Bowdoin Carla P. Gomes(Lead PI Cornell University) Thomas Dietterich(PI Oregon State University) Jon Conrad(Co-PI Cornell University) John Hopcroft(Co-PI

Bowdoin

Carla P. Gomes (Lead PI Cornell University) Thomas Dietterich (PI Oregon State University)Jon Conrad (Co-PI Cornell University)John Hopcroft (Co-PI Cornell University)

Computational Sustainability:Computational Methods for a Sustainable

Environment, Economy, and Society

2

Sustainability and Sustainable Development

The 1987 UN report, “Our Common Future” (Brundtland Report):

Raised serious concerns about the State of the Planet.

Introduced the notion of sustainability and sustainable development:

Sustainability: “development that meets the needs of the present without compromising the ability of future

generations to meet their needs.”

Stated the urgency of policies for sustainable development.

Gro BrundtlandNorwegian Prime Minister

Chair of WCED UN World Commission on Environment and Development,1987.

3

Follow-Up Report: Intergovernmental Panel on Climate Change (IPCC)

(2007)

[Nobel Prize with Gore 2007]+130

countries

Global Warming

Erosion of Biodiversity

Examples:

•The biomass of fish is estimated to be 1/10 of what it was 50 years ago and is declining.

•At the current rates of human destruction of natural ecosystems, 50% of all species of life on earth will be

extinct in 100 years.

”There are no major issues raised in Our Common Future for which the foreseeable trends are favourable.”

4

Computer scientists can — and should — play a key role in increasing the efficiency and effectiveness of the way we manage and allocate our natural resources, while enriching and transforming Computer Science.

Vision

5

I Research Goals and Agenda

II Our Team

III Project Management, and Collaboration Plan

IV Knowledge Transfer, Education, and Outreach

V Value-Added as an Expedition

Outline

6


II Our Team




Outline

7

Overall Goal of our Expedition

To establish a new field, Computational Sustainability:

Focused on computational methods for balancing environmental, economic, and societal needs for a sustainable future.

To inject Computational Thinking into Sustainability that will provide:

– new insights into sustainability questions;– new challenges and new methodologies in Computer Science (Analogous to Computational Biology)

We will create the Institute for Computational Sustainability (ICS),

– to perform and foster research in Computational Sustainability– to establish a vibrant research community, reaching far beyond the

participating members of this Expedition.

8

Sustainability Themes

9

E.g. Wildlife Corridors

I Conservation and Biodiversity

II Balancing Socio-economic Demands and the Environment

Sustainability Themes

E.g. Policies for harvestingrenewable resources

III Renewable Energy

E.g. Biofuels

Ethanol Refinery

Farmers

Fueldistributors

Environmentalimpact

ConsumersEnergycrops

Non-energycrops

Social welfare

Foodsupply

Gasolineproducers

Water quality

Soil quality

Local airpollution

Biodiversity

Energymarket

Economicimpact

10

Computational Challenges

11

Challenges in Constraint Reasoning and Optimization: Conservation and Biodiversity: Wildlife Corridors

Wildlife Corridors link core biological areas, allowing animal, seed, and pollen movement between areas;

Typically: low budgets to implement corridors.

Computational problem Connection Sub-graph Problem

Find a sub-graph of G that:

contains the reserves; is connected; with cost below a given budget;

and with maximum utility

Connection Sub-Graph - NP-Hard

Given a graph G with a set of reserves:

Connection Sub-graph Problem

Worst Case Result --- Real-world problems possess hidden structure that can be exploited allowing

scaling up of solutions Science of Computation.

Our Expedition is partnering with the Conservation Fund which works with US Fish & Wildlife Service and the Nature Conservancy on preserving habitat in

the Northern Rockies .

