Sampling biases shape our view of the natural world

Massive study describes extensive taxonomic and geographic biases in available species occurrence data in terrestrial and marine realms, highlighting need for increased sampling in less developed, distant and inaccessible regions of the world

Sturnus vulgaris
A murmuration of starlings (Sturnus vulgaris Linnaeus, 1758) observed near Diss, South Norfolk, UK by Wendy Herniman (CC BY 4.0)

Our knowledge of biodiversity and ability to protect life on Earth is based on data from species observations and collection specimens. Biases in this data, however, limits the analyses possible, and resulting perspectives will be determined by the quality of the available data.

In a comprehensive global analysis of more than 740 million species occurrences from GBIF and OBIS, researchers explored spatial and taxonomic coverage of a representative selection of vertebrate and invertebrate groups to identify biases and their potential drivers.

The results showed that at a five km resolution, less than seven per cent of the Earth's surface—five percent of oceans and 11 per cent of land—has been sampled. When disregarding birds, the coverage dropped to just four per cent of oceans and seven per cent of land.

In terms of taxonomic bias, less than half a per cent of genera in OBIS data and a mere 100 bird species in GBIF, accounted for more than 50 per cent of all records, respectively. Geographically, nearly 80 per cent of GBIF data came from just 10 countries, and coverage was strongly correlated with GDP per capita.

The study also found that eighty per cent of all records were within 2.5 km of roads. Between 22 (for mammals) and 47 (for arachnids) per cent of records were found within 1 km of a city. Despite representing only nine per cent of land, temperate broadleaf forests contained around 50 per cent of terrestrial records.

In the marine realm, 18 per cent of records and 41 per cent of species coincided with the busiest shipping routes covering only two per cent of the oceans. Representing 65 per cent of the total area, open oceans had 32 per cent of all marine records.

The study calls for increased support for museum specimen digitization, strategic standardized surveys and data sharing—highlighting GBIF initiatives such as BID and BIFA as examples—to improve coverage in less-accessible areas, such as high mountains and deep seas, where citizen science approaches are insufficient.

Hughes AC, Orr MC, Ma K, Costello MJ, Waller J, Provoost P, Yang Q, Zhu C and Qiao H (2021) Sampling biases shape our view of the natural world. Ecography. Wiley 44(9): 1259–1269. Available at: