Advances in computing resources are gradually moving regional as well as global numerical forecasting simulations towards high resolutions. However, long-term simulations at global high resolutions remain a challenge. The recently developed variable-resolution global modeling framework, Model for Prediction Across Scales (MPAS), provides an opportunity to mitigate some computational limitations by using regional grid refinement. For effective use of such a modeling framework in climate studies, the impacts of resolution-dependent physics and dynamics on climate simulations should be thoroughly examined and alleviated. This study assesses the dependence of model sensitivity to resolution on the choice of model physics packages. To this end, several three-year numerical simulations are performed with the non-hydrostatic MPAS coupled to the CAM4 and CAM5 physics at multiple global quasi-uniform resolutions of 240 km, 120 km, 60 km, and 30 km and a global variable resolution configuration at 120 km resolution with a regional mesh refinement at 30km resolution over the tropics. While MPAS-CAM4 appears less sensitive to resolution than MPAS-CAM5 for individual components of precipitation, both exhibit similar increase in total precipitation with increased resolution. The simulations also exhibit sensitivity of the subtropical jet position that affects the sensitivity of atmospheric river (AR) frequency to resolutions. Hence model sensitivity to resolution depends on the physics parameterizations used. The simulations with regional mesh refinement also demonstrate significant impacts of resolution-dependent physics.