Application of parallel computing to stochastic parameter estimation in environmental models

Vrugt, Jasper A.; Nualláin, Breanndán Ó; Robinson, Bruce A.; Bouten, W.; Dekker, Stefan C.; Sloot, Peter M. A.

doi:10.1016/j.cageo.2005.10.015

Cited by 64 publications

(42 citation statements)

References 40 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In this study, we will restrict ourselves to the DrEAM algorithm ("Differential Evolution Adaptive Metropolis", [28]), which is a generalization of DE-MC ("Differential Evolution Markov Chain", [24]). Both DE-MC and DrEAM are related to a class of Population Monte Carlo methods (see the literature review in [24]), which have been successfully used in hydrology research [25][26][27]. In DrEAM, one simultaneously releases Markov chains from an over-dispersed set of points in the parameter space.…”

Section: Multichain Methodsmentioning

confidence: 99%

Efficient uncertainty quantification methodologies for high-dimensional climate land models

Sargsyan

Safta

Berry

et al. 2011

View full text Add to dashboard Cite

Issu ed by Sa n dia Na t ion a l La bor a t or ies, oper a t ed for t h e Un it ed St a t es Depa r t m en t of E n er gy by Sa n dia Cor por a t ion . NOTICE:Th is r epor t wa s pr epa r ed a s a n a ccou n t of wor k spon sor ed by a n a gen cy of t h e Un it ed St a t es Gover n m en t . Neit h er t h e Un it ed St a t es Gover n m en t , n or a n y a gen cy t h er eof, n or a n y of t h eir em ployees, n or a n y of t h eir con t r a ct or s, su bcon t r a ct or s, or t h eir em ployees, m a ke a n y wa r r a n t y, expr ess or im plied, or a ssu m e a n y lega l lia bilit y or r espon sibilit y for t h e a ccu r a cy, com plet en ess, or u sefu ln ess of a n y in for m a t ion , a ppa r a t u s, pr odu ct , or pr ocess disclosed, or r epr esen t t h a t it s u se wou ld n ot in fr in ge pr iva t ely own ed r igh t s. Refer en ce h er ein t o a n y specific com m er cia l pr odu ct , pr ocess, or ser vice by t r a de n a m e, t r a dem a r k, m a n u fa ct u r er, or ot h er wise, does n ot n ecessa r ily con st it u t e or im ply it s en dor sem en t , r ecom m en da t ion , or favor in g by t h e Un it ed St a t es Gover n m en t , a n y a gen cy t h er eof, or a n y of t h eir con t r a ct or s or su bcon t r a ct or s. Th e views a n d opin ion s expr essed h er ein do n ot n ecessa r ily st a t e or r eflect t h ose of t h e Un it ed St a t es Gover n m en t , a n y a gen cy t h er eof, or a n y of t h eir con t r a ct or s. P r in t ed in t h e Un it ed St a t es of Am er ica . Th is r epor t h a s been r epr odu ced dir ect ly fr om t h e best ava ila ble copy.Ava ila ble t o DOE a n d DOE con t r a ct or s fr om U.S. Depa r t m en t of E n er gy Office of Scien t ific a n d or der s@n t is.fedwor ld.gov On lin e or der in g: h t t p://www.n t is.gov/h elp/or der m et h ods.a sp?loc=7-4-0#on lin e DE P A R T M EN T OF E NEfficient uncertainty quantification methodologies for high-dimensional climate land models Khachik Sargsyan, Cosmin Safta, Robert Berry, Jaideep Ray, Bert Debusschere, Habib Najm Sandia National Laboratories, Livermore, CA {ksargsy,csafta,rdberry,jairay,bjdebus,hnnajm}@sandia.gov AbstractIn this report, we proposed, examined and implemented approaches for performing efficient uncertainty quantification (UQ) in climate land models. Specifically, we applied Bayesian compressive sensing framework to a polynomial chaos spectral expansions, enhanced it with an iterative algorithm of basis reduction, and investigated the results on test models as well as on the community land model (CLM). Furthermore, we discussed construction of efficient quadrature rules for forward propagation of uncertainties from high-dimensional, constrained input space to output quantities of interest. The work lays grounds for efficient forward UQ for high-dimensional, strongly non-linear and computationally costly climate models. Moreover, to investigate parameter inference approaches, we have applied two variants of the Markov chain Monte Carlo (MCMC) method to a soil moisture dynamics submodel of the CLM. The ev...

