Download - Bowdoin
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
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
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
ConsumersEnergycrops
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
ConsumersEnergycrops
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
Science of Computation
Science of Computation
20
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
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
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
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
Science of Computation
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
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
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
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
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!