Adventures in Urban Informatics - Steven Koonin

7/24/2019 Adventures in Urban Informatics - Steven Koonin

1/76

Adventures in Urban InformaticsUniversity of California

February, 2016

Dr. Steven E. Koonin, CUSP Director

[email protected]://cusp.nyu.edu


2/76

Big Cities + Big Data

All cities must be better for global issues

Individual cities need to be best for competitiveness in

talent, capital, Be efficient, resilient, sustainable

Address citizen quality of life, equity, engagement

The world is urbanizing

Cities are the loci of

consumption, economic

activity, and innovation

Cities are the cause of our problems

and the source of the solutions


3/76

Big Cities + Big Data

Informatics

capabilities areexploding Storage, transmission,

analysis

Proliferation of staticand mobile sensors

Internet of things

Global network traffic, 30% CAGR


4/76

JUST HOW DID A PHYSICIST WIND

UP IN THIS BUSINESS?


5/76

Properly acquired, integrated, and analyzed, data can Take government beyond imperfect understanding

Better (and more efficient) operations, better planning, better policy

Improve governance and citizen engagement Enable the private sector to develop new services for citizens,

governments, firms

Enable a revolution in the social sciences

Environment

Meteorology, pollution,

noise, flora, fauna

People

Relationships, location,

economic /communications

activities, health, nutrition,opinions, organizations,

Infrastructure

Condition, operations

What does it mean to instrument a city?


6/76

Urban Data Urban data have been collected for millenniastatistics (sttstks) n.1. The mathematics of the collection, organization, andinterpretation of numerical data, especially the analysis of populationcharacteristics by inference from sampling

From German Statistik,political science, from New Latin statisticus, of state affairs, fromItalian statista,person skilled in statecraft, from stato, state, from Old Italian, from Latinstatus,position, form of government.

Sparseness and quality have limited urban science difficult to usefully measure the urban system, test hypotheses

But new data technologies completely recast the study of cities digital records, sensors, computing power, analytical techniques

unprecedented granularity, variety, coverage, and timeliness

When you can measure what you are speaking about, and express it in numbers,

you know something about it; when you cannot express it in numbers, your knowledge

is of a meager and unsatisfactory kind. Lord Kelvin, 1883


7/76

Organic data flows Administrative records (census, permits, )

Transactions (sales, communications, )

Operational (traffic, transit, utilities, health system, )

Twitter feeds, blog posts, Facebook,

Sensors Personal (location, activity, physiological)

Fixedin situ sensors

Crowd sourcing (mobile phones, )

Choke points (people, vehicles)

Opportunities for novel sensor technologies

Visible, infrared and spectral imagery RADAR, LIDAR

Gravity and magnetic

Seismic, acoustic

Ionizing radiation, biological, chemical

Urban Data Sources


8/76

N = 1,150

Mean = 219.5

s.d. = 101.7

0

20

40

60

80

Fr

equency

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

200

400

600

80

0

Fr

equency

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


9/76

Local Law 84 Benchmarking Data

Kontokosta, 2013


10/76


11/76

Wang, P., Hunter, T., Bayen, A.M., Schechtner, K. & Gonzalez, M.C.

Understanding Road Usage Patterns in Urban Areas. Nature, Sci. Rep. 2, 1001; DOI:10.1038/srep01001(2012).

Cell Tower Records for Traffic Analysis


12/76

Daily commute patterns

from phone records

Survey Chicago, Paris

Phone 4X104 in Paris

Model Chicago, Paris


13/76

Taxis as Sensors for NYC

Taxis are sensorsthat 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?


