A limited area model (LAM) intercomparison study is conducted based on a tropical monsoonal deep convection case observed during the Tropical Warm Pool - International Cloud Experiment (TWP-ICE). The LAM simulations are compared with the variational analyses (VA) based on the Atmospheric Radiation Measurement (ARM) observations and the cloud resolving model (CRM) simulations forced by the VA. Driven by the ECMWF analyses or global model forecasts, LAMs are able to produce the large-scale thermodynamic field reasonably well compared with the VA. However, the LAM simulated dynamic fields, such as the large-scale horizontal divergence, vertical velocity, and cyclogenesis in the monsoonal trough, have a large inter-model spread and deviate substantially from observations. Despite the differences in large-scale forcing, there is certain consistency between the CRM and LAM simulations: stratiform (w ≤ 1 m s-1) ice clouds dominate the cloud fraction and convective (w > 3 m s-1) clouds occupy less than 3 percent of the total cloudy area. But the hydrometeor content of stratiform ice clouds is only one tenth of that of convective and transitional (1 m s-1 < w ≤ 3 m s-1) ice clouds. However, there is a large inter-LAM spread in the simulated cloud fraction and hydrometeor mixing ratios. The inter-LAM difference in solid phase hydrometeors (cloud ice, snow, and graupel) can be up to nearly a factor of 10. Among all the hydrometeor types, the stratiform ice clouds are simulated least consistently by the LAMs. The large inter-LAM spread suggests that obtaining consistent and reliable dynamic and cloud fields remains a challenge for the LAM approach.