Given all the advantages of FSOC, Northrop Grumman has been investigating ways to optimize the performance of space-based optical communications systems in the presence of clouds and other atmospheric effects. A major component of this work focuses on improving the forecasts of clouds in the line-of-sight between a ground site and satellites in geosynchronous orbit. One potential ground site is the summit of Haleakala located on the Island of Maui, Hawaii. Haleakala is a very desirable location for ground-based telescopes such as the Mees Solar Observatory and the Daniel K. Inouye Solar Telescope (DKIST). Its summit is usually cloud-free (>70% of the time), but daily variations in the height of the trade wind inversion combine with complex terrain to make it difficult to predict the occurrence of clouds on the summit. To address this, a modified version of the Weather Research & Forecast (WRF) model is used to simulate the atmospheric conditions over the Hawaiian Islands to be able to anticipate link outages due to clouds. In order to capture the complex terrain and the resulting fine-scale atmospheric circulations and vertical structure, WRF is cycled at ultra-high resolution (1/3-km, 81 vertical levels) over Haleakala using 300m terrain data. The WRF three-dimensional variational (3DVAR) data assimilation system is used to combine observations from NCEP prepbufr files with a first-guess field from the previous WRF run to provide a 48 hour forecast. To evaluate the impact of the warm-start runs, vertical profiles of temperature and moisture from the warm-start WRF cycles are compared with those produced by non-cycled, cold-start runs. The impact of horizontal resolution is assessed by comparing the data from the 1-km, 3-km, and 9-km WRF domains to the data produced by the 1/3-km WRF nest. Results indicate that cycling using the WRF Data Assimilation (WRFDA) system captures the diurnal cycle of clouds in the boundary layer well and has a positive impact on the ability to forecast clouds at the summit, typically beating persistence. High-end visualizations of the cloud forecasts will be shown at the conference.