14/76

May 1st

7th

2011

3.6 Million Trips

Train Stations

Airports

Studying Taxi Patterns


15/76

Taxi Rides in Manhattan, October 28 November 3, 2012(Superstorm Sandy)

Juliana Freire, Claudio Silva, et al, NYUPoly


16/76

Lottery Vis

Correlate sales with

Weather Sports team wins

Twitter mood


17/76

From the Willis Tower, Chicago


18/76

Photo by Tyrone Turner/National Geographic

Other synoptic modalities: Hyperspectral, RADAR, LIDAR,

Manhattan in the Thermal IR

199 Water Street

Built 1993 :: 998,000 sq ftelectricity, natural gas, steam

LEED Certified


19/76

The view from CUSPs Urban Observatory in Brooklyn


20/76

Borough Block & Lots (BBL)

Standard UO view

colored by distance


21/76

Picture merges image captured from video, 3D LIDAR map of NYC, PLUTO

(Primary Land Use Tax Lot Output) database, and LL84 Energy Benchmarking data

Source: Dobler, et al. 21


22/76

OBJECTIVES

Develop a fundamentally new modality forstudying the city from a distance

Identify aggregate patterns of light in thetime-dependent brightnesses of city lights

Leverage these patterns into foundationalcontributions to urban science and urbanfunctioning

Proof of Concept

IMPACT

Urban Science

Determine the underlying drivers of the pulse of the city

Understand the effects of perturbations

City Life

Monitor energy consumption by proxy using light patterns as a measure of buildingoccupancy

Evaluate the effects of disturbances (e.g., light/noise pollution) on public health

Camera: Point Grey Flea 3 USB ; 8.8 Mega-pixels ; Raw image output ; 25mm focal length lens

Observations:

1 image every 10 seconds from Oct 26 to Nov 16, 2013; 3 color images at 25MB each

Total data volume ~4.5TB; Custom data processing pipeline

City Lights Project

Dobler et al.;doi:10.1016/j.is.2015.06.002


23/76

Institutional Review Board

approval of all projects involving

non-open data

Close oversight by CUSP ChiefData Officer

Limited # of pixels per window

(but atmosphere/instrument

effects typically dominate) Aggregate and de-identified

analysis only

Privacy Protections


24/76


25/76

Dynamics of the Urban Landscape

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 toas its light curve


26/76

Pulse of the City Lights


27/76

Daytime Phenomenology

11:00 AM

11:01 AM


28/76

Daytime Phenomenology: Subtle Variations


29/76

Daytime Phenomenology: Subtle Variations (animation)


30/76

Background subtraction: registration to reference image

form 10 absolute difference images from

surrounding frames

construct the minimum difference image pixel by

pixel

Subtracting the Cityraw image

background subtracted


31/76

Crossbuilding view of a boiler plume

Such plumes may not be visible from street level.


32/76

Plumes of Opportunity

Background subtraction: registration to reference image

form 10 absolute difference images from

surrounding frames

construct the minimum difference image pixel by

pixel

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

Upcoming use cases: plume rate

repeaters urban winds

carbon vs steam emissions

TOO (triggered) observations

raw image

background subtracted


33/76

Thoughts on the big science questions Can we document the pulse of the city in its various dimensions?

Normal? Variability? Correlations? Response to perturbations? Predictability? Precursors?

How do the macro observables arise from micro behavior? Santa Fe scaling?

Physical structure of cities?

Decision rules in agentbased models

Role of geography? Culture? Policies?


34/76

The CUSP PartnershipNational Laboratories

Lawrence Livermore

Los Alamos

Sandia

Brookhaven

Industrial Partners

IBM

Microsoft

Xerox

AECOM, Arup, IDEO

University Partners

NYU/ NYUPoly

University of Toronto

University of Warwick

CUNY IITBombay

Carnegie Mellon University

City & State Agency Partners

The City of New York

Metropolitan Transit Authority

Port Authority of NY & NJ

Buildings

City Planning

Citywide Administrative

Services

Design and Construction

Economic Development

Environmental Protection Finance

Fire Department

Health and Mental Hygiene

Information Technology

and Telecommunications

Parks and Recreation

Police Department

Sanitation Transportation

Cisco

Con Edison

Lutron

National Grid

Siemens


35/76

Education Programs at CUSP

Master of Science in Applied UrbanScience & Informatics

F/T (One Year)

P/T (Two Year)

Civics Analytics Track

Advanced Certificate in Applied UrbanScience & Informatics


36/76

CENTER FOR URBANSCIENCE+PROGRESS

Graduate Programs at CUSP Interdisciplinary and cutting edge approach that links data science,

statistics and analytics, and mathematics with complex urban

systems, urban management, and policy.

