better cities through data axquisition and analysis › 2017 › graphics › uploads › plenary...
TRANSCRIPT
Better Cities Through Data Axquisition and AnalysisHPC Saudi Arabia
KAUSTMarch 13, 2017
Dr. Steven E. Koonin, NYU CUSP DirectorNYU University Professor Stern School of BusinessTandon School of Engineering [email protected]://cusp.nyu.edu
Cities as a venue for data science
• Cities matter (most people live there)
• The data are plentiful and diverse
– Digitization (and opening) of records, proliferation of sensors
• There are things to do
– Develop technologies/methodologies to acquire, integrate, and analyze data
– Understand the “science of cities”
– Applications for government, citizens, private-sector
CUSP is Part of the NYC Applied Sciences Initiative
Mayoral Announcement April 23, 2012
Environment PeopleInfrastructure
What does it mean to instrument a city?
There’s a lot of data out there
Novel Technologies
• Visible, infrared and
spectral imagery
• RADAR, LIDAR
• Gravity and magnetic
• Seismic, acoustic
• Ionizing radiation,
biological, chemical
• …
Sensors
• Personal (location,
activity, physiological)
• Fixed in situ sensors
• Crowd sourcing
(mobile phones, …)
• Choke points (people,
vehicles)
Organic Data Flows
• Administrative records
(census, permits, …)
• Transactions (sales,
communications, …)
• Operational (traffic,
transit, utilities, health
system, …)
• Social media (Twitter,
Facebook, blogs, …)
Research strategy
• Focus on data
– Technology, science, applications
• Focus on New York City (mostly)
• “Own” the data
– Institutional familiarity, differential access, unique sensor systems
• Create facilities and capabilities
Unique
User
Facilities &
Major
Initiatives Data Facility Quantified Community Urban Observatory
Research
Groups &
Themes
Building Informatics Mobility Sounds of NYC Urban Energy
Projects ACCESS NYC Data Analysis Data Profiling and Integration Impacts of Urban Land Use
Interactive Exploration of NYC Bus Data Local Law 84 Analysis Local Law 87 Analysis
New York Open Government NYC Economic Map NYC Energy and Water Performance
Map NYPD Emergency Call Analysis Parks Quality Assessment Particulate Matter
Exposure Distribution Quantitative Analyses of Urban Topography Urban Waste Analytics
Water Street Data Aggregator 3D Urban GIS Interactive Visualization (Urbane)
Class of 2016 Capstones and more
Research structure
NYC DataBridge
Tweets!
❑New Yorkers tweet a lot:
2.6 Million Accounts2.37% of all tweets(NYU - 2011)
Geotagged Tweets By Eric Fisher - 2012
Lane et. al
The Quantified Community – Red HookIn partnership with
Quantified Community / Neighborhood Innovation Labs Network
Locations of phase I
sensor deployments:
5 <2 months,
15 <6 months
Pilot
sensor
Kontokosta et. al
Quantified Community: Red Hook, Brooklyn
3/21/2017
14
Toward a Real-Time Census of the City -Lower Manhattan WiFi pings
KONTOKOSTA 2016 - NOT FOR DISTRIBUTION
Source: Kontokosta and Johnson, forthcoming
N = 1,150
Mean = 219.5
s.d. = 101.7
020
40
60
80
Fre
qu
en
cy
0 200 400 600 800Weather Normalized Source EUI (kBtu/sq.ft./yr.)
Source: Local Law 84 Disclosure Data, Kontokosta 2013
Source Energy Use Intensity, Office Buildings, New York City
N = 7,505
Mean = 137.9
s.d. = 46.8
0
20
040
060
080
0
Fre
qu
en
cy
0 100 200 300 400Weather Normalized Source EUI (kBtu/sq.ft./yr.)
Source: Local Law 84 Energy Disclosure Data, Kontokosta 2013
Source Energy Use Intensity, Multi-Family Buildings, New York City
Building Energy Efficiency
Kontokosta 2013
Local Law 84 Benchmarking Data
Kontokosta, 2013
Taxis as Sensors for NYCTaxis are sensors that can provide unprecedented insight into city life: economic activity, human behavior, mobility patterns, …
“What is the average trip time from Midtown to the airports during weekdays?'’
“How the taxi fleet activity varies during weekdays?’’
“How was the taxi activity in Midtown affected during a presidential visit?'’
“How did the movement patterns change during Sandy?”
“Where are the popular night spots?”
May 1st – 7th
20113.6 Million Trips
Train Stations
Airports
Studying Taxi Patterns
CONFIDENTIAL: for intended recipient only 19June 27, 2016
Identify Traffic Conflicts
Using “Surrogates”
• “Surrogate” Safety Measures: Indicators that describe the scenarios in which a vehicle would collide with another vehicle if they did not change their current intentions.
