Towards an automated protocol for wildlife density estimation using camera-traps

Andrea Zampetti,Davide Mirante,Pablo Palencia,Luca Santini

DOI: https://doi.org/10.1101/2024.08.05.606345

2024-08-10

Abstract:Camera-traps are valuable tools for estimating wildlife population density, and recently developed models enable density estimation without the need for individual recognition. Still, processing and analysis of camera-trap data are extremely time-consuming. While algorithms for automated species classification are becoming more common, they have only served as supporting tools, limiting their true potential in being implemented in ecological analyses without human supervision. Here, we assessed the capability of two camera-trap based models to provide robust density estimates when image classification is carried out by machine learning algorithms. We simulated density estimation with Camera-Traps Distance Sampling (CT-DS) and Random Encounter Model (REM) under different scenarios of automated image classification. We then applied the two models to obtain density estimates of three focal species (roe deer Capreolus capreolus, red fox Vulpes vulpes, and Eurasian badger Meles meles) in a reserve in central Italy. Species detection and classification was carried out both by the user and machine learning algorithms (respectively, MegaDetector and Wildlife Insights), and all outputs were used to estimate density and ultimately compared. Simulation results suggested that the CT-DS model could provide robust density estimates even at poor algorithm performances (down to 50% of correctly classified images), while the REM model is more unpredictable and depends on multiple factors. Density estimates obtained from the MegaDetector output were highly consistent for both models with the manually labelled images. While Wildlife Insights performance differed greatly between species (recall: badger = 0.15; roe deer = 0.56; fox = 0.75), CT-DS estimates did not vary significantly; on the contrary, REM systematically overestimated density, with little overlap in standard errors. We conclude that CT-DS and REM models can be robust to the loss of images when machine learning algorithms are used to identify animals, with the CT-DS being an ideal candidate for applications in a fully unsupervised framework. We propose guidelines to evaluate when and how to integrate machine learning in the analysis of camera-trap data for density estimation, further strengthening the applicability of camera traps as a cost-effective method for density estimation in (spatially and temporally) extensive multi-species monitoring programs.

Ecology

What problem does this paper attempt to address?

The problem that this paper attempts to solve is: how to use machine - learning algorithms to automate wildlife density estimation, in order to reduce the time and workload required for manual processing of camera - captured data, and ensure the reliability of density estimation results when automatically classifying images.

Specifically, the authors evaluated two camera - trap - based models - Camera - Traps Distance Sampling (CT - DS) and Random Encounter Model (REM) - for their ability to provide robust density estimates when using machine - learning algorithms for image classification in different scenarios. Through simulation experiments and actual case studies, they explored the impact of machine - learning algorithm performance on density estimation, aiming to provide an automated framework for ecological analysis that is completely free of human supervision.

### Key issues:
1. **Effect of automated image classification**: Evaluate the classification performance of machine - learning algorithms on different species, especially whether the density estimate is still reliable when the classification accuracy is low.
2. **Model sensitivity to error**: Compare the performance differences between the CT - DS and REM models when facing image classification errors, and understand which model is more tolerant of classification errors.
3. **Performance in practical applications**: Verify the feasibility of the automated method in the real environment through actual cases, and ensure that it can be applied to a wide range of multi - species monitoring projects.

### Formula summary:
- **Density estimation formula for CT - DS model**:
\[
D=\frac{\sum_{k = 1}^{K}n_k}{\pi\cdot w^2\cdot\sum_{k = 1}^{K}e_k\cdot p}
\]
where:
- \(n_k\) is the number of animal observations at the \(k\) - th point.
- \(w\) is the truncation distance.
- \(e_k\) is the sampling effort at the \(k\) - th point.
- \(p\) is the probability of capturing an animal within the range of \(\theta\) and \(w\).

- **Density estimation formula for REM model**:
\[
D=\frac{Y}{H\cdot\pi\cdot v\cdot r\cdot(2 + \theta)}
\]
where:
- \(Y\) is the number of independent animal capture events.
- \(H\) is the total survey effort of all camera stations.
- \(v\) is the daily activity range of the animal.
- \(r\) and \(\theta\) describe the radius and angle of the effective detection area of the camera.

Through the application of these models and formulas, the authors hope to prove that the automated method can significantly improve work efficiency while maintaining high precision, thus promoting broader application and development.