The extent to which evolution can rescue a species from extinction, or facilitate range expansion, depends critically on the rate, duration, and geographical extent of the evolutionary response to natural selection. Adaptive evolution can occur quickly, but the duration and geographical extent of contemporary evolution in natural systems remains poorly studied. This is particularly true for species with large geographical ranges and for timescales that lie between ‘long-term’ field experiments and the fossil record. Here, we introduce the Virtual Common Garden (VCG) to investigate phenotypic evolution in natural history collections while controlling for phenotypic plasticity in response to local growing conditions. Reconstructing 150 years of evolution in Lythrum salicaria (purple loosestrife) as it invaded North America, we analyze phenology measurements of 3,429 herbarium records, reconstruct growing conditions from more than 12 million local temperature records, and validate predictions across three common gardens spanning 10 degrees of latitude. We find that phenology clines have evolved along parallel climatic gradients, repeatedly throughout the range, during the first century of evolution. Thereafter, the rate of microevolution stalls, recapitulating macroevolutionary stasis observed in the fossil record. Our study demonstrates that preserved specimens are a critical resource for investigating limits to evolution in natural populations. Our results show how natural selection and trade-offs measured in field studies predict adaptive divergence observable in herbarium specimens over 15 decades at a continental scale.

Between August 2016 and December 2017, we analyzed 3,429 digitized herbarium specimens obtained from five sources: (i) the Global Biodiversity Information Facility (GBIF, http://www.gbif.org/), (ii) the Regional Networks of North America Herbaria accessed through the Arizona-Mexico Chapter (http://swbiodiversity.org ), (iii) the New York Botanical Garden (NYBG, http://www.nybg.org/), and (iv) the Database of Vascular Plants of Canada (VASCAN) (20)(42) and v) correspondence with 19 university herbaria to obtain more images (see acknowledgements in main text). We included only specimens with both a full collection date (year, month, day) and location information that could be georeferenced (see Supplementary Methods).

Inflorescence development in L. salicaria occurs acropetally from the base of the stem to the tip of the apical meristem, resulting in three distinct regions of the inflorescence: fruits (basal), flowers, and buds (distal). To capture variation in floral phenology, we measured the length of each segment on each specimen using the segmented tool in ImageJ (43). We did this for the primary meristem, unless it was clearly damaged in which case the longest inflorescence was measured. Unpollinated flowers occur early in the phenology when compatible mates or pollinators are unavailable. As a result, they senesce and fall from the stem, leaving distinctive callouses at the point of attachment that represent the same phenological stage as fully-formed fruits. Using these measurements, we calculated a scale-free phenological stage (φ) for each herbarium specimen i:

Values of φ range from 0 (early) to 1 (late) as a measure of phenological stage at the time of collection. Thus, on any given collection date, specimens with a phenology stage closer to 0 represent phenotypes early in their phenology whereas those with a stage closer to 1 represent phenotypes sampled later in phenological development.

Files include measurement data and code used for analysis and figure generation.