Regression analysis was used to develop models for estimating mean annual suspended-sediment loads for streams in Colorado. Mean annual suspended-sediment loads at 81 selected streamflow-gaging stations in the central, northwestern, and southwestern regions of Colorado were expressed as functions of geomorphic and hydrologic variables. A multiple-regression model that included mean basin elevation, mean annual streamflow, and drainage-basin area explained 78% of the variance in mean annual suspended-sediment load when all sites were analyzed together. The state was divided into four regions to decrease variance from spatial differences in geography and climate, and multiple-regression models were recomputed for each region. The best multiple-regression models for the central, northwestern, and southwestern regions of Colorado included mean annual streamflow and mean basin elevation. A multiple-regression model was not developed for eastern Colorado because few sites in this region had adequate sediment-load records. Regionalization of mean annual suspended-sediment loads resulted in improved multiple-regression models for the central, northwestern, and southwestern regions of Colorado. The regional multiple-regression models can be used to estimate mean annual suspended-sediment loads for other streams in these regions when mean annual streamflow and mean basin elevation are known. Regional regression models based only on drainage area also were developed, and they can be used to estimate mean annual suspended-sediment load when annual streamflow is unknown. (USGS)