Comparison of joint versus postprocessor approaches for hydrological uncertainty estimation accounting for error autocorrelation and heteroscedasticity

The paper appraises two approaches for the treatment of heteroscedasticity and autocorrelation in residual errors of hydrological models. Both approaches use weighted least squares (WLS), with heteroscedasticity modeled as a linear function of predicted flows and autocorrelation represented using an AR(1) process. In the first approach, heteroscedasticity and autocorrelation parameters are inferred jointly with hydrological model parameters. The second approach is a two-stage “postprocessor” scheme, where Stage 1 infers the hydrological parameters ignoring autocorrelation and Stage 2 conditionally infers the heteroscedasticity and autocorrelation parameters. These approaches are compared to a WLS scheme that ignores autocorrelation. Empirical analysis is carried out using daily data from 12 US catchments from the MOPEX set using two conceptual rainfall-runoff models, GR4J, and HBV. Under synthetic conditions, the postprocessor and joint approaches provide similar predictive performance, though the postprocessor approach tends to underestimate parameter uncertainty. However, the MOPEX results indicate that the joint approach can be nonrobust. In particular, when applied to GR4J, it often produces poor predictions due to strong multiway interactions between a hydrological water balance parameter and the error model parameters. The postprocessor approach is more robust precisely because it ignores these interactions. Practical benefits of accounting for error autocorrelation are demonstrated by analyzing streamflow predictions aggregated to a monthly scale (where ignoring daily-scale error autocorrelation leads to significantly underestimated predictive uncertainty), and by analyzing one-day-ahead predictions (where accounting for the error autocorrelation produces clearly higher precision and better tracking of observed data). Including autocorrelation into the residual error model also significantly affects calibrated parameter values and uncertainty estimates. The paper concludes with a summary of outstanding challenges in residual error modeling, particularly in ephemeral catchments.

[1]  G. Kuczera.  Improved parameter inference in catchment models: 1. Evaluating parameter uncertainty , 1983 .

[2]  Peter Reichert,et al.  Analyzing input and structural uncertainty of nonlinear dynamic models with stochastic, time‐dependent parameters , 2009 .

[3]  D. Kavetski,et al.  Confronting Input Uncertainty in Environmental Modelling , 2013 .

[4]  C. Perrin,et al.  Improvement of a parsimonious model for streamflow simulation , 2003 .

[5]  Bettina Schaefli,et al.  Quantifying hydrological modeling errors through a mixture of normal distributions , 2007 .

[6]  Dmitri Kavetski,et al.  The open source RFortran library for accessing R from Fortran, with applications in environmental modelling , 2011, Environ. Model. Softw..

[7]  George Kuczera,et al.  Toward a reliable decomposition of predictive uncertainty in hydrological modeling: Characterizing rainfall errors using conditional simulation , 2011 .

[8]  Alberto Montanari,et al.  Estimating the uncertainty of hydrological forecasts: A statistical approach , 2008 .

[9]  Keith Beven,et al.  So just why would a modeller choose to be incoherent , 2008 .

[10]  Q. Shao,et al.  Experimental evaluation of the dynamic seasonal streamflow forecasting approach , 2011 .

[11]  P. Mantovan,et al.  Hydrological forecasting uncertainty assessment: Incoherence of the GLUE methodology , 2006 .

[12]  Soroosh Sorooshian,et al.  Model Parameter Estimation Experiment (MOPEX): An overview of science strategy and major results from the second and third workshops , 2006 .

[13]  Demetris Koutsoyiannis,et al.  A blueprint for process‐based modeling of uncertain hydrological systems , 2012 .

[14]  Dmitri Kavetski,et al.  Reply to comment by K. Beven et al. on “Pursuing the method of multiple working hypotheses for hydrological modeling” , 2012 .

[15]  Alberto Montanari,et al.  What do we mean by ‘uncertainty’? The need for a consistent wording about uncertainty assessment in hydrology , 2007 .

[16]  Tyler Smith,et al.  Development of a formal likelihood function for improved Bayesian inference of ephemeral catchments , 2010 .

[17]  Jery R. Stedinger,et al.  Appraisal of the generalized likelihood uncertainty estimation (GLUE) method , 2008 .

