1. It is often assumed that long-lived woody perennials with canopy-stored seed banks steadily accumulate seeds over time since fire. Trends in flowering and fruiting have usually been inferred from synchronic studies of sites of different post-fire ages or counting stored seeds in apparent age classes, and mostly in obligate-seeder species. Long-term longitudinal studies on a broader range of species are needed to fully understand the dynamics of flower and fruit production and accumulation of viable seeds. Key questions are: What is the annual trend in flowering and cone production? Is this matched by accumulation of cones and seeds? How is seed germinability related to cone age at time of a fire?

2. We counted inflorescences and tagged cones produced annually by the resprouting shrub Banksia spinulosa in 315 plants over 13 years at one site and in 46 plants over 20 years at another. At the end of the study, we harvested all accumulated cones, burned them and assessed seed viability using germination trials.

3. We detected enormous inter-plant variation in reproductive effort and output. 50% of inflorescences were produced by only 10-15% of plants. There was potential for accumulation of massive seed banks. However, (i) only 8-10% of inflorescences became cones; (ii) only 44-50 % of these were retained until harvest; (iii) many retained cones suffered seed predation; and (iv) viability of retained seeds declined with cone age. The result of these processes meant that the accumulated seed bank was only 2 to 4 seeds per plant, 82% to 94% of the viable seeds had been produced in 6 years prior to harvest, and only 12% to 26% of plants contributed to this viable seed bank.

4. Synthesis: Cone and seed losses and declining seed viability in Banksia spinulosa mean that almost all viable seeds come from the past few years of flowering even though apparently intact cones may be retained for decades. If this is typical for resprouting serotinous shrubs, it is important to understand trends in flowering and fruit set over time because the magnitude of recruitment will depend on fecundity in the few years prior to a fire.

We established a long-term study using two populations of Banksia spinulosa var. spinulosa in south eastern Australia: Barren Grounds Nature Reserve (34° 40’ S, 150° 43’ E) and Kioloa State Forest (35° 34’ S, 150° 19’ E). At Barren Grounds (BG), a random sample of 49 mature plants was selected from a population of about 1,000. The Kioloa State Forest population (KSF) was added nearly 10 years later and comprised a sample of 349 mature plants from within a similarly sized population. All study plants were tagged and mapped. Towards the end of the flowering season each year we counted the numbers of new inflorescences and the numbers of cones resulting from the previous year’s flowering. Inflorescences that failed to produce any follicles were removed to simplify future censuses and each cone was tagged with a stainless-steel tag identifying the flowering year. We harvested all accumulated cones on the BG plants in December 2005 and on the KSF plants in mid 2006.

No data on numbers of cones were recorded for the 1995 flowering season at KSF, though flowering data were recorded in all years.

The timing of this harvest for BG meant that the most recent cones in the sample were from the 2004 flowering season; the mid-2006 harvest for KSF captured the fruit production from the 2005 lowering season.

By the time of harvest, one of the 49 BG plants had died, which we excluded from the data set. There were two instances of plants initially tagged as discrete individuals but later discovered to be emanating from the same lignotuber; the data from these pairs of plants were combined, thus producing a total data set of 46 plants.  At KSF, there were 315 extant plants at the time of the harvest. Losses of initially tagged plants were attributable to three main causes: (i) plants initially tagged as discrete individuals but later discovered to be emanating from the same lignotuber – these were merged in the data set; (ii) track-widening works at KSF resulting in the removal of about 20 plants near the edges of two of the plots; and (iii) crushing of plants by large tree trunks or branches falling in storms. The latter plants could not contribute to flowering and fruiting for several years although most eventually resprouted from their lignotubers. We excluded plants within all three categories from the final data set.