Rear-end conflict
Right-angle conflict
Time to Collision (TTC): The time required
for two vehicles to collide if they continue at
their present speed and on the same path.
Post-Encroachment Time (PET): The time
difference between the arrival of two
vehicles at the potential conflict point.
Potential conflict point
Kaan Ozbay et. al
CONFIDENTIAL: for intended recipient only 20June 27, 2016
Jay St Demo: Conflicts
Conflicts: TTC<1.5 seconds
Estimated Surrogates based on Automatic Tracking Results
Kaan Ozbay et. al
• Cyber-physical system for large-scale continuous monitoring of noise pollution
21
• Custom acoustic sensor based on MEMS microphone technology
22
Current SONYC deployment
3/21/2017
Current deployment
• Ten nodes deployed• ~100 by year’s end• 24/7 SPL data• Random sample of
audio snippets• Real-time telemetry
from all
• State-of-the-art machine listening technology for real-time sound source identification
25
Photo by Tyrone Turner/National Geographic
A single sensor with: synoptic non-permissive coverage, persistence, high granularity
Manhattan in the Thermal IR
199 Water StreetBuilt 1993 :: 998,000 sq ft
electricity, natural gas, steamLEED Certified
hyperspectral
infrared
visible
• Unique user facility for persistent and
synoptic observations of cities
• Advance “urban science” and applications to
enhance public well-being, city operations,
and future urban design
• Instrumentation to include both broad band
and hyper-spectral from visible to infrared
wavelengths
• Combine correlative data including
administrative records, in situ measurements,
topography, etc.
persistent synoptic granular
CUSP Urban
Observatory
The view from CUSP’s Urban Observatory in Brooklyn
Borough Block & Lots (BBL)
Standard UO view
colored by distance
Picture merges image captured from video, 3-D LIDAR map of NYC, PLUTO (Primary Land Use Tax Lot Output) database, and LL84 Energy Benchmarking data
Source: Dobler, et al. 30
Dynamics of the Urban Lightscape
• Each frame is “registered” to a common frame by spatial correlations
• 4,200 window apertures are identified by hand (out of approximately 20,000 windows in the scene)
• For each frame, the average brightness of each source is calculated in
3 bands (RGB)
• The brightness of a given source as a function of time is referred to
as its “light curve”
Dobler et al.;doi:10.1016/j.is.2015.06.002
Pulse of the City Lights
Proxies for energy consumption, occupation, sleep cycle, …
Sleep-wake rhythms
JAWBONE groundtruth validation
Detection of Soot Plumes
Plume identification and tracking:• denoise background subtracted image
• identify excess/deficit in luminosity space
• cross check object location in color space
• localization and probability weighted tracking of
centroids
Use cases:• Plume ID
• Plume rate, repeaters
• urban winds
• carbon vs steam emissions
• TOO (triggered) observations for composition
raw image
background subtracted
Source: Dobler, et al.
Hyperspectral Imaging of Manhattan Urban Lights
36
Short-term variability in thethermal IR
Reference image
~ 2 min later
Differenceimage
A cooling towerlit up.
Proxy for energyconsumption• Occupation ?• Activity ?
Temperature/ Emissivity Separation
T E
R
Materials
Reflections
Broadband Thermography
For building envelops
Chlorodifluoromethane (Freon 22)
Chlorodifluoromethane (Freon 22)
Chlorodifluoromethane (Freonne)
Thermal Image
• Freon 22 appears in absorptionand emission
• Ozone-destroying moleculebeing phased out in the US
Acetone Plume April 7, 2015 Midtown Manhattan
(1.5km distance) 4mx6m
Acetone is used extensively in dry cleaning
A proxy for economic activity?
