Accurate and timely operational river forecasts are critical for prediction of both flood levels and water supply in the western United States. Although operational hydrologic models vary in complexity, nearly all the models need values for particular parameters. The traditional method of obtaining parameters is manual calibration, requiring extensive training, time, and labor. For the past few years, through the collaboration of the National Weather Service (NWS), researchers at the University of Arizona, and now at UCLA, have been able to combine components of manual and automatic calibration techniques to produce a Multi-step Automatic Calibration Scheme (MACS). This scheme differs from previously tested automatic routines in two respects: its global search algorithm and its multi-step approach designed to mimic NWS calibration procedures. Originally, the MACS approach was tested on basins in the North Central River Forecast Center (NCRFC) and South East River Forecast Center (SERFC). The results compared favorably with the RFC manual calibration results for these “lumped systems”. The MACS procedure was then also tested on multi-level basins in the western United States. Original work on basins in the Colorado Basin River Forecast Center (CBRFC) revealed that additional research is needed to adapt and implement an automated procedure for the complex multi-level basins in the western United States. This study focuses on several forecast points in the San Juan River system in southwest Colorado. Various approaches are being tested to estimate the numerous parameters (~80) needed for a single forecast point. Preliminary results show improvement when using a grouping of levels (or tiers) as opposed to a level-by-level approach. Results from these various methods will be presented along with current efforts on integration of automatic calibration procedures into RFC operations.