[18]  P. Reichert,et al.  Linking statistical bias description to multiobjective model calibration , 2012 .

[19]  Martyn P. Clark,et al.  Framework for Understanding Structural Errors (FUSE): A modular framework to diagnose differences between hydrological models , 2008 .

[20]  Q. Duana,et al.  Model Parameter Estimation Experiment (MOPEX): An overview of science strategy and major results from the second and third workshops , 2006 .

[21]  Dmitri Kavetski,et al.  Pursuing the method of multiple working hypotheses for hydrological modeling , 2011 .

[22]  George Kuczera,et al.  Critical evaluation of parameter consistency and predictive uncertainty in hydrological modeling: A case study using Bayesian total error analysis , 2009 .

[23]  Stefania Tamea,et al.  Verification tools for probabilistic forecasts of continuous hydrological variables , 2006 .

[24]  J. Stedinger,et al.  Appraisal of the generalized likelihood uncertainty estimation (GLUE) method , 2008 .

[25]  George Kuczera,et al.  Pitfalls and improvements in the joint inference of heteroscedasticity and autocorrelation in hydrological model calibration , 2013 .

[26]  J. Vrugt,et al.  A formal likelihood function for parameter and predictive inference of hydrologic models with correlated, heteroscedastic, and non‐Gaussian errors , 2010 .

[27]  G. Villarini,et al.  Empirically-based modeling of spatial sampling uncertainties associated with rainfall measurements by rain gauges , 2008 .

[28]  George Kuczera,et al.  Understanding predictive uncertainty in hydrologic modeling: The challenge of identifying input and structural errors , 2010 .

[29]  Dmitri Kavetski,et al.  Ancient numerical daemons of conceptual hydrological modeling: 1. Fidelity and efficiency of time stepping schemes , 2010 .

[30]  Keith Beven,et al.  Comment on “Pursuing the method of multiple working hypotheses for hydrological modeling” by P. Clark et al. , 2012 .

[31]  Benjamin Renard,et al.  Evaluation of statistical models for forecast errors from the HBV model , 2010 .

[32]  C. Perrin,et al.  Does a large number of parameters enhance model performance? Comparative assessment of common catchment model structures on 429 catchments , 2001 .

[33]  S. Sorooshian,et al.  Stochastic parameter estimation procedures for hydrologie rainfall‐runoff models: Correlated and heteroscedastic error cases , 1980 .

[34]  K. Bogner,et al.  Post‐processing hydrological ensemble predictions intercomparison experiment , 2013 .

[35]  M. Clark,et al.  Ancient numerical daemons of conceptual hydrological modeling: 2. Impact of time stepping schemes on model analysis and prediction , 2010 .

[36]  B. Bates,et al.  A Markov Chain Monte Carlo Scheme for parameter estimation and inference in conceptual rainfall‐runoff modeling , 2001 .

[37]  Chong-Yu Xu,et al.  Systematic evaluation of autoregressive error models as post-processors for a probabilistic streamflow forecast system , 2011 .

[38]  Anthony J. Jakeman,et al.  Performance of conceptual rainfall‐runoff models in low‐yielding ephemeral catchments , 1997 .

[39]  Khaled H. Hamed,et al.  A modified Mann-Kendall trend test for autocorrelated data , 1998 .

[40]  C. Diks,et al.  Improved treatment of uncertainty in hydrologic modeling: Combining the strengths of global optimization and data assimilation , 2005 .

[41]  David R. Brillinger,et al.  Consistent detection of a monotonic trend superposed on a stationary time series , 1989 .

[42]  V. Singh,et al.  The HBV model. , 1995 .

[43]  Francesca Pianosi,et al.  Dynamic modeling of predictive uncertainty by regression on absolute errors , 2012 .

[44]  Patrick M. Reed,et al.  Reducing uncertainty in predictions in ungauged basins by combining hydrologic indices regionalization and multiobjective optimization , 2008, Water Resources Research.

[45]  Keith Beven,et al.  The future of distributed models: model calibration and uncertainty prediction. , 1992 .

[46]  J. Stedinger,et al.  Multisite ARMA(1,1) and Disaggregation Models for Annual Streamflow Generation , 1985 .