Keywords

Integrated Assessment and Modeling, Integrated Environmental Modeling, Interprocess Communication

Location

Session A3: Innovative Architectures and Approaches of Cloud and Mobile Technology for Environmental Modeling

Start Date

18-6-2014 9:00 AM

End Date

18-6-2014 10:20 AM

Abstract

Integration of environmental system models across multiple disparate disciplines has been an area of growing interest because such models typically only represent knowledge and advancements in one discipline or several, similar disciplines. Many challenges have arisen, though, in integrating such models due to varied programming expertise amongst modelers, dependence on specific frameworks or operating systems in addition to different programming languages and requirements for two-way feedbacks that are crucial for improving model representation of the physical reality. in response to these challenges, this paper presents a novel approach based on a publish-subscribe type system to both simplify and improve the model integration process. The approach allows access to legacy model variables at any point during simulation feedbacks between models. Additionally, a level of abstraction is given to the model integration process that allows researchers and other technical personnel to perform more detailed and interactive modeling, visualization, optimization, calibration, and uncertainty analysis without requiring deep understanding of interprocess communication. Utilizing this approach may significantly decrease developer time and potentially computation time, and enhance interdisciplinary research by providing detailed two-way feedback mechanisms between various simulation models with minimal changes to legacy code. A simple test case of a connection between a biogeochemical model and a finite element mass and heat transfer model demonstrates the utility and ease of the new model integration approach.

COinS
 
Jun 18th, 9:00 AM Jun 18th, 10:20 AM

MODPI: A parallel model data passing interface for integrating legacy environmental system models

Session A3: Innovative Architectures and Approaches of Cloud and Mobile Technology for Environmental Modeling

Integration of environmental system models across multiple disparate disciplines has been an area of growing interest because such models typically only represent knowledge and advancements in one discipline or several, similar disciplines. Many challenges have arisen, though, in integrating such models due to varied programming expertise amongst modelers, dependence on specific frameworks or operating systems in addition to different programming languages and requirements for two-way feedbacks that are crucial for improving model representation of the physical reality. in response to these challenges, this paper presents a novel approach based on a publish-subscribe type system to both simplify and improve the model integration process. The approach allows access to legacy model variables at any point during simulation feedbacks between models. Additionally, a level of abstraction is given to the model integration process that allows researchers and other technical personnel to perform more detailed and interactive modeling, visualization, optimization, calibration, and uncertainty analysis without requiring deep understanding of interprocess communication. Utilizing this approach may significantly decrease developer time and potentially computation time, and enhance interdisciplinary research by providing detailed two-way feedback mechanisms between various simulation models with minimal changes to legacy code. A simple test case of a connection between a biogeochemical model and a finite element mass and heat transfer model demonstrates the utility and ease of the new model integration approach.