FishNet2 (www.fishnet2.net) is a global network of fish collection databases that gives researchers access to  data on roughly 4 million fish species lots, representing over 30 million specimens. In 2012 a collaboration involving investigators and staff associated with fish collections at 12 institutions - Academy of Natural Sciences of Drexel University, California Academy of Sciences, Cornell University, Field Museum, Natural History Museum of Los Angeles County, Tulane University, University of Florida, Illinois Natural History Survey (University of Illinois at Urbana-Champaign), University of Kansas, University of Michigan, University of New Mexico, University of Texas at Austin - was funded to improve the state of geospatial referencing of data within the network.  The specific goals of the project were to georeference (determine latitude and longitude coordinates for) all sampling locality descriptions lacking these data and repatriate the results back to the institutions that provided the data.  Before the project began, only about two-thirds of the records in Fishnet2 were georeferenced.  Investigators at Tulane University began work on the project by harvesting all records within Fishnet2 lacking geographic coordinates and building 12 geographically delimited datasets (regions of the world pre-assigned to each of the collaborating institutions for georeferencing).  Records in need of georeferencing came from 58 Fishnet2 provider institutions, including fish collections at the 12 collaborating institutions. The Collaborative Georeferencing Client (CoGe) of the GEOLocate Platform was used to evaluate the roughly 250,000 localities in FishNet2 lacking coordinates. Each of the twelve collaborating institutions hired a full-time georeferencing technician who, after a brief training period, worked 18 months verifying or correcting the CoGe generated geographic coordinates for localities assigned to his or her institution.  At the completion of the georeferencing phase of the project, 282,221 localities (equivalent to 1.4 million specimen lots) were evaluated and 247,501, localities (equivalent to 1.2 million specimen lots) were assigned coordinates and estimates of uncertainty. We are now in the data repatriation phase of the project, wherein the map coordinates generated in the project are being written to the 58 provider databases.  At the conclusion of the project, it will be possible to map 1.2 million additional specimen occurrences (1/3 of the data in Fishnet2) with a fairly high degree of accuracy.  We also conducted experiments of how well unpaid and minimally trained student volunteers were able to georeference localities compared to experienced project technicians.  The results of the volunteer experiments suggest that minimally trained volunteers are able to georeference specimen localities, within acceptable error distances of results of experienced technicians, high percentages of the time.  The volunteer experiments are especially encouraging considering the very high volume of specimen collection localities in need of georeferencing in U.S. natural history collections.

Last Modified: 11/15/2016
Modified by: Mark Westneat