Downscaling techniques have been used in a wide range of weather and climate applications from severe storms to hurricane forecast. Its computational inexpensive nature allows us to obtain detailed atmospheric information comparing to running fine resolution numerical forecast model with sufficient spinning time periods. It provides critical location specific and terrain dependent flow information for fine scale analysis and nowcasting. A Multiscale dynamic downscaling technique has recently been developed and evaluated at NOAA Earth System Research Laboratory. It is based on a Multigrid technique and minimizes the temporal tendency of full Euler equations. It combines information from a limited area (weather prediction) model forecast, observational data, and complex terrain to produce high-resolution analysis, and has demonstrated its superiority to the single grid analysis for many applications. In this presentation, we will demonstrate the robust performance of this multigrid-based downscaling technique, and its application in providing realistic, fine resolution wind and precipitation forecast for typhoon and fire weather.