show abstract

Section: Multichain Methodsmentioning

confidence: 99%

Efficient uncertainty quantification methodologies for high-dimensional climate land models

Sargsyan

Safta

Berry

et al. 2011

View full text Add to dashboard Cite

show abstract

“…MCMC also have difficulty with narrow and skewed posterior distributions; these may be resolved by either transforming the unknowns yielding a better-behaved distribution (i.e., more circular) [66] or adapting the proposal distribution [72], particularly when dealing with high-dimensional problems [73]. An intuitive way to deal with higher dimensional problems (where a single chain may have difficulty visiting the entire space in a reasonable number of iterations) or problems where evaluating the posterior distribution is computationally expensive (where a single chain may not be able to take many steps in a reasonable amount of time) is to have multiple chains distributed among multiple CPUs in a parallel supercomputer; Population Monte Carlo methods to do so have been investigated [74,75].…”

Section: Markov Chain Monte Carlo Methodsmentioning

confidence: 99%

Distributed micro-releases of bioterror pathogens : threat characterizations and epidemiology from uncertain patient observables.

Wolf

Marzouk²,

Adams³

et al. 2008

View full text Add to dashboard Cite

Terrorist attacks using an aerosolized pathogen preparation have gained credibility as a national security concern since the anthrax attacks of 2001. The ability to characterize the parameters of such attacks, i.e., to estimate the number of people infected, the time of infection, the average dose received, and the rate of disease spread in contemporary American society (for contagious diseases), is important when planning a medical response. For non-contagious diseases, we address the characterization problem by formulating a Bayesian inverse problem predicated on a short time-series of diagnosed patients exhibiting symptoms. To keep the approach relevant for response planning, we limit ourselves to 3-5 days of data. In computational tests performed for anthrax, we usually find these observation windows sufficient, especially if the outbreak model employed in the inverse problem is accurate. For contagious diseases, we formulated a Bayesian inversion technique to infer both pathogenic transmissibility and the social network from outbreak observations, ensuring that the two determinants of spreading are identified separately. We tested this technique on data collected from a 1967 smallpox epidemic in Abakaliki, Nigeria. We inferred, probabilistically, different transmissibilities in the structured Abakaliki population, the social network, and the chain of transmission. Finally, we developed an individual-based epidemic model to realistically simulate the spread of a rare (or eradicated) disease in a modern society. This model incorporates the mixing patterns observed 3 in an (American) urban setting and accepts, as model input, pathogenic transmissibilities estimated from historical outbreaks that may have occurred in socio-economic environments with little resemblance to contemporary society. Techniques were also developed to simulate disease spread on static and sampled network reductions of the dynamic social networks originally in the individual-based model, yielding faster, though approximate, network-based epidemic models. These reduced-order models are useful in scenario analysis for medical response planning, as well as in computationally intensive inverse problems.

show abstract

“…A. Vrugt (vrugt@lanl.gov) jectives to create the initial sample of points. The results presented in Vrugt et al (2003) have demonstrated that this alternative search strategy provides a computational efficient and robust alternative to multiobjective optimization.It is surprising however that TRW have not considered this second or alternative optimization strategy in their paper. I believe that this alternative sampling approach should have been used in conjunction with the various algorithms, to appropriately disseminate and implement the ideas presented in Vrugt et al (2003).…”

mentioning

confidence: 89%

mentioning

confidence: 89%

Comment on "How effective and efficient are multiobjective evolutionary algorithms at hydrologic model calibration?" by Y. Tang et al., Hydrol. Earth Syst. Sci., 10, 289–307, 2006

Vrugt

2007

Hydrol. Earth Syst. Sci.

