Plant phenology is difficult to capture in tropical forests compared to other ecosystems (e.g. temperate forest), due to the diversity and asynchrony in phenology among different species and individuals. To understand ecosystem-level tropical leaf phenology dynamics, which are important for many ecosystem processes (e.g. carbon and water cycling), such variations must be accounted for.
We used Unmanned Aerial Vehicles (UAVs) to monitor phenology in tropical moist forest, a 50-ha plot in Barro Colorado Island (BCI) in Panama. Combining use of ground-based crown mapping, UAV image assessment, and a machine learning algorithm, we quantified leaf cover (%) for 2,000-individual tree crowns at weekly to biweekly intervals from 2014 to 2015. Utilizing this crown-level leaf cover dataset, we quantified species-specific summary in leaf phenology of 53 canopy species in BCI.
Results/Conclusions
There were strikingly different seasonal trajectories of different species in the timing of leaf shedding and flushing, despite some similarities in species-level leaf phenology. For example, the majority of deciduous species started to lose their leaves near the end of the wet season in December 2014. However, some species did not start shedding leaves until late February or even April 2015 (e.g. Zanthoxyllum belinzense; Tabeuia rosia). Also, the extent of leaf loss varied widely across species, ranging from 23%(Virola sebifera) to 83%(Dipteryx panamensis). The duration for which species were completely or partially deciduous also varied widely, from 1 month (Sterculia apetala) to 6 months (Cavanillesia platanifolia).
Canopy species also spanned a gradient from highly synchronous species, in which nearly all individuals showed similar phenology, to highly asynchronous species, in which different individuals showed variable phenology patterns. An example of a synchronous species is Zanthoxyllum belinzense: 50 out of 58 individuals (86%) of lost their leaves for 3 months between March and July. In contrast, an example of an asynchronous species is Cordia alliodora: the 12 individuals varied greatly in the extent (30% to 98%) and the length of deciduousness (1 to 6 months).
Our work demonstrates large variation within and among co-occurring tropical tree species in the timing and degree of deciduousness. These results are a step towards understanding variation in tropical forest leaf phenology, which is the first-order control of primary productivity. Future work will combine the patterns shown here with data on plant functional traits, edaphic conditions, and tree-neighborhoods to better understand the mechanisms driving inter- and intraspecific variation in tropical forest leaf phenology.