Drivers of metacommunity dynamics in river-floodplain fish: A path modeling approach

Feng, Kai 1 ; Czeglédi, István1; Funk, Andrea2; Hein, Thomas2; Pont, Didier2; Meulenbroek, Paul2; Valentini, Alice3; Erős, Tibor1

Published Aug 14, 2025 on Dryad. https://doi.org/10.5061/dryad.wstqjq301

Data files

Aug 14, 2025 version files 68.04 MB

All_Variables_Standardized_Matrix_Distance.RData

11.21 MB
Beta_Components_SEM.R

29.88 KB
Convert_Pairwise_to_Matrix.R

13.14 KB
Function_Vector_to_Matrix.R

2.82 KB
Pairwise_Site_Data_Raw.csv

8.02 MB
Permutation_Based_PathAnalysis.R

1.66 KB
README.md

12.71 KB
SEM_Results.RData

48.75 MB
Site_Time.csv

3.80 KB

Abstract

Metacommunity theory offers a compelling framework for understanding the processes that govern biodiversity patterns across space and time. Yet, a persistent challenge remains: integrating the wide array of ecological drivers into a unified model using observational data from complex, dynamic ecosystems. In this study, we present a novel, process-explicit path modelling approach that bridges recent theoretical advances in metacommunity ecology with empirical data. Focusing on fish communities in the floodplains of the Danube River, we leverage environmental DNA (eDNA) metabarcoding to characterize community composition across a spatio-temporal network of sites. We partition beta diversity into its species replacement and richness difference components and apply structural equation modelling to evaluate the relative influence of multiple ecological drivers—including spatial and temporal dispersal, demographic stochasticity, abiotic filtering and interspecific interactions. Our results reveal that river-floodplain fish metacommunities are shaped by a complex web of interacting processes. Notably, we find that species replacement is primarily driven by spatial distance and environmental filtering, while richness differences are more influenced by biotic interactions and community size. Lateral hydrological connectivity emerged as a pivotal landscape feature, governing beta diversity both directly and indirectly through its modulation of local environmental conditions. This connectivity acted as a structural conduit, mediating dispersal, environmental heterogeneity, and biotic interactions. By disentangling the contributions of multiple processes, our model underscores the dominant role of spatial structuring and abiotic filtering over temporal dynamics and biotic interactions in shaping metacommunity assembly. The model also demonstrates improved explanatory power and stronger model fit, outperforming previous studies. These findings underscore the need for integrative frameworks that consider the simultaneous influence of multiple ecological processes, particularly in highly dynamic systems like river-floodplains. Our conceptual and modelling approach advances metacommunity theory by offering a robust, data-driven means to assess complex assembly mechanisms and by emphasizing the critical role of connectivity and habitat complementarity in sustaining biodiversity within dynamic landscapes.

Dataset DOI: 10.5061/dryad.wstqjq301

Description of the data and file structure

Project Overview

This project processes and transforms pairwise site data from lake ecosystems for ecological analysis, focusing on environmental filtering, demographic stochasticity, patch dynamics, interspecific interactions, and beta-diversity components.
The workflow covers data preparation, transformation, standardization, and conversion to symmetric matrices, producing an RData file ready for Structural Equation Modeling (SEM) and other downstream analyses.

Workflow Summary

Data preparation — Pairwise site-time dataset and reference metadata.
Data transformation — Long-to-wide conversion, skewness check, variable transformation, and standardization.
Matrix generation — Conversion of variables into symmetric distance matrices.
Statistical analysis — SEM and permutation-based path analysis to quantify ecological process contributions.
Output generation — Final SEM results stored in RData format.

Data Notes

All variables in the pairwise dataset are expressed as differences or similarity indices between two site-time combinations.
Reference metadata (e.g., site IDs, sampling dates) is provided for interpretation but not directly used in analyses.
Outputs are organized to keep data processing and analysis steps modular, enabling reproducibility and easier review.

File Descriptions

1. Pairwise_Site_Data_Raw.csv

This dataset contains pairwise observations between sampling sites and occasions, used as input for SEM analysis.

Column name	Description
`Site1`, `Site2`	Site-time codes for the two sites in the pair. Codes follow the format `AT_##_##` or `HU_##_##`, where the prefix denotes the country (AT = Austria, HU = Hungary), the first number is the site number, and the second number is the sampling occasion.
`tax_total_df`	Pairwise beta diversity index (total dissimilarity in community composition based on relative abundance).
`tax_repl_df`	Beta diversity component representing species replacement.
`tax_rich_df`	Beta diversity component representing richness differences.
`abundance_df`	Mean community size (average number of fish DNA copies per unit water volume between the two sites).
`abundance_dist_df`	Absolute difference in community size between the two sites.
`CHP1_df`	Difference in chemical properties (first principal component score from PCA).
`LUP1_df`	Difference in land use composition within 500 m buffer (first principal component score from PCA).
`PHP1_df`	Difference in physical habitat characteristics (first principal component score from PCA).
`MAP1_df`	Difference in hydrophyte coverage (first principal component score from PCA).
`RPP1_df`	Difference in riparian coverage (first principal component score from PCA).
`SES_corQ_df`	Difference in abundance‐weighted species association strength, derived from residual correlation patterns in joint species distribution models (HMSC).
`dry_df`	Difference in proportion of drying period (% of days per year dry).
`LHC_df`	Difference in lateral hydrological connectivity (% of days per year connected to main channel).
`regime_df`	Difference in water level.
`volume_df`	Difference in estimated waterbody volume (km² × m).
`distance_df`	Streamline distance between sites (km).
`time_df`	Absolute time distance (days since study start).
`month_df`	Cyclic time distance (day of year difference).

