Evolutionary significance of awn length variation in a clonal grass of fire‐prone savannas

We investigated the selective advantage of continuous variation of awn length in the clonal grass Hyparrhenia diplandra (Poaceae) in a West African tropical savanna (Ivory Coast) that burns every year. This species is apomictic, so all seeds produced by one mother plant have the same genotype. H. diplandra produces diaspores with a hygroscopic awn whose length varies within and among plants. In other grass species, awns play a significant role in secondary dispersal on the ground and in seed burial. Using an experimental approach, we quantified (1) the environmental and genetic components of awn length variation, (2) the influence of awn length variation on seed burial depth, (3) the effect of awn length on seed germination and seedling survival rates in burned and unburned plots, and (4) the potential role of fire intensity in selecting diaspores with different awn lengths. Awn length varied from 35 to 70 mm and was highly heritable (h² = 0.63). Variation within individuals was not related to flowering phenology or flower position on the inflorescence. Awn length was significantly and positively correlated with seed burial depth. The range of depths from where buried seeds were recovered varied between 0 and 2.4 cm; this represents the layer where high soil temperature was recorded during fires in both field and experimental conditions. Thus, short‐awned seeds were more likely to be affected by soil temperature elevation during a fire than were long‐awned seeds. During two consecutive years, a significant relationship between mean seedling survival and mean awn length was only found in 3 ha of naturally burned plots. No such significant relationship existed in experimentally unburned plots in 1997 or 1998. In burned plots, the relationship between mean awn length and mean proportion of surviving seedlings was negative in 1997, while in 1998 the same relationship was positive. In natural conditions, fire intensity was significantly different between years, and variation in fire intensity could be related to between‐year differences in climatic conditions and the amount of biomass that was available for burning. Furthermore, seed germination probability increased when seed burial depth increased and when elevation of the soil temperature was moderate (<100°C). We suggest that, depending on fire intensity, different awn lengths are selected because of different soil temperatures, with very intense fires selecting for long‐awned diaspores. Since fire intensity is unpredictable from year to year, this maintains the continuous variation of awn length within populations. The spatial and temporal heterogeneity select for different clones that have contrasting awn morphologies.