PUB 2012: Theme 3

PUB SYMPOSIUM 2012

Theme 3: Uncertainty analysis and model diagnostics - Tuesday, October 23, 2012


Convener: Hoshin Gupta ([email protected])
Co-Conveners: Thorsten Wagener, Saket Pande, Jim Freer ([email protected]; [email protected]; [email protected])

During the past decade (since the challenge was posed at the 1999 IAHS meeting in Birmingham), considerable progress has been made in characterizing uncertainties in hydrological model simulations. New methods now try to disentangle the different uncertainty sources (data, parameters, and model structure) in gauged and ungauged basins. And while these methods span a range of approaches, there may be scope to further exploit rich mathematical theories on uncertainty assessment. Concurrently, progress has been made on methods for transferring, extrapolating and regionalising models across basins - a necessary requirement for Predictions in Ungauged Basins, where model calibration is not feasible. While our ability to make such predictions has improved, few approaches produce estimates of predictive uncertainty at ungauged locations. Also, methods for discriminating between alternative model hypotheses remain weak (it has become common to talk about multiple 'equally good' models and parameter sets), and continue to rely heavily on the comparison to observed streamflow for model evaluation. Therefore, attention has turned to the problem of how to characterize, detect, and diagnose model structural inadequacies, and to use this knowledge to develop better models.
In this session, we will report on, and discuss, the past decade of scientific achievements in uncertainty analysis and diagnostic model evaluation in the context of PUB, the insights gained, and the major challenges still facing us. We invite contributions from all scientists that have worked in this area and anticipate a vigorous and productive discussion.

Oral Programme (15 minute talks)

Time Speaker Title
13:00-13:15 Neil McIntyre Aligning hydrological response spaces for prediction in ungauged catchments (invited)
13:15-13:30 Zhe Li An informal Bayesian model for diagnose of hydrological model structural uncertainty
13:45-14:00 Charles Perrin Model testing for improving flow simulation and uncertainty quantification: what can we learn from large catchment sets? (invited)
13:30-13:45 Tanja Euser A framework to assess the realism of model structures using hydrological signatures
15:30-15:45 Hilary McMillan Using data from research basins to identify appropriate model stuctures (invited)
14:15-14:30 Steven Weijs Bayes in terms of description lengths and surprises
14:30-15:00 Discussion and Future Directions

15:00-15:30 Coffee, Tea & Posters

15:45-16:00 Shervan Gharari Calibration validation revisited or how to make better use of available data: Sub-period calibration (invited)
14:00-14:15 Patrick Willems Data-based approach to analyse response behaviour of hydrological models at catchment scale
16:00-16:15 Juan Munera Estrada Predictive uncertainty estimation at ungauged basins in a Bayesian framework
16:15-16:30 Francois Bourgin Can model uncertainty estimates be transferred from gaged to ungaged catchments?
16:30-16:45 Ida Westerberg Constraining the predicitve uncertainty of a Central-American water-balance model using regionalised flow-duration curves
16:45-17:00 Jan Seibert Gauging the ungauged catchment: which data are most valuable? (invited)
17:00-17:30 Discussion and Future Directions

Poster Programme

Author Title
Susana Almeida Importance of prior distributions for regionalisation
Martijn Booij Uncertainty in high and low flows due to parameter and model structure errors: a case study for Qiantang River basin, East China
David Ferras Evaluation of different concepts of spatially distributed hydrological models in the Upper Blue river basin using remote sensing data
Ian Littlewood Effects of data time-step on the accuracy of calibrated rainfall-streamflow model parameters: practical aspects of uncertainty reduction
Thibault Mathevet Improving hydrological model calibration and selection via a better use of evaluation metrics and streamflow sub-samples for operational applications
Micah Mukolwe Uncertainty communication of probablilistic flood mapping: a stakeholder driven approach (poster requested)
Jane Tanner Surface-groundwater interactions: process understanding and using models to test hypotheses
  back to top