Testing for temporal dependence of pollen cone production in jack pine (Pinus banksiana Lamb.).

Abstract

Nine-year records of the presence or absence of pollen cones (microsporangiate strobili) in 1299 jack pine (Pinus banksiana) trees growing in 38 locations (plots) across northern Ontario, Canada, were tested for temporal independence. The null hypothesis of a zero-order Markov chain was tested with three statistics at the tree level and at the location level. Tree-level test results were generally in agreement. Only ∼5% of the trees had records indicating a significant departure from the null hypothesis. Approximately one-third of the trees carried pollen cones every year or not at all. After adjustment for multiple comparisons, no trees violated the null hypothesis. Location-level tests relied on Monte-Carlo reference distributions. Depending on the test statistic, we found 2-15 locations with significant departures from the null hypothesis. Conditional (on location) temporal independence was accepted as a reasonable working model. A β binomial model described reasonably well the overall frequency distribution of the number of years (out of nine) with pollen cones. The intra-location correlation coefficient of pollen cone production was strong (0.40).