Efficient stochastic approaches for sensitivity studies of an Eulerian large-scale air pollution model

Reliability of large-scale mathematical models is an important issue when such models are used to support decision makers. Sensitivity analysis of model outputs to variation or natural uncertainties of model inputs is crucial for improving the reliability of mathematical models. A comprehensive experimental study of Monte Carlo algorithms based on Sobol sequences for multidimensional numerical integration has been done. A comparison with Latin hypercube sampling and a particular quasi-Monte Carlo lattice rule based on generalized Fibonacci numbers has been presented. The algorithms have been successfully applied to compute global Sobol sensitivity measures corresponding to the influence of several input parameters (six chemical reactions rates and four different groups of pollutants) on the concentrations of important air pollutants. The concentration values have been generated by the Unified Danish Eulerian Model. The sensitivity study has been done for the areas of several European cities with different geographical locations. The numerical tests show that the stochastic algorithms under consideration are efficient for multidimensional integration and especially for computing small by value sensitivity indices. It is a crucial element since even small indices may be important to be estimated in order to achieve a more accurate distribution of inputs influence and a more reliable interpretation of the mathematical model results.Reliability of large-scale mathematical models is an important issue when such models are used to support decision makers. Sensitivity analysis of model outputs to variation or natural uncertainties of model inputs is crucial for improving the reliability of mathematical models. A comprehensive experimental study of Monte Carlo algorithms based on Sobol sequences for multidimensional numerical integration has been done. A comparison with Latin hypercube sampling and a particular quasi-Monte Carlo lattice rule based on generalized Fibonacci numbers has been presented. The algorithms have been successfully applied to compute global Sobol sensitivity measures corresponding to the influence of several input parameters (six chemical reactions rates and four different groups of pollutants) on the concentrations of important air pollutants. The concentration values have been generated by the Unified Danish Eulerian Model. The sensitivity study has been done for the areas of several European cities with different...

[1]  Wojciech Jarosz,et al.  Efficient Monte Carlo methods for light transport in scattering media , 2008 .

[2]  I. Dimov,et al.  Multidimensional Sensitivity Analysis of Large-Scale Mathematical Models , 2013 .

[3]  Ivan Tomov Dimov,et al.  Latin Hypercube Sampling and Fibonacci Based Lattice Method Comparison for Computation of Multidimensional Integrals , 2016, NAA.

[4]  Y. Wang,et al.  An Historical Overview of Lattice Point Sets , 2002 .

[5]  Stefano Tarantola,et al.  Sensitivity Analysis in Practice: A Guide to Assessing Scientific Models , 2004 .

[6]  Zahari Zlatev,et al.  Computer Treatment of Large Air Pollution Models , 1995 .

[7]  Giray Ökten,et al.  Generating low-discrepancy sequences from the normal distribution: Box-Muller or inverse transform? , 2011, Math. Comput. Model..

[8]  A. Saltelli,et al.  Importance measures in global sensitivity analysis of nonlinear models , 1996 .

[9]  Zahari Zlatev,et al.  Variance-Based Sensitivity Analysis of the Unified Danish Eulerian Model According to Variations of Chemical Rates , 2012, NAA.

[10]  Budiman Minasny,et al.  A conditioned Latin hypercube method for sampling in the presence of ancillary information , 2006, Comput. Geosci..

[11]  I. A. Antonov,et al.  An economic method of computing LPτ-sequences , 1979 .

[12]  I. Sloan,et al.  Lattice methods for multiple integration: theory, error analysis and examples , 1987 .

[13]  Zahari Zlatev,et al.  Air pollution modelling, sensitivity analysis and parallel implementation , 2011 .

[14]  B. Minasny,et al.  Conditioned Latin Hypercube Sampling for Calibrating Soil Sensor Data to Soil Properties , 2010 .

[15]  I. Sloan.  Lattice Methods for Multiple Integration , 1994 .

[16]  Makoto Matsumoto,et al.  SIMD-Oriented Fast Mersenne Twister: a 128-bit Pseudorandom Number Generator , 2008 .

[17]  Frances Y. Kuo,et al.  Constructing Sobol Sequences with Better Two-Dimensional Projections , 2008, SIAM J. Sci. Comput..

[18]  Stefano Tarantola,et al.  Trends in sensitivity analysis practice in the last decade. , 2016, The Science of the total environment.

[19]  Zahari Zlatev,et al.  Advanced algorithms for multidimensional sensitivity studies of large-scale air pollution models based on Sobol sequences , 2013, Comput. Math. Appl..

[20]  Zahari Zlatev,et al.  Parallel Computation of Sensitivity Analysis Data for the Danish Eulerian Model , 2011, LSSC.

[21]  Stefano Tarantola,et al.  Estimating the approximation error when fixing unessential factors in global sensitivity analysis , 2007, Reliab. Eng. Syst. Saf..

[22]  Sean McKee,et al.  Monte Carlo Methods for Applied Scientists , 2005 .

[23]  Zahari Zlatev,et al.  Advanced Sensitivity Analysis of the Danish Eulerian Model in Parallel and Grid Environment , 2011 .

[24]  Mark M. Meerschaert,et al.  Mathematical Modeling , 2014, Encyclopedia of Social Network Analysis and Mining.