Falcons reduce pre-harvest food safety risks and crop damage from wild birds
Data files
Oct 21, 2025 version files 375.34 KB
-
Data_S1_Pathogen_testing_data.csv
140.42 KB
-
Data_S2_Transect_survey_data.csv
5.77 KB
-
Data_S3_Percent_branches_with_feces.csv
117.92 KB
-
Data_S4_Damage_data.csv
36.84 KB
-
README.md
74.39 KB
Abstract
Foodborne illness outbreaks have heightened pressures on growers to improve food safety, including mitigating possible threats from wildlife. Among wildlife, birds are particularly challenging to deter, and the risks they pose to pre-harvest food safety remain unclear. Further, deterrence efforts can jeopardize conservation and biological control, necessitating strategies that effectively lead to co-management of farmlands for conservation, pest control, and pre-harvest food safety. Promotion of birds of prey with nest boxes may be one promising strategy to promote species of conservation concern that can deter pest birds that damage crops and introduce foodborne pathogens. Here, we evaluate if the American kestrel (Falco sparverius), a small falcon, can concurrently reduce crop damage and pre-harvest food safety risks from birds in sweet cherry orchards in Michigan, USA. In orchards with and without active kestrel nest boxes, we conducted avian transect surveys, estimated the percentage of cherries with bird damage, and estimated the percentage of branches and cherries with feces. We collected fecal samples directly from birds and crop surfaces. We tested feces for Campylobacter, the most common foodborne pathogen in birds, using both culturing and PCR. Fewer birds were present in fields with nest boxes, which translated into reduced bird damage (0.47% vs. 2.50%) and fewer branches with feces (2.33% vs. 6.88%). Feces on individual cherries were rare (4/15,890 [0.025%] cherries across all sites). We detected one or more species of Campylobacter using culturing and/or PCR in 10.65% (33/310) of bird feces collected from crops and in 19.67% (24/122) of samples collected directly from birds. Detection rates were similar in fields with and without nest boxes. Despite the somewhat high overall detection, cultivable Campylobacter were only detected in 0.97% of feces collected from crops. Pre-harvest food safety and wildlife conservation are often thought to be in conflict, and produce growers have few tools to effectively manage birds. However, our findings suggest that the promotion of birds of prey using nest boxes may be one way for growers to conserve a declining species, reduce crop damage, and reduce in-field fecal contamination that could cause foodborne illness.
Dataset DOI: 10.5061/dryad.9w0vt4bv8
Description of the data and file structure
Data S1. Pathogen testing data: This file contains data pertaining to testing avian fecal samples for Campylobacter using both culturing and PCR approaches. It also provides bird identification data linked to each sample.
Data S2. Transect survey data: This file contains avian density estimates from transect surveys.
Data S3. Percent branches with feces: This file contains data on the number of avian fecal droppings on focal branches on cherry trees. It also provides data on the number of droppings with Campylobacter.
Data S4. Damage data: This file provides estimates on the percentage of cherries with vertebrate damage.
Files and variables
File: Data_S1_Pathogen_testing_data.csv
Description: This file contains data pertaining to testing avian fecal samples for Campylobacter using both culturing and PCR approaches. It also provides bird identification data linked to each sample.
