Recognition of trace element hyperaccumulation based on empirical datasets derived from XRF scanning of herbarium specimens

Background and aims Hyperaccumulation is generally defined as plants exhibiting concentrations of metal(loid)s in their shoots at least an order of magnitude higher than that found in ‘normal’ plants, but this notional threshold appears to have limited statistical underpinning. The advent of massive (handheld) X-ray fluorescence datasets of herbarium specimens makes it increasingly important to accurately define threshold criteria for recognising hyperaccumulation of metal(loid)s such as manganese, cobalt, nickel, zinc, arsenic, selenium, and rare earth elements. Methods We use an extensive dataset of X-ray fluorescence elemental data of ~ 27,000 herbarium specimens together with Inductively Coupled Plasma Atomic Emission Spectroscopy (ICP-AES) elemental data of 1710 specimens to corroborate threshold values for hyperaccumulator plants. The distribution of elemental data was treated as a Gaussian mixture model due to subpopulations within the dataset and sub-populations were clustered in ‘normal’ and ‘hyperaccumulator’ classes. The historical hyperaccumulator thresholds were compared to the concentrations corresponding to the value for which the cumulative distribution function of the Gaussian model of the hyperaccumulator class reaches a probability of 99%. Results Our analysis of X-ray fluorescence data indicates that the historical thresholds for manganese (10,000 µg g −1 ), cobalt (300 µg g −1 ), nickel (1000 µg g −1 ), zinc (3000 µg g −1 ), arsenic (1000 µg g −1 ), and selenium (100 µg g −1 ) are substantially higher than then the concentrations required to have a 99% probability of falling in the hyperaccumulator class at 1210 µg g −1 for manganese, 32 µg g −1 for cobalt, 280 µg g −1 for nickel, 181 µg g −1 for zinc, 8 µg g −1 for arsenic, and 10 µg g −1 for selenium. All of the historical hyperaccumulation thresholds exceed the mean concentration of the hyperaccumulator populations and fall in the far-right tail of the models. Conclusions The historical thresholds for manganese, cobalt, nickel, zinc, arsenic, and selenium are considerably higher than necessary to identify hyperaccumulators. Our findings provide a more precise understanding of the statistical underpinnings of the phenomenon of hyperaccumulation, which will ensure consistency in reporting on these plants.