Restoration treatments in dry forests of the western US often attempt silvicultural practices to restore the historical characteristics of forest structure and fire behavior. However, it is suggested that a reliance on non-spatial metrics of forest stand structure, along with the use of wildland fire behavior models that lack the ability to handle complex structures, may lead to uncharacteristically homogeneous rather than heterogeneous forest structures following restoration. In our study, we used spatially explicit forest inventory data and a physics based fire behavior model to investigate the effects of restoration driven, variable retention harvests on structural complexity, both of horizontal and vertical dimensions, and potential fire behavior. Structural complexity was assessed at stand and patch scales using a combination of point pattern analyses, a patch detection algorithm, and nearest-neighbor and tree patch indices of height variation. The potential fire behavior before and after treatment was simulated across a range of open wind speeds using a 3-D physics based fire behavior model, the Wildland-urban interface Fire Dynamics Simulator (WFDS). Our results show that treatments resulted in an aggregated spatial pattern of trees consisting of a matrix of individual trees, clumps and openings similar to descriptions of historical dry forests. Treatments had inconsistent effects on vertical complexity across sites likely due to differences in treatment of ladder fuels; lack of reference conditions hinder evaluation of this structural aspect. Simulation modeling using WFDS suggest that treatments moderated fire rate of spread, fireline intensity and canopy consumption across all wind speeds tested and shifted potential fire behavior towards historical ranges. Our findings suggest that current restoration-based variable retention harvests can simultaneously fulfill objectives of altering structural complexity and of reducing fire behavior, though we recommend further research on desired ranges of vertical complexity to inform treatment design.