Institute for Mathematical Sciences Event Archive
Online registration form
Enquiries |
|
Data-driven and Physically-based Models for Characterization of Processes in Hydrology, Hydraulics, Oceanography and Climate Change
(6 - 28 January 2008)
Jointly organized with Pacific Institute for Mathematical Sciences, UBC
Organizing Committee · Visitors and Participants · Overview · Activities · Venue · Funding for Young Scientists
Co-chairs
- Sylvia Esterby (University of British Columbia)
- Hans-Rudolf Künsch (ETH Zurich)
- Shie-Yui Liong (National University of Singapore)
Members
- Vladan Babovic (National University of Singapore)
- Wolfgang Kinzelbach (ETH Zurich)
- Pavel Tkalich (National University of Singapore)
- Jim Zidek (University of British Columbia)
Special thanks to Singapore - Delft Water Alliance for sponsoring a number of visitors.
The 3-week program will consist of a full week of seminars/lectures, and two weeks of workshops and research discussions aimed at developing research collaboration. Three main topics are covered in the program. They are:
- “Development of a fully integrated data driven and physical-based models for water resources management”
- “Dynamic and Statistical Downscaling on Climate Change Study”
- “Nonlinear Wave Dynamics and Tsunami Modeling”
Physically based modeling maps natural phenomena to a computer simulation program. There are two basic processes in this mapping: mathematical modeling and numerical solution. The mathematical modeling concerns the description of the natural phenomena by mathematical equations. The numerical solution involves computing an efficient and accurate solution of the mathematical equations. Models are essential tools for synthesizing observations, theory, and experimental results in order to investigate the physical phenomena which govern the behavior of water in the system under study, and to understand how the system is affected by human activities. Models can be used in both a retrospective sense, to test the accuracy of modeled changes in the system by comparing model results with observations of past change, and in a prognostic sense, for calculating the response of the system to projected future changes.
This part of the program will focus on improvements of description of physical, environmental and water quality processes through hydrodynamics, morphology, hydrology, water quality, ecology as well as numerical methods and techniques such as finite difference methods, finite element methods and boundary element methods, with applications to physically based modeling of lakes and reservoirs, prediction of runoff in poorly gauged catchments using physically based models, and flood modeling.
Data driven modeling and computational intelligence: In situations when knowledge about the processes to be modelled is limited, physically based model cannot be built, or they are inadequate. There are situations, however, when at least some of the variables characterising a particular process have been measured, and there is enough data to represent the input-output relationships associated with the process. In such cases data-driven models (DDM) can be built that make it possible to model and forecast some output variables. An example is the modelling of a rainfall-runoff relationship using statistical models or artificial neural networks. Typically, in order to build a DDM, methods of computational intelligence would be used. Research here is concentrated on testing various methods and their combinations in different types of modelling problems, and, developing new modelling methods.
Often, physically based models do not exhibit the needed accuracy, or are inadequate to model particular situations, e.g. those of very high flows for the purpose of flood forecasting. On the other hand, there may not be enough data to train data driven models alone. In this case combinations of models of different types (i.e., hybrid models) could be a solution. Research in hybrid modeling is aimed at developing algorithms to ensure optimal combinations of physically based and data-driven models, and testing the resulting models in various situations. This paradigm will explore a number of approaches and techniques, such as data assimilation based on Kalman filtering, model-error characterization and its correction; data-model integration techniques, data-driven knowledge discovery and finally adaptive and learning modeling environments under which models adapt their internal structure on the basis of observed data.
The program will also consider recent development in statistics relevant to the topical areas described in the following subsections. Considerable efforts have been made to assess uncertainty by comparing and combining different physical models (especially in weather prediction and climate modeling) and on calibrating complex computer models with observations, taking non-identifiability and structural model deficits into account. It should be noted that these topics are currently the object of a program at SAMSI (Statistical and Applied Mathematical Sciences Institute).
The program will concentrate on bridging the gap and establishing the bridges between the two approaches (and two scientific communities) by addressing several specific topical areas: water resources management, down-scaling in climate change and non-linear wave and tsunami modeling.
Development of fully integrated data driven and physical-based models for water resources management
Developing an effective and efficient computational tool for water resources management of water-scarce regions or countries like Singapore is of utmost importance. The program will cover the chain from real time monitoring of storms, reservoir and sea levels via forecasting of runoff and flooding to decision making on reservoir operation. Both water quantity and water quality should be considered in the study and the applications of smart sensing technologies should simultaneously be explored.
This topic will concentrate on forecasting storms, surface runoff and downstream tidal levels in advance of their actual arrival in a holistic and integrated manner. First various deterministic models including atmospheric, rainfall-runoff, reservoir, and coastal hydrodynamics will be fully integrated. After calibrating and validating the system, a database containing simulated relevant data resulting from representative scenarios will be set-up. This database will be used to train some data driven models which are known to be computationally of many orders of magnitude faster than their deterministic counterparts.
Dynamic and Statistical Downscaling in Climate Change Study
There is an emerging scientific consensus that human action, especially the release of man-made greenhouse gases, is leading to global climate change. Some of the most current research activities are the study of dynamic and statistical downscaling of climate parameters (e.g. rainfall, sea level) and extreme weather and climate events. Their impacts particularly on small islands such as Singapore are of grave concerns.
This topic will in particular focus on dynamic and statistical downscaling issues of climate parameters (e.g. rainfall, sea level). Analytical results from different General Circulation Models are known to differ significantly. Taking the more conservative results would result in prohibitively high cost in adaptation measures while the other extreme will certainly be catastrophic for small island states like Singapore. This topic will critically assess the existing dynamic and statistical downscaling methods.
Nonlinear wave dynamics and tsunami modeling
Nonlinear waves are observed in all branches of science and engineering, and are present in different aspects of our daily life. Physics and biology, road traffic control and structure of the universe, electronic and communication systems are affected by the same phenomenon at different spatial and temporal scales, namely nonlinear wave dynamics. Nonlinear waves can be significant in an act of creation or destruction, and be simultaneously fascinating and tragic. Indian Ocean (2004) Tsunami is a pure example of a series of events dominated by nonlinear wave dynamics, starting from tectonic movements and up to tsunami run-up on shore. The focus of the discussions would be placed on the application of the theories to nonlinear wave dynamics in the ocean.
The period of the program is from 6 to
28 Jan 2008. The first week will be dedicated totally on
seminars/lectures on three topics described above. Each of
the following 2 weeks will start with two days of
presentations, by a number of invited speakers, focusing on
the topics described above. The remaining three days of each
of these two weeks will be reserved for work in smaller
multi-disciplinary groups. The groups will address a number
of concrete challenges associated with the three topical
areas. The general idea is to arrive at possible research
collaboration in the immediate future; and to draft
scientific publications by the end of the workshop.
Please click here for a copy of the IMS workshop report (MsWord | PDF)
Students and researchers who are interested in attending these activities are requested to complete the online registration form.
The following do not need to register:
- Those invited to participate.
- Those applying for
financial support.
The Institute for Mathematical Sciences has limited funds to cover partial support for travel and living expenses for young scientists interested in participating in the program. Applications should be received at least three (3) months before the commencement of the program. Application form is available in (MSWord|PDF|PS) format for download.
For enquiries, please email us at ims(AT)nus.edu.sg.
For enquiries on scientific aspects of the program, please
email Yui LIONG at tmslsy(AT)nus.edu.sg.
Organizing Committee · Visitors and Participants · Overview · Activities · Venue · Funding for Young Scientists
