The Series with Elastic Extents Problem (SWEEP)

"Gerrymandering" Urban Time Series

Sam Stehle / @sam_stehle

Building City Dashboards

National Centre for Geocomputation, Maynooth University, Maynooth Ireland

GISRUK 2019, Newcastle University

The Modifiable Areal Unit Problem

Parallels in time: The SWEEP

Dublin Sound monitoring

http://dublincitynoise.sonitussystems.com/applications/api/dublinnoisedata.php?location=7&start=1554218025&end=1554822825

Public Reporting

Dublin Agglomeration Environmental

Noise Action Plan (Nov. 2018)



Threshold (db) Daytime (07:00 - 19:00) nighttime (23:00 - 07:00)
Quiet Area <55 <45
Desirable Low <55 <50
Undesirable High >70 >55

testing for SWEEP effects

  • (log) mean decibel level
  • frequency of observations which breach value limits


  • higher than 'undesirable high'
  • higher than 'desirable low'
  • lower than 'quiet area'
  • for each separate value limit, night and day
    •

    Blessington Street Basin

    Blessington Street Basin

    Blessington Street Basin

    Blessington Street Basin

    mean skewness of 5-minute observations = 0.251

    Implications of scalar SWEEP effects

    except in extreme cases, aggregating yields higher mean db values

    meaning more frequent passing of high thresholds

    and less frequent satisfaction of quiet area thresholds

    1-hour aggregation as 12 segmentations

    Segmentation effects

    segmentation effects

    mean time of day of :00 segment: 11:30

    mean time of day of :55 segment: 12:25

    segmentation and trend

    segmentation and trend



    Stations with quiet area SWEEP effect Stations with undesirable high SWEEP effect observations with increasing trend
    Daytime 67% 20% 49.4%
    Nighttime 80% 73% 79.7%


    implications of segmentation effects



  • Increasing trend influences segmentation effect
  • SWEEP effects depend on how day/night times are defined (boundary effects)
  • Quiet area threshold is more volatile that undesirable high

    • anything more than a theoretical exercise?

    while space is continuous, time is discretised, forming logical boundaries

    considering intent: gerrymandering time series

    • Communication: portray low sound to the public
      • report disaggregated observations
    • Incentive/Disincentive schemes: reward low/punish high sound
      • no more cynical than political gerrymandering

    Concluding recommendations



  • Standards for environmental reporting, especially when incentivised
  • Further reflection on the impact of stationarity, variability, boundary definitions
  • Streaming data for official statistics
  • Measure the impact of day of week - noncontiguous segmentations
  • Uncertain Geographic Context Problem (Kwan 2012)

    • Sam Stehle

    Sam.Stehle@mu.ie

    @sam_stehle

    dublindashboard.ie

    dashboards.maynoothuniversity.ie