Note: All variables are expressed as pairwise differences or similarity indices between two site-time combinations, enabling the analysis of spatial, temporal, and environmental factors shaping community composition.

2. Site_Time.csv

A reference table describing the mapping between site/time codes used in Pairwise_Site_Data_Raw.csv and their actual locations and sampling dates.

Column 1 – Site: Site IDs, ranging from AUT1 to AUT15 (Austria) and HUN1 to HUN15 (Hungary).
Column 2 – Country: Country names corresponding to site codes (Austria = AUT, Hungary = HUN).
Columns 3–13 – Time1 to Time11: Sampling dates (YYYY-MM-DD) for up to eleven sampling occasions per site. Missing values are indicated by NA.

Relationship to Pairwise_Site_Data_Raw.csv
In Pairwise_Site_Data_Raw.csv, site-time codes follow the pattern AT_01_01 or HU_07_08:

Prefix (AT or HU) is derived from the country code (AUT or HUN).
The first number is the site number (matching the numeric part of the Site column in Site_Time.csv).
The second number is the occasion number (matching the relevant Time# column in Site_Time.csv).

This file is not used directly in the analysis but serves as metadata to interpret the site and time identifiers in the pairwise dataset.

3. Function_Vector_to_Matrix.R

Contains a wrapper function convert_to_symmetric_matrix that converts a long-format pairwise dataframe into a symmetric matrix.

Functionality includes:

Transforming long-format data into wide-format
Optionally adding a new column at the beginning and a new row at the end
Symmetrizing the matrix and setting the diagonal to 0

This function is called within Convert_Pairwise_to_Matrix.R to achieve one-step matrix conversion.

4. Convert_Pairwise_to_Matrix.R

An R script that processes Pairwise_Site_Data_Raw.csv and performs the following tasks:

Computes skewness for each variable column and chooses the appropriate transformation/standardization
Standardizes each variable
Calls convert_to_symmetric_matrix to convert each variable into a symmetric matrix
Generates distance matrices for all ecological and environmental variables

Running this script produces All_Variables_Standardized_Matrix_Distance.RData for downstream analyses.

5. All_Variables_Standardized_Matrix_Distance.RData

Output from the Convert_Pairwise_to_Matrix.R script, containing standardized symmetric matrices, including:

Environmental filtering variables (e.g., CHP1, PHP1, MAP1, etc.)
Demographic stochasticity variables (e.g., abundance)
Patch dynamics and beta-diversity metrics
Interspecific interaction indices (e.g., SES correlation matrices)

The RData file is provided separately to keep data processing and data analysis stages modular.

6. Beta_Components_SEM.R

An R script that performs SEM to quantify the contributions of different ecological processes to beta-diversity and its components (total, replacement, richness difference).

Main steps:

Prepare the SEM dataframe from All_Variables_Standardized_Matrix_Distance.RData
Build matrices for permutation-based SEM
Define SEM models incorporating environmental filtering, demographic stochasticity, patch dynamics, spatial/temporal dispersal, and interspecific interactions
Run SEM and permutation-based path analysis
Conduct variance partitioning by sequentially removing predictor groups (patch, environment, spatial, interaction, community size, temporal) and calculating changes in R²
Save final results to SEM_Results.RData

7. Permutation_Based_PathAnalysis.R

Contains a function implementing permutation-based path analysis for matrix SEM, adapted from:

Fourtune et al. (2018) The American Naturalist, 191(4), 491–508
Jabot et al. (2020) Ecology Letters, 23(9), 1330–1339

This function is called within Beta_Components_SEM.R to test the significance of SEM path coefficients in a distance-matrix framework.

8. SEM_Results.RData

Final output from Beta_Components_SEM.R, containing SEM results and permutation test outputs.
Provided for direct inspection without re-running computationally intensive permutation analyses.

Reproducibility

The following sequence reproduces all analyses from raw pairwise data to final SEM results:

Prepare data files
- Ensure Pairwise_Site_Data_Raw.csv and Site_Time.csv are available in the working directory.
Convert pairwise data to symmetric matrices

   source("Function_Vector_to_Matrix.R")
   source("Convert_Pairwise_to_Matrix.R")

This produces All_Variables_Standardized_Matrix_Distance.RData.
3. Run Structural Equation Modeling

source("Beta_Components_SEM.R")

This uses permutation-based path analysis (via Permutation_Based_PathAnalysis.R) and generates SEM_Results.RData.
4. Inspect results
Load SEM_Results.RData directly to explore SEM outputs without re-running the computational steps.

Note: This workflow is fully contained within the provided scripts and datasets, requiring no external raw field data access.