A better understanding of the depth distribution of soil mineralizable nitrogen (N) pools is important to improve prediction of net soil N mineralization. However, our understanding of the depth distribution of these N pools under the semi-arid conditions of western Canada is limited. This study examined the depth distribution of soil mineralizable N pools (k_S, the rate constant of a nondepleting zero-order stable N pool, and N_L, the size of a depleting first-order labile N pool) of six sites in western Canada chosen to vary with respect to soil zone, soil texture, and cropping system. The depth distribution of mineralizable N pools varied substantially among sites, indicating that this distribution needs to be considered in making predictions of net soil N mineralization. A single regression equation including soil total nitrogen (STN), Pool I (a labile mineralizable N pool determined through a 14-day aerobic incubation), and soil pH explained 67% of the variation in k_S across sites and soil depths. In addition, 95% of the variation in N_L was explained by a regression model with Pool I. Thus, although the depth distribution of soil mineralizable N pools can vary substantially among sites, the mineralizable N parameters can be adequately predicted across sites and soil depths from simple soil properties. Comparison with a study using surface soils under humid conditions in New Brunswick suggests that the relationship between N_L and Pool I is applicable across a wide range of soils, climatic zones, and cropping systems, whereas the regression model to predict k_S varied with climatic zone, perhaps reflecting different pedogenic processes stabilizing the organic matter in these climatic zones.