Search

Aim of the study: Models that combine parentage analysis from molecular data with spatial information of seeds and seedlings provide a framework to describe and identify the factors involved in seed dispersal and recruitment of forest species. In the present study we used a spatially explicit method (the gene shadow model) in order to assess primary and effective dispersal in Pinus canariensis.Area of study: Pinus canariensis is endemic to the Canary Islands (Spain). Sampling sites were a high density forest in southern slopes of Tenerife and a low density stand in South Gran Canaria. Materials and methods: We fitted models based on parentage analysis from seeds and seedlings collected in two sites with contrasting stand density, and then compared the resulting dispersal distributions.Main results: The results showed that: 1) P. canariensis has a remarkable dispersal ability compared to other pine species; 2) there is no discordance between primary and effective dispersals, suggesting limited secondary dispersal by animals and lack of Janzen-Connell effect; and 3) low stand densities enhance the extent of seed dispersal, which was higher in the low density stand.Research highlights: The efficient dispersal mechanism of P. canariensis by wind inferred by the gene shadow model is congruent with indirect measures of gene flow, and has utility in reconstructing past demographic events and in predicting future distribution ranges for the species.Keywords: Bayesian inference; Canary Islands; gene shadow model; microsatellites; parentage analysis.

International audience; 1. The seed dispersal kernel is a major determinant of spatial population dynamics and spatial distribution of genetic diversity. Among the main methods to estimate it, inverse modelling (IM) and gene shadow model (GSM) rely on seed counts in traps, whereas competing source model (CSM) and spatially explicit mating models (SEMMs) rely on compositions of seed pools. Moreover, GSM, CSM and SEMM exploit genetic information from molecular markers, whereas IM only exploits seed counts ignoring seed origins. These methods were also applied to established seedlings. 2. In the presence of post-dispersal density-dependent mortality (DDM), the effective dispersal kernel, describing the spatial distribution of established seedlings relatively to the seed source, is notoriously different from the basic dispersal kernel, describing the spatial distribution of seed deposition sites relatively to the source. Using simulated data sets, we investigated whether IM, GSM, CSM and SEMM applied to established seedlings estimate the basic or the effective dispersal kernel. In our simulations, DDM resulted in a shift of the mean basic dispersal distance (10m) towards substantially higher effective mean dispersal distances (15m and 208m). 3. We demonstrated that CSM and SEMM estimate the basic seed dispersal kernel, independently from the presence of post-dispersal mortality. By contrast, GSM estimates the effective dispersal kernel. IM failed to provide satisfactory estimates in the presence of DDM in our sampling design. Besides, for all methods, seed migration was inflated in the presence of DDM, due to lower mortality among randomly distributed immigrants relatively to local seedlings. 4. It could seem intuitive that estimates based on seedlings or seeds provide effective or basic dispersal kernels respectively. Our results showed that it is not true for estimates obtained with CSM or SEMM because they rely on the composition of seed/seedling pools and not seed/seedling counts such as IM or GSM. This has important consequences for life stage studies where the discordance of dispersal kernels estimated from different cohorts is used to investigate post-dispersal density-dependent mortality.