Most global climate models, such as CMIP5 models, ignore the falling ice (snow) radiative effects (FIREs). Extended from our previous works, we explore the impacts of FIREs on the geographical distribution changes of sea ice concentration (SIC), sea ice thickness and skin temperature (Ts) under 1% per year increase of atmospheric CO2 concentration. We perform a pair of 140-year experiments including FIREs (SoN) and excluding FIREs (NoS) using CESM1-CAM5. These two simulations are compared with each other and against CMIP5 multi-model mean without FIREs (CMIP5-NoS). The results show that the changes of SIC, thickness and radiation fields in NoS minus SoN largely match the changes between CMIP5-NoS and SoN in winter but less so in summer and annual mean. Both NoS and CMIP5-NoS simulate less downward longwave and net radiative warming (~20 - 30 W m-2), resulting in colder Ts over Arctic and adjacent lands (~5 - 8 K colder) and stronger meridional temperature gradient, leading to more SIC and thicker sea ice (~30 - 40 cm) over the Arctic ocean. The inclusion of FIREs produces stronger changes in geographic patterns and magnitudes of Ts, SIC and thickness from the first to middle 20 years but less from the middle to last 20 years. The SIC and thickness changes in SoN are associated with warmer Ts, increasing downward surface longwave warming and thus net warming relative to NoS and CMIP5-NoS. With FIREs, the model shows faster warming-driven sea ice retreats and thinning over the entire Arctic ocean, resulting in a sea ice-free Arctic ocean 30 years earlier, as well as much warmer Ts (up to 5 K) over the adjacent lands than in NoS case.