Photographs, datasets and code supporting ‘An accurate and efficient semiautomated approach to counting birds: estimating Northern Gannet colony size in Canada'
Bibliographic record
Abstract
ABSTRACTImproving the efficiency of population monitoring and conservation programs is beneficial, so long as the accuracy of the information collected is not diminished. The need to expeditiously estimate the population size of seabird colonies is especially acute during mass mortality events when aerial surveys can provide information quickly on the extent of effects and total mortality. In 2022, the Highly Pathogenic Avian Influenza virus caused outbreaks at most Northern Gannet Morus bassanus colonies worldwide, killing tens of thousands of gannets in eastern Canada. In this study, we evaluated the accuracy and efficiency of a semiautomated method using the free software CountEm for counting Northern Gannet nests by reanalysing thirteen years of aerial photographs from past population surveys (2009–2020 and 2022). We developed a protocol that generated population estimates that are accurate enough to support population management objectives (i.e., within 2–5% of manual counts) and outline additional ways to improve CountEm accuracy. Additionally, using CountEm was 1100% more efficient than manually counting based on counting time. Since CountEm relies on human identification of objects to be counted, our methods, results, and conclusions are transferable to any taxa that form large aggregations and can be identified and counted in photographs.About this repositoryThis repository can be cited as follows:Walker, Jacob, Trevor S. Avery, Francis St-Pierre, Jean-François Rail, Danielle E. A. Quinn, Matthew English, and Stephanie Avery-Gomm. 2024. “Photographs, datasets and code supporting ‘An accurate and efficient semiautomated approach to counting birds: estimating Northern Gannet colony size in Canada’.” Figshare. https://doi.org/10.6084/m9.figshare.25483174It contains photographs, datasets and code supporting the peer reviewed publication Walker, J., Avery, T. S., St‐Pierre, F., Rail, J., Quinn, D. E. A., English, M., & Avery‐Gomm, S. (2025). An accurate and efficient semiautomated approach to counting birds: Estimating Northern Gannet colony size in Canada. Ecosphere, 16(2), e70183. https://doi.org/10.1002/ecs2.70183PhotographsThis repository contains data associated with 52 composite photographs of Northern Gannet colonies at Ile Bonaventure and Rochers aux Oiseaux taken between 2009 and 2022. See the manuscript above for details.DatasetsThe raw data are provided in alldata.csv. Results of repeated CountEm runs (n = 11 photographs) are found in multipleruns.csv. The provided variable key describes variables in both data files (VariableKey.xlsx). To reproduce the analyses performed in this study, use the code provided in reproducible_analysis.R.CodeThe code required to reproduce the double count analysis is in reproducible_analysis.R, using the input files data/rawdata.csv and data/multipleruns.csv. The code, including detailed comments, is organized into 8 sections:Load Packages: loads the required packages (see Sofware requirements, below)Import, Restructure, and Subset Data: five subsets of the available data are created to faciliate analyses in sections 3-8df: requires data/rawdata.csv; results from the first CountEm run for each photo using 300 quadrats, only considering AOTs (52 rows, 22 columns)df_500: requires data/rawdata.csv; CountEm results from a subset of 12 photos using 300 and 500 quadrats, only considering AOTs (12 rows, 10 columns)df_dead: requires data/rawdata.csv; results from the first CountEm run for each photo using 300 quadrats, only considering dead birds (4 rows, 22 columns)mdf: requires data/multipleruns.csv; results from ten CountEm runs for a subset of 11 photos, only considering AOTs (110 rows, 22 columns)mdf_dead: requires data/multipleruns.csv; results from ten CountEm runs for a subset of 11 photos, only considering dead birds (30 rows, 22 columns)Sections 3-8 are used to generate the results found in the corresponding Results subheaders of the text:Results: CountEm Accuracy: uses the data object dfto assess the use of CountEm to estimate the number of AOTs; produces Figures 3 and 4Results: Increasing CountEm Quadrats: uses the data object df_500 to assess the impact of increasing the number of CountEm quadrats from 300 to 500Results: Estimating the Number of Dead Birds: uses the data object df_dead to assess the use of CountEm to estimate the number of dead birdsResults: Accuracy of Multiple CountEm Runs: uses the data object mdf to run a resampling routine and assess the use of multiple CountEm runs to estimate the number of AOTsResults: CV to Inform Number of CountEm Runs: uses the results of the simulation in section 6 to determine if the coefficient of variation (CV) can be used to determine the number of CountEm runs that could be summarised to generate estimates within 5% of the manual count of AOTsResults: Efficiency: uses the data objects df and mdf to summarise the user time required to apply CountEm to generate estimates of the number of AOTsSoftware requirementsScripts are written for R v4.3.1. See scripts and manuscript for packages and software citations.Required R packages can be installed in R with: install.packages(c("tidyverse", "readxl", "BSDA", "boot", "ggdist", "ggeffects", "emmeans", "marginaleffects"))
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.003 | 0.016 |
| Meta-epidemiology (narrow) | 0.001 | 0.001 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.006 | 0.012 |
| Science and technology studies | 0.002 | 0.001 |
| Scholarly communication | 0.003 | 0.002 |
| Open science | 0.004 | 0.003 |
| Research integrity | 0.001 | 0.001 |
| Insufficient payload (model declined to judge) | 0.107 | 0.056 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".