The Extended Coherent Flamelet Model (ECFM) is limited to lower order upwinding schemes to minimize the numerical discrepancy between species and tracers, which can lead to inaccurate estimates of the progress variable and consequently negative conditional mass fractions in the burned gases after ignition. The recently developed Species-Based ECFM (SB-ECFM) removes the species tracers from the definition of the progress variable, and allows the use of higher order schemes. In this study, SB-ECFM is coupled with the Imposed Stretch Spark Ignition Model (ISSIM) to simulate a spark-ignition engine, the transparent combustion chamber (TCC) engine. To examine the spatial discretization effect and demonstrate the improvement due to using higher order schemes, Reynolds-Averaged-Navier-Stokes (RANS) simulations performed with a first-order upwinding scheme and a second-order central differencing scheme are compared. In addition, with second-order scheme, the effect of grid refinement around the spark plug on the predicted cycle-to-cycle variability (CCV) is investigated. To efficiently simulate multiple engine cycles, concurrent engine cycles are run following the parallel perturbation method (PPM). The simulation results are compared to experimental data. Three observations result from this study. First, the comparison shows that the second-order scheme is able to capture the CCV of in-cylinder pressure to some extent, while results from the first-order scheme do not capture variation after the first few cycles. Second, for this TCC case, the grid refinement around the spark plug does not improve the prediction of coefficient of variation (COV). Third, in this TCC case, the predicted in-cylinder peak pressure is strongly correlated to the turbulent kinetic energy at the beginning of spark.