Question: Are flower production and associated phenological variables (onset, end, duration, and three measures of flowering synchrony) randomly distributed in space or, alternatively, is there a neighbourhood structure (spatial autocorrelation) in the values of these variables? To which extent does spatial autocorrelation affect the correlation tests between phenological traits?

Location: A tree savanna reserve in Southeastern Brazil (22°15′ S, 47°08′ W).

Methods: The flowering season of Chromolaena odorata was followed for all (96) individuals in a completely mapped area of 3000 m². The phenological traits were estimated by counting flower heads in anthesis on individual plants every seven days for 14 weeks.

Results: Flowering time was unimodally distributed, but with different peak dates depending on whether individual flower heads or plants were counted. Three phenological traits and canopy closure above the plants showed some degree of spatial autocorrelation, which caused loss of up to 35% of degrees of freedom in nine of 36 correlation tests. Such a decrease in the degrees of freedom resulted in loss of significance for correlations in three pairs of variables.

Conclusions: We hypothesize that the spatial autocorrelation in phenological traits between C. odorata neighbours may be driven by genetic similarity among neighbouring plants and/or spatial structuring of environmental factors. Because location and distance between samples may affect their statistical independence, we suggest that spatial autocorrelation should be taken into account in future studies of plant phenology, e.g. by using effective sample size in statistical tests.

Abbreviations: SA = Spatial autocorrelation; S_AG= Augspurger's (1983) synchrony index; S_M = Marquis' (1988) synchrony index; S_EL-K = English-Loeb & Karban's (1992) synchrony index.