Data for: Critical habitat thresholds for effective pollinator conservation in agricultural landscapes
Data files
Sep 29, 2025 version files 358.18 KB
-
Bishopetal_DataCollated_v04.xlsx
341 KB
-
README.md
17.18 KB
Abstract
Biodiversity in human-dominated landscapes is declining, but evidence-based conservation targets to guide international policies for such landscapes are lacking. We present a framework for informing habitat conservation policies based on the enhancement of habitat quantity and quality and define thresholds of habitat quantity at which it becomes effective to also prioritize habitat quality. We applied this framework to pollinators, an important part of agroecosystem biodiversity, by synthesizing 59 studies from 19 countries. Given low habitat quality, hoverflies had the lowest threshold at 6% semi-natural habitat cover, followed by solitary bees (16%), bumble bees (18%), and butterflies (37%). These figures represent minimum habitat thresholds in agricultural landscapes, but when habitat quantity is restricted, marked increases in quality are required to reach similar outcomes.
https://doi.org/10.5061/dryad.nk98sf82b
Data supporting "Critical habitat thresholds for effective pollinator conservation in agricultural landscapes", published in Science on 25 Sept 2025, https://www.doi.org/10.1126/science.adr2146.
We have included in the file "Bishopetal_DataCollated_v04.xlsx" eleven sheets of data that support the models and figures in our paper, and an R script detailing how the models were run and how the figures can be created with this data. These data include insect, flower, and landscape information as input data for the models, and the model predictions and coefficients as input data for the figures.
Description of the data and file structure
Sheet "BB" - Data for bumble bee model
| Study_Year | Study ID and year of study, separated by an underscore. This variable is the grouping variable used as a random effect. Each line represents one landscape within the study_year. |
|---|---|
| DensBB | Standardized density of bumble bees (per 150m2 and 15min). This variable is the response variable. |
| AvgTransectArea.sc | Log transformed, centered, and scaled survey effort (area in m2). Covariate |
| AvgTransectMinutes.sc | Log transformed, centered, and scaled survey effort (time in min). Covariate |
| FlwrRich.sc | Log transformed, centered, and scaled flower richness. Fixed effect and random slope |
| FlwrCovPerc.sc | Log transformed, centered, and scaled flower cover (%). Fixed effect and random slope |
| PercSNH.sc | Centered and scaled landscape semi-natural habitat cover (%). Fixed effect |
| MFCBloom | Presence ("Yes") or absence ("No") of a blooming mass-flowering crop in the study landscape at any point during the study period. Fixed effect |
| TotalSamples | Number of sampling events per landscape. Model weight |
Sheet "SB" - Data for solitary bee model
| Study_Year | Study ID and year of study, separated by an underscore. This variable is the grouping variable used as a random effect. Each line represents one landscape within the study_year. |
|---|---|
| DensSB | Standardized density of solitary bees (per 150m2 and 15min). This variable is the response variable. |
| AvgTransectArea.sc | Log transformed, centered, and scaled survey effort (area in m2). Covariate |
| AvgTransectMinutes.sc | Log transformed, centered, and scaled survey effort (time in min). Covariate |
| FlwrRich.sc | Log transformed, centered, and scaled flower richness. Fixed effect and random slope |
| FlwrCovPerc.sc | Log transformed, centered, and scaled flower cover (%). Fixed effect and random slope |
| PercSNH.sc | Centered and scaled landscape semi-natural habitat cover (%). Fixed effect |
| MFCBloom | Presence ("Yes") or absence ("No") of a blooming mass-flowering crop in the study landscape at any point during the study period. Fixed effect |
| TotalSamples | Number of sampling events per landscape. Model weight |
Sheet "HF" - Data for hoverfly model
| Study_Year | Study ID and year of study, separated by an underscore. This variable is the grouping variable used as a random effect. Each line represents one landscape within the study_year. |
|---|---|
| DensHF | Standardized density of hoverflies (per 150m2 and 15min). This variable is the response variable. |
| AvgTransectArea.sc | Log transformed, centered, and scaled survey effort (area in m2). Covariate |
| AvgTransectMinutes.sc | Log transformed, centered, and scaled survey effort (time in min). Covariate |
| FlwrRich.sc | Log transformed, centered, and scaled flower richness. Fixed effect and random slope |
| FlwrCovPerc.sc | Log transformed, centered, and scaled flower cover (%). Fixed effect and random slope |
| PercSNH.sc | Centered and scaled landscape semi-natural habitat cover (%). Fixed effect |
| MFCBloom | Presence ("Yes") or absence ("No") of a blooming mass-flowering crop in the study landscape at any point during the study period. Fixed effect |
| TotalSamples | Number of sampling events per landscape. Model weight |
Sheet "BF" - Data for butterfly model
| Study_Year | Study ID and year of study, separated by an underscore. This variable is the grouping variable used as a random effect. Each line represents one landscape within the study_year. |
|---|---|
