Abstract
Seed harvesting from wild plant populations is key for ecological restoration, but it may threaten the persistence of the source populations. Consequently, several countries have set guidelines limiting the proportions of harvestable seeds. However, these guidelines are inconsistent and lack a solid empirical basis. We used matrix population models based on 280 wild plant species, stored in he COMPADRE Plant Matrix Database, to model the demographic consequences of seed harvesting. Current guidelines do not protect populations of annuals and short‐lived perennials because maximal allowed harvest drew all annual species included in our study to extinction. In contrast, current guidelines are overly restrictive for long‐lived plants because these plants could tolerate even higher seed removal than currently allowed. The maximum possible fraction of seed production that can be harvested without compromising the long‐term persistence of populations was strongly related to generation time of the target species. When harvesting seeds every year, the fraction of seeds that was safe to harvest (safe seed fraction) ranged from 80% in long‐lived species to 2% in most annuals. Less frequent seed harvesting substantially increased the safe seed fraction. In the most vulnerable annual species, it was safe to harvest 5%, 10%, or 30% of a population's seed production when harvesting every 2, 5, or 10 years, respectively. Our results provide a quantitative basis for seed harvesting legislation, based on the generation times of species and harvesting regime.