Variables
| Variable | Description |
|---|---|
| ID | Fecal sample identifier used for sorting and merging datasets during sample processing |
| Sample type | Type of fecal sample. “Environmental” = collected from cherry trees. “Mist-net” = collected directly from wild birds after we captured them by mist-netting. |
| Treatment | Kestrel = site with active kestrel nest box; control = site without active kestrel nest box |
| Year | Year row corresponds to. 2021 or 2022. |
| Date | Date of survey corresponding to that row |
| Farm | Name of the “cherry orchard” corresponding to that row. Sites were delineated as separate based on being contiguously managed areas of cherry trees. No real farm names are used in this file for grower privacy. |
| Transect | Transect identifier. Transects were each 100 m in length and contained 10 focal trees. There were two transects per site. |
| Tree ID | Tree identifier. Each transect had 10 focal trees, labelled "a" through "j". For fecal log only: if fecal samples were found on non-focal trees in the transect, the trees were given alternative unique identifiers. |
| Sample | Fecal sample identifier given in the field |
| Appearance | Appearance of fecal samples collected from cherry trees. “Fresh” = apparently freshly deposited bird feces (e.g., moist, not degraded). All samples from mist-netted birds were scored “fresh”. “Fresh?” = sample was moist but could have been rehydrated from rain. “Medium” = between fresh and old. “Old/dry” = obviously old and dry. Frequently partially washed away from crop surfaces. “Old/wet” = sample appeared old but was wet, apparently from rain or dew. |
| Appearance_binary | Binary fresh/old score binned from “appearance”. “Fresh” includes “fresh”, “fresh?”, and “medium”. Old includes “old/dry” and “old/wet”. |
| Location on tree | Part of cherry tree the fecal sample in that row was collected off of. “Branch” = woody branch part of tree. “Cherry” = sweet cherry fruit. “Leaf” = leaf of cherry tree. “Leaf and branch” = part of fecal sample was on a leaf or leaves and part was on a branch. “Leaf and cherry” = part of fecal sample was on a leaf or leaves and part was on a cherry or cherries. “Stump” = feces on a pruning cut/cut stub (wooded surface where a branch or tree had been pruned). NA for mist-netted samples. |
| On transect branch | If sample was collected in one of our randomly selected 1-m focal branch sections. This was used to estimate density of feces with pathogens in the paper. Y = sample was from within the bounds of a focal branch. N = sample was not in a focal branch section. "Y but beyond 1 m" = sample was on focal branch but past the 1-m focal section. Y/N = sample was partially on 1-m focal section and partially outside of it. |
| Field collection notes | Notes pertaining to sample collection or appearance in the field |
| Campy | If sample was culturally positive for Campylobacter using regular and/or enrichment culture methods. Yes = Campylobacter detected by culturing. No = no Campylobacter detected by culturing. |
| Regular culture | If Campylobacter was detected using regular direct culture methods. If Campylobacter was detected, the species is given. Species were identified using MALDI-TOF MS Microflex LT Biotyper (Bruker Daltonics, Germany). |
| Enrichment culture | If Campylobacter was detected using enrichment culture methods. If Campylobacter was detected, the species is given. Species were identified using MALDI-TOF MS Microflex LT Biotyper (Bruker Daltonics, Germany). |
| Regular culture other | Any bacterial species identified using regular direct culture methods that was not Campylobacter. These data were only recorded for 2021. Nil = none detected. NA = did not assess (2022 data). These data were not used in the manuscript. |
| Enrichment culture other | Any bacterial species identified using enrichment culture methods that was not Campylobacter. These data were only recorded for 2021. Nil = none detected. NA = did not assess (2022 data). These data were not used in the manuscript. |
| DNA_concentration | DNA concentration obtained via Nanodrop |
| DNA260_280 | DNA 260/280 wavelength absorption ratio, obtained via Nanodrop |
| Notes_extraction | Notes on any divergence of methodology for DNA extraction |
| Gel_Picture | File name corresponding to the electrophoresis gel photo where the call was made for a positive or negative detection of Campylobacter, according to size and band expected for each multiplex PCR reaction used in this study (please see methods for further detail). The gel images are not included in this archived but can be accessed by emailing Pedro Augusto Da Pos Rodrigues (pedro.rodrigues@uga.edu). |
