The goal of this study is to examine the impact of GLM data on RAP/HRRR data assimilation and forecasts. Programs were developed to derive flash rates form real-time GLM data feed and convert them to representative reflectivity profiles. These proxy reflectivity profiles are then transferred to temperature tendencies and assimilated by RAP/HRRR through Diabatic Digital Filter or latent heating nudging. A case study on the Caribbean area was carried. At this area, there is very limited radar coverage and limited ground-based lightning observation. Therefore, GLM data will play more vital role in making quality short-term storm forecasts for this area. The characteristics of GLM data distribution is compared with GLD360 data which is from a ground-based lightning observing network. The legacy L2R (Lightning to Reflectivity) algorithm is updated to best represent the GLD data characteristics. The comparison experiments showed that the GLM data improves the data assimilation and short-term forecasts up to 6h.