On Thin Ice: Towards Explainable Conservation Monitoring via Attribution and Perturbations

2025-10-24

AI Summary

Black-box neural networks lack interpretability, hindering trust among conservation practitioners in ecological monitoring. This paper addresses aerial harbor seal detection and introduces the first explainability evaluation framework tailored to conservation applications, quantifying explanation quality along three dimensions: localization fidelity, confidence faithfulness, and diagnostic utility. Leveraging Faster R-CNN, we integrate HiResCAM, LayerCAM, LIME, and perturbation-based analysis to generate attribution maps. Results show the model attends primarily to seal torsos—not background ice—and occlusion of salient regions substantially reduces prediction confidence. Systematic error analysis reveals consistent misclassifications of black ice and rocks as seals. Beyond diagnosing failure modes, our framework informs targeted data curation strategies, advancing auditable, trustworthy decision-support tools for wildlife conservation. (149 words)

Technology Category

• Computer Vision: Interpretability, Explainability, and Transparency
• Machine Learning: Calibration & Uncertainty Quantification
• Knowledge Representation and Reasoning: Diagnosis and Abductive Reasoning

Application Category

• Search and Retrieval-Augmented AI: Web evaluation methodologies and metrics
• User Modeling, Personalization and Recommendation: Explainable and interpretable methods for personalization
• Responsible Web: Algorithmic accountability and transparency on the web

Abstract

Computer vision can accelerate ecological research and conservation monitoring, yet adoption in ecology lags in part because of a lack of trust in black-box neural-network-based models. We seek to address this challenge by applying post-hoc explanations to provide evidence for predictions and document limitations that are important to field deployment. Using aerial imagery from Glacier Bay National Park, we train a Faster R-CNN to detect pinnipeds (harbor seals) and generate explanations via gradient-based class activation mapping (HiResCAM, LayerCAM), local interpretable model-agnostic explanations (LIME), and perturbation-based explanations. We assess explanations along three axes relevant to field use: (i) localization fidelity: whether high-attribution regions coincide with the animal rather than background context; (ii) faithfulness: whether deletion/insertion tests produce changes in detector confidence; and (iii) diagnostic utility: whether explanations reveal systematic failure modes. Explanations concentrate on seal torsos and contours rather than surrounding ice/rock, and removal of the seals reduces detection confidence, providing model-evidence for true positives. The analysis also uncovers recurrent error sources, including confusion between seals and black ice and rocks. We translate these findings into actionable next steps for model development, including more targeted data curation and augmentation. By pairing object detection with post-hoc explainability, we can move beyond"black-box"predictions toward auditable, decision-supporting tools for conservation monitoring.

Problem

• Enhancing trust in black-box neural networks for conservation monitoring
• Evaluating explanation methods for seal detection in aerial imagery
• Identifying systematic model errors to improve ecological monitoring tools

Innovation

• Uses gradient-based activation mapping for model explanations
• Applies perturbation tests to verify detection confidence changes
• Combines object detection with post-hoc explainability methods