| 23S | Detection of bands corresponding to PCR products amplified by the general Campylobacter 23S rRNA primer set designed by Wang et al. (2002). Note that this set of primers is known to amplify bacteria other than Campylobacter (see methods for further detail). |
| CF | Detection of bands corresponding to Campylobacter fetus subsp. fetus. If not noted otherwise in “Notes_Campy”, results based on the multiplex PCR designed by Wang et al. (2002). |
| CJ | Detection of bands corresponding to Campylobacter jejuni. If not noted otherwise in “Notes_Campy”, results based on the multiplex PCR designed by Wang et al. (2002). |
| CL | Detection of bands corresponding to Campylobacter lari. If not noted otherwise in “Notes_Campy”, results based on the multiplex PCR designed by Wang et al. (2002). |
| CU | Detection of bands corresponding to Campylobacter upsaliensis. If not noted otherwise in “Notes_Campy”, results based on the multiplex PCR designed by Wang et al. (2002). |
| CC | Detection of bands corresponding to Campylobacter coli. If not noted otherwise in “Notes_Campy”, results based on the multiplex PCR designed by Wang et al. (2002). |
| 16S | Detection of bands corresponding to PCR products amplified by the general Campylobacter 16S rRNA primer set designed by Yamazaki-Matsune et al. (2007) |
| Campy_any_PCR | If any of the Campylobacter primers (23S, CF, CJ, CL, CU, and/or CC) were amplified by PCR by either the Wang et al. (2002) or Yamazaki-Matsune et al. (2007) protocol. The Yamazaki-Matsune et al. (2007) protocol was only used on samples that had amplified the 23S rRNA primers from Wang et al. (2002) but did not amplify any of the species-specific ones. Yes = at least one primer was amplified; no = no primers were amplified |
| Campy_non23S_PCR | If any of the species-level Campylobacter primers (CF, CJ, CL, CU, and/or CC) were amplified by PCR in either the Wang et al. (2002) protocol or Yamazaki-Matsune et al. (2007) protocol. The ones that were identified through the Yamazaki-Matsune et al. (2007) protocol are noted in "Notes_Campy". Yes = at least one primer was amplified; no = no primers were amplified |
| Notes_Campy | Notes pertinent to the Campylobacter PCRs, such as detection of species-specific primer bands in the multiplex PCR design by Yamazaki-Matsune et al. (2007) |
| campy_any_PCRculturing | Sample was positive for Campylobacter using cultural methods and/or PCR. This column also includes samples that were only positive for the Campylobacter 23S rRNA primer. |
| species_campy_ID_PCRculturing | Sample was positive for Campylobacter using cultural methods and/or PCR. This column only includes samples for which a specific Campylobacter species was identified. |
| BlastBirdID | Identification obtained from BLAST (Megablast, e-value < 1e-10) on sequences that were subject to a less stringent quality control |
| TreeBirdID | Identification obtained via a more stringent set of quality-filtering criteria, and placement in a phylogenetic tree (see SI methods for further details) |
| PrimerRegion | For environmental fecal samples, primer set used in the final bird identification. "KBird" = sample identified using the K_Bird_F1 and K_Bird_R1 primers from the Joo & Park (2012) protocol, which targets the avian COI gene. "ND2" = sample identified using the L5216 and H6313 from the Sorenson et al. (1999) protocol, which targets the ND2 gene. We only conducted this second set of PCRs when the first COI approach failed to yield an avian identification after Sanger Sequencing PCR products and conducting a BLAST search. "BirdF1" = sample identified using an alternative set of primers targeting the COI gene (BirdF1-COIBirdR2 and BirdF1-AvMiR by Kerr et al. (2009)). See SI Appendix, Supplementary Methods and "ID_Notes" for the sample for details. "Kbird/ND2/BirdF1" = sample was identified using all sets of primers above. This was only used on one sample that was *Campylobacter *positive via culture. See SI Appendix, Supplementary Methods and "ID_Notes" for the sample for details. |
| ID_Notes | Notes on identification methods, in particular for cases where the identification obtained via BLAST and “TreeBirdID” method were not a match. Usually denotes quality aspects of the sequence in question, such as nucleotide assignment quality scores and sequence length. |