Scaling up Solutions by Exploiting Structure:

Typical Case AnalysisIdentification of Tractable Sub-problemsStreamlining for OptimizationStatic/Dynamic Pruning

5 km grid (12788 land parcels):

minimum cost solution

5 km grid (12788 land parcels):

+1% of min. cost

Glacier Park

Our approach reduced corridor cost from $1 Billion to $10 Million [Conrad et al. 2007]

Yellowstone

Salmon-Selway

Interdisciplinary Research Project (IRP):Wildlife Corridors (Amundsen, Conrad, Gomes, Selman + postdoc + 2 Ph.D. students )

Real world instance:

Corridor for grizzly bears in the Northern Rockies, connecting:

YellowstoneSalmon-Selway EcosystemGlacier Park

(12788 nodes)

13

• Multiple species (hundreds or thousands), with interactions (e.g. predator/prey).

• Highly stochastic environments

• Different models of land acquisition (e.g., purchase, conservation easements, auctions) typically over different time periods

• Dynamical modelsDynamics of species; Movements and migrations;

Themes - Transformative Synthesis: Optimization and Machine Learning

Dynamics and Machine LearningDynamics and Optimization

Additional Levels of Complexity: Stochasticity, Uncertainty, Large-Scale Data Modeling

• Spatially-explicit aspects within-species

IRP Collective Inference on

Markov Models for Modeling Bird Migration

IRPNisource:

Conservation Fund (14 states)

IRP Native Plant Habitat

Recovery Victoria (Australia)

Combinatorial Auctions

IRP Joint Public/Private Management for Biodiversity

Negative observationsPositive observations

Ruby-throated Hummingbird Sightings

OutreachCornell’s eBirdCitizen Science

10M+ bird observations since 2002, by general public

14

Challenges in Dynamic Models and Optimization :

Balancing Socio-Economic Demands and Environment:

Economy

Increasing Complexity: multiple renewable resources interactions

Example of a Biological Growth Function F(x):Logistics map: x t+1 = r xt (1 - xt), r is the growth rate

Non

-line

ar d

ynam

ics

Key Issues in Dynamical Models: multiple scales and multistability

IRP: Rotational Management of Fishing GroundsIRP: Joint Public/Private Management for Biodiversity

We are interested in problems that involve policy decisions (e.g. when to open/close a fishery ground over time). Uncharted territory: Combinatorial optimization problems with an underlying dynamical model.

Transformative Synthesis: New Class of Hybrid Dynamic Optimization Models

IRP: Fire Management in Forests

15

Challenges in Highly Interconnected Multi-Agent Systems:Renewable Energy

Ambitious mandatory goal of36 billion gallons of renewable fuels by 2022

(five-fold increase from current level)

Energy Independence and Security Act(Signed into law in Dec. 2007)

Farmers

Fueldistributors

Environmentalimpact


Non-energycrops

Social welfare

Foodsupply

BiofuelRefineries

Water quality

Soil quality

Local airpollution

Biodiversity

Energymarket

Economicimpact

IRP Large Scale Logistics Planning for Biofuels

Farmers

Fueldistributors

Environmentalimpact


Non-energycrops

Social welfare

Foodsupply

Gasolineproducers

Water quality

Soil quality

Local airpollution

Biodiversity

Energymarket

Economicimpact

Current approaches limited in scope and complexity E.g. based on general equilibrium models (e.g., Nash style) Strong convexity assumptions to keep the model simple enough

for analytical, closed-form solutions (unrealistic scenarios) Limited computational thinking

Transformative research directions More realistic computational models in which meaningful solutions can be computed Large-scale data, beyond state-of-the-art CS techniques Study of dynamics of reaching equilibrium — key for adaptive policy making!

IRP Impact of Biofuels: Dynamic Equilibrium Models

IRP Impact of Land-use on Climate

Realistic Computational Economic Models

17

Science of Computation

Our approach:

The study computational problems as natural phenomena in

which principled experimentation, to uncover hidden

structure, is as important as formal analysis

Science of Computation,

Small world phenomenon

Pioneered science of networks

Phase Transitions in ComputationLed to interactions between CS,

statistical physics, and math

[Selman & Kirkpatrick][Watts & Strogatz] [Gomes et al.]

Heavy tails in ComputationLed to randomization and

portfolios in combinatorial search

Our team has a track record of making compelling scientific discoveries using such an approach.