Couture Cleaners679 Washington Street
Bringing multiple data sources together
Urbane: A 3D Framework to Support Data Driven Decision Making in Urban Development
Nivan Ferreira, Marcos Lage, Harish Doraiswamy, Huy Vo, Luc Wilson, Heidi Werner, Muchan Park, Claudio Silva
Urbane
• 3D based framework that enables a data-driven approach for decision making in urban development
– Interactive exploration of the data at multiple scales
– Test of “what-if” scenarios and compute impact of proposed changes
– Use of computer graphics techniques to achieve interactivity
– Designed in collaboration with architects
Huy Vo et. al
Urbane in action
CUSP EDUCATIONMS in Applied Urban Science and Informatics (MSAUSI)
MSAUSI - Urban Informatics Track (One Year)
PRE-FALL
Urban Computing
Skills Lab
City Challenge Week
FALL
Principles of Urban
Informatics
Civic Analytics &
Urban Intelligence
Select 2 from:
Urban Spatial
Analytics
Applied Data Science
Urban Decision
Models
SPRING
Machine Learning for
Cities
Urban Science
Intensive I: City
Operations & Applied
Informatics
Select 2 electives:
Data Science Elective
Domain Application
Elective
SUMMER
Data Governance,
Ethics, and Privacy
Urban Science
Intensive II: Practicum
Select 1 from:
Science of Cities
Research Seminar
Civic Technology
Strategy
Advance Topics in
Urban Informatics
Winter
Week
Spring Break
Data Dive
Informatics Core
Urban Core
Optional Programs
PRE-FALL FALL 2015 SPRING 2016 SUMMER 2016
Urban Computing Skills
Lab
4001 Civic Analytics &
Urban Intelligence
5006 Machine Learning
for Cities
1007 Data Governance
Ethics & Privacy
City Challenge Week5003 Principles of Urban
Informatics
9001 Urban Science
Intensive I: City
Operations & Applied
Informatics
6001 Science of Cities
Seminar
5004 Applied Data
Science
3007 Urban Spatial
Analytics
6004 Advanced Topics in
Informatics: Managing IT
5005 Urban Decision
Models
5008 Big Data
Management & Analysis
6004 Advanced Topics in
Informatics: Machine
Sensing & Learning
3007 Urban Spatial
Analytics
6003 Civic Technology
Strategy & Management
9002 Urban Science
Intensive II: Practicum
6008 Operations
Research for Cities
INFORMATICS CORE
URBAN CORE
ELECTIVES
CUSP 2015 - 2016 Classes Offered
Inaugural Class of 2014 ProfileInaugural Academic Year: September 2013 – July 2014
24 21% 27 36% 3.5Inaugural Class
(including 1 Adv. Cert.)
Selectivity Years
Average Age
Female Average
Undergraduate GPA
20 48% 9 4 28%Undergraduate
Disciplines
International Countries
Represented
Years Average
Work Experience
With Graduate Degree
Fall 2014 CohortClass of 2015 Profile – 2nd Cohort (64 students)
19 COUNTRIES 45% FEMALE 28 AVERAGE AGE
5YRS. AVG. WORK
EXPERIENCE32%GRAD
DEGREES completed/in-progress
6NYC
EMPLOYEES
Fall 2014 CohortClass of 2016 Profile – 3rd Cohort (82 students)
18 COUNTRIES 35% FEMALE 27 AVERAGE AGE
3.5 AVG. YRS. WORK
EXPERIENCE30% W/ GRAD
DEGREES 7
NYC
EMPLOYEES
CONFIDENTIAL: for intended recipient only 52June 27, 2016
Sample Capstone Projects 2016
Urban Science Intensives
- Capstones
Public/Nonprofit 40%
Technology 27%
Consulting 10%
Start Own Business
10%
Other 13%
Construction3% Educational
Services7%
Finance or Insurance
11%
Information32%
Management of Companies or Enterprises
3%
Other Services (Except Public Administration)
7%
Professional, Scientific or Technical Services
29%
Retail Trade4%
Transportation or Warehousing
4%
Talent Pipeline: CUSP Graduate Placement
• 17zuoye
• Accenture
• AECOM
• American Express
• Apple Inc.
• Bloomberg LP
• Booz Allen Hamilton
• Capital Energy Consultants
• Case Commons
• Cityzenith
• ClearGrid
• Coalition for Queens
• Collectivei
• CUNY Office of Research,
Evaluation, and Program Support
• Data & Society Fellow
• Datapolitan
• Descartes Labs
• DMI-Digital Management Inc.
• Enigma
• ENSTOA
• HR&A Advisors
• Huawei Technologies Co. Ltd.
• IBM Watson
• Memorial Sloan Kettering
• Motion Global
• New Lab
• New York City Department of City
Planning
New York City Department of
Education
• New York City Department of
Education
• New York City Economic
Development Corporation
• New York City Police Department
• New York Metropolitan
Transportation Authority
• New York State Office of the
Attorney General
• New York University Center for
Urban Science + Progress
• Phillips Innovation Group
• Prism Consultants
• Salzburg University of Applied
Science
• San Francisco Municipal
Transportation Agency
• simulmedia
• Taipei Cultural Center in New York
• Thinkful
• Two Sigma
• U.S. Department of the Treasury
Office of Financial Research
• Urban Compass
• 4 startups
Class of 2015 ▪ Cohort Size: 64 (includes part-time)
Placement rate: 86% (6 months after graduation)
Average salary: $94,000
Class of 2014 and 2015 Placements to Date
Class of 2014 ▪ Cohort Size: 23
Placement rate: 100%
Average salary: $90,000
Talent Pipeline: CUSP Graduates
Further points
• Urban ontology? Data standards?
• Data hoarding
• Who are the customers?
– City governments? Firms? Citizens?
• Citizen acceptance (Privacy? Control?)
• Putting the “citizen” in smart cities