Weather systems that affect Hawaii are largely convective in nature and are organized on the mesoscale. However, the central Pacific region which encompasses the Hawaiian Island chain is a comparatively data sparse area. Operational in situ data are available from only 9 synoptic stations located on 4 islands, 2 radiosonde sites, 4 buoys, aircraft reports and a few maritime and local (mesonet) observations. Remotely sensed data (derived from satellite-based instruments, WSR-88D radar, GPS, sferics detectors, and UV lidar) represent a varied resource that can provide mesoscale information on atmospheric structures over the region. To make best use of these data resources in the central Pacific the Local Analysis and Prediction System (LAPS) has been implemented and is being tested at University of Hawaii. LAPS, designed at NOAA Forecast Systems Laboratory (FSL), is used to assimilate varied data sets into a coherent analysis of the atmosphere at high spatial and temporal resolution to provide (i) a powerful nowcasting tool, and (ii) as input to the Penn State/NCAR mesoscale model (MM5) for an explicit initialization of clouds and precipitation. MM5 has run operationally at the University of Hawaii Mauna Kea Weather Center since 1999, providing daily weather simulations, with a resolution of 9 km over the Hawaiian Island chain and up to 1 km over the summit of Mauna Kea. Preliminary results from cases study are presented here. In particular the impact of satellite data assimilation on initialization of MM5 is investigated with reference to the forecast distribution of clouds, water vapor, and precipitation.