Lev Tarasov, Memorial University of Newfoundland
Department of Physics and Physical Oceanography
Professor
St. John's, Newfoundland and Labrador
lev@mun.ca
Office:
(709) 864-2675
Bio/Research
I'm generally interested in the modelling of complex systems, with an expertise in glacial systems (combining ice, climate, and earth). Modelling is well suited to those of us who like to build/create things. It offers the opportunity to explore virtual worlds, probe myriads of "what ifs", and cr...
Click to Expand >>
Click to Expand >>
Bio/Research
I'm generally interested in the modelling of complex systems, with an expertise in glacial systems (combining ice, climate, and earth). Modelling is well suited to those of us who like to build/create things. It offers the opportunity to explore virtual worlds, probe myriads of "what ifs", and create piles of data. The challenge is to come up with meaningful results with limited computational resources. Limited resources and limited understanding implies that models of complex physical systems will invariably require simplifications and parameterizations. Along with uncertainties in initial and boundary conditions, the analysis and interpretation of model results becomes a major challenge.
A key point in this regard (for which ice-sheet and climate modelers have been generally deficient) is the need to create meaningful error bars, or better yet probability distributions, for the results of models when used in the context of prediction or retrodiction. The determination of meaningful probability distributions by means of Bayesian calibration of models against observational constraints has therefore become a central focus of my work.
I am also very interested in improving our ability to constrain the changing variability of systems and associated potential thresholds. In the context of climate change, this is arguably both the greatest scientific challenge and the aspect that carries the highest potential impacts. My current approach to this challenge involves three key steps. First, identify the bounds on dynamical processes potentially controlling variability and thresholds. Second, constrain the spatial and temporal scale sensitivities of the representation of these key dynamical processes and their interactions. Finally, develop probability distributions of potential response through a combination of large ensemble data-calibrated modelling with stochastic probing of the bounding critical dynamics.
Click to Shrink <<
A key point in this regard (for which ice-sheet and climate modelers have been generally deficient) is the need to create meaningful error bars, or better yet probability distributions, for the results of models when used in the context of prediction or retrodiction. The determination of meaningful probability distributions by means of Bayesian calibration of models against observational constraints has therefore become a central focus of my work.
I am also very interested in improving our ability to constrain the changing variability of systems and associated potential thresholds. In the context of climate change, this is arguably both the greatest scientific challenge and the aspect that carries the highest potential impacts. My current approach to this challenge involves three key steps. First, identify the bounds on dynamical processes potentially controlling variability and thresholds. Second, constrain the spatial and temporal scale sensitivities of the representation of these key dynamical processes and their interactions. Finally, develop probability distributions of potential response through a combination of large ensemble data-calibrated modelling with stochastic probing of the bounding critical dynamics.
Click to Shrink <<