Networks in aquatic communities collapse upon neonicotinoid-induced stress
Data files
Apr 02, 2025 version files 272.95 KB
-
1.RawWaterchemistry__ecosystem_parameters.xlsx
64.78 KB
-
2.Rawbiodiversityassay.xlsx
46.37 KB
-
3.Random_Interactions_Option1_ZIPoverdisp_TotalAbundancePerTreatment_cor000-100_run6.csv
19.37 KB
-
4.Random_Interactions_Option1_ZIPoverdisp_TotalAbundancePerTreatment_cor000-100_run6_Null.csv
7.53 KB
-
5.Biodiversity_biomass_feeding_guild_pSEM.xlsx
79.10 KB
-
6.RawData.csv
8.09 KB
-
7.Taxa_ClassFamiliesDisplayName.csv
5.88 KB
-
8.Taxa_OrderingChordDiagram.csv
3 KB
-
Co-occurrence_Figure3_and_S5.R
12.57 KB
-
Networks_Figure4_and_S6.R
20.94 KB
-
README.md
5.32 KB
Abstract
Freshwater ecosystems worldwide are under pressure from neonicotinoid insecticides. While recognized that communities of species are responsible for ecosystem functioning, it remains unknown if neonicotinoid-induced community transformations negatively affect ecosystem functioning. Therefore, we employed an experimental approach with 36 naturally established freshwater ecosystems exposed to increasing field realistic concentrations of the neonicotinoid thiacloprid. Upon exposure, we found severe degradation of ecosystem functioning in the form of loss of organic matter consumption and dramatic shifts in primary productivity. This functional decline coincides with strongly eroded species co-occurrence networks to the point that these are indistinguishable from randomized assemblages of species. Together, these findings show how current environmental concentrations of a neonicotinoid can strongly disrupt freshwater ecosystem functioning via degradation of the invertebrate food web. Since this dramatic ecosystem degradation occurs below nearly all identified ecotoxicological risks, we call here for the reconsideration of the use of these insecticides.
https://doi.org/10.5061/dryad.b2rbnzsnk
The full datasets for water chemistry, biodiversity assessment, ecosystem functioning and accompanying codes for co-occurence network development are provided.
Description of the data and file structure
Please see the Readme per excel file. The .csv files:
- 6.Rawdata.csv
- 7.Taxa_ClassFamiliesDisplayName.csv
- 8.Taxa_OrderingChordDiagram.csv
- 3.Random_Interactions_Option1_ZIPoverdisp_TotalAbundancePerTreatment_cor000-100_run6.csv
- 4.Random_Interactions_Option1_ZIPoverdisp_TotalAbundancePerTreatment_cor000-100_run6_Null.csv
are used to perform the modeling behind and making of Figure 3, 4, S5 and S6. Please note that the files 6.Rawdata.csv, 7.Taxa_ClassFamiliesDisplayName.csv, and 8.Taxa_OrderingChordDiagram.csv are simply reorded datafiles to assist with coding of the file ‘Co-occurrence_Figure3_and_S5.R’. The actual raw data of these files are given in a more logical format ‘2.Rawbiodiversityassay.xlsx’.
File
- 6.Rawdata.csv
A reordered file containing all abundance data per taxon (Column F to CK). Column A to C give the dates when the sample was taken. Column D the respective test ditch. Column E the nominal test concentration of Thiacloprid in μg/L.
File
- 7.Taxa_ClassFamiliesDisplayName.csv
This file contains all taxonomic information (columns A-F). Cells where no data is filled in means that this taxonomic level was not identified. For ease of use, leave these cells empty. Column G ‘Name’ is an unique identifier based on the lowest level of taxonomic identification available.
File
- 8.Taxa_OrderingChordDiagram.csv
This file helps to keep a consistent order in taxon-taxon correlation within co-occurrence figures. Column A and B indicate the two correlating taxa and their respective classes are given in column E and F. Column C is simply to indicate correlations between the taxa and column D is used for coloring of the correlation, please see the code in Co-occurrence_Figure3_and_S5.
File
- 3.Random_Interactions_Option1_ZIPoverdisp_TotalAbundancePerTreatment_cor000-100_run6.csv
is an optional step to test networks versus null models produced per test concentration of Thiacloprid. Readme:
The code for these null models is given in the R file ‘Networks_Figure4_and_S6.R’. The null models run from Spearman’s correlation strength 0 - 100. In case NA is given, this means that no correlations were produced at this correlation strength. If ‘Inf’ is given, this means infinite, these are removed later on in figure production with the mentioned R script.
column A: step of the model (with steps of 0.01 | rho | ). Leave A1 empty for coding. |
column B: ‘conc’ = Nominal concentration Thiacloprid in water in (µg/L)
column C: ‘rho’ = Absolute Spearman’s correlation strength
column D: ‘randcors_null_median’ = median number of correlations of 1000 generated random communities
column E to F contain confidence intervals around the 1000 generated random communities to plot the associated uncertainties around the null model.
File
- 4.Random_Interactions_Option1_ZIPoverdisp_TotalAbundancePerTreatment_cor000-100_run6_Null.csv
is the null model as used in the paper. But note that new null models can also be produced following file ‘Networks_Figure4_and_S6.R’. Readme:
This datafile contains the used randomized null model in the paper. The code for this null model is given in the R file ‘Networks_Figure4_and_S6.R’. The null model runs from Spearman’s correlation strength 0 - 100. In case NA is given, this means that no correlations were produced at this correlation strength. Do not remove NA as this information is needed for figure production. If ‘Inf’ is given, this means infinite, these are removed later on in figure production with the mentioned R script. This null model data is used to reproduce figures as given. However, one can opt to create a new null model with the codes provided in the R file mentioned in row 5. This null model is based on the control of the experiment (see paper section ‘Statistics on community network and structure’).
column A: step of the model (with steps of 0.01 | rho | ). Leave A1 empty for coding. |
column B: ‘conc’ = Nominal concentration Thiacloprid in water in (µg/L)
column C: ‘rho’ = Absolute Spearman’s correlation strength
column D: ‘randcors_null_median’ = median number of correlations of 1000 generated random communities
column E to F contain confidence intervals around the 1000 generated random communities to plot the associated uncertainties around the null model.
Code/Software
All codes are provided in R scripts format. These codes can be used to reproduce co-occurrence and network results and figures.
The codes provided are the R files:
- Co-occurrence_Figure3_and_S5.
Which is used to reproduce co-occurrence matrices and associated statistics and Figures 3 and S5.
- Networks_Figure4_and_S6.R
Which is used to built null models and test networks per test concentration of Thiacloprid statistically against the produced null models. And to construct figures 4 and S6.
Accompanying packages are:
tidyr
dplyr
Hmisc
pscl
VGAM
fitdistrplus
circlize