The study of air-coupled rain/droplet flow behavior on surfaces with different roughness conditions is critical for the water management problem in automotive industry, which has become more significant nowadays due to the fast development of e-vehicles. In this work, a robust weakly-compressible SPH formulation is presented to facilitate the numerical study of typical rain/droplet flow phenomena involved in the vehicle driving environment. More specifically, an adaptive and low-dissipative Riemann solver is employed to ensure a regularized pressure field of the flow. A modified wall boundary conditions is developed to take into account the roughness effect of the geometry surfaces. In order to capture the droplet flows, a pair-wised particle force (PF) model is introduced to handle the surface tension effect and to obtain a prescribed static contact angle. The advantage of the PF model is the inclusion of air phase is not required. To predict the influence of transient air flow on droplets, an off-line air-fluid coupling method is developed. Moreover, a new rain inlet boundary condition is proposed to generate realistic rain flow with distributed droplet diameters and exact volume flow rate. Last, to speedup the simulation, the proposed method is implemented on a novel GPU SPH in-house code, which allows us to simulate up to 107 particle on a single NVIDIA A100 card. Various numerical experiments are carried out to evaluate the proposed SPH formulation, which varies from single droplet on a flat plate to wheel hydroplaning in a wind tunnel. Our numerical results are validated with experimental results captured by high-speed camera provided by CAERI. Comparison results suggest that a good agreement is achieved.