[Paper Review] AI-based Mapping of the Conservation Status of Orchid Assemblages at Global Scale

Joaquim Estopinan, Maximilien Servajean+3 more|arXiv (Cornell University)|Jan 9, 2024

Species Distribution and Climate ChangeEnvironmental Science3 citations

TL;DR

This study introduces a deep learning-based Species Distribution Model trained on 1 million orchid occurrences to map the global conservation status of orchid assemblages at 1 km resolution. It identifies Madagascar and Sumatra as high-threat regions and demonstrates that protected areas align poorly with predicted threat levels, revealing significant gaps in current IUCN assessments.

ABSTRACT

Although increasing threats on biodiversity are now widely recognised, there are no accurate global maps showing whether and where species assemblages are at risk. We hereby assess and map at kilometre resolution the conservation status of the iconic orchid family, and discuss the insights conveyed at multiple scales. We introduce a new Deep Species Distribution Model trained on 1M occurrences of 14K orchid species to predict their assemblages at global scale and at kilometre resolution. We propose two main indicators of the conservation status of the assemblages: (i) the proportion of threatened species, and (ii) the status of the most threatened species in the assemblage. We show and analyze the variation of these indicators at World scale and in relation to currently protected areas in Sumatra island. Global and interactive maps available online show the indicators of conservation status of orchid assemblages, with sharp spatial variations at all scales. The highest level of threat is found at Madagascar and the neighbouring islands. In Sumatra, we found good correspondence of protected areas with our indicators, but supplementing current IUCN assessments with status predictions results in alarming levels of species threat across the island. Recent advances in deep learning enable reliable mapping of the conservation status of species assemblages on a global scale. As an umbrella taxon, orchid family provides a reference for identifying vulnerable ecosystems worldwide, and prioritising conservation actions both at international and local levels.

Motivation & Objective

• Address the lack of global, high-resolution maps showing the conservation status of species assemblages.
• Overcome the Wallacean shortfall and IUCN assessment gap by predicting threat levels for unassessed orchid species.
• Develop spatially explicit indicators of conservation status at multiple scales using deep learning and environmental predictors.
• Evaluate the alignment between predicted threat levels and existing protected areas, particularly in Sumatra.
• Provide a scalable framework for identifying biodiversity hotspots and informing conservation prioritization using orchids as an umbrella taxon.

Proposed method

• Trained a Deep Species Distribution Model (Deep-SDM) on 1 million occurrence records of 14,000 orchid species using environmental predictors from remote sensing and climate data.
• Used a multi-label classification approach to predict the presence of multiple orchid species per 1 km grid cell, with model calibration to control for false positives.
• Defined two key conservation indicators: (i) the proportion of threatened species in each assemblage, and (ii) the threat status of the most at-risk species present.
• Applied a conditional probability threshold to balance recall and precision, optimizing for both sensitivity and specificity in predictions.
• Validated model performance using a held-out test set, measuring error as the absence of true labels in predicted assemblages.
• Generated global and interactive maps of conservation status indicators, with spatial resolution at 1 km, enabling multiscale analysis.

Experimental results

Research questions

• RQ1 To what extent can deep learning models predict the conservation status of orchid assemblages at global scale with high spatial resolution?
• RQ2 Where are the highest concentrations of threatened orchid assemblages globally, and how do they relate to known biodiversity hotspots?
• RQ3 How well do existing protected areas in Sumatra align with predicted threat levels for orchid assemblages?
• RQ4 To what degree do current IUCN assessments underestimate the true level of threat to orchid species in regions like Sumatra?
• RQ5 Can species assemblage-level indicators derived from AI models serve as reliable proxies for ecosystem-level conservation risk?

Key findings

• The highest global threat levels for orchid assemblages are concentrated in Madagascar and neighboring Indian Ocean islands, with high proportions of predicted threatened species.
• In Sumatra, protected areas show moderate correspondence with predicted threat indicators, but the model reveals widespread, unassessed threat levels across the island.
• The model achieved high recall in validation, with error defined as missing true labels in predicted assemblages, and precision was maximized by tuning the conditional probability threshold.
• Orchid assemblages in the Neotropics and Southeast Asia show high richness and threat levels, consistent with known biodiversity hotspots.
• The study’s 1 km resolution maps reveal sharp spatial variations in threat indicators, enabling fine-scale conservation planning.
• The approach outperforms existing global indicators like the Red List of Ecosystems and global plant extinction probability maps in spatial resolution and integration of species occurrence probabilities.