Modeling sediment yield is a complex task because of the nonlinearity of natural processes intervening at slope and basin scale. In this study slope steepness, vegetation cover, and soil properties along with sediment yield were studied in 20 pasture micro‐catchments in a semi‐arid region, NW Iran in order to understand and predict sediment yield. The micro‐catchments included only one first‐order gully and drain toward a rock check dam in the outlet. The sediment yield of each micro‐catchment was calculated using the measurement of sediment mass in the check dams for a 16‐year period (1994–2010). Relationships between sediment yield and drainage characteristics were analyzed using correlation matrix and multiple linear regression method. Based on the results, sediment yield in the micro‐catchments varied from 0.29 Mg ha−1 y−1 to 14.81 Mg ha−1 y−1, with an average of 5.04 Mg ha−1 y−1. It was significantly related to slope steepness, vegetation cover, and soil organic matter using a linear regression equation (R2 = 0.87, p < 0.001). The slope, vegetation, and soil organic matter explained about 44%, 23%, and 20% of total variance in sediment yield, respectively. The spatial validation of the model using data from eight different micro‐catchments located nearby showed that the model efficiency is 0.94. Therefore, the model can be used for predicting sediment yield in this and similar study area, with a high degree of accuracy. Copyright © 2016 John Wiley & Sons, Ltd.