A new study from University of Liverpool researchers reveals that the species hardest to detect—those rarely seen, recorded, or included in scientific monitoring—are also the most vulnerable to human-driven habitat change.
The findings published in Global Ecology and Biogeography suggest that because these elusive species are underrepresented in global biodiversity databases, current biodiversity indicators likely underestimate the true scale of biodiversity loss.
The study found that:
- Hard-to-sample species decline more sharply as land-use intensity increases.
- Intensive agriculture may support only 18% of original biodiversity, compared to previous estimates of 47% that omit poorly recorded species.
- A new statistical method corrects for ‘recordability bias,’ allowing biodiversity indicators to better represent the full range of species affected by human activity.
Addressing data gaps
Land-use change is a major driver of biodiversity decline, yet many of the species most at risk—particularly those infrequently recorded—are missing from the datasets that underpin global biodiversity metrics. This research shows that excluding these elusive species leads to substantial underestimation of biodiversity loss.
Importantly, the study demonstrates that this gap can be corrected without new fieldwork. By leveraging open, crowd-sourced data from around the world, researchers can make biodiversity indicators more accurate and more representative—an essential step for effective conservation policy and for tracking progress toward international biodiversity targets.
About the study
The team combined abundance data from the PREDICTS (Projecting Responses of Ecological Diversity in Changing Terrestrial Systems) database with occurrence records from the Global Biodiversity Information Facility (GBIF). Together, these resources include scientific surveys and millions of open, crowd-sourced biodiversity observations, many contributed by the public through platforms such as iNaturalist.
Using a statistical approach, researchers estimated how less-recorded species respond to different land uses. They analysed more than 4,000 species of plants, birds, and spiders from PREDICTS and applied the model to over 270,000 species worldwide with occurrence records in GBIF, covering data from 1600 to 2023. The findings suggest far greater biodiversity loss in heavily modified landscapes than previously recognised.
iNaturalist alone has generated more than 118 million verified observations and is one of the largest contributors of open biodiversity data. As iNaturalist Executive Director Scott Loarie notes: “iNaturalist is unique as a GBIF publisher because it has generated tens of millions of records distributed across hundreds of thousands of species. This is important from a conservation perspective because, unless we rely on a relatively small number of species such as birds (~11k species globally) as proxies for global biodiversity, reducing extinction rates is going to require insights across a large portion of the ~2M named species.”
“Correcting for biases in existing datasets gives us a clearer picture of biodiversity loss and helps design more effective recovery strategies,” said the University of Liverpool’s Dr Claudia Gutierrez-Arellano, Corresponding Author of the study. “This work shows how vital open data and public participation in science are for revealing species that would otherwise remain invisible.”
Next steps
The researchers recommend that other biodiversity indicators adopt similar extrapolation techniques to ensure conservation policies reflect the true magnitude of biodiversity decline. By incorporating hard-to-detect species, global biodiversity metrics can more accurately guide efforts to protect and restore ecosystems.
The full study Hard-to-Sample Species Are More Sensitive to Land-Use Change: Implications for Global Biodiversity Metrics is available here: Hard‐To‐Sample Species Are More Sensitive to Land‐Use Change: Implications for Global Biodiversity Metrics – Gutiérrez‐Arellano – 2025 – Global Ecology and Biogeography – Wiley Online Library