SESSIONS - COMPUTER SCIENCES
C1. Advances in Neural Networks and Machine Learning and Applications in Land, Water, Environmental, Biological and Human SystemsOrganiser: Sandhya Samarasinghe (sandhya@lincoln.ac.nz)
This sessions invites papers describing recent theoretical/algorithmic advances as well as applications of neural networks, machine learning, statistical and adaptive computational methods. Applications to complex systems including land and water resources; systems biology/bioinformatics; natural, environmental and human systems and engineering where these methods are partcularly useful for providing efficient solutions are welcome.
This sessions invites papers describing recent theoretical/algorithmic advances as well as applications of neural networks, machine learning, statistical and adaptive computational methods. Applications to complex systems including land and water resources; systems biology/bioinformatics; natural, environmental and human systems and engineering where these methods are partcularly useful for providing efficient solutions are welcome.
C2. Adaptive Model StructuresOrganiser: David Swayne, University of Guelph (dswayne@cis.uoguelph.ca)
Many government and private sector modellers are required by legal strictures to use so-called "gold standard" models (eg. SWMM to assess urban stormwater issues and SWAT for watershed modelling, as new housing estates are constructed). There is a need for computer science input on issues such as: dealing with legacy code, lowering the knowledge requirement for effective model use, "wrappers" for heterogeneous inputs and ancillary tools such as visualization and GIS software). Multiple models "strung together" and jointly calibrated require significant new research into autocalibration and related issues (time- and spatial-step matching).
Many government and private sector modellers are required by legal strictures to use so-called "gold standard" models (eg. SWMM to assess urban stormwater issues and SWAT for watershed modelling, as new housing estates are constructed). There is a need for computer science input on issues such as: dealing with legacy code, lowering the knowledge requirement for effective model use, "wrappers" for heterogeneous inputs and ancillary tools such as visualization and GIS software). Multiple models "strung together" and jointly calibrated require significant new research into autocalibration and related issues (time- and spatial-step matching).
Availability of massively parallel computation (either clusters or grids) has created the need for separation of GUI (and even basic file structures) from the code kernel, since every little bit of disk file input "wastes" supercomputer time. Thus, innovative structures are required to achieve massive, coarse-grained parallelism.
This session set would provide applied researchers with a forum to exchange ideas and solutions for these various practical problems.
C3. Development and Application of Environmental Modelling Frameworks Organiser: Peter Krause (p.krause@uni-jena.de) & Sven Kralisch (sven.kralisch@uni-jena.de)
Model development and application for Integrated Water Resources Management (IWRM) of river basins is facing the challenge to reflect and combine approaches from natural as well as social science. Methods, modelling techniques and model design differ in the various involved scientific disciplines and address diverse issues of system complexity and different scales in multiple ways. Integration and coupling of such modelling approaches applied by social and natural sciences for IWRM has led to intensified research activities related to the development of generic modelling frameworks and modular modelling approaches. This session will provide a comprehensive overview and a discussion platform for recent advances and trends currently under development in this context. Papers dealing with development and concepts of modelling frameworks as well as those presenting applications with focus on the role of the used framework system are very welcome.
Model development and application for Integrated Water Resources Management (IWRM) of river basins is facing the challenge to reflect and combine approaches from natural as well as social science. Methods, modelling techniques and model design differ in the various involved scientific disciplines and address diverse issues of system complexity and different scales in multiple ways. Integration and coupling of such modelling approaches applied by social and natural sciences for IWRM has led to intensified research activities related to the development of generic modelling frameworks and modular modelling approaches. This session will provide a comprehensive overview and a discussion platform for recent advances and trends currently under development in this context. Papers dealing with development and concepts of modelling frameworks as well as those presenting applications with focus on the role of the used framework system are very welcome.
C4. Sensor Network Architectures and ModelsOrganiser: Gerald Schimak, Department of Information Technology, Austrian Research Centre (gerald.schimak@arcs.ac.at)
Envisaged is a session dealing with sensor networks architectures and related models. Emphasise will be put clearly on simulations models to simulate sensor network behaviours on the one hand and but on the other hand discussing also models that provide input-data to sensor networks - which we usually call soft-sensors. Beside the architectural and organisational aspects of sensor networks (e.g. self-organising, self-healing, self-configuring), aspects dealing with data propagation, alarming and data quality issues shall be discussed. Therefore we invite you to send in abstracts/papers on a broader but innovative and future oriented view (e.g. your scientific view) on sensor networks & models.
