Background and aims: Seed desiccation sensitivity limits the ex situ conservation of up to 47 % of plant species, dependent on habitat. Whilst desirable, empirically determining desiccation tolerance levels in seeds of all species is unrealistic. A probabilistic model for the rapid identification of woody species at high risk of displaying seed desiccation sensitivity is presented.
Methods: The model was developed using binary logistic regression on seed trait data [seed mass, moisture content, seed coat ratio (SCR) and rainfall in the month of seed dispersal] for 104 species from 37 families from a semi-deciduous tropical forest in Panamá.
Key results: For the Panamanian species, only seed mass and SCR were significantly related to the response to desiccation, with the desiccation-sensitive seeds being large and having a relatively low SCR (i.e. thin 'seed' coats). Application of this model to a further 38 species, of known seed storage behaviour, from two additional continents and differing vegetation types (dryland Africa and temperate Europe) correctly predicted the response to desiccation in all cases, and resolved conflicting published data for two species (Acer pseudoplatanus and Azadirachta indica).
Conclusions: This model may have application as a decision-making tool in the handling of species of unknown seed storage behaviour in species from three disparate habitats.