Subject: JCOMP-D-08-00674 From: Journal of Computational Physics Date: Fri, 12 Dec 2008 10:09:28 -0500 To: "Luo, Li-Shi" Re: JCOMP-D-08-00674 Dear Prof. Luo, The Editorial Office has just received the decision on the paper entitled "Comparison of the Lattice Boltzmann and Pseudo-Spectral Methods for Decaying Turbulence. Part I. Low-Order Statistics". The reviewers' comments are as follows: ************************************************************ Reviewer #1: This is a very interesting and relevant paper that compares the de-facto standard method for DNS of turbulence (the pseudo spectral method) to a relatively novel approach to numerical simulation of fluid flow (the lattice-Boltzmann method). It is relevant in the sense that it adds to the value of the LB method for turbulence simulations. The LB method has great potential in the sense that is easily parallelizable and can handle complexly shaped flow domains. The specific case studied is the decay of homogeneous isotropic turbulence in a fully periodic, three-dimensional domain. The paper does not contain new computational approaches; it compares (in a very detailed manner) two existing ones. In my view it is suitable for publication as is. The only general question/comment I have is that it surprises me that the strong pressure waves travelling through the domain (apparently) does not have significant impact on the velocity (and vorticity) fields. The authors' opinion on this aspect would be highly appreciated. The remarks on resolution (grid size as compared to Kolmogorov scale, eqs 28 and 29) are not fully clear to me. Does it mean that LB grids should be (in a linear sense) roughly twice as large to capture the detailed structures of turbulence? If so, shouldn't this come back when discussing computational efficiency? For an explicit scheme a twice as fine grid implies a 16 times bigger computational effort. ************************************************************ Reviewer #2: Comparison of the Lattice Boltzmann and Pseudo-Spectral Methods for Decaying Turbulence. PartI. Low-Order Statistics by Yan Peng, Wei Liao, Li-Shi Luo, and Lian-Ping Wang The authors compare two methods, Lattice Boltzmann (LBE) and pseudo-spectral (PS) on a three-dimensional numerical simulations of a turbulent flow with random initial data with prescribed spectrum. The tests performed are standard (temporal data on energy and dissipation, spectral data, third and fourth-order moments, and pressure). The Mach number (since LBE is a compressible method) is also monitored. I leave to the Editor to decide whether this type of work is acceptable to JCP (no new method, just comparisons) and will comment on the quality of the tests. The comparison is detailed (perhaps too much) and careful but I have several reservations concerning this work. The Taylor Reynolds number is very modest (ca. 25), with linear resolutions of 128 points, a factor of 32 from the world record of 4096 performed in Japan, i.e. 10**6 below in computer resources. It is thus not surprising that the dissipation (Fig. 5) increases but barely; in other words this flow is barely turbulent. So, carrying the comparison to 30 turn over times is not really meaningful unless one is interested in quasi laminar flows. This Reynolds number is lower than what was published in 2005 by L.S. Luo et al. (ref. 21) (see also ref. 24)! This provides a second challenge for the Editor: can this comparison be considered sufficiently informative in the turbulence regime? My answer is negative. Some tests are performed at higher Reynolds number (see Table 2), but with a resolution criterium that is not entirely satisfactory. One can follow the authors in saying that LBE behaves satisfactorily at these low Reynolds numbers, which is information indeed. In view of preceding works by the authors or a subset thereof on similar investigations, and including in JCP, I think this paper should be rejected. I must say that in principle, what happens for high-order statistics which provide more stringent tests of numerical methods, is to be seen although at these Reynolds number, not much can be inferred. ************************************************************ In view of these comments made, we regret to inform you that the Associate Editor who guided your paper, Professor Pouquet, has decided that we are unable to publish your work in Journal of Computational Physics. Yours sincerely, Basil Nyaku on behalf of the Editors of Journal of Computational Physics Editorial-Production Department, Elsevier Radarweg 29 1043 NX Amsterdam The Netherlands Tel: +31 20 485 3426 Fax: +31 20 485 2521 E-mail: jcp@elsevier.com