Envisaged is a session dealing with sensor networks architectures and related models. Emphasise will be put clearly on simulations models to simulate sensor network behaviours on the one hand and but on the other hand discussing also models that provide input-data to sensor networks - which we usually call soft-sensors. Beside the architectural and organisational aspects of sensor networks (e.g. self-organising, self-healing, self-configuring), aspects dealing with data propagation, alarming and data quality issues shall be discussed. Therefore we invite you to send in abstracts/papers on a broader but innovative and future oriented view (e.g. your scientific view) on sensor networks & models.
C5. High Performance Computing and SimulationOrganiser: Dadong Wang (Dadong.Wang@csiro.au)
Scientific and technical advances in High Performance Computing (HPC) and large data storage have yielded an era in which solutions for complex compute- and data-intensive problems can be provided in a more efficient way. With the low-cost HPC components becoming widely available such as multi-core CPUs and GPUs, new challenges have emerged in computing, modelling and simulation.
The session will consist of presentations from various fields related to HPC applications and case studies, such as image analysis, visualization, environmental modelling, large scale simulations etc. The goal of the session is to encourage innovations in addressing HPC challenges and to stimulate the use of the HPC technologies in these areas.
Scientific and technical advances in High Performance Computing (HPC) and large data storage have yielded an era in which solutions for complex compute- and data-intensive problems can be provided in a more efficient way. With the low-cost HPC components becoming widely available such as multi-core CPUs and GPUs, new challenges have emerged in computing, modelling and simulation.
The session will consist of presentations from various fields related to HPC applications and case studies, such as image analysis, visualization, environmental modelling, large scale simulations etc. The goal of the session is to encourage innovations in addressing HPC challenges and to stimulate the use of the HPC technologies in these areas.
C6. Advanced Numerical and Computational Techniques for Efficient Construction and Execution of Complex Dynamic Environmental ModelsOrganisers: Peter Vanrolleghem, modelEAU, Université Laval (Peter.Vanrolleghem@gci.ulaval.ca),
Filip Claeys, MOSTforWATER (fc@mostforwater.com)
In many domains of environmental science, complex dynamic models are adopted that are mostly represented by large sets of differential and/or algebraic equations. The complexity of these models is caused by several factors, such as the sheer number of equations and their properties in terms of dynamics and stiffness. Moreover, to support decision-making, there is a growing trend towards increasingly complex procedures based on the execution of these models, such as dynamic simulation over long time horizons, single and multi-objective optimization, Monte Carlo analysis, sensitivity analysis, exhaustive scenario analysis, etc. In order to allow for the efficient (i.e. speedy) construction and execution of complex models, advanced numerical and computational techniques are required that are based on recent advances in mathematics and computer science. Examples include model compilation (flattening), computer algebra, distributed execution (cluster and grid computing), stiffness analysis and elimination, constraint exploitation, etc. This session aims to provide a forum to researchers and practitioners to exchange information on problems encountered in the scope of the construction and execution of complex dynamic environmental models, and on solutions that have recently been studied and/or implemented to solve these problems.
Filip Claeys, MOSTforWATER (fc@mostforwater.com)
In many domains of environmental science, complex dynamic models are adopted that are mostly represented by large sets of differential and/or algebraic equations. The complexity of these models is caused by several factors, such as the sheer number of equations and their properties in terms of dynamics and stiffness. Moreover, to support decision-making, there is a growing trend towards increasingly complex procedures based on the execution of these models, such as dynamic simulation over long time horizons, single and multi-objective optimization, Monte Carlo analysis, sensitivity analysis, exhaustive scenario analysis, etc. In order to allow for the efficient (i.e. speedy) construction and execution of complex models, advanced numerical and computational techniques are required that are based on recent advances in mathematics and computer science. Examples include model compilation (flattening), computer algebra, distributed execution (cluster and grid computing), stiffness analysis and elimination, constraint exploitation, etc. This session aims to provide a forum to researchers and practitioners to exchange information on problems encountered in the scope of the construction and execution of complex dynamic environmental models, and on solutions that have recently been studied and/or implemented to solve these problems.