Seed germination of a South American forest tree described by linear thermal time models

Thermal time models may describe and compare seed germination, providing information useful to explain species distribution. However, the capacity of such threshold models to describe germination of tropical native species has been less studied. We evaluated seed germination of a legume tree species (Peltophorum dubium), typical from South American seasonal forests, as described by two linear thermal time models: probit model and graphic model. Seeds were provided from four different provenances in a latitudinal gradient in Brazil, and their physical dormancy mechanically released before the trials. Graphic model and probit regression were used to determine thermal parameters (cardinal temperatures and thermal time requirement) on sub- and supra-optimal ranges for the different seed provenances. Germination rate mainly followed linear relationship with temperature, and regression lines of different germination fractions converged base temperature in the x-axis. Therefore, probit model assumed a single-value of base temperature in the sub-optimal range and a normal distribution of thermal time. Base temperature tended to be higher in the Northern provenance, Porto Velho, showing slower germination under cooler temperatures. Supra-optimal temperatures have shown similar thermal time requirements and different values for ceiling temperature, according to model predictions. No clear patterns have been found between seed provenances and thermal time requirement. Both probit and graphic models have provided reasonable predictions of germination times (t10 and t50), except under the coolest temperatures. Probit regression always described at least 70% of seed germination. Thermal time assumptions, linear models applicability and their limitations are discussed.