18

Constraint Satisfaction and Optimization

Data and Uncertainty

DistributedHighly

InterconnectedComponents / Agents

Complexity levels in Computational Sustainability Problems

Increasing

Com

plexity

Sci

ence

of

Com

puta

tion

Dynamics

Study computational problems as natural phenomena Science of Computation

Many highly interconnected components; From Centralized to Distributed:

Computational Resource Economics

Multiple time scales From Statics to Dynamics: Dynamic Models

Large-scale data and uncertainty Machine Learning, Statistical Modeling

Overall Research Vision:Transformative Computer Science Research:

Driven by Deep Research Challenges posed by Sustainability

Design of policies to manage natural resources translating into large-scale optimization and learning problems, combining a mixture of discrete and continuous effects, in a highly dynamic and uncertain environment increasing levels of complexity

Complex decision models Constraint Reasoning and Optimization

19

Transformative Computer Science Research:Driven by Deep Research Challenges posed by Sustainability

Design of policies to effectively manage Earth’s naturalresources translate into large-scale decision/optimization and learning problems, combining a mixture of discrete and continuous effects, in a highly dynamic and uncertain environment increasing levels of complexity

Data & Machine

LearningDietterichPI

WongCo-PI

ChavarriaH

Environmental &Socioeconomic

Needs

Ren

ewab

le E

nerg

y

Biodiversity

Dynamical Models

Guckenheimer

Strogatz

ZeemanPI,W

YakubuAA

Constraint Reasoning

& OptimizationGomes PI,W

, Hopcroft Co-PI

SelmanCo-PI, ShmoysCo-PI

Resource Economics,Environmental

Sciences & EngineeringAlbersCo-PI,W, Amundsen, Barrett, Bento,

ConradCo-PI, DiSalvo, MahowaldW, MontgomeryCo-PI,W

Rosenberg,SofiaW, WalkerAA

Transformative Synthesis:

Dynamics &Learning

Tran

sfor

mat

ive

Syn

thes

is:

Dyn

amic

s an

d

Opt

imiz

atio

n

Transformative

Synthesis:

Optim

ization

& Learning

Study computational problems as natural phenomena Science of Computation

Many highly interconnected components; From Centralized to Distributed:

Computational Resource Economics

Multiple scales (e.g., temporal, spatial, geographic) From Statics to Dynamics: Dynamic Models

Large-scale data and uncertainty Machine Learning, Statistical Modeling

Complex decision models Constraint Reasoning, Optimization, Stochasticity



20


II Our Team




Outline

21

Multi-institutional, Multidisciplinary Research Team 6 Institutions, 7 colleges, 12 departments

Institute for ComputationalSustainability

GomesCS

Cornell

ZeemanAppl. MathBowdoin

DietterichCS

OSU

MahowaldEarth &

Atmos. Sci.Cornell

AmundsenConservation

PlanningCons. Fund

StrogatzAppl. Math

Cornell

GuckenheimerAppl. Math

Cornell

BentoRes. & Env. Economics

Cornell

ConradRes. and Env.

EconomicsCornell

WongCS

OSUDiSalvo

ChemistryCornell

ShmoysCS & ORCornell

YakubuAppl. Math

Howard

SelmanCS

Cornell

HopcroftCS

Cornell

WalkerBiology

& Eng.

Cornell

AlbersRes. & Env.

Econo.OSU

ChavarriaHPC. PNNL

RosenbergConservation

BiologyCornell

MontgomeryRes. & Env.Economics

OSU

SofiaBiologyCornell

BarrettRes. & Env. Economics

Cornell

Cornell University,Ithaca, NY

Oregon State UniversityCorvallis, OR

Bowdoin CollegeBrunswick, ME

Howard UniversityWashington, DC

Liberal arts undergraduate college

Historically Black University

Leading producer of African-American Ph.D.s

Focus on Science, Technology, Engineering, and Mathematics

Culture of undergraduate research

.

Tradition of public service as part of its land-grant mission (private & state university)

Several centers focusing on sustainability research and dissemination results to community and policy makers

Strong programs in computer science, agriculture, natural resources, biology, renewable energy, and environment

Conservation Fund Nationwide

Since 1985, the Fund and partners have safeguarded over 6 million acres of wildlife habitat, forests, and community greenspace.

