Improving data access democratizes and diversifies science

Abstract

The foundation of the scientific method rests on access to data, and yet such access is often restricted or costly. We investigate how improved data access shifts the quantity, quality, and diversity of scientific research. We examine the impact of reductions in cost and sharing restrictions for satellite imagery data from NASA's Landsat program (the longest record of remote-sensing observations of the Earth) on academic science using a sample of about 24,000 Landsat publications by over 34,000 authors matched to almost 3,000 unique study locations. Analyses show that improved access had a substantial and positive effect on the quantity and quality of Landsat-enabled science. Improved data access also democratizes science by disproportionately helping scientists from the developing world and lower-ranked institutions to publish using Landsat data. This democratization in turn increases the geographic and topical diversity of Landsat-enabled research. Scientists who start using Landsat data after access is improved tend to focus on previously understudied regions close to their home location and introduce novel research topics. These findings suggest that policies that improve access to valuable scientific data may promote scientific progress, reduce inequality among scientists, and increase the diversity of scientific research.

Keywords: Landsat; data access; diversity; inequality; science of science.

Fig. 1.

Landsat-related publications before, during, and after the Landsat commercialization era. This figure shows the number of Landsat publications over time for three different types of publications. In all three panels, the bars in blue to the right of the vertical dashed line indicate publications after the Landsat program was transferred back to the US government. (A) All publications, (B) publications with 100 or more cites as of 2017, and (C) all publications in about 80 journals that represent the top two percentiles of journals ranked by citation score metrics. In A, the dashed line shows the general trend for all earth and environmental science publications. In all three panels, note that trends in the number of publications are mostly steady during the commercial era, after which there is a rapid increase in publications in the open data era.

Fig. 2.

How data access affects who participates in Landsat research. This figure explores the effects of lowering costs of data access on authors’ locations. (A) A map where each light gray dot represents the presence of at least one author institution that has published a paper using Landsat data before data access costs were reduced. The dots in black represent locations where an author institution published a paper using Landsat data only after data access costs were reduced. A graph depicting this change is found in SI Appendix, Fig. S6. (B) Total number of Landsat publications separated by institutional rank (top 50 vs. 50 to 200) as per the Quacquarelli Symonds (QS) World top university rankings. (C) Total number of publications separated by the authors’ country income categories. For publications with authors from different country income groups, we sort the publication based on the minimum country income group. Overall, the data suggest that lowering costs of data access was particularly helpful for authors in lower-ranked institutions and in non–high-income countries.

Fig. 3.

How data access affects study locations. This figure explores the effects of lowering costs of data access on study locations. (A) A map where each light gray dot represents at least one Landsat publication that studies the region before data access costs were reduced. The dots in black represent at least one Landsat publication that studies the region only after data access costs were reduced. (B) Total number of unique locations studied by Landsat publications separated by country income groups. (C) The distribution of distances between authors and study locations for all study locations that had not been explored in the commercial era and are located outside the United States. A version of C that includes US locations is in SI Appendix, Fig. S8. Overall, these findings suggest that easing data access restrictions particularly helped increase the number and range of study locations.

Fig. 4.

Topical diversity in Landsat science. (A) The total number of first-time abstract words used in Landsat publications. (B) The distribution of the relatedness index by incumbent (dark) versus newcomer (light) authors. The higher the value of the index, the more related a focal word is to other newly introduced words. The distribution of newcomer words is clearly shifted to the right, which implies that new words introduced by newcomers are more likely to be related to other words introduced by newcomers (compared to new words introduced by incumbents).