| Alpha code | Four letter alpha code for the species that row corresponds to. Codes follow the American Ornithological Society checklist (https://www.birdpop.org/docs/misc/Alpha_codes_tax.pdf). This column has species identity for all mist-netted birds and all environmental fecal samples for which the species could be identified. For the feces from crops, codes are given for the species identified using the tree-based approach given in “TreeBirdID”. In 16 samples from crops, the tree approach identified species to the genus but not species level. For four of these genera, there was only one possible species in the identified genus at our sites during the study period, so we assigned those to the only possible species (n = 15 samples). NA if we could not assign a sample to the species-level. |
| Common name | Species common name following the 2023 American Ornithological Society’s Check-list of North American Birds. This column has species identity for all mist-netted birds and all environmental fecal samples for which the species could be identified. For the feces from crops, codes are given for the species identified using the tree-based approach given in “TreeBirdID”. In 16 samples from crops, the tree approach identified species to the genus but not species level. For four of these genera, there was only one possible species in the identified genus at our sites during the study period, so we assigned those to the only possible species (n = 15 samples). NA if we could not assign a sample to the species-level. |
| Latin name | Species Latin name following the 2023 American Ornithological Society’s Check-list of North American Birds. This column has species identity for all mist-netted birds and all environmental fecal samples for which the species could be identified. For the feces from crops, codes are given for the species identified using the tree-based approach given in “TreeBirdID”. In 16 samples from crops, the tree approach identified species to the genus but not species level. For four of these genera, there was only one possible species in the identified genus at our sites during the study period, so we assigned those to the only possible species (n = 15 samples). The genus alone for the remaining sample is given. NA if we could not assign a sample to the genus level. |
| Banding_Age | For mist-net dataset only. Age of bird row corresponds to. "After Hatching Year" = bird hatched before the calendar year of banding but year of hatch otherwise unknown. "After Second Year" = a bird known to have hatched earlier than the calendar year preceding the year of banding but hatch year otherwise unknown. "Hatching Year" = a bird capable of sustained flight and known to have hatched during the calendar year in which it was banded. "Second Year" = a bird known to have hatched in the calendar year preceding the year of banding and in its second calendar year of life. "NA" = unknown (two mist-netted birds escaped prior to aging; NAs used for environmental sample dataset). |
| Banding_Age_Binary | Banding_Age binned into "After Hatching Year" (“AHY”) and "Hatching Year" (“HY”). AHY includes after hatching year, second year, and after second year birds. "NA" if unknown (including all environmental fecal samples). |
| Banding_Sex | Sex of banded bird corresponding to this row. Female, male, or unknown. NA for all environmental fecal samples. |
| Banding_Wing_Chord | For mist-net dataset only. Length in mm from the most prominent point of the wrist joint to the most prominent point of the longest primary feather. NA for three mist-netted birds that escaped prior to wing chord measurements. NA for environmental fecal samples. |
| Banding_Weight | For mist-net dataset only. Bird weight in grams. NA for three mist-netted birds that escaped prior to weighing. NA for environmental fecal samples. |
| Landscape_nat | Percent seminatural cover in the landscape (1.0 km radius from each transect center). Seminatural cover included forest, scrubland, herbaceous, and wetland categories of the 2019 NLCD. 0 = no seminatural cover, 1 = complete seminatural cover |
| Nat_dist | Distance in meters from the tree the data row corresponds to, to the closest seminatural cover (e.g., hedge, forest) |
| Dist_box | Given for kestrel sites only. Distance in meters from the tree the data row corresponds to, to the active kestrel nest box on that site. Control sites = NA |
File: Data_S2_Transect_survey_data.csv
Description: This file contains avian density estimates from transect surveys.
