The Community of Practice for Integrated Environmental Modeling (CIEM) hosted several sessions and workshops at the 2010 International Congress on Environmental Modelling and Software. Workshop 6 was about “The Future of Science and Technology of Integrated Modeling.”

This wiki will be used for the community to further discuss the items that were covered during the workshop. Members are invited to add their thoughts or clarifications to the meeting “notes” in the sections below.

The original outline of the workshop can be found here: iemss2010workshop6outline.pdf . In the sections below you are invited to contribute your thoughts to the topics from the workshop.

Please note that the line item bullets below were just my (Gabriel Olchin) notes from the workshop. They are summary statements of the discussion that was taking place; posted here in a deliberative forum

Workshop Abstract:

This workshop will provide participants with the opportunity to discuss advancing the science and technology of integrated modeling for environmental assessment and decision making. In addition, the workshop will seek to identify software and computational technology trends and how these may impact the development of integrated modeling.

Introduction and Presentation of the Objectives of Workshop 06

Speaker: Gerry Laniak

The attached file includes the slides used to introduce the workshop, it’s objectives, motivation, content, structure, etc.

workshop6intro.pdf

Discussion Topic 1: Technology : Spouse, friend, or acquaintance?

Facilitator: Michael Hiscock

Prompt: The last decade has seen a dramatic evolution of technologies that facilitate analysis and communication. New technologies include web services, cloud computing, facebook, GIS, modeling infrastructures, iPad. New software designs and architectures are providing for increased computational complexity and efficiencies. Hardware has a seemingly unlimited potential for storing and moving data. As we move into the future of integrated modeling what will be our individual/collective relationship with technology? Will the community of integration modelers be strategic and form intimate relationships with the diverse community of technologists (e.g., influencing directions in technology) or will we utilize technologies on an individual as needed/as available basis? This discussion will focus on what technologies are emerging that can play a significant role in IM and how can/should the IM community interact with the technologies and the community of technology developers.

SUMMARY OF DISCUSSION – contribute your thoughts below!

What are our needs?

• Software frameworks
• Run-time technologies
• Usability of new models and training for new users or interested stakeholders?
  • How can we make this process easier?
  • Can we provide web-based training to other modelers?
  • What technologies can facilitate this?
• How to balance development costs with the needed utility of the model
• Parallel networks/computing
• Mobile computing
• Ease of use – documentation – stand alone documentation

Are we effectively using the existing technologies?

• Do community based approaches help to develop models that match the needs of the community?
• Registry of transferable components rather than frameworks?
• Perhaps the technology exists, but we should focus on developing new components rather than additional frameworks?
• Industrial simulations vs environmental modeling —> granularity issues
  • Components would be framework independent
  • We then need databases that serve all these components

A CSIRO Perspective:

• How are new technologies incorporated into day to day operations
• What are the new paradigms to focus on?
• Strategic directions – community approaches

Community Approaches

• Community should work together to filter new technologies
• Technology roadmaps to make consistent decisions, with buy-in and comprehensive
  • Are there other roadmaps we can use that?
  • iemHUB – hosting these types documents?
  • Does OpenMI have an example?
  • Technology roadmaps are consequences of the technologies being used
• Frameworks
• Wrappers
• Models
• softIM
• Where we want to be with users, science, technologies in the near future?
• Business application framework
• Support
• Community testing of new software / peer review

Technology

• Cannot access models easily – we are putting own restrictions on them
• Where can we catalog all the existing information
• IM should aslo collaboration between research scientists and computer scientists

How can we utilize existing models?

• Community platform for sharing expertise and technologies?
• 3rd party involvement during participatory modeling

Summary Points from the Discussion_

1. Usability! who is the user?
   • Consciousness requirements are different for the user groups
   • Possible incentives
2. Technology roadmaps / science / standards
   • How to employ collectively
3. How to get community to branch out of their comfort zones and use existing models
4. Problems —> requirements —> roadmap

Discussion Topic 2: Standards : Curse or course?

Facilitator: Daniel Ames

Prompt: Standards essentially establish designs and protocols that facilitate the interoperability of system components. Every model and modeling system reflects a standard, i.e., the methods for data transfer to and from the components is set and unchanging. As the need for integration increases so does the need to connect a myriad of disparate sources of data with a similarly disparate set of modeling components. Establishing these connections will require reconciliation of the myriad standards employed by the myriad data sources and models. Developing and implementing universal standards, e.g., for data representation and transfer, can greatly enhance information/knowledge exchange. Standards can also be arduous and time consuming to implement. They may limit one’s ability to “get a job done in a timely manner”. Thus, there are many reasons to pursue standards within the world of IM (e.g., interoperability, reuse, quality assurance, etc.) and there are reasons to keep standards at arms length. This discussion will focus on what types of standards are needed, at what scale should they be implemented, what standards are available now and should be adopted, what standards are emerging that could/should be assimilated?

There was also a lot of discussion regarding “Open Source”…

SUMMARY OF DISCUSSION – contribute your thoughts below!

