Identifying knowledge gaps about context dependency in freshwater species invasions
Data files
Feb 11, 2026 version files 2.30 MB
-
Final_Meta_Analysis_Code.html
1.48 MB
-
Final_Meta_Analysis_Code.Rmd
65.16 KB
-
Master_Data.csv
259.03 KB
-
Paper_Rejection_Reasons.csv
474.27 KB
-
README.md
6.93 KB
-
signswitching.csv
14.22 KB
Abstract
The abiotic template of freshwater ecosystems is being altered by human activity on a global scale, often promoting the establishment of new invasive taxa and increasing the likelihood of facilitation between invaders. These positive interactions represent a substantive threat to freshwater ecosystems by promoting the establishment and persistence of invasive taxa. To assess the threat posed by freshwater invasive species under global change, we leveraged the existing literature at the intersection of freshwater invasive species and positive biotic interactions using systematic review and meta-analysis. Our findings revealed critical knowledge gaps in the study of interactions between invasive taxa, which may lead to an underestimation of their threat. These gaps included a lack of studies that address indirect interactions, particularly over long temporal scales. Such interactions were understudied for invasive taxa but were often strongly positive. Additionally, almost a third of the publications examined documented context-dependent interactions, but only a single publication did so between invasive organisms. Common forms of anthropogenic change, including alterations to habitat quality, nutrient availability, and hydrology, were among the most frequent drivers of context dependency. Our analysis supports the conclusion that 1) the threat of facilitation between freshwater invaders may be underestimated by the current literature, and 2) global change may drive context-dependent interactions between freshwater invasive species, but there is a lack of research addressing this possibility.
https://doi.org/10.5061/dryad.2z34tmpwh
Description of the data and file structure
The data associated with this project were collected in order to better understand factors which contextualize the outcome of interactions between freshwater species. This includes interactions between both invasive and non-invasive taxa. The intent of this project was to leverage the growing literature on positive biotic interactions in freshwater to highlight knowledge gaps in the study of freshwater invasions.
The file structure consists of three .CSV files, one .RMD file, and one .HTML file. The raw data for the analysis are contained in the .CSV files. The .RMD file contains the code needed to reproduce the analysis. The .HTML file is a knitted rendering of the analysis produced by the .RMD file.
Files and variables
File: Master_Data.csv
Description: This file houses the raw data on the outcome of interactions between freshwater species. The variables used for the analysis and creation of figures include:
- Author. The last name of the lead author on a publication.
- PubYear. The publication year.
- Journal. The journal in which the publication appeared.
- Experiment. A numerical identifier for each unique experiment or study within a publication.
- Manipulation. A numerical identifier for distinct manipulations or treatments within a given experiment.
- Metric. A numerical identifier for distinct metrics used to measure the outcome of a given manipulation.
- ComparitiveManipulative. Whether the observation was collected from a comparative or manipulative study or experiment.
- DirectIndirect. Whether the interaction in question followed a direct pathway, an indirect pathway, or was the net effect of both.
- MasterResponse. Broad, reader assigned category for the type of response measured.
- ResponseVar. More detailed information on the response measured.
- NumFacilitator and NumBeneficiary. The number of facilitator and beneficiary taxa involved in the interaction. Either an exact number or "multiple" if the exact number is greater than one and ambiguous.
- FacilitatorTaxon and BeneficiaryTaxon. The taxonomic identity of the facilitator and beneficiary in an interaciton.
- FacilitatorType and BeneficiaryType. Abbreviations for broad taxonomic categories of the facilitator and beneficiary. Abbreviations are corrected to full words in the .Rmd file.
- FacilitatorInvasionStat and BeneficiaryInvasionStat. Whether the author of the original work considered the facilitator and beneficiary to be invasive. Left as NA if ambiguous.
- ResponseUnits. Not used in the analysis, but provides the original units for group means.
- Days. How long a manipulation or difference between natural control and experimental groups had been occurring when the measurement was taken.
- PatchSize_m2. The spatial extent over which the measurement occurred in square meters.
- nT and nC. The sample sizes for each treatment and control group.
- T and C. Group means for each sample and control group.
- SE_T and SD_C. Standard errors for each treatment and control group, where reported.
- Var_T and Var_C. Variance for each treatment and control group, where reported.
- 95CI_T and 95CI_C. The 95% confidence interval for each treatment and control group, where reported.
- SD_T and SD_C. Standard deviations for each treatment and control group, where reported.
File: signswitching.csv
Description: This file houses information on potential instances of context dependency documented by the publications included in the meta-analysis:
Variables used for joining this data to the master data include:
- Author, PubYear, and Journal. As described above, these variables are used only for joining the data.
Variables used for the analysis include:
- Sign Switching. Whether or not the publication documents a possible instance of context dependency.
- Comparative_Manipulative. Whether the context dependency was observed in a comparative or manipulative setting.
- Direct_Indirect. Whether the context dependent interaction followed a direct pathway, an indirect pathway, or involved the net result of both.
- Facilitator and Beneficiary. The taxonomic identity of both the facilitator and beneficiary.
- Facilitator_Invasive and Beneficiary_Invasive. Whether the facilitator and beneficiary were invasive according to the authors of the original publication.
