Data from: Long-term monitoring reveals how pondscape connectivity shapes the early spread of a biological invasion
Data files
Feb 17, 2026 version files 221.41 KB
-
data_ms.csv
217.02 KB
-
README.md
4.39 KB
Abstract
Biological invasions lead to loss of biodiversity and ecosystem degradation, and a better understanding of their drivers is urgently needed. This is particularly true in the initial stages of an invasion, when the transition from local establishments to regional advancing fronts occurs. The abundances of the local population, demographic rates, and spatial conditions for dispersal have been proposed to interact in determining invasive species expansion, but little empirical evidence has been accumulated. This study examines the population dynamics of the American bullfrog (Aquarana catesbeiana), a globally invasive aquatic anuran, during the initial stages of its invasion process in a pond landscape within the Pampas grasslands of Uruguay. We evaluated how the spatial spread of the invasion was influenced by connectivity between invaded and uninvaded ponds, bullfrog abundance in invaded ponds, elevation gradients, and pond area. This analysis was based on an 11-year monitoring program (2012-2022) that captured the onset and initial stage of spatial expansion. Throughout the study period, the number of invaded ponds increased at a rate of 7.7 % per year. A model of invasion probability revealed that connectivity to invaded ponds and the population status of those ponds were key determinants of spread. Connectivity to previously invaded ponds interacted with bullfrog abundance to determine invasion probability. Ponds connected to invaded locations with intermediate bullfrog abundances showed the highest invasion likelihood, with colonization odds more than twice those associated with connections to ponds with low bullfrog abundances, whereas connections to ponds with high abundances had more moderate effects. Our results highlighted how ponds with intermediate bullfrog abundance play a crucial role in facilitating the invasion spread, this is most likely due to higher propagule release. Ponds with high bullfrog abundance were likely constrained by density-dependent effects, reducing the survival and dispersal of metamorphs. Our study highlights that prioritizing bullfrog eradication in systems with intermediate abundances may be more effective to prevent the expansion of the invasion. We note that landscape features and population demography within invaded areas could be more interrelated than commonly assumed and should be jointly considered in invasive species management strategies.
Dataset DOI: 10.5061/dryad.5dv41nskw
Description of the data and file structure
This dataset was collected to analyze the spatial and temporal dynamics of the American bullfrog (Aquarana catesbeiana) invasion in a pondscape in Aceguá, Uruguay, between 2012 and 2022. The dataset includes a unique identifier for each pond (pond_id), its invasion status (invaded; 0 = non-invaded, 1 = invaded), and a categorical bullfrog abundance level (abundance_level), defined as 0 for non-invaded ponds and as low (1), intermediate (2), or high (3) for invaded ponds. Abundance levels were derived from annual visual and auditory surveys, classifying invaded ponds based on the number of individuals observed and/or heard during each survey. For each year and non-invaded pond, the dataset includes weighted in-degree centrality values derived from spatial networks constructed separately for each abundance level (degree_ab_level_1, degree_ab_level_2, and degree_ab_level_3), which quantify potential connectivity with invaded ponds as a function of spatial proximity. Additionally, the dataset includes environmental and spatial attributes for each pond, including pond area (m2), elevation (m a.s.l.), and an elevation gradient calculated using the five nearest neighboring ponds.
Files and variables
File: data_ms.csv
Description:
Variables
- pond_id: Unique identifier for each pond.
- invaded: Binary variable indicating invasion status of the pond (0 = non-invaded, 1 = invaded).
- abundance_level: Categorical variable describing bullfrog abundance at invaded ponds. Values are defined as 0 = non-invaded pond, 1 = low abundance, 2 = intermediate abundance, and 3 = high abundance. Abundance levels were derived from annual visual and auditory surveys.
- year: Year of observation (2012–2022).
- degree_ab_level_1: Weighted in-degree centrality of each non-invaded pond derived from the network including invaded ponds with low bullfrog abundance. This metric quantifies potential connectivity based on spatial proximity.
- degree_ab_level_2: Weighted in-degree centrality of each non-invaded pond derived from the network including invaded ponds with intermediate bullfrog abundance.
- degree_ab_level_3: Weighted in-degree centrality of each non-invaded pond derived from the network including invaded ponds with high bullfrog abundance.
- area: Pond area, expressed in square meters (m2).
- elevation: Elevation of the pond centroid, expressed in meters above sea level (m a.s.l.).
- elevation_gradienent: Elevation gradient calculated as the mean elevation difference between the focal pond and its five nearest neighboring ponds. Positive values indicate a higher relative position and negative values indicate a lower relative position.
Missing values
- Missing values are indicated as NA.
Code/software
The data can be viewed and analyzed using the free and open-source software R (R Core Team, 2024), and can also be opened with standard spreadsheet software.
Spatial network construction and analysis were conducted in R using the igraph package. Annual networks were built separately for each bullfrog abundance level (low, intermediate, and high), including all non-invaded ponds and only invaded ponds of the focal abundance level. Networks were directed and bipartite, linking invaded to non-invaded ponds, with links weighted by the inverse of Euclidean distance. Weighted in-degree centrality was calculated for each non-invaded pond and abundance level and compiled into the final dataset.
Statistical analyses were performed in R using generalized linear models (GLMs). Count and binomial models were fitted using appropriate error distributions (including negative binomial and Conway–Maxwell–Poisson), primarily implemented with the glmmTMB package. Model diagnostics were conducted using DHARMa, and model selection was based on Akaike’s Information Criterion (AIC). No proprietary software is required to view or analyze the data.
Access information
Not applicable. The data are original and have not been derived from other publicly accessible sources. No alternative repositories or external data sources are associated with this dataset.