Combining non-invasive survey methods increases cumulative detection probability for breeding Harlequin Ducks (Histrionicus histrionicus)
Data files
Feb 26, 2026 version files 392.49 KB
-
camera_mt.csv
11.64 KB
-
camera_tl.csv
10.97 KB
-
camera_trap_models.Rmd
2.76 KB
-
Camera_trap_results.Rmd
3.02 KB
-
Camera_with_all_data.csv
14.74 KB
-
combining_methods_cummulative_detection.Rmd
4.77 KB
-
distance_sampling_lorelogram.R
6.43 KB
-
eDNA_brood_final.csv
41.71 KB
-
eDNA_clean_QeDNA.csv
125.98 KB
-
eDNA_final_data.csv
73.98 KB
-
eDNA_models.Rmd
3.50 KB
-
eDNA_results.Rmd
3 KB
-
eDNA_single_final.csv
36.18 KB
-
GBS_analysis.Rmd
2.87 KB
-
GBS_brood.csv
9.14 KB
-
GBS_final.csv
9.82 KB
-
GBS_results.Rmd
13.37 KB
-
GBS_single.csv
6.76 KB
-
metadata_camera.csv
1.56 KB
-
metadata_eDNA.csv
1.51 KB
-
metadata_GBS.csv
1.28 KB
-
metadata_lorlegram.csv
2.38 KB
-
README.md
5.13 KB
Abstract
The effective implementation of new technologies for wildlife population monitoring is limited by knowledge of factors that impact their efficacy. Population monitoring of Harlequin Ducks (Histrionicus histrionicus) on their breeding streams in Montana and Idaho in the Northern Rocky Mountains, has historically relied on ground-based foot surveys (GBS), which have produced variable count data, making it hard to decipher trends in stream occupancy and population size. We quantified the detection probability of GBS and compared it to two additional non-invasive technologies, environmental DNA (eDNA), and camera traps (time-lapse and motion detection) for detecting breeding Harlequin Ducks. We surveyed for Harlequin Duck presence on 10 occupied streams during incubation and brood rearing in 2022 and 2023 using GBS, eDNA, and camera traps. We found that a single, five km, GBS had a mean detection probability of 0.51 (95% CI: 0.31 – 0.71) and varied with the relative abundance of Harlequin Ducks, streamflow, and survey pace, but not between incubation and brood-rearing periods. Both eDNA and camera traps proved effective in detecting Harlequin Ducks, particularly when replicated across space (multiple samples of eDNA along a stream reach) or time (camera traps deployed over several weeks). It is important to note that eDNA and camera traps only provide presence / absence data, while GBS provides relative abundance. Combining methods proved particularly effective; we estimated a cumulative detection probability >0.95 for a single-day survey effort by collecting eDNA samples spread throughout a stream reach in tandem with a GBS. Utilizing and combining these non-invasive survey techniques could also be used to effectively detect other low density, cryptic, waterbirds occupying stream habitats.
Dataset DOI: 10.5061/dryad.2v6wwq039
Description of the data and file structure
Final data and model code that was used in the paper Combining non-invasive survey methods increases cumulative detection probability for breeding Harlequin Ducks (Histrionicus histrionicus) by Holmes, H.A., et al. 2026. DOI: 10.1002/wlb3.01610.
All data in this research project was collected by personal connected with Glacier National Park, Montana Fish, Wildlife and Parks, Idaho Department of Fish and Game, the USDA Forest Service, Northern Region, and Montana Audubon between 2022 - 2024. All data was gathered and cleaned, and code was written and analyzed by Holli Holmes.
Files and variables
File: metadata_camera.csv
Description: Metadata with variable descriptions for Camera_with_all_data.csv, camera_mt.csv, and camera_tl.csv.
File: Camera_with_all_data.csv
Description: All game camera data used in both the time-lapse and motion-detection game camera models.
Variables: See metadata_camera.csv for full details.
File: camera_mt.csv
Description: Game camera data used in the motion-detection game camera model.
Variables: See metadata_camera.csv for full details.
File: camera_tl.csv
Description: Game camera data used in the time-lapse game camera model.
Variables: See metadata_camera.csv for full details.
File: metadata_eDNA.csv
Description: Metadata with variable descriptions for eDNA_final_data.csv, eDNA_single_final.csv, and eDNA_brood_final.csv
File: eDNA_final_data.csv
Description: All eDNA data used in both the incubation and brood rearing models.
Variables: See metadata_eDNA.csv for full details.
File: eDNA_single_final.csv
Description: All eDNA data used in the incubation detection model.
Variables: See metadata_eDNA.csv for full details.
File: eDNA_brood_final.csv
Description: All eDNA data used in the brood rearing detection model.
Variables: See metadata_eDNA.csv for full details.
File: metadata_GBS.csv
Description: Metadata with variable descriptions for GBS_final_clean_data_for_model.csv, GBS_single.csv, and GBS_brood.csv.
File: GBS_final.csv
Description: All ground-based foot survey data used in both the incubation and brood rearing models.
Variables: See metadata_GBS.csv for full details.
File: GBS_single.csv
Description: All ground-based foot survey data used in the incubation detection model.
Variables: See metadata_GBS.csv for full details.
File: GBS_brood.csv
Description: All ground-based foot survey data used in the brood rearing detection model.
Variables: See metadata_GBS.csv for full details.
File: metadata_lorlegram.csv
Description: Metadata with variable descriptions for eDNA_clean_QeDNA.csv. This data was used in the lorlegram extra analysis.
File: eDNA_clean_QeDNA.csv
Description: All eDNA data that includes the quantitative eDNA results. This data was used in the lorlegram extra analysis.
Variables: See metadata_lorlegram.csv for full details.
Code/software
File: camera_trap_models.Rmd
Description: Code for both the time-lapse and motion detection camera trap detection models.
File: Camera_trap_results.Rmd
Description: Code showing how we extrapolated to the number of cameras needed to achieve detection probabilities of 80, 90, and 95%.
File: eDNA_models.Rmd
Description: Code for both the incubation and brood rearing detection eDNA models.
File: eDNA_results.Rmd
Description: Code showing how we extrapolated to the number of eDNA samples needed to achieve detection probabilities of 80, 90, and 95%.
File: GBS_analysis.Rmd
Description: Code for both the incubation and brood rearing detection ground-based foot survey models.
File: GBS_results.Rmd
Description: Code showing how we extrapolated to the number of ground-based foot surveys needed to achieve detection probabilities of 80, 90, and 95%.
File: combining_methods_cummulative_detection.Rmd
Description: Code showing how we extrapolated the detection probability using different combinations of camera traps, eDNA, and ground-based foot surveys.
File: distance_sampling_lorelogram.R
Description: Code for how we conducted the lorelogram extra analysis.
Access information
Other publicly accessible locations of the data:
- NA
Data was derived from the following sources:
- All data was collected in the field by personal connected with Glacier National Park, Montana Fish, Wildlife and Parks, Idaho Department of Fish and Game, the USDA Forest Service, Northern Region, and Montana Audubon
