The quality of specimen records in publicly accessible biodiversity databases varies enormously. When using these data, researchers commonly expunge vaguely-georeferenced records before conducting biogeographic analysis. However, discarding these specimens risks underestimating the range and climatic tolerances of species. We evaluated the effect of removing poorly-georeferenced specimens on estimates of range size and climatic niche breadth using a well-curated dataset of 32 species in the genus Asclepias (milkweeds) in Mexico and the US obtained from the TROPICOS plant database. We designated records “accurate” if they could be located with confidence to an area smaller than a county (or equivalent), and “vague” if they could only be locatable to a county. To estimate range size, we calculated the area of the minimum convex polygon for accurate specimens. We then compared this to the area of the minimum convex polygon for the combined set of accurate and vague specimens. To ensure the polygon for all records was conservatively estimated, we geolocated the vague specimens to the point in each county that was closest to the geographic centroid of the polygon from the accurate points. Climatic niche breadth was estimated in a similar manner using the occupied range of mean annual temperature and precipitation.
Results/Conclusions
Excluding vaguely-georeferenced specimen records led to a dramatic underestimation of range size and niche breadth. On average 64% of the records for a species comprised vaguely-georeferenced specimens (range: 2 to 90%). Range size calculated only with accurate records was 30.9% (median) of the range size calculated with accurate plus vague records (range: 0.5 to 98%). Reductions in niche breadth were similar to reductions in range size. Our results demonstrate that researchers should re-consider habitually expunging poorly georeferenced specimens before conducting biogeographic analyses. Even if specimens cannot be accurately geolocated, they can still represent large portions of a species’ range that otherwise have scarce documentation. Furthermore, climatic tolerances can be severely underestimated if vague specimens are expunged, thereby leading to an overestimation of vulnerability to anticipated climate change and responses to past climate change. We encourage researchers to take advantage of the growing suite of methods that can account for spatial uncertainty in analysis using specimen data rather than discarding vaguely-georeferenced yet highly valuable specimen records.