1: \begin{abstract}

2: In this paper we present a rigorous cost and error analysis of a

3: multilevel estimator based on randomly shifted

4: Quasi-Monte Carlo (QMC) lattice rules for lognormal diffusion problems.

5: These problems are motivated by uncertainty quantification problems in

6: subsurface flow. 

7: We extend the convergence analysis in [Graham et al.,

8:   Numer.~Math.~2014]

9: to \emph{multilevel} Quasi-Monte Carlo finite element

10: discretizations and give a constructive proof of the dimension-independent

11: convergence of the QMC rules. More precisely, we provide suitable

12: parameters for the construction of such rules that yield 

13: the required variance reduction

14: for the multilevel scheme to achieve an $\varepsilon$-error with a cost of $\mathcal{O}(\varepsilon^{-\theta})$ with $\theta < 2$, and in practice even $\theta \approx 1$, for sufficiently fast decaying covariance kernels 

15: of the underlying Gaussian random field inputs. This confirms that

16: the computational gains due to the application of multilevel sampling

17: methods and the gains due to the application of QMC methods, 

18: both demonstrated in earlier works for the same model problem,

19: are complementary. A

20: series of numerical experiments confirms these gains.

21: The results show that in practice the

22: multilevel QMC method consistently outperforms both the multilevel MC method and

23: the single-level variants 

24: even for non-smooth problems.

25: \end{abstract}

26: