Bernacci L.C.,Instituto Agronomico Of Campinas Iac |
De Siqueira M.F.,Institute Pesquisas Do Jardim Botanico Do Rio Of Janeiro Jbrj |
Rocha F.S.,Federal Rural University of Rio de Janeiro
Natureza a Conservacao | Year: 2012
Biological collections evoke contrasting feelings for being such a vast source of biodiversity data which is prone to all sorts of errors and uncertainties. The situation is not different for Brazilian herbaria, currently sharing more than two million easily accessible records on the Web. Properly dealing with this reality is a crucial task when using this kind of data for ecological niche modelling (ENM), so that errors and uncertainties do not generate misleading results in conservation. Here we investigate some of the issues that can be found in herbarium specimen data, describing a set of automatic procedures that can be used for a prior selection of records for ENM. In total, 11531 records for 135 species of Passifloraceae that natively occur in Brazil were analyzed considering different spatial resolutions, ranging from 30 arc-seconds to 10 arc-minutes. After applying the procedures, the proportion of spatially unique records was 9.3% for the highest resolution considering all species, with an average number of 8 records selected per species. These numbers increased to 17% and 16, respectively, for all other resolutions. This scenario highlights the importance of using data quality filters and further developing ENM presence-only methods that can work with a low number of records per species. Automatic procedures still cannot discard expert review, but they can greatly facilitate it by drawing attention to a much smaller number of records potentially useful for ENM. Most of the data quality procedures described here can also be applied to other taxonomic groups, regions and specimen data sources. © 2012 ABECO.