Finding birds in the archives: Expanding recent bird occurrence data with historical observation records

Abstract Over a century's worth of avian occurrence data are preserved in paper records in libraries and archives around the United States. These older records, many generated by citizen scientists, are not typically accessible through widely used observation databases and are thus largely inaccessible to ornithologists. Species occurrence records are valuable as they provide insights into changes in species distributions over time, population dynamics, and ecological relationships. In this work, we ask (1) if it is feasible given typical digitization project labor capacities to extract usable occurrence data from archival collections when digitization is designed with both historical and scientific users in mind, and (2) to what extent archival data can complement existing biodiversity databases, such as the Global Biodiversity Information Facility (GBIF), the Breeding Bird Survey (BBS) and Christmas Bird Count (CBC). We used the Avian Archives of Iowa Online (AAIO), an occurrence dataset generated through a digitization project using paper collections from an academic archive in Iowa, USA. These records contain the entire corpus of the Iowa Ornithologist Union’s records of observations from 1819–2002 along with observations from several prominent ornithologists and complete record books from other bird associations in Iowa. We were able to extract over 2,000 new usable bird occurrence records, including 37 new species for the study area. Academic archives and library digitization projects are typically designed for historians, not scientists. Our analysis revealed that using ecological metadata during digitization of the source records can successfully serve scientific inquiry, but that dates for observations must be handled carefully in the metadata and the focus must be on describing occurrences, not on describing the source documents. We developed 3 recommendations for digitization of historical science records focusing on collaborative approaches, appropriate DarwinCore metadata, and selection of historical records that are more suitable for extracting occurrence data.