From experimental spark ignition (SI) engine studies, it is known that the slow-down of early flame kernel development caused by the (Le > 1)-property of common transportation-fuel/air mixtures tends to increase cycle-to-cycle variations (CCV). To improve the fundamental understanding of the complex phenomena inside the flame structure of developing flame kernels, an engine-relevant DNS database is investigated in this work. Conclusive analyses are enabled by considering equivalent flame kernels and turbulent planar flames computed with Le > 1 and Le = 1. In Part I of the present study (Falkenstein et al., Combust. Flame, 2020), a reduced representation of the local mixture state was proposed for the purpose of this analysis. Fluctuations in heat release rate attributed to differential diffusion were found to be governed by the parameters local enthalpy, local equivalence ratio, and H-radical mass fraction. Here, a coupling relation for the diffusion-controlled mixture parameter local enthalpy with local flame geometry and structure is derived, characterized by the key parameters k and vertical bar del c vertical bar/vertical bar del c vertical bar(lam). The analysis shows that the large positive global mean curvature intrinsic to the flame kernel configuration may detrimentally affect the local mixture state inside the reaction zone, particularly during the initial flame kernel development phase. External energy supply by spark ignition may effectively bridge over this critical stage, which causes the impact of global mean flame kernel curvature to be small under the present conditions compared to the overall effect of Le not equal 1 observed in a statistically planar flame. Once ignition effects have decayed, the mixture state inside the reaction zone locally exhibits an identical dependence on vertical bar del c vertical bar as in a strained laminar flame. This implies that differential diffusion effects at Karlovitz numbers representative for part-load conditions are not weakened by small-scale turbulent mixing, which is undesirable for the engine application, but can be favorable in terms of modeling. (C) 2020 The Combustion Institute. Published by Elsevier Inc. All rights reserved.