The Predictive Power of Ground‐Motion Prediction Equations

Although model calibration should be always performed alongside model validation, this is rarely the case when deriving new ground‐motion prediction equations (GMPEs). Explanatory modeling (Shmueli, 2010) is often preferred to data‐driven predictive approaches, and the analysis of the residual distribution is generally used to support the model qualification. Following previous studies (Kuehn et al. , 2009; Scherbaum et al. , 2009), this work aims to again stress the importance of validation for assessing the predictive power of GMPEs and for avoiding, or at least limiting, data overfitting. Considering the strong‐motion data and models recently analyzed by Roselli et al. (2016), I exemplify the application of standard validation approaches based on predictive metrics (Akaike and Bayesian information criteria), resample techniques (cross validation and bootstrap), and validation against new data. The results confirm that GMPEs overfitting the calibration data have a limited predictive power, whereas, regarding validation against new data, the selection of the validation data set should take into account possible regional effects in the ground motion.

[1]  Leo Breiman,et al.  Statistical Modeling: The Two Cultures (with comments and a rejoinder by the author) , 2001 .

[2]  H. Akaike,et al.  Information Theory and an Extension of the Maximum Likelihood Principle , 1973 .

[3]  G. Schwarz.  Estimating the Dimension of a Model , 1978 .

[4]  F. Scherbaum,et al.  Model Selection in Seismic Hazard Analysis: An Information-Theoretic Perspective , 2009 .

[5]  Galit Shmueli,et al.  To Explain or To Predict? , 2010 .

[6]  F. Scherbaum,et al.  Deriving Empirical Ground-Motion Models: Balancing Data Constraints and Physical Assumptions to Optimize Prediction Capability , 2009 .

[7]  Lucia Luzi,et al.  ITACA (ITalian ACcelerometric Archive): A Web Portal for the Dissemination of Italian Strong-motion Data , 2008 .

[8]  Jonathan P. Stewart,et al.  Comparison of NGA-West2 GMPEs , 2014 .

[9]  David R. Anderson,et al.  Multimodel Inference , 2004 .

[10]  J. Kaklamanos,et al.  Model Validations and Comparisons of the Next Generation Attenuation of Ground Motions (NGA-West) Project , 2011 .

[11]  R Core Team,et al.  R: A language and environment for statistical computing. , 2014 .

[12]  Dino Bindi,et al.  Partially non-ergodic region specific GMPE for Europe and Middle-East , 2016, Bulletin of Earthquake Engineering.

[13]  Pierre-Yves Bard,et al.  Adapting the Neural Network Approach to PGA Prediction: An Example Based on the KiK-net Data , 2012 .

[14]  R. Tibshirani,et al.  Improvements on Cross-Validation: The 632+ Bootstrap Method , 1997 .

[15]  R. Mcguire.  Seismic Hazard and Risk Analysis , 2004 .

[16]  K. Goda,et al.  Ground‐Motion Prediction Models for Arias Intensity and Cumulative Absolute Velocity for Japanese Earthquakes Considering Single‐Station Sigma and Within‐Event Spatial Correlation , 2015 .

[17]  D. Bates,et al.  fitting linear mixed effects models using lme 4 arxiv , 2014 .

[18]  Response to Discussion by P. Malhotra on “NGA-West2 Research Project” , 2015 .

[19]  F. Scherbaum,et al.  Testing the Global Applicability of Ground‐Motion Prediction Equations for Active Shallow Crustal Regions , 2012 .

[20]  Galit Shmueli,et al.  To Explain or To Predict? , 2010, 1101.0891.

[21]  Julian J. Bommer,et al.  Challenges of Building Logic Trees for Probabilistic Seismic Hazard Analysis , 2012 .

[22]  Praveen K. Malhotra.  Discussion of “NGA-West2 Research Project” , 2015 .

[23]  M. Stone.  Cross‐Validatory Choice and Assessment of Statistical Predictions , 1976 .

[24]  Julian J. Bommer,et al.  On the Selection of Ground-Motion Prediction Equations for Seismic Hazard Analysis , 2010 .

[25]  J. Cavanaugh,et al.  The Bayesian information criterion: background, derivation, and applications , 2012 .

[26]  Thomas H. Jordan.  The Prediction Problems of Earthquake System Science , 2014 .

[27]  D. Bindi,et al.  Ground motion prediction equations derived from the Italian strong motion database , 2011 .

[28]  Warner Marzocchi,et al.  Toward a New Probabilistic Framework to Score and Merge Ground‐Motion Prediction Equations: The Case of the Italian Region , 2016 .

[29]  J. Bommer,et al.  Comparisons among the five ground-motion models developed using RESORCE for the prediction of response spectral accelerations due to earthquakes in Europe and the Middle East , 2014, Bulletin of Earthquake Engineering.

[30]  Malcolm R. Forster,et al.  How to Tell When Simpler, More Unified, or Less Ad Hoc Theories will Provide More Accurate Predictions , 1994, The British Journal for the Philosophy of Science.

[31]  W. Silva,et al.  NGA-West2 Database , 2014 .

[32]  Michael R. Hagerty,et al.  Comparing the predictive powers of alternative multiple regression models , 1991 .

[33]  Danijel Schorlemmer,et al.  Validating Intensity Prediction Equations for Italy by Observations , 2015 .

[34]  E. Faccioli,et al.  Broadband (0.05 to 20 s) prediction of displacement response spectra based on worldwide digital records , 2008 .

[35]  F. Cotton,et al.  Towards fully data driven ground-motion prediction models for Europe , 2014, Bulletin of Earthquake Engineering.

[36]  Leo Breiman,et al.  Statistical Modeling: The Two Cultures (with comments and a rejoinder by the author) , 2001, Statistical Science.

[37]  K. Campbell,et al.  NGA-West2 Ground Motion Model for the Average Horizontal Components of PGA, PGV, and 5% Damped Linear Acceleration Response Spectra , 2014 .

[38]  M. Stone.  An Asymptotic Equivalence of Choice of Model by Cross‐Validation and Akaike's Criterion , 1977 .