The paper aims at assessing a hypothesis that resolution required to evaluate fuel consumption and heat release rates by directly (i.e., without a subgrid model of unresolved influence of small-scale turbulent eddies on the local flame) processing filtered fields of density, temperature, and species mass fractions should be significantly finer than resolution required to directly compute flame surface density by processing the same filtered fields. For this purpose, box filters of various widths Δ are applied to threedimensional Direct Numerical Simulation data obtained earlier by Dave et al. (Combust. Flame 196 (2018) 386-399) from a statistically one-dimensional and planar, moderately lean H2/air complexchemistry flame propagating in a box under conditions of sufficiently intense small-scale turbulence (Karlovitz number is larger than unity and a ratio of laminar flame thickness δL to Kolmogorov length scale is about 20). Results confirm this hypothesis and show that the mean flame surface density and area can be predicted with acceptable accuracy by processing filtered combustion progress variable fields computed using a sufficiently wide filter, e.g., ∆⁄δL=4⁄3. Such an approach does not require a model of the influence of subgrid turbulent eddies on flame surface density provided that ∆ and δL are of the same order of magnitude. Good performance of this approach is attributed to inability of smallscale (when compared to δL) turbulent eddies to substantially change the local flame structure, which, nevertheless, is significantly perturbed by larger turbulent eddies that strain the local flame.