In this dissertation a wind erosion model for sparsely vegetated rangelands has been developed which uses a small set of key input parameters while also incorporating stochastic methods to capture the inherent variability of threshold shear velocities. The model addresses the effect of vegetation on the threshold of sediment entrainment using the Raupach 'et al.' (1993) shear stress partitioning approach. In this study, a field research component complements the computational modeling. A portable field wind tunnel was used to quantify threshold shear velocity (u*t) and horizontal (q) and vertical (F) sediment flux for representative rangeland soils of the Jornada Experimental Range, New Mexico. Field wind tunnel data are combined with crust strengths, measured using a pin penetrometer, to assess the effect of crust strength on wind erosion. Rice 'et al.' (1997; 1999) developed a conceptual model describing the relationship between grain impact energy and the energy required to rupture the crust. Several studies have been conducted in laboratory wind tunnels testing this hypothesis, but few have been conducted in the field. The energy required to rupture these natural crusts was found to exceed the energy of any individual saltating grains, suggesting that the surface would be resistant to wind erosion. However, weak, statistically significant power relationships were found between crust strength and both the horizontal sand flux and the abrasion efficiency. To estimate the impact energy of the saltating grains, a numerical particle trajectory model was further developed and calibrated for use in this dissertation research. For each representative soil texture a range of u*t values was observed, hence a range of u*t values more accurately represents this parameter. Stochasticity is incorporated into the wind erosion model by using an algorithm which assigns a u*t value based on the mean value along with the associated uncertainty specified by the standard deviation. Thus the link between empirical research and model development is illustrated. The output of the model shows that the spatial pattern of erosion 'hot spots' can be predicted using this methodology and supports the concept that surface heterogeneity can be addressed by employing stochastic methods.