Erratum: Large contribution from anthropogenic warming to an emerging North American megadrought (American Association for the Advancement of Science (2020) DOI: 10.1126/science.aaz9600)

Abstract

After the Report, “Large contribution from anthropogenic warming to an emerging North American megadrought” was published, the authors discovered two minor errors in their database of tree-ring width index (RWI) chronologies. They have corrected these errors and repeated all of the calculations in the study; the main text, Figs. 1 to 3, the text in sections 3.1 and 3.3 of the when compared with all reconstructed megadroughts since 800 CE. Additionally, the contribution of anthropogenic climate trends to the 2000–2018 SWNA soil-moisture anomaly was 46%, nearly identical to the originally calculated 47%. Specific details on the corrections to the RWI database are described below. The originally published soil-moisture reconstructions were based on a network of 1586 RWI chronologies from western North America, but 65 of these chronologies should not have been included in the reconstruction. In Mexico, 59 chronologies came from a previous reconstruction effort in 2010. These chronologies had since been updated with new tree-ring records and standardization techniques, which were also included in the database that the authors used. Thus, most of the mistakenly included chronologies were largely duplicates of other more recently developed chronologies that were also used in the reconstructions. Additionally, one of the mistakenly included chronologies was listed with an incorrect location, 3° latitude south of the correct location. Similar duplication mistakes occurred in Nevada and Oregon. In Nevada, five pairs of nearly identical chronologies were included in the database, with each pair comprising results from an older or more recent standardization. In Oregon, one identical chronology appeared twice in the original database. Duplicate chronologies are, in principle, problematic because they double the ability of a single chronology, and/or weaken the ability of other chronologies, to affect the reconstruction at nearby grid locations. All of these duplicate chronologies have been removed from the RWI database, which now comprises 1521 RWI chronologies. Before conducting the soil-moisture reconstruction, the authors performed a forward-reconstruction procedure in which the ends of the RWI chronologies were extended forward in time based on other RWI chronologies in the surrounding area (described in section 3.3 of the materials and methods). This method occasionally assigned negative RWI values in years when the forward reconstruction predicted very small amounts of growth. This affected 89 of the RWI chronologies in the original study. In most RWI standardization approaches, negative values are not allowed. In the revised RWI database, negative RWI values estimated from the forward reconstruction have been set to zero. In addition, the original article’s Acknowledgments section inadvertently omitted acknowledgment of data contributions from J. Littell. This is corrected in the revision.