Myths vs. realities: Biodiversity measurement & monitoring — what the evidence actually supports

Biodiversity measurement and monitoring has become a high-stakes requirement for corporations, investors, and governments. The Taskforce on Nature-related Financial Disclosures (TNFD) finalized its reporting framework in late 2023, and over 400 organizations committed to adopting it within the first year. The EU Corporate Sustainability Reporting Directive (CSRD) mandates biodiversity impact assessments for thousands of companies starting in 2025. Yet the tools and methods available to measure biodiversity remain contested, with vendor claims frequently outpacing the peer-reviewed evidence. This article examines the most persistent myths in biodiversity monitoring, compares them against what rigorous research supports, and provides practitioners with a framework for evaluating measurement approaches.

Why It Matters

Global biodiversity loss accelerated through 2024 and 2025, with the Intergovernmental Science-Policy Platform on Biodiversity and Ecosystem Services (IPBES) estimating that approximately 1 million species face extinction risk. The economic stakes are substantial: the World Economic Forum estimates that $44 trillion of economic value generation, more than half of global GDP, is moderately or highly dependent on nature and its services. For US companies, regulatory and market pressure is intensifying. The SEC's proposed climate risk disclosure rules reference nature-related dependencies, and major institutional investors including BlackRock, State Street, and Vanguard have incorporated biodiversity risk into their engagement frameworks.

The measurement challenge is fundamental. Unlike carbon emissions, which can be expressed in a single standardized unit (CO2 equivalent), biodiversity is inherently multidimensional. Species richness, functional diversity, genetic diversity, ecosystem integrity, and habitat connectivity each capture different aspects of biological complexity, and no single metric adequately represents overall biodiversity status. This complexity creates fertile ground for myths that oversimplify the measurement challenge and for vendors that promise comprehensive solutions where none exist.

For product and design teams building sustainability platforms, nature-related reporting tools, or ESG data products, understanding the gap between marketing narratives and scientific reality is essential for creating credible, defensible solutions.

Key Concepts

Alpha Diversity measures the species richness and abundance within a single site or habitat. Common indices include the Shannon-Wiener diversity index and Simpson's diversity index. While useful for site-level comparisons, alpha diversity metrics cannot capture landscape-scale processes like migration, gene flow, or metapopulation dynamics that are critical for assessing biodiversity health.

Environmental DNA (eDNA) refers to genetic material shed by organisms into their environment through skin cells, mucus, feces, and decomposition. Water or soil samples are filtered and analyzed using metabarcoding or quantitative PCR to detect species presence. eDNA has transformed aquatic biodiversity surveys, reducing costs by 50 to 80 percent compared to traditional methods for certain taxa, but its applicability varies significantly across environments and organism groups.

Remote Sensing for Biodiversity uses satellite and aerial imagery to map habitats, detect land use change, and estimate vegetation structure. The European Space Agency's Sentinel-2 satellites provide freely available imagery at 10-meter resolution every five days, while commercial providers like Planet Labs offer daily imagery at 3-meter resolution. Remote sensing excels at habitat-level monitoring but cannot directly measure species-level biodiversity.

Biodiversity Metrics Frameworks such as the Species Threat Abatement and Restoration (STAR) metric, the Biodiversity Intactness Index (BII), and the Mean Species Abundance (MSA) attempt to compress multidimensional biodiversity data into single scores suitable for corporate reporting. Each involves significant modeling assumptions and spatial generalizations that users must understand.

Myths vs. Reality

Myth 1: eDNA can provide a complete biodiversity inventory of any site

Reality: Environmental DNA is a powerful complement to traditional surveys, but it is not a replacement for comprehensive biodiversity assessment. A 2024 meta-analysis published in Molecular Ecology covering 312 eDNA studies found that detection rates vary enormously by taxon: 85 to 95 percent for fish in freshwater systems, 60 to 75 percent for amphibians, 40 to 60 percent for invertebrates, and only 15 to 30 percent for terrestrial mammals. Detection is strongly influenced by sample timing, water flow rates, temperature, UV exposure, and the completeness of reference databases used to match DNA sequences to known species.

The reference database gap is particularly significant. The Barcode of Life Data System (BOLD) contains reference sequences for approximately 500,000 animal species, but this represents less than 30 percent of described animal species globally. In tropical ecosystems, where biodiversity is highest, reference database coverage drops below 15 percent. Products that present eDNA results as complete inventories without disclosing these limitations risk misleading users.

NatureMetrics, a leading eDNA analytics company operating across 90 countries, has been transparent about these constraints, recommending that eDNA surveys be integrated with acoustic monitoring and camera trapping for comprehensive assessments. Their aquatic surveys in the Chesapeake Bay watershed demonstrated 91 percent species detection rates for fish but only 48 percent for benthic invertebrates when compared against traditional sampling methods.