- Switching_Mechanisim. The underlying mechanism which the original authors proposed as the driver of context dependency.
- Switching_Category. A broad category for the switching mechanisim.
- Switch_Type. The signs taken on by the interaction in question.
File: Paper_Rejection_Reasons.csv
Description: Publications which were selected for additional screening after the literature search, with notes on reasons for whether the publication was accepted or rejected for the final dataset. Variables include:
- Authors. The complete author list for the publication.
- Title. The title of the publication.
- DOI. The DOI of the publication.
- Use_Y_N. Whether or not the publication was passed on to the full research team for data extraction.
- Reason_R_Summary. A summary of the reason for rejecting the publication.
File: Final_Meta_Analysis_Code.Rmd
Description: The code needed to reproduce the analysis described in the publication. The code is designed to reproduce the full analysis if all chunks are run in order. This file should be in the same directory as the .CSV files detailed above in order to run the analysis correctly. The file will save multiple .CSV tables and large image files to the working directory when run.
File: Final_Meta_Analysis_Code.html
Description: A knitted version of the .Rmd described above. This file contains all relevant graphs, results, and comments on the analysis.
Code/software
The analysis was run using R version 4.4.1. Meta-regression modeling was conducted using the* metafor* package version 4.6-0. Data wrangling was conducted using* tidyverse* 2.0.0, readxl version 1.4.3, openxlsx version 4.2.6.1, and car version 3.1-2. Data visualization was conducted using* ggpubr* version 0.6.0,* ggpp* version 0.5.8-1,* ggdist* version 3.3.2,* GGally* version 2.2.1,* *and lemon version 0.4.9.
To reproduce the workflow used to run the analysis, extract the included .zip file to your chosen dirrectory. Then, open the file "Meta_Analysis_for_Distribution.Rmd" in R Studio (version 2023.12.1), then choose knit to PDF.
Access information
Other publicly accessible locations of the data:
- NA
Data was derived from the following sources:
- Various publications described in the full data.
We based our analysis on a systematic search of the available literature on positive species interactions in freshwater ecosystems. The core of our corpus was formed by publications that had previously been identified by Albertson et al. (2021). These publications were augmented by an updated search conducted in ISI Web of Science on April 1, 2022, using the following search terms:
(“aquatic” OR “stream$” OR “river$” OR “freshwater$” OR “wetland$” OR “lake$” OR “pond$” OR “vernal”) AND ((“positive interact*” OR “mutualis*” OR “commensal*” OR “facilitation” OR “symbios*” OR “habitat ameliorat*” OR “stress ameliorat*” OR “ecosystem engineer*” OR “stress gradient*”) OR ((“positive interact*” OR “mutualis*” OR “commensal*” OR “facilitation” OR “symbios*” OR “habitat ameliorat*” OR “stress ameliorat*” OR “ecosystem engineer*” OR “stress gradient*”) AND (“invas*” OR “non-native” OR “invad*” OR “non-indigenous”))) NOT “blood"
The objectives of the updated search were to expand the previous corpus to include any new articles that had not been published at the time of the Albertson et al. (2021) search, and to conduct a more thorough search for publications that examined interactions between invasive species.
Publication Vetting and Data Collection
Following the literature search, we examined each publication to determine if it met our criteria for inclusion in the meta-analysis. Publications were included if they met the following criteria:
1. A measure of effect size for a beneficiary species from both a treatment group (with potential facilitator species present) and a control group (no potential facilitator species present), reported as the mean response of the beneficiary in both groups. These means were derived from observations. Effect sizes derived from model estimates (e.g., marginal means obtained from regression parameters) or estimated means arising from extrapolation were not included. We also excluded observations for which one or both group means were zero, since such observations interfered with our calculation of effect size (see Eqs. 1 and 2 below). Although we use the terminology “beneficiary” and “facilitator” here based on the original authors’ treatment assignments, a goal of our paper is to describe how the potential facilitative interaction ultimately took place (see number 5 below).
2. A sample size (n) for both groups.
3. A measure of uncertainty for both the treatment and control group means. This could be reported as standard deviation, standard error, sample variance, 95% confidence interval of the mean, etc., but had to be able to be transformed to yield the within-group standard deviation of the mean.
4. At least one of the participants in the interaction had to be an obligate freshwater species.
5. At least one interaction documented in the publication had to be positive. We defined positive interactions as those for which the mean value of the response variable in the treatment group exceeded the mean value in the control group, regardless of uncertainty around the mean. We therefore included publications for which the difference between the means did not meet author assigned thresholds for statistical significance.
If a publication met the criteria for inclusion, we collected the group means, uncertainty, and sample size for any species interactions that reported the relevant values. We collected values when reported in text or tables directly, or in cases for which values of group mean, uncertainty, or moderators were reported only in figures, we obtained values using the Figure Calibration plugin for ImageJ (Schindelin et al., 2012). When available, we collected information on several moderating variables. These moderators included the temporal scale over which the measurement was carried out, the spatial scale of the measurement, whether the measurement was collected from a manipulative or observational study, the interaction pathway (i.e., direct interaction, indirect interaction, or the net result of both types), and the invasion status of the participants.