Data Status Report for MICA - Management of Invasive Coypu and muskrAt in Europe
Research Institute for Nature and Forest (INBO), Wageningen University, and Unie van Waterschappen
Tim Adriaens, Kristof Baert, Warre Baert, Abel De Boer, Gust Boiten, Dimitri Brosens, Emma Cartuyvels, Jim Casaer, Bram D’hondt, Manon Debrabandere, Peter Desmet, Sander Devisscher, Dennis Donckers, Sanne Van Donink, Silke Dupont, Wouter Franceus, Heiko Fritz, Lilja Fromme, Friederike Gethöffer, Jan Gouwy, Sanne Govaert, Casper Herbots, Frank Huysentruyt, Leo Kehl, Liam Letheren, Lydia Liebgott, Yorick Liefting, Jan Lodewijkx, Claudia Maistrelli, Björn Matthies, Kelly Meijvisch, Dolf Moerkens, Axel Neukermans, Brecht Neukermans, Jelle Ronsijn, Kurt Schamp, Dan Slootmaekers, Linda Tiggelman, and Danny Van der beeck
June 18, 2026
Abstract
This report provides an overview of the camera trap dataset and the preprocessing steps used for automated report generation. It enables users to quickly assess the quality of their data through a concise summary generated by a series of automated checks. The report is structured into six main sections: Setup, Data Availability, Species Records, Validation, Annotation, and Observation Types by Capture Method, each offering a concise assessment of dataset integrity and readiness. At the conclusion of the report, you will receive an overall data quality classification, which will be one of the following:
- Perfect: All key checks passed; no issues detected.
- Acceptable: Minor issues found; the dataset remains usable, although corrections are recommended.
- Needs Improvement: Major issues identified; corrections are mandatory before continuing.
If the status is Perfect or Acceptable, you may proceed with generating the final report. However, if the status is Needs Improvement, report generation is not possible due to major issues identified in the dataset, and the dataset must be improved before continuing.
Note: To ensure the highest quality automated report generation from your camera trap data, we strongly recommend improving your dataset whenever possible. If the status is Acceptable, consider refining it to reach the ideal Perfect level. This may require only minor adjustments, but it can make a significant difference in the final report. If the status is Needs Improvement, we advise resolving the identified issues to bring the dataset to at least the Acceptable level, and ideally to the Perfect standard.
Chapter 1: Setup
This chapter covers spatial and temporal checks. Dataset spans Belgium, Netherlands, and Germany (MICA - Management of Invasive Coypu and muskrAt in Europe) with time zone UTC+2 and coordinate range 50.699°–53.407°N and 3.518°–8.330°E. All required CSV files (deployments.csv, media.csv, observations.csv) and JSON file are available in the dataset. The 49 distinct camera locations are distributed within a minimum convex polygon (MCP) of 43608.26 km².
Spatial Data
This section summarizes the spatial quality of the camera-trap data extracted from the dataset files. It focuses specifically on location data, assessing issues such as missing values, duplicate entries, spatial outliers, and overall spatial structure. The complete results are presented in Table 1. If any issues are identified, we strongly recommend correcting the data before proceeding with report generation for your study site.
Duplicate entries (by coordinates, LocationName, or LocationID) are collapsed to the first occurrence, and any rows with missing values in any column are excluded from the report-generation analysis unless you edit them. If any issues are identified in the table below, we strongly recommend correcting the data before proceeding with report generation for your study site.
Note: If your dataset contains rows where the coordinates (longitude and latitude) are complete but either the LocationID or LocationName is missing, these records are still treated as missing. This is because, for spatial quality checks and analysis, both the geographic coordinates and their associated metadata (LocationID and LocationName) are equally important.