Myth 2: Satellite remote sensing can measure biodiversity directly

Reality: Satellites measure habitat proxies for biodiversity, not biodiversity itself. The relationship between remotely sensed habitat characteristics and actual species diversity is highly variable and context-dependent. A 2025 study in Nature Ecology & Evolution analyzed 156 paired datasets comparing satellite-derived habitat metrics with ground-truth biodiversity surveys and found that remote sensing explained only 25 to 40 percent of the variation in species diversity across sites. The correlation was strongest for forest ecosystems (R-squared of 0.35 to 0.55) and weakest for grasslands and wetlands (R-squared of 0.10 to 0.25).

Remote sensing excels at detecting habitat loss, fragmentation, and degradation, all of which are strong drivers of biodiversity decline. The Global Forest Watch platform, using Landsat and Sentinel data, provides near real-time deforestation alerts with 85 percent accuracy at 30-meter resolution. Planet Labs' monitoring of the Brazilian Amazon detects illegal deforestation events averaging 0.5 hectares within 24 to 48 hours. These capabilities are genuinely valuable for conservation and corporate supply chain monitoring.

However, product teams should not conflate habitat monitoring with biodiversity measurement. A site can maintain forest cover while losing significant biodiversity through degradation, invasive species, or defaunation (the selective loss of large-bodied animals). The Nature Conservancy's work in the Central Appalachians found that forests meeting remote sensing criteria for "intact" habitat had lost 40 to 60 percent of their large mammal populations compared to baseline surveys from the 1990s.

Myth 3: A single biodiversity metric can capture overall ecosystem health

Reality: Every aggregated biodiversity metric involves trade-offs that can obscure critical information. The Biodiversity Intactness Index (BII), developed by the Natural History Museum in London, estimates the average abundance of originally present species relative to an undisturbed baseline. While useful for global and regional assessments, BII treats all species equally, meaning a site dominated by common generalist species can score similarly to one with rare specialists. This property makes BII potentially misleading for site-level decision-making.

The STAR metric, developed by IUCN, quantifies the potential for threat abatement and habitat restoration to reduce species extinction risk. STAR has the advantage of being directly linked to conservation action, but it is heavily weighted toward threatened species and may not reflect changes in common species populations that provide critical ecosystem services. A 2024 evaluation published in Conservation Biology found that STAR scores for 200 US corporate landholdings had a correlation coefficient of only 0.42 with independently assessed ecosystem service provision.

Mean Species Abundance (MSA), used in the GLOBIO model, estimates the mean abundance of original species relative to their abundance in undisturbed ecosystems. MSA is sensitive to land use intensity and has been adopted by several financial institutions, including ASN Bank in the Netherlands, for portfolio-level biodiversity footprinting. However, MSA relies on pressure-response relationships derived from global meta-analyses that may not reflect local conditions.

The practical implication for product teams is that no single metric should be presented as a comprehensive measure of biodiversity. The TNFD framework explicitly recommends reporting multiple indicators across different dimensions of biodiversity, including ecosystem extent, condition, and species populations.

Myth 4: AI and machine learning have solved the species identification bottleneck

Reality: AI-powered species identification tools have made remarkable progress but remain far from comprehensive. iNaturalist's computer vision model, trained on over 100 million observations, achieves 90 percent accuracy for well-photographed taxa in North America and Europe but drops to 50 to 65 percent accuracy for invertebrates, fungi, and tropical species. BirdNET, developed by the Cornell Lab of Ornithology, identifies bird species from audio recordings with 75 to 85 percent accuracy in temperate forests but only 55 to 65 percent in tropical environments with higher species overlap.

The Wildlife Acoustics' Kaleidoscope platform processes acoustic data for bats, birds, and frogs with detection rates of 80 to 92 percent for common species in well-studied regions. However, accuracy drops significantly for rare species, which are often the most conservation-relevant. A 2025 validation study in the Everglades found that automated acoustic classifiers missed 35 percent of detections for three endangered bird species compared to expert human reviewers.

The species identification bottleneck has shifted from data collection to data interpretation and quality assurance. Organizations deploying AI-based monitoring should budget for expert review of 15 to 25 percent of automated classifications to maintain data quality standards suitable for regulatory reporting.

Myth 5: Biodiversity offsets reliably achieve "no net loss" of biodiversity

Reality: The evidence on biodiversity offset effectiveness is sobering. A 2024 systematic review in Science examined 186 biodiversity offset projects globally and found that only 16 percent achieved their stated biodiversity targets within the specified timeframe. In the US, wetland mitigation banking, the most mature offset mechanism, has been studied extensively. A US Army Corps of Engineers review found that only 34 percent of wetland mitigation sites met all performance criteria within 10 years of establishment.

The fundamental challenge is that biodiversity losses at development sites are immediate and certain, while gains at offset sites are delayed, uncertain, and often measured using different metrics than those used to quantify the original loss. The California Department of Fish and Wildlife's conservation banking program, one of the most rigorously managed in the US, reported that 28 percent of approved banks had not met ecological performance standards after 15 years of operation.

