1 may 2008page 1 model collaboration, coordination and… the model web for ecological forecasting
TRANSCRIPT
1 May 2008 Page 1
Model collaboration, coordination and…the Model Web
for ecological forecasting
1 May 2008 Page 2
From Yesterday
Need to assess impacts of change…
“A new generation of integrated modeling is needed” (Tony Janetos)
“Major investment in impact models is needed”
“Need integrated research” (Ex: biofuels/food/crop)
Need “Model comparisons…multi-model analyses…coordinated agenda”
“Need to develop an infrastructure”
“Need to investigate alternative futures”
“Need a new NASA program”
1 May 2008 Page 3
Overview
Bring people together
Review Model Web concept
Discuss strawperson approach to doing it
Focus on: Initial Framework Infrastructure
• Right models?
• Right people?
1 May 2008 Page 4
Current Modeling Environment
Very limited model-model communication
• Technical and especially non-technical barriers
Response
• Try a “web of models” analogous to the WWW
• Can’t just plan and build it
• Facilitate and encourage
1 May 2008 Page 5
Demonstration System
Global ClimateDataset
RegionalClimate Dataset
Species Distribution
Model
Species distribution
TOPS
1 May 2008 Page 6
System Growth
Global ClimateDataset
RegionalClimate Dataset
FireModelTOPS
Species distribution
Species Distribution
Model
1 May 2008 Page 7
System Growth
Global ClimateDataset
RegionalClimate Dataset
FireModel
LandscapeModelTOPS
Species distribution
Fire risk,Intensity, etc
PossibleLandscapes
Species Distribution
Model
1 May 2008 Page 8
Distributed network of interoperating models (and datasets and sensors)
Using web services
Models are harmonized
Multi-disciplinary scope
Model Web: 5-10+ Year Vision (1/2)
1 May 2008 Page 9
Model Web: 5-10+ Year Vision (2/2)
Grows organically within framework of broad goals and data exchange standards
Models and datasets maintained, operated, and served independently
Web access provided to researchers, managers, policy makers, public
1 May 2008 Page 10
Model and Data Space
Model A
Model B
Model CDataset 1
Dataset 2
Model D
Model E
Model F
Dataset 3
Dataset 4
Model G
1 May 2008 Page 11
Model Web "Traditional"
Harmonization of components a lot of work Harmonization of components a lot of work
Communication by web services Communication by API/system calls
Dynamic, interoperable system of systems Static, isolated, integrated systems
Loosely coupled components Tightly coupled components
Distributed system lacking centralized control Centralized system controlled by developers
Open system. New components added without permission Closed system. New components only added by system developers
Organic and opportunistic growth, similar to WWW (but guided) All growth planned
High level of component reuse—once harmonized with other components a component is available to everyone
Low level of reuse; once harmonized with other components remains within the tightly coupled system
Generally high level of data sharing Generally lower level of sharing; focus is on specific questions and tightly coupled components so that only final products are shared
Indeterminate growth Determinate growth as defined by developers
Long term evolutionary process that gradually converges on higher levels of interoperability
Shorter term development process with complete interoperability at delivery
Leads to a shared modeling infrastructure accessible by all Leads to isolated model systems available to a few
Untrained users may misuse model or outputs Misuse rarer because users and developers are typically the same
Unsuitable for systems that require high levels of data exchange between components
High levels of data exchange less of a problem due to co-location of components
1 May 2008 Page 12
Next Step
Think about an initial “framework infrastructure”
Expand demo system
Include additional basic components
1 May 2008
Strawperson Initial Framework Infrastructure
Landscape•Ecotype•Tree density•Biomass•Disturbance•Succession•Composition
Species Distribution
Fire•Risk•Frequency• Intensity•Behavior•Emissions
Environment(eg, TOPS)
Carbon•Accounting•Allocation•Flows•Sequestration
EcosystemServices
Species Abundance
All kindsof data
Invasive Species
Community-levelBiodiversity
(Ferrier)
Role TBD
Public HealthSocio-economic
1 May 2008 Page 14
Selection Criteria
Model developer is engaged
Model developer is broad thinker and shares vision of a model infrastructure
Models have potential for harmonization
1 May 2008 Page 15
Model or Category Modeler Model information
NASA TOPS Rama Nemani (NASA) Environmental information past, present, future, including temperature, rainfall, radiation, LAI, soil moisture, stream outflow, ET, vegetation stress, primary productivity
GEOSS GBIF IP3 demo system
Stefano Nativi (Italian National Research Council)
Uses OpenModeler and GBIF to produce range maps based on the provided climate
Community models Simon Ferrier (CSIRO, Australia)
TBD
Carbon Yiqi Luo (Univ of Kansas) TBD
Fire TBD Many options; Nativi has a Mediterranean model, TOPS will likely add an emissions model; USFS has many models; risk, intensity, frequency, coverage, emissions are of interest
Landscape eg, LANDIS--Rob Scheller (Conservation Biology Institute)
LANDIS has a fire module as well as other components that are loosely coupled and available as DLLs. Should also check into the Ecosystem Demography model (Paul Moorcroft, George Hurtt)
Socioeconomic Robert Costanza's group (Gund Inst for Env Economics)
Ecosystem services, predicted landcover change; roads; etc
Public health EPA--Gary Foley’s group TBD
Ecosystem services EPA—Rick Linthurst’s group, with John Johnston
TBD
Invasive species Jeff Morisette and John Schnase (NASA)
NASA Invasive Species Forecasting System; predicts likely spread of invasives
1 May 2008 Page 16
Related Topic
Chris Field and Modeling Breakout group
• Coordination in assessing impact of change
• Would enhance application of model results
• Form a working group?
1 May 2008 Page 17
Next Steps
Expand demo system to Initial framework system
Finalize models and participants
Submit proposals?
Plan Model Web Workshop
Impact modeling WG for coordination?
1 May 2008 Page 18
Principle of Gradual Convergence
Vision cannot be achieved quickly
• Gradually converged upon
Cannot be achieved completely
• Not necessary
Follow path of organic growth
• Organized around general framework
• Facilitated by tools and technologies
• Lubricated by interested sponsors
1 May 2008 Page 19
Barriers to Interoperability
Data format
Data terminology
Missing data
Temporal/spatial gridding
Standards
Sponsor goals
Model purpose
Effort required
Proprietariness
Acceptance
Cultural differences
}}
Technical
Non-technical
1 May 2008 Page 20
Larger infrastructure
GlobalClimate
Public Health
Human Population Density and Distribution
Socioeconomic•Poverty•Technology•Economy•Behavior
Regional Environmental Change
• Habitat Alteration• Deforestation/Disturbance• Agricultural Intensification• Urbanization• Fire• Carbon• Landscape• Ecosystem Services
Species & Habitat•Range•Abundance•Functions• Interactions
Local/RegionalClimate
Theseinteract
All kindsof data
Includes vectors & hosts