| Metric | Result | Status |
|---|---|---|
| Number of locations | 54 | Before any filtering |
| Number of duplication in coordinates | 5 | 🔴 Duplicate coordinates found in 5 groups; 5 duplicate rows. |
| Number of duplication in LocationID | 0 | 🟢 No duplicated locationIDs |
| Number of duplication in LocationName | 0 | 🟢 No duplicated locationNames |
| Missing data | 0 | 🟢 No missing data found |
| Total distinct coordinates | 49 | After removing duplicates/missing values |
| Mean distance between locations (m) | 4499.4 | Mean inter-location distance |
| Max distance between locations (m) | 88620.03 | D_LD_lake Dümmer and B_DM_val 1_weerstation |
| Min distance between locations (m) | 69.92 | B_DL_val 3_dikke boom and B_DL_val 2_emissaire |
| Spatial pattern | targeted (indicated explicitly in metadata) | Clustered (using point-pattern analysis) |
| Outliers |
High risk outlier = 1 Medium risk outlier = 0 Low risk outlier = 0 Non-Terrestrial = 0 |
🔴 High-risk (1): D_LD_lake Dümmer | 🟢 All locations are on land. |
Temporal Data
This section summarizes the temporal patterns of the camera-trap dataset, including coverage over time, potential gaps, and any detected anomalies or outliers, which are presented in Table 2.
| Metric | Result |
|---|---|
| Deployment year coverage | 2019 – 2023 (🟢 complete) |
| Observation year coverage | 1999 – 2023 (🔴 missing: 2000–2018) |
| Deployment first/last setup | 2019-09-18 07:11:08 – 2023-09-27 08:23:15 |
| Observation first/last record | 1999-12-31 – 2023-05-03 |
| Temporal consistency | 🔴 Temporal inconsistency (Observations exist, but deployments are missing for: 1999) |
| First/last date check | 🔴 Earliest observation is earlier than the first deployment start date (check timestamps or timezone). 🚨 |
| Month coverage span |
2019: Sep–Dec 2020: Jan–Dec 2021: Jan–Dec 2022: Jan–Dec 2023: Jan–May, Sep |
| Calendar coverage | 1325 of 1471 days (90.1%) |
| Max gap between deployments | 146 days (from 2023-05-04 to 2023-09-26) |
| Min gap between deployments | 1 days (from 2020-09-02 to 2020-09-02) |
| Missing deployment intervals | None 🟢 |
| Zero-length deployments | 64 zero-length interval(s) 🟡 |
| Temporal outliers | None 🟢 |
| Invalid timestamp format | 602 timestamp(s) have invalid format 🔴 (rows: 7540, 7558, 7577, 7595, 7616, 7640, 7659, 7678, 7700, 7719, …) |
| Future observation timestamps | None 🟢 |
Chapter 2: Data Availability
Review Essential Data Availability
In this chapter, we review all essential data components required for generating the report. These include both mandatory fields and necessary supporting fields or files that contribute to the completeness and overall quality of the final output. Table 3 provides a structured overview of the availability and completeness of these elements. While all fields listed in the table are important for generating a high-quality report, the bolded fields are mandatory. Their status must be marked as Complete; otherwise, any associated records with missing or partial values will be automatically excluded from the final output.
| Category | Field | Status |
|---|---|---|
| Locations | locationID | 🟢 Complete |
| locationName | 🟢 Complete | |
| longitude | 🟢 Complete | |
| latitude | 🟢 Complete | |
| Deployment | deploymentID | 🟢 Complete |
| locationID | 🟢 Complete | |
| deployment_interval | 🟢 Complete | |
| deploymentStart | 🟢 Complete | |
| deploymentEnd | 🟢 Complete | |
| habitat | 🔴 Incomplete (1539 of 1539 missing; 100%) | |
| setupBy | 🟡 Partial (451 of 1539 missing; 29.30%) | |
| baitUse | 🔴 Incomplete (1539 of 1539 missing; 100%) | |
| cameraHeight | 🟢 Complete | |
| Observations | timestamps | 🟡 Partial (602 of 183813 missing; 0.33%) | invalid format: 602 |
| observationType | 🟡 Partial (16102 of 183813 missing; 8.76%) | |
| count | 🟢 Complete | |
| classifiedBy | 🟢 Complete | |
| taxonID | 🟢 Complete (taxonID recorded for 116529 of 116529 animals; 100%) | 100 unique | |
| behavior | 🔴 Incomplete (behavior recorded for 0 of 116529 animals; 0%) | |
| sex | 🟡 Partial (sex recorded for 9850 of 116529 animals; 8.45%) | |
| lifeStage | 🟡 Partial (lifeStage recorded for 39374 of 116529 animals; 33.79%) | |
| angle | 🔴 Incomplete (angle recorded for 0 of 116529 animals; 0%) | |
| radius | 🔴 Incomplete (radius recorded for 0 of 116529 animals; 0%) | |
| speed | 🔴 Incomplete (speed recorded for 0 of 116529 animals; 0%) | |
