Ecological processes of low-productivity ecosystems have long been considered to be driven by abiotic controls with biotic interactions playing an insignificant role. However, existing studies present conflicting evidence concerning the roles of these factors, in part due to the short temporal extent of most data sets and inability to test indirect effects of environmental variables modulated by biotic interactions. Using structural equation modeling to analyze 65 years of perennial vegetation change in the Sonoran Desert, we found that precipitation had a stronger positive effect on recruitment beneath existing canopies than in open microsites due to reduced evaporation rates. Variation in perennial canopy cover had additional facilitative effects on juvenile recruitment, which was indirectly driven by effects of density and precipitation on cover. Mortality was strongly influenced by competition as indicated by negative density-dependence, whereas precipitation had no effect. The combined direct, indirect, and interactive facilitative effects of precipitation and cover on recruitment were substantial, as was the effect of competition on mortality, providing strong evidence for dual control of community dynamics by climate and biotic interactions. Through an empirically derived simulation model, we also found that the positive feedback of density on cover produces unique temporal abundance patterns, buffering changes in abundance from high frequency variation in precipitation, amplifying effects of low frequency variation, and decoupling community abundance from precipitation patterns at high abundance. Such dynamics should be generally applicable to low-productivity systems in which facilitation is important and can only be understood within the context of long-term variation in climatic patterns. This predictive model can be applied to better manage low-productivity ecosystems, in which variation in biogeochemical processes and trophic dynamics may be driven by positive density-dependent feedbacks that influence temporal abundance and productivity patterns.