Fig. 2: Herbarium records exhibit less temporal and spatial bias than iNaturalist observations.

Temporal bias (a) is illustrated as the accumulation of new records over time. Spatial bias (b) is quantified using nearest neighbor index (NNI) which is a measure of spatial autocorrelation. When log transformed, a negative NNI indicates clustering, a positive NNI indicates dispersion, with an NNI of 0 indicating points are randomly distributed in space. NNI was calculated at the species level and only significant (p < 0.05) estimates of NNI were retained for this visualization (n = 2823). Herbarium records exhibited a mean log NNI of −0.88 while the mean for iNaturalist records was significantly (p = 0.006) lower at −0.93 (Table S1). Boxes show the quantiles (Q1-3) with the horizontal line representing the median and the whiskers representing the minima and maxima (calculated as 1.5 times the difference between Q1-3). We also tested for spatial bias (c) by estimating the relationship between density of records and human population density using negative binomial generalized linear models for which we report R² (Kullback–Leibler) and p values (Tables S2-S3).