Species occurrence information can help inform many National Park resource management decisions. What species of concern occur, or potentially occur, in a project footprint, and need to be addressed in planning documents? What sensitive or invasive species occurred in a fire polygon, which can inform post-burn recovery efforts? Where in a park does species x occur? Is an observation of species y unusual? NPS has several datasets such as the evidence data behind park species lists, vegetation plot data, bird point counts, and stream macroinvertebrates, that contain species occurrence records. Large biodiversity portals such as gbif, iDigBio, eBird, and citizen science initiatives such as iNaturalist and BioBlitzes, contain over 500 million occurrence records potentially relevant to park management. If these existing occurrence records can be leveraged to address park management issues, NPS can focus funding on filling in gaps in the existing data. Substantial data querying, de-duplicating, checking, and cleaning is necessary, and good tools exist for those steps.
Results/Conclusions
A more fundamental issue is that these records were generated by fieldwork conducted for specific purposes, and suitability for those purposes does not imply suitability for other purposes or questions or spatial scales. The properties of occurrence data differ widely, depending on how and why they were collected. The obvious differences are presence-only vs presence/absence, and random vs convenience or targeted locations. But less obvious differences matter for suitability for many questions. Most multi-species surveys record only the first individual of each species in the surveyed area. Therefore, counts of records from surveys, iNaturalist, and BioBlitzes do not reflect relative abundance. Further, common species tend to be first encountered and thus recorded nearer access points or trailheads, even if they are abundant in the backcountry. Therefore, occurrences from surveys are problematic for questions at scales smaller than the survey unit. Vegetation plots aimed at characterizing vegetation classes often are non-randomly located, but also shifted short distances in the field to exclude either sensitive species or clumps of non-native invasive species. This has slightly different implications than simple record masking or coordinate fuzzing for sensitive species.
Valid answers to specific questions require careful consideration of data suitability issues, especially at small spatial scales. The Events tables in eBird and gbif Darwin Core make species occurrence records suitable for addressing a wider set of questions, but more information about the data collection process is needed for other questions.