The model also used soil parameters including a Poisson’s ratio of 0.4, Young’s modulus of 80 MPa, and shear modulus of 28.6 MPa.Source: Materials and Methods · p. 12
  • For each channel and day, a 20-second template built from the median of about 200 STA/LTA detections was cross-correlated with the time series; values above 0.7 were labeled vehicles.The STA/LTA threshold was set to yield about 200 detections per day and was validated against human picks on short segments.Source: Materials and Methods · pp. 12–13

Figures and Tables

The figures locate the DAS fiber and show channel-specific spectral variation, vehicle signals, and vehicle detections. They also compare daily traffic changes with mobile-phone activity and relate surface-wave strain to detected vehicle counts.

Figures and Tables

  • Figure 1 maps the Palo Alto optical-fiber route and presents daily median strain power spectra for selected channels, with curves progressing from March 1 through the end of April.The channels include a quiet reference, Stanford Hospital, and upper and lower Sand Hill Road; the figure highlights spatial variability and changes during quarantine.Source: Figure 1 · p. 18, Figures and Tables · p. 18
  • Figure 2 contrasts 3–30 Hz vehicle surface waves with 0.1–1 Hz roadbed strain responses, then shows example detections against modeled strain.The example includes three STA/LTA detections and describes subsequent template matching using the median of approximately 200 detected vehicle signals.Source: Figure 2 · p. 19, Figures and Tables · p. 19
  • Figure 3 plots Lower Sand Hill Road morning vehicle counts on the left axis and relative percentage change on the right, alongside mobile-phone activity comparisons.The comparison includes Google and Apple statistics; the figure also shows a Stanford Hospital channel.Source: Figure 3 · p. 20, Figures and Tables · p. 20
  • Figure 4 compares daytime horizontal strain with detected vehicle counts in 4.0–10.0 Hz and 15.0–30.0 Hz bands; correlation improves at higher frequencies.Squares mark the 4.0–10.0 Hz band and triangles mark the 15.0–30.0 Hz band.Source: Figure 4 · p. 21
Loading 2005.04861v1…