Cumulus clouds are of great interest in numerical weather prediction. However, the scarcity of observed data on the Tibetan Plateau (TP) has not allowed correct interpretation of their development. The Third TP Atmospheric Science Experiment provided experimental data to address this challenge. This study used a combined weather research and forecasting large-eddy simulation (WRF-LES) model and final reanalysis data from the Global Forecast System to simulate cumulus clouds over southern TP on July 19, 2014. We applied observation nudging and one-way nesting strategies to influence the optimality of WRF-LES runs. The study performed simulations with six different scenarios in comparison with observation data. Results showed that cumulus clouds locally initiated and grew upscale but were however influenced by large-scale forcing. Compared to the observations, simulations with observation nudging provided more accurate and reliable results than the simulations without nudging. LES with mesoscale forcing yielded a relatively good atmospheric boundary-layer (ABL) water vapor profile and a similar microphysical evolution to the observations but misled the observed surface variables. Without mesoscale forcing, LES provided the best ABL water vapor and sensible heat flux, however, failed to provide a good microphysics field. In this aspect large-scale forcing played an important role in cumulus development on July 19, 2014. The study recommended observation nudging and one-way nesting strategies in separate iterations to be improved by focusing on the model’s response to the terrain and boundary conditions.