Primary Biodiversity Data (PBD) are defined as the basic attributes of observations or records of the occurrences of species. PBD is a fundamental concept of biodiversity informatics since it is substantial in quantity and provides the links to organize other large and independent bodies of data concerning species (= taxonomic information) and environments. In fact, PBD is at the core of the exploding field of biodiversity informatics, which in some sense now underlies biogeography, macroecology, landscape ecology and several other subdisciplines of biology.
A principal - and rapidly growing - class of research that can be performed using PBD is the estimation of a species' environmental requirements and the projection of these in both environmental and geographic spaces to estimate niches or distributional ranges, generally by using models of ecological niches and species' distributions (often called ENMs or SDMs, respectively).
The largest point of access to PBD in the world is the Global Biodiversity Information Facility (GBIF), and hundreds of papers have now used GBIF-mediated data to fit and apply ENM/SDM.
Experience has shown that GBIF, like other aggregated data research infrastructures, holds a number of potential problems related to incomplete or difficult access to all the fields in its schema, inconsistent information among fields, or simply erroneous or incomplete data. These drawbacks complicate ENM/SDM analyses considerably, and detract from the enormous scientific value of this information storehouse.
Three overlapping communities participate in GBIF's data process: providers (museums, herbaria, and observer's networks), users (scientists, analysts working for governments, NGOs or the private sector, the public) and the technical staff managing the huge databases, web services and servers at GBIF. Each can play a different role in fixing data issues of GBIF.
Our main recommendations for the GBIF Secretariat are the following:
 GBIF.org should serve indicators of precision, quality, and uncertainty of data that can be calculated practically, and preferably "on the fly", as well as summaries and metrics of completeness of inventories, at scales and for regions defined by the user. The summaries should display maps and graphs of completeness by region, time-period and taxa.
 The implementation of the GBIF information resource should go beyond unique identifiers of queries (DOIs for downloads, including the capability to re-run queries, http://www.gbif.org/publishing-data/summary#supporteddatasettypes), and to include identifiers of the individual data that make up the queried data.
 GBIF.org should include applications or functionalities enabling users to annotate errors or problems, and communicate those changes directly to providers, as it may be practical and appropriate. This point may need to be discussed with providers.
 A procedure enabling users to make accessible versions of their databases that have been improved and annotated should be supported, but this functionality should not lose the vital tie back to the original data records and the actual data provider.
 GBIF should partner with and/or support initiatives to do more for training and guiding users on the proper use of the data; such initiatives should incorporate actual expert uses in ENM/SDM to assure that current best practices are followed.