Corecurriculum

Urban core Foundational understanding of the theories of urban planning andthe application of data-driven approaches to urban challenges.

Informatics core Fundamentals of data science/computer science, data management,

data mining, visualization, model selection, and machine learning

tools to urban problems and datasets.

Tracks UrbanInformatics For students who are looking for deep training in data science andinformatics as applied to cities.

Civic Analytics For students who will utilize analytics and data-driven decision-

making techniques to inform urban operations and policy decisions.

Length One Year Full-

Time Program

A research- and project-intensive environment

Two Year Part-

Time for

Working

Professionals

Evening courses with numerous opportunities for networking with

peers, faculty, and experts in the industry.


37/76


MSAUSI - URBAN INFORMATICS TRACK (One Year)

PRE-FALLPRE-FALL

1001 Urban

Computing Skills Lab

1001 Urban

Computing Skills Lab

1000 City Challenge

Week

1000 City Challenge

Week

FALLFALL

5003 Principles of

Urban Informatics

5003 Principles of

Urban Informatics

4001 Computational

Urban Policy &

Planning

4001 Computational

Urban Policy &

Planning

Select 1 from:

7007 Urban Spatial

Analytics

9002 Urban Decision

Models

SPRINGSPRING

5006 Machine

Learning for Cities

5006 Machine

Learning for Cities

9001 Urban Science

Intensive I: City

Operations & Applied

Informatics

9001 Urban Science

Intensive I: City

Operations & Applied

Informatics

SUMMERSUMMER

1007 Data

Governance, Ethics,

and Privacy

9002 Urban Science

Intensive II: Practicum

9002 Urban Science

Intensive II: Practicum

Select 1 from:

6001 Science of Cities

Research Seminar

6003 Civic Technology

Strategy

Winter

Week

Winter

Week

Spring Break

Data Dive

Spring Break

Data Dive

Informatics Core

Urban Core

Optional Courses

5004 Applied DataScience

6004 Advanced Topics

in Urban Informatics

Year 1

Data Science Elective

Domain Application

Elective

Ad i i S Cl f 2014


38/76


Admissions Summary, Class of 2014Inaugural 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 Cohort

2015 Class Highlights


39/76


Fall 2014 Cohort2015 Class Highlights

19COUNTRIES

(111% ) 45%FEMALE

(275% ) 28AVERAGE

AGE (3% )

5YRS. AVG. WORK

EXPERIENCE

(25% )32%

GRAD

DEGREEScompleted/in-progress

66NYC

EMPLOYEES

F ll 2015 C h t


40/76

Fall 2015 Cohort

NEW STUDENTS

August 2015

8787

F ll 2015 C h t


41/76

Fall 2015 Cohort

NYC Employees

August 2015

1010

S d R h (GRA P 19 j )


42/76

Student Research (GRA Program, 19 projects)

Aerial Analytics/Hyperspectral imagingBluetooth Tracking Researcher

Buildings Informatics Energy Index

Dynamics of the City Lights

Economic Impacts of Public Parks and Greenspaces

Efficiently Indexing the New York City Open Data For Spatial-

temporal-keyword QueriesEmotion Sensing

Garbage Identification

Machine Learning and Computational Statistics for NYPD

MTA Project

Parks Utilization and Attendance

Pedestrians and Vehicles: Interactions at Intersections

Quantified Community Research Initiative

Social Cities Initiative

SONYC

TaxiVis

Traffic Safety from Video Recordings

Understanding the spatial structure of crime

Urban ThermodynamicsUsing MTA bus-time data to determine traffic conditions

2015 Capstone Projects


43/76

2015 Capstone Projects ACCESS NYCData Analysis

Analysis of Citibike Data and Modeling of TimeDependent OriginDestination Matrices

