Calibration and Uncertainty Assessment of Environmental Models: Methods and Applications
Jasper A. Vrugt and Bruce A. Robinson
Earth and Environmental Sciences Division, Los Alamos National Laboratory, Mail Stop T003, Los Alamos NM, 87545
The field of earth sciences is experiencing rapid changes as a result of the growing understanding of environmental physics, along with recent advances in measurement technologies, and dramatic increases in computing power. More complex, spatially explicit computer models are now possible, allowing for a more realistic representation of the system of interest. However, application of these models is not straightforward: many of the spatially distributed parameters and internal states in these models require calibration/updating before meaningful predictions can be made. Classical parameter estimation and data assimilation methods, originally developed for low-dimensional lumped problems, are frequently used. However, strong model nonlinearity, high dimensionality, and significant measurement and model structural uncertainty, hamper the use of such methods in complex simulation models. Also, spatially explicit models typically simulate several output fluxes for which measurement data are available and must be properly assimilated. In this talk I will present methods for improved calibration and uncertainty assessment of environmental models. In particular, I will discuss different methods for probabilistic forecasting, and highlight new concepts of model averaging and genetically adaptive multi-method search strategies for single and multi-objective model calibration. The various methods are illustrated using examples taken from basin scale groundwater modeling, surface hydrology, meteorology, biology, and applied mathematics.
AAPG Search and Discover Article #90066©2007 AAPG Hedberg Conference, The Hague, The Netherlands