| DensBF | Standardized density of butterflies (per 300m2). This variable is the response variable. |
| AvgTransectArea.sc | Log transformed, centered, and scaled survey effort (area in m2). Covariate |
| FlwrRich.sc | Log transformed, centered, and scaled flower richness. Fixed effect and random slope |
| FlwrCovPerc.sc | Log transformed, centered, and scaled flower cover (%). Fixed effect and random slope |
| PercSNH.sc | Centered and scaled landscape semi-natural habitat cover (%). Fixed effect |
| MFCBloom | Presence ("Yes") or absence ("No") of a blooming mass-flowering crop in the study landscape at any point during the study period. Fixed effect |
| TotalSamples | Number of sampling events per landscape. Model weight |
Sheet "Tropics" - Data for model of bees from tropical regions
| Study_Year | Study ID and year of study, separated by an underscore. This variable is the grouping variable used as a random effect. Each line represents one landscape within the study_year. |
|---|---|
| DensBees | Standardized density of bees (per 150m2 and 15min). This variable is the response variable. |
| AvgTransectArea.sc | Log transformed, centered, and scaled survey effort (area in m2). Covariate |
| AvgTransectMinutes.sc | Log transformed, centered, and scaled survey effort (time in min). Covariate |
| FRwithin.sc | Log transformed, within-study centered, and scaled flower richness. Fixed effect |
| FCwithin.sc | Log transformed, within-study centered, and scaled flower cover (%). Fixed effect |
| PercSNH.sc | Centered and scaled landscape semi-natural habitat cover (%). Fixed effect |
| MFCBloom | Presence ("Yes") or absence ("No") of a blooming mass-flowering crop in the study landscape at any point during the study period. Fixed effect |
| TotalSamples | Number of sampling events per landscape. Model weight |
Sheet "FigS5_ModelCoefs" - model coefficients underlying fig. S5
| Model | Name of model; SB=solitary bee, HF=hover fly, BF=butterfly, TB=bees in tropical regions |
|---|---|
| Term | Name of fixed effect |
| Coef | Value of model coefficient |
| SE | Standard error of model coefficient |
| Low | Lower bound of 95% CI of model coefficient |
| High | Upper bound of 95% CI of model coefficient |
Sheet "BB_predictions" - Model predictions of local bumble bee densities extrapolated to landscape-level abundances that underly Figs. 2, 3, and S6. Can be used to validate and construct these figures.
| Var1 | Floral resource quantile, Q1-Q20 from low to high |
|---|---|
| Var2 | Landscape semi-natural habitat (%), representing 20 quantiles |
| value | Landscape-level bumble bee abundance at each quantile combination, representing back-transformed model-predicted density extrapolated to total landscape habitat area |
Sheet "SB_predictions" - Model predictions of local solitary bee densities extrapolated to landscape-level abundances that underly Figs. 2, 3, and S7. Can be used to validate and construct these figures.
| Var1 | Floral resource quantile, Q1-Q20 from low to high |
|---|---|
| Var2 | Landscape semi-natural habitat (%), representing 20 quantiles |
| value | Landscape-level solitary bee abundance at each quantile combination, representing back-transformed model-predicted density extrapolated to total landscape habitat area |
Sheet "HF_predictions" - Model predictions of local hoverfly densities extrapolated to landscape-level abundances that underly Figs. 2, 3, and S8. Can be used to validate and construct these figures.
| Var1 | Floral resource quantile, Q1-Q20 from low to high |
|---|---|
| Var2 | Landscape semi-natural habitat (%), representing 20 quantiles |
| value | Landscape-level hoverfly abundance at each quantile combination, representing back-transformed model-predicted density extrapolated to total landscape habitat area |
Sheet "BF_predictions" - Model predictions of local butterfly densities extrapolated to landscape-level abundances that underly Figs. 2, 3, and S9. Can be used to validate and construct these figures.
| Var1 | Floral resource quantile, Q1-Q20 from low to high |
|---|---|
| Var2 | Landscape semi-natural habitat (%), representing 20 quantiles |
| value | Landscape-level butterfly abundance at each quantile combination, representing back-transformed model-predicted density extrapolated to total landscape habitat area |
Sheet "Tropics_predictions" - Model predictions of local bee densities (tropical regions) extrapolated to landscape-level abundances that underly Figs. 2, 3, and S10. Can be used to validate and construct these figures.
| Var1 | Floral resource quantile, Q1-Q20 from low to high |
|---|---|
| Var2 | Landscape semi-natural habitat (%), representing 20 quantiles |
| value | Landscape-level bee abundance at each quantile combination, representing back-transformed model-predicted density extrapolated to total landscape habitat area |
Sharing/Access information
The location data associated with this dataset is available at https://doi.org/10.5281/zenodo.15497131
The open access version of the article that this dataset supports is available at https://zenodo.org/records/17210898
The Zenodo records contain further information on how to access the location data.
Code/Software
The R script "Code_Bishopetal_v02.R" was created using R 4.1.2 and contains code and annotated instructions on how the models were run and how the figures were created.