Summary Runoff in arid and semiarid areas is characterized by high spatial and temporal variability. The spatial component is largely associated with the high spatial heterogeneity of soil surface attributes, such as vegetation and rock fragment covers, topography, and soil crust typology. Biological soil crusts (BSCs) are a common soil cover in arid and semiarid areas, and they play an essential role in local hydrological processes, since they affect many soil surface attributes associated with hydrologic properties. Although several publications have reported on the influence of BSCs on runoff at microplot spatial scales, only a few have examined their influence on larger spatial scales. Moreover, very few studies have analyzed the effect of BSCs on runoff under natural rain conditions. This is difficult, since a complex pattern of interactions among rainfall properties, BSC characteristics and some local characteristics, such as topography or type of soil is expected. In addition, in order to achieve a realistic model of how BSCs and rainfall affect runoff, it would be necessary to consider the level of human-driven degradation of BSCs. In this study, runoff was analyzed in plots with varying cover of cyanobacterial and lichen BSCs at microplot and small hillslope scales for two hydrological years in badlands in SE Spain. Structural equation modelling (SEM) was applied to test both direct and indirect relationships of BSC cover, slope gradient, rainfall characteristics and runoff. Our model showed that rainfall characteristics were the main factors controlling runoff yield. The slope positively affected runoff at small hillslope scales, but did not influence runoff at microplot scales. Runoff decreased at both scales with increased lichen-BSC cover. However, this effect was only significant during low-intensity events. Under high rainfall intensities, neither the BSC cover nor the slope had a causal effect on runoff. Our results suggest that incorporation of BSC-crusted surfaces in models should improve their capabilities for low-intensity rainfall events in semiarid areas similar to ours, but would be less important for high-intensity events. In addition, our approach, which takes direct and indirect relationships of factors affecting runoff into consideration, provides a very accurate picture of the process.