• Standards for data
• Standards for interoperability / communicating between models
• OGC has standards that could be adopted
• Open source for IEM
  • Not going to make $$
  • Increase transparency
  • Bad decisions or bad applications – who is liabile?
  • QA concerns with open source and future applications
  • What does it mean to be free?
  • What about version control?
• When should we be concerned about intellectual property rights?
• Licensing to keep intellectual property
• Standards should make open source more feasible
• Do standards and open source go hand in hand?
  • What about standardizing how data files are called?
  • What about standardizing how models use the data?
• Ontologies can be mapped – formalized knowledge behind the data
• Standardized I/O
• Standardize the descriptions and assumptions (scales) – these are needed when we want to maximize interoperability
• Standard for model meta-data
  • Reusable Asset Specification (version 2.2)
• OpenMI – standards/protocols come first then frameworks

Discussion Topic 3: Modeling science and Data : Integrated or INTEGRATED?

Facilitator: Alexey Voinov

Description: As modeling systems become more complex, involving models from several science disciplines, a host of issues surface that must be addressed. Sharing information across geospatial and temporal scales, integrating multi-disciplinary science components, characterizing, propagating, and quantifying uncertainty, and matching data with model with problem represent only a few of the issues that will pose significant challenges to the IM community. The pure dimensionality associated with this new problem domain and scale may require that the word integration itself be redefined. This discussion will focus first on enumerating the issues and then on developing a strategic community sense of a path forward that will keep IM scientifically clear, credible, and responsive.

SUMMARY OF DISCUSSION – contribute your thoughts below!

Calibration

• Not included in integrated modeling efforts
• Distinction between over parameterized models and data-intensive modeling
• Emipirical models need to be calibrated for each site – these concerns are more important when we get away from well known domains
• Is site-specific calibration okay?
  • What is the process?
• Are the requirements put upon us to make integrated models causing us to apply calibrated models at a larger scale or did we say we could do that?
• Cannot assume that site-specific applications can be extrapolated SHOULD THIS BE: Must assume that site-specific recommendations cannot be extrapolated (R. Sojda)
  • Thanks for your comments. In the context of this discussion, the term “application” was being used in place of “model”. In this sense, we were discussing whether models that have been calibrated for a specific site could applied at larger scales when integrated (i.e. linked) with other models. However, your comments are correct in that site-specific applications cannot be extrapolated. -G. Olchin
• Can calibration methods be incorporated into the integrated modeling process?
• Can model integration help model calibration??
• Can integrated models be calibrated as a whole? Or individually?
• Can calibration(s) be integrated?
• Knowledge-based models have many of the same issues, avoid some issues like over parameterization, but add others related to needing to articulate assumptions (R. Sojda)

Definition of Integrated Modeling (IM) as a Procedure Of Collaboration

I find it important to clarify what “integrated modeling” means. Then, the role of technical solutions(or cyber-infrastructure) can be defined within such IM process. (rt.arnold).

• By definition, IM methods support the analysis of cause-effect-chains, connecting concepts from a wide range of scientific disciplines.
• Scientific disciplines are a construct of today’s academic organizations, not properties of the system. As a “reversal of reductionism”, integration re-connects divergent concepts of today’s disciplines and their methods using models.
• The IM process includes integration of concepts, technical implementation of integrated models, calibration and analysis of the full cause-effect-chain, uncertainty assessment and learning.
• IM is performed either by individuals, or by teams with different academic training (knowledge, methods, culture and ontology). Individuals are limited in their resources, and larger teams are limited by increasing transaction costs (communication and technological problems). Methods of IM, including cyber infrastructure, must operate within this institutional environment.

Within such learning cycle that cross multiple disciplines (and institutions), the core challenges are (a) access to detailed knowledge at any time, and (b) minimizing transaction costs for this knowledge transfer. Thus, I ask “Which roles should institutions play to facilitate an IM process? How would an institution look like that optimizes IM effectiveness?” (rt.arnold)

Other Discussions__

• Cannot assume that level of detail is consistent
• It is more feasible to build an integrated model (multiple domains, simple empirical relationships) rather than integrating existing models
• Misuse of models
• Illusion of technology advances enabling us to solve complex problems
• Generality of models – should they become applicable at larger scales
• What data do I have available? What data is needed?
• Can model outputs be coupled with data to deduce results?
• Data quality?

These issues are not being addressed in new frameworks!

Summary Points from the Discussion_

• Need functionality of plugging models into a framework AND an adequate choice of models
• What are the questions we are trying to resolve?
• What methods are to be used?
• How are models to be integrated?
  • Should there be guidance and protocols?
• What are the limitations of IEM?
• Technology is outpacing the modeling…

Discussion Topic 4: Integrated Modeling Applications and Uses: Part or apart?

Facilitator: Gene Whelan

Prompt: A primary driver in the emergence of integrated modeling is the need to inform a wide range of decisions that address transdisciplinary environmental problems. These problems are complex and solutions must reflect the interests and value systems of a diverse stakeholder community. We are seeing that the “information” associated with these complex problems can be voluminous and the relationship among the pieces of information often lacks clarity and structure. The interface between science/analysis and decision is critical. Synthesizing knowledge and information and presenting it in a form that is relevant and consumable by decision makers will become a science unto itself. This discussion will focus on the interface between modeling science and decision making and, in particular, what the role and responsibility of IM professionals will be in the context of aligning problems with solutions and synthesizing results of analyses.

SUMMARY OF DISCUSSION – contribute your thoughts below!

• Transparency
• Evaluation / Uncertainty
• Communication
• Visualization for familiarity
• Community of interest
• Need to formalize and integrate mental models

Created on 21 May 2012, Last modified on 27 Jan 2011