| individualID | 🔴 Incomplete (individualID recorded for 0 of 116529 animals; 0%) | |
| Media | media timestamp | 🟡 Partial (657 of 3996924 missing; 0.02%) | invalid format: 657 |
| file.path | 🟢 Complete | |
| comments | 🟢 Complete | |
| favourite | 🟢 Complete | |
| Sequences | nrphotos | 🟢 Complete |
| captureMethod | 🟢 Complete | |
| Taxonomy | taxonID | 101 unique taxonID identified |
| scientificName | 🟢 Complete | |
| vernacularNames.nld | 🟢 Complete | |
| vernacularNames.eng | 🟢 Complete | |
| Additional Files | habitat.csv | 🟢 Complete |
| spatial boundary (e.g., shapefile) | 🟢 Complete |
Chapter 4: Validation
In this chapter, we summarize the classification and validation results of the observation data. The ‘CaptureMethod’ column indicates the method used to capture each image, which may include motion detection, time-lapse, or other mechanisms depending on the camera trap setup. The columns ‘Human’, ‘Machine’, and ‘NA_Classification’ represent the number and percentage of observations classified manually by a human, by an automated machine model, or left unclassified (NA), respectively. NA values often result from blank images, errors, or missing data. The ‘Machine_Animal’ column reports how many machine-classified observations were identified as animals, along with their percentage out of all machine classifications for that method. Finally, ‘Validated_Animal’ shows how many of those machine-identified animals were subsequently validated by a human, indicating human-confirmed correctness of the machine prediction. Collectively, this summary helps evaluate both the extent and reliability of the classification process across capture methods. In cases where validation rates are low or discrepancies are noted, users are encouraged to revisit the raw data for quality control, potential retraining of models, or targeted human review of specific subsets.
| captureMethod | Human | Machine | NA_Classification | Total | Machine_Animal | Validated_Animal |
|---|---|---|---|---|---|---|
| motionDetection | 75670 (43.4%) | 50198 (28.8%) | 48439 (27.8%) | 174307 | 34705 (69.1%) | 370 (1.1%) |
| timeLapse | 8 (0.1%) | 4384 (46.1%) | 5114 (53.8%) | 9506 | 33 (0.8%) | 0 (0%) |
| TOTAL | 75678 (41.2%) | 54582 (29.7%) | 53553 (29.1%) | 183813 | 34738 (63.6%) | 370 (1.1%) |
Chapter 5: Annotation
In this chapter, we provide an overview of annotation confidence in the camera-trap observations. The summary distinguishes between machine-generated and human-generated classifications and highlights the variation in confidence scores across annotation methods. This section helps assess annotation reliability, identify records below selected confidence thresholds, and evaluate whether additional quality control or review may be needed before ecological reporting.
Machine Classification Confidence
| Statistic | Value |
|---|---|
| Minimum Confidence Score | 0.00 |
| Maximum Confidence Score | 1.00 |
| Mean Confidence Score | 0.75 |
| Below Threshold (<0.8) | 0 |
| Total Annotations | 35124 |
Human Classification Confidence
| Statistic | Value |
|---|---|
| Minimum Confidence Score | 0.50 |
| Maximum Confidence Score | 1.00 |
| Mean Confidence Score | 0.99 |
| Below Threshold (<1) | 1456 |
| Total Annotations | 75678 |
Chapter 6: Observation Types by Capture Method
This figure summarizes observation types by capture method (Motion, TimeLapse) and overall. Each pie shows the within-method percentage of each observation type—animal, human, vehicle, blank, unknown, and unclassified; providing a quick view of what was observed and its relative weight in the dataset.
Motion
TimeLapse
Total
Conclusion
Based on results across five key sections—Spatial, Temporal, Data Availability, Validation, and Annotation, this dataset is classified as Acceptable — Minor issues found; the dataset remains usable, although corrections are recommended.
Acknowledgment
This report was generated using the camtrapReport R package, developed by Elham Ebrahimi at Wageningen University & Research and Utrecht University, the Netherlands. The development of camtrapReport was supported by Biodiversa+ through the Big Picture project. We also gratefully acknowledge the European Observatory of Wildlife network for its contribution to package testing. Users are kindly requested to cite the package when using camtrapReport or publishing results derived from it.