Data from: Inferring riverscape dispersal processes from fish biodiversity patterns
Data files
Mar 17, 2025 version files 141.55 KB
-
Data_JAE2024.RData
138.02 KB
-
README.md
3.54 KB
Abstract
Dispersal patterns are recognized as a main determinant of biodiversity structure, particularly in rivers. The dendritic organization of rivers, waterflow direction, large distance immigrants from the outlet, and fragmentation by dams combine to produce a complex dispersal scenario. Unraveling the role, magnitude and spatial scale at which these different dispersal sources determine metacommunity diversity is challenging and demands a large amount of spatiotemporal information, which is rarely available. Here, we incorporated alternative dispersal hypotheses in coalescent and lottery models, contrasting their predictions with observed fish diversity in the Negro River basin of Uruguay, South America. Evidence from different models but supporting the same hypothesis was combined, finding support for i) a dispersal constrained by the riverscape—with no role for direct immigration from the metacommunity; ii) an asymmetric neighbor dispersal, sharply decaying in upward but not in downward river directions; iii) an outlet as a source of individuals that affects diversity even at distant communities; and finally iv) a strong effect of dams, indicating that fish diversity at the riverscape level is now affected by dam constraints on individual movement. Both observed alpha (r=0.58) and beta (r=0.54) diversity were well predicted by the selected model, with a positive and a U-shaped association with communities centrality. Process stochasticity—variation in diversity metrics along realizations of the same assembly process— in alpha and beta diversity were negatively and hump-shaped associated with communities centrality. The ongoing fragmentation of rivers worldwide demands a mechanistic understanding of the role of large-scale dispersal on ecosystem diversity and stability. Selecting models and combining evidence from different models allows the present research to advance on the relative support for hypotheses with a comparatively limited amount of empirical information, thus highlighting the potential of theoretical-empirical approaches for identifying the mechanisms that shape biodiversity.
Access this dataset on Dryad: DOI:10.5061/dryad.kh18932jm
Description of the data and file structure
These files contain R-codes to reproduce metacommunity assembly combining a coalescent-based approach and a lottery dynamic run on a spatially explicit landscape to estimate alfa and beta diversity, considering different hypothesis of dispersal that could take place in a riverscape. That is, upstream dispersal, downstream dispersal, and dispersal from the outlet. It also considers the size of the communities, a regional pool, and the potential effect of fragmentation by dams.
1. hill.kernel.to.all.J.2nd.model.migration.corrected.R
This file contains the code to estimate metacommunity diversity based on a coalescent assembly and a lottery simulation. Note that the first function called “fit.Coalescent.and.lottery.hill.all.J” corresponds to a parallelisation to optimise the simulation. The second one called “Coalescent.and.lottery.hill.all.J” is the “master” function that makes up the metacommunity assembly, considering an explicit graph (G) and directed dispersal kernels. In this case, since the graph is a river, three kernels of dispersal are considered: upstream dispersal, downstream dispersal and dispersal from the outlet, and are represented in the code by the parameters d50.range.in, d50.range.out, d50.range.outlet (d50 refers to the dispersal ability). In addition, the size of the communities (Ji), a regional pool and the potential effect of fragmentation by dams are considered. Briefly, communities are filled with Ji individuals according to the relative strength of the different dispersal sources. When all communities are filled, the lottery dynamic begins with a deletion and replacement of individuals in each community, through it iterations. A final metacommunity matrix N with elements representing the abundance of each species in each community is obtained; from this matrix the alfa and beta diversity is estimated. Then, a Generalized Linear Model (GLM) is fitted to the observed and expected species richness from the different simulated scenario (dispersal sources). The observed diversity is taken from https://doi.org/10.6084/m9.figshare.12476393. The model with the largest weight of evidence is identified and the difference in AIC between the best models was assessed.
2. iterations_best_models.R
This code allows to account for the stochastic nature of lottery model results. Thus, once the best model was identified, a number of replicates (“replicas”) were performed with the same parameters, retaining the median richness and beta diversity expected in each community.
3. Data_JAE2024.RData
This file contains the R space with the graph used that represents the riverscape, and the vector with the empirical records of fish diversity used in this study, representing in the argument “Obs”.
Access information
Other publicly accessible locations of the data:
- Not apply
Data was derived from the following sources:
- All data is generated from the codes presented above. Only, the empirical records of fish diversity used in this study (called “Obs”) as well as the position of communities in the riverscape are taken from https://doi.org/10.6084/m9.figshare.12476393. This information is also in the R space load herein (Data_JAE2024.RData).