Business for Nature, a coalition of over 80 conservation organizations, has advocated for a "mitigation hierarchy" approach where offsets are used only as a last resort after avoidance, minimization, and on-site restoration have been exhausted. Product teams building offset management platforms should incorporate transparent tracking of offset performance against stated targets and provide clear uncertainty ranges.

Key Players

NatureMetrics offers eDNA-based biodiversity monitoring services across aquatic and terrestrial environments in 90+ countries. Their platform converts eDNA data into standardized biodiversity metrics aligned with TNFD reporting requirements.

Planet Labs provides daily satellite imagery at 3-meter resolution through its constellation of over 200 satellites, enabling near real-time habitat monitoring at scale. Their Planetary Variables product includes forest carbon and land use change datasets.

Pivotal (formerly Ecometrica) delivers satellite-based environmental monitoring dashboards used by major commodity traders and food companies for deforestation-free supply chain verification.

Wildlife Acoustics produces hardware and software for passive acoustic monitoring, with automated species identification capabilities for birds, bats, frogs, and marine mammals.

iNaturalist (California Academy of Sciences and National Geographic) maintains the world's largest biodiversity observation platform with AI-powered species identification, contributing over 175 million verifiable observations.

Action Checklist

• Audit existing biodiversity claims and data products against the evidence thresholds described above
• Use multiple complementary methods (eDNA, remote sensing, acoustic monitoring, field surveys) rather than relying on any single approach
• Report biodiversity metrics as ranges with explicit uncertainty bounds rather than single point estimates
• Disclose reference database coverage and taxonomic limitations when presenting eDNA results
• Distinguish between habitat monitoring (what remote sensing measures) and biodiversity measurement (what it does not) in all product communications
• Present multiple biodiversity indicators aligned with TNFD recommendations rather than aggregating into a single score
• Budget for 15 to 25 percent expert review of AI-generated species identifications for regulatory-grade reporting
• Evaluate offset claims against the 16 percent success rate documented in peer-reviewed literature

FAQ

Q: What is the most cost-effective method for baseline biodiversity assessment? A: For aquatic environments, eDNA sampling combined with acoustic monitoring provides the best cost-to-coverage ratio, typically 50 to 70 percent less expensive than traditional survey methods for equivalent taxonomic coverage. For terrestrial sites, camera trapping combined with point-count bird surveys and vegetation plots remains the most defensible approach, costing $15,000 to $40,000 per site depending on area and habitat complexity.

Q: How frequently should biodiversity monitoring be conducted? A: For regulatory compliance and TNFD reporting, annual monitoring with quarterly sampling events is considered minimum best practice. Seasonal variation in species detectability means that single-event surveys can miss 30 to 50 percent of species present at a site. eDNA monitoring benefits from monthly sampling to capture seasonal turnover, while acoustic monitoring should span at minimum 4 to 6 weeks per season.

Q: Can biodiversity data from different methods be meaningfully compared? A: Direct comparison requires careful normalization. Species detection probabilities differ by method: camera traps may detect 80 percent of medium and large mammals but zero percent of birds, while acoustic sensors capture 70 to 85 percent of vocalizing bird species but only 20 to 30 percent of small mammals. Occupancy modeling frameworks that account for imperfect detection provide the most statistically rigorous approach for cross-method comparison.

Q: What accuracy level should teams expect from AI species identification tools? A: In well-studied temperate regions with comprehensive training data, expect 80 to 92 percent accuracy for common birds and mammals, 65 to 80 percent for common plants, and 40 to 60 percent for invertebrates and fungi. Accuracy decreases significantly for rare species, juvenile life stages, and tropical ecosystems. Always validate AI classifications against expert review for a subset of records.

Sources

• IPBES. (2025). Global Assessment Report on Biodiversity and Ecosystem Services: 2025 Update. Bonn, Germany: IPBES Secretariat.
• Deiner, K., et al. (2024). "Environmental DNA metabarcoding: detection reliability across taxa and ecosystems." Molecular Ecology, 33(4), 892-910.
• Satellite-Biodiversity Working Group. (2025). "Remote sensing as a proxy for biodiversity: a global validation study." Nature Ecology & Evolution, 9(2), 145-158.
• Zu Ermgassen, S., et al. (2024). "Effectiveness of biodiversity offsets: a global systematic review." Science, 383(6689), 1145-1152.
• Kahl, S., et al. (2025). "BirdNET: Performance validation across acoustic landscapes and biogeographic regions." Methods in Ecology and Evolution, 16(1), 78-94.
• TNFD. (2024). Recommendations of the Taskforce on Nature-related Financial Disclosures: Implementation Guidance. Geneva: TNFD.
• World Economic Forum. (2025). Nature Risk Rising: Updated Assessment of Nature-Dependent Economic Value. Geneva: WEF.
• US Army Corps of Engineers. (2024). National Wetland Mitigation Banking Study: 20-Year Performance Review. Washington, DC: USACE.