Variables
| Variable | Description |
|---|---|
| Farm | Name of the “cherry orchard” corresponding to that row. Sites were delineated as separate based on being contiguously managed areas of cherry trees. No real farm names are used in this file for grower privacy. |
| Transect | Transect identifier. Transects were each 100-m in length and contained 10 focal trees. There were two transects per site. |
| Date | Date of survey corresponding to that row |
| Survey | If the survey was the 1st (1), 2nd (2), or 3rd (3) replicate for that transect |
| Abundance | Number of individual birds seen and/or heard in the avian transect survey corresponding to that row |
| Abundance_nokestrels | Number of individual birds (NOT including American kestrels) seen and/or heard in the avian transect survey corresponding to that row. This column was used in analyses in the manuscript. |
| Treatment | Kestrel = site with active kestrel nest box; control = site without active kestrel nest box |
| Landscape_nat | Percent seminatural cover in the landscape (1.0-km radius from each transect center). Seminatural cover included forest, scrubland, herbaceous, and wetland categories of the 2019 NLCD. 0 = no seminatural cover, 1 = complete seminatural cover |
| Nat_dist | Distance from the transect center that row corresponds to, to the closest seminatural cover. |
| Dist_box | Given for kestrel sites only. Distance from the transect center that row corresponds to, to the nearest active kestrel nest box. Control sites = NA |
File: Data_S3_Percent_branches_with_feces.csv
Description: This file contains data on the number of avian fecal droppings on focal branches on cherry trees. It also provides data on the number of droppings with Campylobacter.
Variables
| Variable | Description |
|---|---|
| Farm | Name of the “cherry orchard” corresponding to that row. Sites were delineated as separate based on being contiguously managed areas of cherry trees. No real farm names are used in this file for grower privacy. |
| Transect | Transect identifier. Transects were each 100-m in length and contained 10 focal trees. There were two transects per site. |
| Tree ID | Tree identifier. Each transect had 10 focal trees, labelled “a” through “j”. |
| Date | Date of survey corresponding to that row |
| Year | Year row corresponds to. 2021 or 2022. |
| Survey | Survey for that transect. 1 = 1st one for 2021, 2 = 2nd one for 2021, 3 = 1st one for 2022, 4 = 2nd one for 2022 |
| Survey.1 | Survey for that transect coded into early = 1st survey for 2021 or 2022 (i.e., survey 1/3 in the prior column), and late = 2nd survey for 2021 or 2022 (i.e., survey 2/4 in the prior column). Used in analyses on percent branches with feces. |
| Total cherries | Total number of cherries counted on the tree’s focal branch. We examined the outer 1-m of one randomly selected branch per focal tree. |
| Bird feces on cherry | Number of cherries with bird feces on them on the focal branch for that tree. NA = cherries harvested by the time of the survey, so we didn't use the cherries in this analysis. |
| N_feces_on_leaves | Number of leaves with bird feces on them on the focal branch for that tree. |
| Branch_feces | Whether the focal section of the branch had bird feces detected on it. This includes feces on leaves, stems, and/or cherries. 0 = no feces detected, 1 = 1+ feces detected. This was the main column used for the manuscript. |
| Cultural yes | Number of samples from the focal branch on the tree corresponding to that row that were positive for Campylobacter using cultural methods. |
| species_campy_ID_PCRculturing | Number of samples from the focal branch on the tree corresponding to that row that were positive for Campylobacter using cultural methods and/or PCR. This column only includes samples for which a specific Campylobacter species was identified. |
| campy_any_PCRculturing | Number of samples from the focal branch on the tree corresponding to that row that were positive for Campylobacter using cultural methods and/or PCR. This column also includes samples that were only positive for the Campylobacter 23S rRNA primer. |
| N feces tested | Number of feces tested for Campylobacter collected from the focal branch on the tree corresponding to that row |
| Feces on focal branch | Number of feces observed on the focal branch on the tree corresponding to that row. A discrepancy between “N feces tested” and “Feces on focal branch” was caused by one sample falling on the ground prior to collection |
| Treatment | Kestrel = site with active kestrel nest box; control = site without active kestrel nest box |
| Landscape_nat | Percent seminatural cover in the landscape (1.0-km radius from each transect center). Seminatural cover included forest, scrubland, herbaceous, and wetland categories of the 2019 NLCD. 0 = no seminatural cover, 1 = complete seminatural cover |
| Nat_dist | Distance in meters from the focal tree the row corresponds to, to the closest seminatural cover (e.g., hedge, forest). |
| Dist_box | Given for kestrel sites only. For damage, fecal log, and percent branches with feces files: distance in meters from the focal tree the row corresponds to, to the nearest active kestrel nest box on that site. For avian transect survey dataset: distance from the transect center that row corresponds to, to the nearest active kestrel nest box. Control sites = NA |
| Field collection notes | Notes about field-collected data in that row |
File: Data_S4_Damage_data.csv
Description: This file provides estimates on the percentage of cherries with vertebrate damage.
