Functional principal component analysis for global sensitivity analysis of model with spatial output

被引:24
作者
Perrin, T. V. E. [1 ]
Roustant, O. [2 ]
Rohmer, J. [3 ]
Alata, O. [4 ]
Naulin, J. P. [5 ]
Idier, D. [3 ]
Pedreros, R. [3 ]
Moncoulon, D. [5 ]
Tinard, P. [5 ]
机构
[1] Univ Clermont Auvergne, Inst Henri Fayol, CNRS, Mines St Etienne,UMR LIMOS 6158, F-42023 St Etienne, France
[2] Univ Toulouse, Inst Math Toulouse, INSA, 135 Ave Rangueil, F-31077 Toulouse 4, France
[3] Bur Rech Geol & Minieres, 3 Av Claude Guillemin,BP 36009, F-45060 Orleans 2, France
[4] Univ Lyon, Lab Hubert Curien, UMR CNRS 5516, UJM St Etienne,IOGS, F-42023 St Etienne, France
[5] CCR, 157 Blvd Haussmann, F-75008 Paris, France
关键词
Global sensitivity analysis; Spatial data; Functional principal component analysis; Wavelet; B-splines; SIMULATIONS; DEPENDENCE;
D O I
10.1016/j.ress.2021.107522
中图分类号
T [工业技术];
学科分类号
08 ;
摘要
Motivated by risk assessment of coastal flooding, we consider time-consuming simulators with a spatial output. The aim is to perform sensitivity analysis (SA), quantifying the influence of input parameters on the output. There are three main issues. First, due to computational time, standard SA techniques cannot be directly applied on the simulator. Second, the output is infinite dimensional, or at least high dimensional if the output is discretized. Third, the spatial output is non-stationary and exhibits strong local variations. We show that all these issues can be addressed all together by using functional PCA (FPCA). We first specify a functional basis, such as wavelets or B-splines, designed to handle local variations. Secondly, we select the most influential basis terms, either with an energy criterion after basis orthonormalization, or directly on the original basis with a penalized regression approach. Then FPCA further reduces dimension by doing PCA on the most influential basis coefficients, with an ad-hoc metric. Finally, fast-to-evaluate metamodels are built on the few selected principal components. They provide a proxy on which SA can be done. As a by-product, we obtain analytical formulas for variance-based sensitivity indices, generalizing known formula assuming orthonormality of basis functions.
引用
收藏
页数:13
相关论文
共 35 条
[1]  
[Anonymous], 2009, CROSS VALIDATION
[2]  
[Anonymous], 2015, Journal of Statistical Software, DOI [DOI 10.18637/JSS.V065.I11, 10.18637/jss.v065.i11]
[3]  
BJORCK A, 1994, LINEAR ALGEBRA APPL, V198, P297
[4]   IMPLEMENTING SOBOLS QUASIRANDOM SEQUENCE GENERATOR [J].
BRATLEY, P ;
FOX, BL .
ACM TRANSACTIONS ON MATHEMATICAL SOFTWARE, 1988, 14 (01) :88-100
[5]   Storm-induced marine flooding: Lessons from a multidisciplinary approach [J].
Chaumillon, Eric ;
Bertin, Xavier ;
Fortunato, Andre B. ;
Bajo, Marco ;
Schneider, Jean-Luc ;
Dezileau, Laurent ;
Walsh, John Patrick ;
Michelot, Agnes ;
Chauveau, Etienne ;
Creach, Axel ;
Henaff, Alain ;
Sauzeau, Thierry ;
Waeles, Benoit ;
Gervais, Bruno ;
Jan, Gwenaele ;
Baumann, Juliette ;
Breilh, Jean-Francois ;
Pedreros, Rodrigo .
EARTH-SCIENCE REVIEWS, 2017, 165 :151-184
[6]   Efficient meta-modelling of complex process simulations with time-space-dependent outputs [J].
Chen, Tao ;
Hadinoto, Kunn ;
Yan, Wenjin ;
Ma, Yifei .
COMPUTERS & CHEMICAL ENGINEERING, 2011, 35 (03) :502-509
[7]   Global sensitivity analysis with dependence measures [J].
Da Veiga, Sebastien .
JOURNAL OF STATISTICAL COMPUTATION AND SIMULATION, 2015, 85 (07) :1283-1305
[8]   ORTHONORMAL BASES OF COMPACTLY SUPPORTED WAVELETS [J].
DAUBECHIES, I .
COMMUNICATIONS ON PURE AND APPLIED MATHEMATICS, 1988, 41 (07) :909-996
[9]   Sensitivity analysis with dependence and variance-based measures for spatio-temporal numerical simulators [J].
De Lozzo, Matthias ;
Marrel, Amandine .
STOCHASTIC ENVIRONMENTAL RESEARCH AND RISK ASSESSMENT, 2017, 31 (06) :1437-1453
[10]  
FFSA, 2011, LA TEMPETE