Non-profit org. focused on conserving natural resources.

Pacific NorthwestNational Laboratory

Department of Energy's (DOE's) laboratory: focus onenvironmental science, climate change, remediation, and security.

Access to HPC Systems

Provost’s initiative in Ecosystem InformaticsProvost’s campus wide initiative in Sustainability

Tradition in agricultural extension service & research stations

#1 ranked Forestry School

Strong Ecosystem Informatics (IGERT, NSF Summer Inst.)

Strong programs in computer science, fisheries & wildlife, atmospheric Sciences, oceanography, and environmental sciences

23

Institutional Support

Strong institutional support:

• From top level management (president, provost)

• Researchers eager to share data, problems, and anxious for new computational models to solve their sustainability problems

• Cornell Center for a Sustainable Future (CCSF) with organizational support and matching funds (5% non-federal money).

Institute for Computational Sustainability

And this space!

24


II Our Team




Outline

25

Integrated Management

Institute for Computational Sustainability:

Director Carla P. Gomes (PI Cornell University)Associate Director David Shmoys (Co-PI Cornell Univeristy)Deputy Director Thomas Dietterich (PI Oregon State University)Deputy Director Mary Lou Zeeman (PI Bowdoin University)

Plan, coordinate, and evaluate exciting and aggressive research, education, and outreach program. Disseminate computational sustainability.

Executive Committee: PI’s + Co-PI’s Meets once a month.

Advisory Board: Prominent researchers in fields related to computational sustainability, both from the application side (2 people) and the computer science side (2 people). Will also invite an NSF representative. Meets once a year.

Collaboration Plan: Through Interdisciplinary Research Projects (IRPs, shared personnel), executive committee meetings, exchange students and faculty, three-day annual meetings, meetings at regular CS conferences.

26

Interdisciplinary Research Projects (IRPs):The Building Blocks of our Expedition

IRP Name Faculty Team

1. Wildlife Corridors for Grizzly BearsAmundsen, Conrad, Gomes, Selman, Shmoys

2. BiofuelsBento, Gomes, Mahowald, Shmoys, Strogatz, Walker, Wong

3. Bird ConservationRosenberg, Conrad, Dietterich, Gomes, Hopcroft, Strogatz, Zeeman

4.Native Plant Habitat Recovery in Victoria, Australia

Dietterich, Gomes, Selman

5.Joint Public/Private Management for Biodiversity

Amundsen, Montgomery, Dietterich, Gomes, Hopcroft

6. Fire Management in Forests Albers, Conrad, Guckenheimer, Selman

7. Rotational Management of Fishing Grounds Conrad, Guckenheimer, Yakubu, Zeeman

8. Pastoral Systems in East Africa Barrett, Conrad, Wong

Seedling IRPs

27

Nurturing New IRPs

Application/Synthesis Facilitators


Dynamics & Optimization

Optimization & Learning

Learning & Dynamics

Agent Integration

Dissemination, Outreach, &

Diversity

Conservation and Biodiversity ✔ ✔ ✔ ✔

Conrad, Montgomery,

Gomes

Balancing Socio-Economic Demands and Environment

✔ ✔ ✔ ✔ ✔ ✔

Albers, Conrad,

Guckenheimer

Renewable Energy ✔ ✔ ✔ ✔ ✔

Bento, Gomes, Shmoys

Dietterich, Hopcroft, Selman

Guckenheimer, Shmoys,

Zeeman

Dietterich, Gomes, Hopcroft

Guckenheimer, Wong

Bento, Gomes, Shmoys

Hopcroft, Zeeman

Synthesis Facilitators

Computational Themes

Ap

plic

ati

on

F

ac

ilita

tors

Ap

plic

ati

on

Are

as

28


II Our Team




Outline

Institute Activities

Web Portal

Summer REU program targeting minority students

Annual symposium

Host visitingScientists

CatalogOf

ProblemsPanels & tutorials at major conferences

Computational Sustainability follow-up to State of the Planet course

Research seminarseries Cornell

Cooperative Extension

Citizen Science Project

ICS

Research

Outreach

Building Research Community

Postdocs

Doctoral students Honors

projects

Coordinating transformative

synthesis collaborations

Conservation Fund

Cornell Center for a Sustainable Future

OSU Alliance for Computational Sustainability

Education

Interdisciplinary Research

Projects (IRPs)

30

Education and Outreach

Research on Computational Sustainability will significantly broaden the field of computer science and attract a new generation of students who traditionally may not have considered studying computer science --- thus contributing to a revitalization of CS education.