View full text Add to dashboard Cite

In a recent paper by Tang, Reed and Wagener (2006, hereafter referred to as TRW) a comparison assessment was presented of three state-of-the-art evolutionary algorithms for multiobjective calibration of hydrologic models. Through three illustrative case studies, TRW demonstrate that the Strength Pareto Evolutionary Algorithm 2 (SPEA2) and Epsilon Dominance Nondominated Sorted Genetic Algorithm (ε-NSGAII) achieve a better performance than the Multiobjective Shuffled Complex Evolution Metropolis algorithm (MOSCEM-UA), previously developed by us and presented in Vrugt et al. (2003). I would like to congratulate TRW with their paper, which I believe provides a strong and valuable contribution to the field of hydrologic model calibration. However, I wish to differ in opinion about some of the main conclusions presented in their paper, especially with respect to the seemingly inferior performance of the MOSCEM-UA algorithm.The results presented in TRW were obtained using uniform random sampling of the initial parameter space. Such a sampling strategy is widely used within the water resources and computational science literature, and expresses a situation where very little prior information is available about the location of the Pareto optimal solution set. The initial sample is subsequently iteratively improved using the various algorithmic steps in the employed evolutionary algorithm. It is however possible to significantly improve the efficiency and robustness of evolutionary search for case studies (2) and (3) reported in TRW if we first attempt to create an initial sample that approximates the Pareto tradeoff surface as closely as possible. In our original paper (Vrugt et al., 2003) we suggest such an alternative sampling strategy by first locating the theoretical ends of the Pareto set using classical single objective optimization, and to use traditional first-order statistical theory around the optimal estimators for these individual obCorrespondence to: J. A. Vrugt (vrugt@lanl.gov) jectives to create the initial sample of points. The results presented in Vrugt et al. (2003) have demonstrated that this alternative search strategy provides a computational efficient and robust alternative to multiobjective optimization.It is surprising however that TRW have not considered this second or alternative optimization strategy in their paper. I believe that this alternative sampling approach should have been used in conjunction with the various algorithms, to appropriately disseminate and implement the ideas presented in Vrugt et al. (2003). If correctly used, this alternative search strategy would have provided the entire Pareto tradeoff surface as depicted in Fig. 5, at far less computational costs than the SPEA2 and ε-NSGAII algorithms. For example, preliminary analyses of the identification problem discussed in Fig. 5, suggest that state-of-the-art single objective search algorithms can identify the single criterion solutions of RMSE(R) and RMSE(T ) in less than 20 000 function evaluations. Experience further sug...

show abstract

Application of parallel computing to stochastic parameter estimation in environmental models

Cited by 64 publications

References 40 publications

Efficient uncertainty quantification methodologies for high-dimensional climate land models

Efficient uncertainty quantification methodologies for high-dimensional climate land models

Distributed micro-releases of bioterror pathogens : threat characterizations and epidemiology from uncertain patient observables.

Comment on "How effective and efficient are multiobjective evolutionary algorithms at hydrologic model calibration?" by Y. Tang et al., Hydrol. Earth Syst. Sci., 10, 289–307, 2006

Contact Info

Product

Resources

About

Application of parallel computing to stochastic parameter estimation in environmental models

Cited by 64 publications

References 40 publications

Efficient uncertainty quantification methodologies for high-dimensional climate land models

Efficient uncertainty quantification methodologies for high-dimensional climate land models

Distributed micro-releases of bioterror pathogens : threat characterizations and epidemiology from uncertain patient observables.

Comment on &quot;How effective and efficient are multiobjective evolutionary algorithms at hydrologic model calibration?&quot; by Y. Tang et al., Hydrol. Earth Syst. Sci., 10, 289–307, 2006

Contact Info

Product

Resources

About

Comment on "How effective and efficient are multiobjective evolutionary algorithms at hydrologic model calibration?" by Y. Tang et al., Hydrol. Earth Syst. Sci., 10, 289–307, 2006