Spatial modeling of biological soil crusts to support rangeland assessment and monitoring

Biological soil crusts are a diverse soil surface community, prevalent in semiarid regions, which function as ecosystem engineers and perform numerous important ecosystem services. Loss of crusts has been implicated as a factor leading to accelerated soil erosion and other forms of land degradation. To support assessment and monitoring efforts aimed at ensuring the sustainability of rangeland ecosystems, managers require spatially explicit information concerning potential cover and composition of biological soil crusts. We sampled low disturbance sites in Grand Staircase-Escalante National Monument (Utah, USA) to determine the feasibility of modeling the potential cover and composition of biological soil crusts in a large area. We used classification and regression trees to model cover of four crust types (light cyanobacterial, dark cyanobacterial, moss, lichen) and 1 cyanobacterial biomass proxy (chlorophyll a), based upon a parsimonious set of GIS (Geographic Information Systems) data layers (soil types, precipitation, and elevation). Soil type was consistently the best predictor, although elevation and precipitation were both invoked in the various models. Predicted and observed values for the dark cyanobacterial, moss, and lichen models corresponded moderately well (R 2 = 0.49, 0.64, 0.55, respectively). Cover of late successional crust elements (moss + lichen + dark cyanobacterial) was also successfully modeled (R2 = 0.64). We were less successful with models of light cyanobacterial cover (R2 = 0.22) and chlorophyll a (R2 = 0.09). We believe that our difficulty modeling chlorophyll a concentration is related to a severe drought and subsequent cyanobacterial mortality during the course of the study. These models provide the necessary reference conditions to facilitate the comparison between the actual cover and composition of biological soil crusts at a given site and their potential cover and composition condition so that sites in poor condition can be identified and management actions can be taken.