Building & Sustainability Informatics

BusVis: Interactive Exploration of NYC Bus Data

Crime and Policing Analytics in New York City Digital Equality: Sensing, citizen science, data analytics &

visualization

From Light Variability to Energy Consumption

LearnrA Seamless Education Volunteering Platform New York City Economic Map

New York Open Government

Parks Quality Assessment

Quantifying Particulate Matter ExposureDistribution in NYC Quantitative Analyses of Urban Topography

Urban Waste Analytics

Using Social Media to Predict Urban Transportation

Al i P fil


44/76


Alumni Profiles

Advance your CareerAliya Merali (B.S. Physics)Director of Learning and Access, Coalition For Queens

Become a Data Scientist

Warren Reed (B.S. Chemical Engineering)Data Scientist, Office Of Financial Research

Career ChangerAlex Chohlas-Wood (B.A. Studio Arts)

Director of Research & Evaluation, NYPD

CUSP Facilities/Capabilities


45/76

CUSP Facilities/Capabilities

Under Development

Data facility

Quantified Community

SONYC project

Urban Observatory

Data Facility


46/76

Data Facility

Overview

Omnivorous ingestion to a repository for NYCrelated data

Objective and Goals Make data interoperable, with proper multilayered access protocols

Data

Data from City agencies on operations, schedules, maps, etc.

Working with the Mayors Office of Analytics

Start with the open datasets

Includes proprietary data, social media data, CUSPgenerated data

CUSP Chief Data Officer oversees ethical, legal, and social issues

https://datahub.cusp.nyu.edu/dataset


47/76

NYC DataBridge


48/76

Privacy, Big Data, and the Public Good:


49/76

The book identifies ways in which vast new sets of data on human beings can

be collected, integrated, and analyzed to improve urban systems and quality of

life while protecting confidentiality. Sponsored by CUSP, the American Statistical

Association, its Privacy and Confidentiality subcommittee, and the ResearchData Centre of the German Federal Employment Agency.

Editors: Julia Lane, American Institutes for Research; Victoria Stodden, Columbia;

Stefan Bender, The German Federal Employment Agency; Helen Nissenbaum, NYU

Chapter AuthorsSteve Koonin, CUSP; Frauke Kreuter, U-MD and Richard Peng, Johns Hopkins; Alessandro Acquisti, Carnegie

Mellon University; Robert Goerge, UChicago; Helen Nissenbaum, NYU; Kathy Strandberg, NYU;

Paul Ohm, Colorado; Victoria Stodden, Columbia; Alan Karr, National Institute of Statistical Sciences andJerry Reiter, Duke University; John Wilbanks, Sage Bionetworks/Kauffman Foundation;

Cynthia Dwork, Microsoft; Alexander Pentland, et al., MIT; Carl Landwehr, George Washington

University; Peter Elias, University of Warwick.

Privacy, Big Data, and the Public Good:

Frameworks for Engagement

The Quantified Community


50/76

The CUSP Quantified Community (QC) will be

a fully instrumented urban neighborhood that uses

an integrated, expandable sensor network and

citizen engagement to support the measurement,integration, and analysis of neighborhood

conditions.

Through an informatics overlay, data on physical

and environmental conditions and use patterns will

be processed in real-time to maximize

operational efficiencies, improve quality of life

for residents and visitors, and drive evidence-

based planning.

Kontokosta, et al.

The Quantified Community

Understanding the Patterns of Urban Life

Buildings

Resource consumption;

Infrastructure

Solid waste, storm-water


51/76

p ;

indoor air quality;

productivity, health

measures

People

Behavior; mobility;

health; activity; social

networks, metagenomics

Environment

carbon emissions; air

pollution and particulates;

noise; climate

,

management, power

generation/distribution

Safety and SecurityNetwork Security,

Situational Awareness,

Emergency Management

Integration, Event

Forecasting


52/76

Sound of New York City

Cyberphysical systemfor largescale, continuousmonitoring of noise pollution

Custom acoustic sensor(~$100/unit), dB measurement

accurate to city agency standards (+/2 dB)

Stateoftheart machine listeningtechnology forautomatic sound source identification in realtime

