The seasonal and interannual influence of clouds on tropical forest phenology

Background/Question/Methods

Seasonal variation in solar insolation has been shown to control the timing of phenological events in tropical forests, such as the flowering and fruiting of trees.  Particularly in diverse tropical forests, these events have important cascading effects on many organisms that are dependent upon forest functions.  The phenology of seasonally dry tropical forests, which experience a pronounced dry season, is expected to be strongly influenced by precipitation.  In comparison, tropical wet forests, which receive plenty of precipitation throughout the year, should respond strongly to variation in sunlight.  Both experimental and observational studies have shown that increasing cloud cover can limit solar radiation and canopy photosynthesis in tropical forests.  Alternatively, instead of limiting incoming solar radiation, the presence of clouds may have a positive effect on productivity by increasing diffuse radiation, which scatters light more uniformly throughout the canopy.  This study uses a new globally gridded satellite dataset, the NOAA NCDC GridSat-B1 data, which provides visible and infrared reflectance to quantify cloud frequency, to understand how variation in clouds, precipitation, and temperature affect tropical wet forest phenology at Luquillo, Puerto Rico (1992-2007) as opposed to the seasonally dry tropical forest at Barro Colorado Island (BCI), Panama (1987-2009). 

Results/Conclusions

Using regressions with autocorrelated errors, we found that flower production at BCI decreased with cloud cover and increased with maximum temperatures.  At Luquillo, flower production increased with cloud cover and maximum temperatures.  The models without clouds were the worst ranked models based on Akaike’s information criterion (AIC) at BCI, whereas at Luquillo, the models without temperature were the worst ranked models.  The best-ranked models at both BCI and Luquillo included clouds and maximum temperatures.  When regressions were performed on de-seasonalized data to capture year-to-year variations in flower production at BCI, the best model included the positive effect of temperature and rainfall, but no effect of clouds, suggesting that the effect of clouds is primarily seasonal.  Regressions using de-seasonalized data at Luquillo included the negative effects of minimum temperature and rainfall, again suggesting that the role of clouds is seasonal.  In addition to providing evidence for the role of light-limitation on flower production at BCI and diffuse radiation at Luquillo, our results show the importance of temperature on tropical forests, which are not widely thought to be controlled by variation in temperature.