XIX International Conference
The classical approach to model calibration introduced by Carl Friedrich Gauss (1794) is based on a philosophy that involves some longstanding and questionable assumptions that arise from the adoption of classical statistical techniques to fitting empirical regressions models to data. One of these limitations is that structural deficiencies in the model and errors in the input data are ignored and assumed to be either "small" or to be somehow "absorbed" into the error residual. The residuals are then expected to behave statistically in the same manner as the calibration data measurement error. Many published studies have shown that this classical modeling approach is inadequate to identify structural model errors, a prerequisite to improving our theory of how water flows through watersheds. For instance, consider the Sacramento soil moisture accounting (SAC-SMA) model introduced by Burnash et al. in the early 1970s and used by the US National Weather Service for flash-flood forecasting throughout the United States. In about four decades of fitting the SAC-SMA model to (spatially distributed) streamflow data, we have not been able to make any noticeable improvements to the underlying theory and equations of the model. Here, I will discuss an alternative blueprint to hydrologic modeling and uncertainty analysis that explicitly recognizes the role of input data error and model structural deficiencies. This methodology is inspired by recent developments in Markov Chain Monte Carlo (MCMC) simulation and Sequential Monte Carlo (SMC) sampling and provides detailed insights into the probabilistic properties of the model, parameter, input, and calibration data error. The methodology is illustrated using a few preliminary case studies involving simple watershed models.
Jasper Vrugt earned M.S. (1999: Cum Laude) and Ph.D. (2004: Cum Laude) degrees from the University of Amsterdam. He specializes in environmental systems modeling, and regularly develops optimization and uncertainty analysis methods to better analyze the discrepancy between model predictions and actual observations and improve the theory, understanding and predictability of environmental systems. Much of his work is within the realm of surface hydrology and soil physics, but also spans the fields of ecology, hydrogeophysics, hydrometeorology, atmospheric modeling, and geophysics.
He has published 70+ papers in peer reviewed international journals, and is Associate Editor of Water Resources Research, Vadose Zone Journal, Environmental Modeling & Software, and Hydrology and Earth System Sciences. He Fellow of the AGU and GSA, and recipient of the 2011 Donath Medal (GSA), 2010 James B. Macelwane Medal (AGU), 2010 Young Scientist Award (EGU), 2007 Dutch Hydrology Prize, and 2007 Early Career Award Soil Physics (SSSAJ).