Undergraduate and graduate students will be drawn into hands-on computational sustainability research by joining IRP-teams.

The ICS will organize regular seminars and an annual symposium and summer school.

31

State of the Planet Course

“Students unite to create State of the Planet course” - K.L. Rypien et al., Nature, Vol 447, July 2007, p775

- mentors Tom Eisner and Mary Lou Zeeman (co-PI)

250 students, 45 different majors. 95% recommend it to their peers:

• “… the class … has started me thinking about career paths I hadn’t considered before. I find most of the lectures inspiring and hopeful; they leave me feeling that I can actually make a difference in the world.”

• “Amazing course! Very thought-provoking, interesting, varied, and exciting. Best course I've taken at Cornell so far. I have been recommending it to everyone.”

We will create a companion course on Computational Sustainability

Renewable Energy Biodiversity Carbon and climateFood & water

32

Integration of Research and Outreach Example: Citizen Science at the Cornell Laboratory Of Ornithology

Citizen Science empowers everyone interested in birds to contribute to research.

Increase scientific knowledgeGather meaningful data to answer large-scale research questions

Increase scientific literacyEnable participants to experience the process of scientific investigation and develop problem-solving skills

Increase conservation actionApply results to science-based conservation efforts

Additional Channels for Impact and Outreach

We will leverage a host of sustainability centers and programs at our 6 institutions.

Researchcommunity& students

Conservation FundOle Amundsen

CornellCooperative Extension

A. Bento

Institutefor Computational

Sustainability

Community, policy makers,

local and state leaders

Community and RuralDevelopment Institute

Cornell Center forSustainable FutureDirector: F. DiSalvo

Institute for EnvironmentEnergy & Policy

Director: A. Bento

Northeast Sun Grant Institute at CornellDirector: L. Walker

New York Sea-Grant Cornell Extension

Jon Conrad

Lab. of OrnithologyK. Rosenberg

Earth & Atmospheric Sciences

N. Mahowald

African Food and SecurityNatural Resources Mgmt.Prgm

Director: C. Barrett

USDA ForestrySciences Lab

EPA WesternEcology Division

OSU IGERTEco-SystemInformatics

Howard UnA.Yakubu

CornellBiofuels Lab

L.Walker

36


II Our Team




Outline

37

Value-Added as an Expedition

Fundamentally new intellectual territory for computer science

Unique societal benefits

Tremendous potential for recruiting new groups (including under-represented minorities) to computer science

Establishing the field of Computational Sustainability requires critical mass and visibility that cannot be achieved with piece-meal efforts

Our research proposal is fundamentally cross-disciplinary:– IRPs require large teams involving both domain scientists and computer

scientists– Transformative syntheses are also across disciplines in CS

NSF is the agency best positioned to lead this initiative

38

Summary

Expedition will lead to:

1) Foundational contributions to computer science driven by Sustainability questions.

2) Societal and environmental impact --- development of computational methods to alleviate sustainability problems (e.g. significant opportunity for more efficient use of natural resources).

3) Establishment of the field of Computational Sustainability.

4) Integration of research, education, and outreach. New courses and seminars integrated with Interdisciplinary Research Projects.

5) Increasing diversity in computer science. Bringing in a new generation of students, traditionally not drawn to CS; also, broadening the public image of CS.

Thank You!

Documents

Bowdoin Carla P. Gomes(Lead PI Cornell University) Thomas Dietterich(PI Oregon State University) Jon Conrad(Co-PI Cornell University) John Hopcroft(Co-PI