Also includes citizen science & data visualizationcomponents

54

J. Bello, C. Mydlarz, J. Salamon


53/76

Cyberphysical system for largescale

continuous monitoring of noise pollution

55


54/76

Custom acoustic sensor based on MEMS

microphone technology

56


55/76

Stateoftheart machine listening technology

for realtime sound source identification

57

Urban Observatory


56/76

Provisioned urban vantage point(s) include Downtown Brooklyn

Midtown Manhattan

Suite of boresighted instruments Photometric and colorimetric optical imaging

Broadband IR imaging (SWIR, MWIR, and thermal)

Hyperspectral imaging (trace gases)

LIDAR (building motions, pollution)

RADAR (building /street vibrations, building motion, traffic flow)

Correlative data on the urban scenes Meteorology (temperature, winds, visibility)

Scene geometry (distances, directions, identities of features visible)

Parcel and land use data, building characteristics and activities,building utility consumptions, and real estate valuation data

In situpollution data and location/nature of major sources

In situvehicle and pedestrian traffic for the streets visible

Demographic and economic data

Capability to archive, process, and analyze data acquired

Image processing chains Data warehouse, GIS, Visualization tools

Software and procedures to enhance privacy protection

Personnel and funding to create and operate the above

Urban Observatory

Hyperspectral Imaging of Manhattan Bridge Lights


57/76

Source: Dobler, et al.

Hyperspectral Imaging of Manhattan Bridge Lights

59


58/76

Persistent LWIR Imaging of Manhattan CUSP + Aerospace Corporation

April 615 from HobokenWest Side from the Battery to ~59 Street

128 spectral channels covering 7.6 13.6 m

Plume detection and molecular ID

61


59/76

Northern portion of the view from


60/76

Northern portion of the view from

Hoboken


61/76

GBSS set


62/76

GBSS set

up


63/76

Shortterm

variability in the


64/76

y

thermal IR

Reference image

~ 1 min later

Difference

image

Shortterm

variability in the


65/76

y

thermal IR due

to a cooling tower

lighting up.

Temperature/ Emissivity SeparationR


66/76

T EMaterials

Reflections

BroadbandThermography

For building envelops

Thermal ImageSpectral analysisshows diverse,


67/76

Ammonia

Difluoroethane (Freon)

episodic plumes

A typical spectral fit to pixels with NH


68/76

A typical spectral fit to pixels with NH3

TStatistic for fit to each

of

700 compounds in the

library;NH3is a hit with t~10

8

NH3spectral template

Data fit with template

Residual to fit


69/76

CO2at 200 C

Controled Release Diflouroethane

North of Chelsea piers


70/76

One can at each of three locations


71/76

Number of Captures per Molecule in 9 days


72/76

p p y

ID Count of ID

Ammonia 9,816

Chlorodifluoromethane 1,722

Carbon Dioxide (HITRAN 200C) 1,051

Difluoromethane 412

Carbon Dioxide (HITRAN 300C) 211

1,1,1,2Tetrafluoroethane 197

Pentafluoroethane 190

Carbon Dioxide (HITRAN 100C) 162

Methane 83

Acetyl iodide 65Sulfur dioxide 57

Methane (HITRAN 5C) 39

Chlorofluoromethane 36

2(Diisopropylamino)ethanol 29

Cyclohexanol 29Carbon Dioxide (HITRAN 50C) 28

Other components 416

TOTAL 14,543

Acetone Plume

April 7, 2015 Midtown Manhattan


73/76

p ,

(1.5km distance) 4mx6m


74/76

Acetone is used extensively

in dry cleaning

Couture Cleaners

679 Washington Street

Whats success after 5 years?


75/76

Define and elaborate Urban Science

A vibrant worldclass center pursuing suchNucleate an NYU/NYUPoly community

Implement CUSP facilities

Projects that impact the City and its CitizensCUSP established as a trusted partner to NYC

Support public understanding and engagement Train several hundred people in this new field

Commercialization of CUSP technologies

Bring new tools to the social sciences Begin to franchise the brand globally


76/76

Thank You

cusp.nyu.edu

NYUCUSP@NYUCUSP

Documents

Adventures in Urban Informatics - Steven Koonin