Variables
| Variable | Description |
|---|---|
| Farm | Name of the “cherry orchard” corresponding to that row. Sites were delineated as separate based on being contiguously managed areas of cherry trees. No real farm names are used in this file for grower privacy. |
| Transect | Transect identifier. Transects were each 100-m in length and contained 10 focal trees. There were two transects per site. |
| Tree ID | Tree identifier. Each transect had 10 focal trees, labelled “a” through “j”. For fecal log only: if fecal samples were found on non-focal trees in the transect, the trees were given alternative unique identifiers. |
| Date | Date of survey corresponding to that row |
| Year | Year row corresponds to. 2021 or 2022. |
| Survey | Survey for that transect. 1 = 1st one for 2021, 2 = 2nd one for 2021, 3 = 1st one for 2022, 4 = 2nd one for 2022. Bird damage surveys were only conducted for surveys 1 and 3 (1st survey of each year). |
| Total cherries | Total number of cherries counted on the tree’s focal branch. We examined the outer 1-m of one randomly selected branch per focal tree. |
| Bird damage | Number of cherries that had signs of vertebrate damage. We examined cherries for beak marks and looked for stems with missing cherries. This could also include rodent damage. We only collected damage data during our first survey of each year and before harvest occurred. |
| Undamaged | Total cherries minus bird damaged cherries. |
| Treatment | Kestrel = site with active kestrel nest box; control = site without active kestrel nest box |
| Landscape_nat | Percent seminatural cover in the landscape (1.0-km radius from each transect center). Seminatural cover included forest, scrubland, herbaceous, and wetland categories of the 2019 NLCD. 0 = no seminatural cover, 1 = complete seminatural cover |
| Nat_dist | Distance in meters from the focal tree the row corresponds to, to the closest seminatural cover (e.g., hedge, forest). |
| Dist_box | Given for kestrel sites only. Distance in meters from the focal tree the row corresponds to, to the nearest active kestrel nest box on that site. Control sites = NA |
| Field collection notes | Notes about field-collected data in that row |
Code/software
Files were created using Microsoft Excel.
Analyses in the manuscript were run using R version 4.3.2 (R Core Team 2023).
All models in this study were fit using the glmmTMB package in R (Brooks et al. 2017).
Assumptions for all models in this study were checked in the DHARMa package (Hartig 2021).
For visualization and interpretation on the response scale, we used estimated marginal means obtained with the emmeans package in R (Lenth 2021) for treatment comparisons, and the sjPlot package in R (Lüdecke 2023) to visualize predicted effects of continuous covariates.
REFERENCES
Brooks, M.E., Kristensen, K., van Benthem, K.J., Magnusson, A., Berg, C.W., Nielsen, A., et al. (2017). glmmTMB balances speed and flexibility among packages for zero-inflated generalized linear mixed modeling. R J., 9, 378–400.
Hartig, F. (2021). DHARMa: Residual diagnostics for hierarchical (multi-level/mixed) regression models.
Lenth, R.V. (2021). emmeans: Estimated marginal means, aka least-squares means.
Lüdecke, D. (2023). sjPlot: Data visualization for statistics in social science.
R Core Team. (2023). R: A language and environment for statistical computing. R Foundation for Statistical Computing. https:// www. R-project.org/