There is a widespread need for reliable biodiversity databases for science and conservation. Among the many public databases available, we lack guidance as to how their data quality varies. Here, we compare species distribution data for a well known regional reef fish fauna extracted from five global online databases that supply “as is” data (GBIF, OBIS, IDigBio, FishNet2 and FishBase) and our own curated regional database (STRI, Smithsonian Tropical Research Institute) using quantitative criteria, and assess how they affect biogeographical analyses. We first describe the databases and quantify overlap between them. We then describe variation in the geographical distributions of species records and species richness and in the completeness of local species lists. Finally, we assess the consequences of using these different databases in biogeographical analyses by comparing patterns of species turnover (beta diversity) and bioregionalization. The databases vary considerably in size and show high overlap in species lists, but low overlap in georeferenced species records. Levels of completeness of local inventories are spatially heterogeneous and low in most databases. Spatial biases produced artefactual variation in patterns of species turnover and delineation of bioregions in all databases. Although not the largest, STRI database has the most complete geographic coverage of data, showed relatively low turnover and the clearest biogeographic regionality. Incorporating data from a wide range of other sources, curating data to reduce errors, and assessing effects of spatial biases in data is critical to obtaining an accurate picture of the geography of biodiversity and its change.

中文翻译：

---

比较生物多样性数据库：以大加勒比礁鱼为例

普遍需要可靠的生物多样性数据库来进行科学和保护。在众多可用的公共数据库中，我们缺乏有关其数据质量如何变化的指导。在这里，我们比较了从五个提供“原样”数据（GBIF，OBIS，IDigBio，FishNet2和FishBase）的全球在线数据库中提取的著名区域珊瑚礁鱼类动物的物种分布数据，以及我们自己策划的区域数据库（STRI，Smithsonian Tropical研究所）使用定量标准，并评估它们如何影响生物地理分析。我们首先描述数据库并量化它们之间的重叠。然后，我们描述物种记录的地理分布和物种丰富度以及本地物种清单的完整性的变化。最后，我们通过比较物种更新（β多样性）和生物区域化的模式来评估在生物地理分析中使用这些不同数据库的后果。数据库的大小差异很大，并且在物种清单中显示出高度重叠，而在地理参考物种记录中则表现出低重叠。在大多数数据库中，本地清单的完整性级别在空间上异质且较低。空间偏见在所有数据库中都在物种更新和生物区划界的模式上产生了人工假象变化。虽然不是最大，但STRI数据库具有最完整的数据地理覆盖范围，显示出相对较低的周转率和最清晰的生物地理区域性。整合来自其他多种来源的数据，整理数据以减少错误，