Question: Does the seed bank filter annual plant composition and determine cover at the species level?

Location: 510 m a.s.l., central Spain.

Methods: Seven transects and 136 quadrats were established in a semi-arid gypsum system. Seed bank samples were collected in each quadrat in September. The community was sampled the following April. For each quadrat we measured slope, microslope, landform, elevation, perennial cover and crust cover. Seed bank was estimated using the direct emergence method in glasshouse. Relationship among seed bank and annual community was assessed by Mantel correlations. Above-ground cover for the five most abundant species was modelled with GLMs.

Results: Seed bank density was the best predictor for annual community cover; perennial cover and landform were also included in the model. Species composition between September seed bank and April annual community cover was also highly related according to the Mantel test. This relationship was constant, even when the effect due to other abiotic (landform, microslope) or biotic (perennial cover, crust cover) parameters were partialled out. Microslope, elevation and seed bank density were the best parameters to predict spring cover of the five most abundant species.

Conclusions: Above-ground and below-ground community compartments are strongly related in terms of abundance and species composition. This relationship is filtered by several environmental factors (e.g. perennial cover, landform, microslope) that exert a strong control at community and individual levels. Our results support the hypothesis that annual community performance is affected by seed bank pattern.

Nomenclature: Tutin et al. (1964–1980) and Castroviejo et al. (1986–2005), except for Chaenorhinum reyesii (C. Vicioso & Pau) Benedí.

Abbreviation: GLM = General linearized model.