Lessons from a high-impact observatory: The Hubble Space Telescope's science productivity between 1998 and 2008

Almost two decades of continuous operation of the versatile and productive Hubble Space Telescope (HST) provide uniquely well-documented, robust statistics to study the scientific impact of a major astronomical observatory. We compiled a detailed database of refereed articles that use HST data for analysis and show it to be >95% complete. This HST Publication Database is publicly available and searchable: it contains more than 8700 articles, cited more than 300,000 times in the literature. By cross-linking this data set with our extensive proposal database and NASA's ADS service, we are able to trace the evolution of ideas from the proposal stage through the observations and publication steps to the final impact on the astronomical literature. Here we present a detailed study of HST's performance, including the temporal evolution of the publication rate, the citation statistics, the relative contributions from different program types, the time allocation strategy, and the relative contributions of the HST instruments. We also discuss the properties of typical and very highly-cited articles. By analyzing this complete and well-characterized database, we identify five key features that contribute to the productivity and high impact of the observatory: (1) the time allocation policies; (2) the well-characterized HST archive; (3) the breadth of science projects ranging from the solar system to cosmology; (4) the Director's Discretionary time allocations; (5) the large international user community and its involvement in the observatory's functions. In addition, we find the following general characteristics. Following its launch, HST's productivity has been steadily increasing; 8 yr after launch, HST reached equilibrium between the incoming data volume and the number of published articles that are based on those data. The overall productivity, however, is still steadily increasing due to the increasing number of archival articles. We find that small programs produce more citations per orbit than large programs, but only large programs have the potential to lead to very high-impact articles and data sets with lasting legacy value. We find that while typical HST articles receive the largest number of citations 2-3 yr after publication and exhibit a subsequent decline, the most-cited articles show a qualitatively different citation history. Together these results provide a detailed picture of HST's science productivity and identify key characteristics that contribute to making HST a high-impact observatory.