An Ecoinformatic Analysis of the Effect of Seasonal and Annual Variation in Temperature, Precipitation, and Solar Irradiance on Pollen Productivity in Two Neotropical Forests

Thursday, 18 December 2014
Derek Scott Haselhorst, University of Illinois at Urbana Champaign, Program in Ecology, Evolution and Conservation Biology, Urbana, IL, United States, David K Tcheng, University of Illinois at Urbana Champaign, Illinois Informatics Institute, National Center for Supercomputing Applications, Urbana, IL, United States, Jorge Enrique Moreno, Smithsonian Tropical Research Institute, Balboa, Panama and Surangi W. Punyasena, University of Illinois at Urbana Champaign, Department of Plant Biology, Urbana, IL, United States
Observational data provide a powerful source of information for understanding the phenological response of tropical forests to a changing climate. Annual changes in mean temperature, precipitation, and solar irradiance, in part driven by ENSO cycles, provide a natural experiment. However, these time series are often relatively short (several years to several decades), the average climatic variability experienced in that timeframe is relatively small, and the corresponding response is therefore often very weak. As a result, standard statistical approaches may fail in detecting a biological response.

We present an alternative ecoinformatic analysis that demonstrates the power of weak models in the discovery and interpretation of statistically significant signals in short, noisy, ecological time series. We developed a simple response prediction model that uses cross-validation to explore a landscape of models that correlate the phenological behavior of individual taxa (pollen production, flowering, fruiting) to seasonal and annual mean temperature, precipitation, and solar irradiance using multivariate linear regression. We use a sign slope sensitivity analysis of each linear model that tallies positive and negative slope counts of a taxon’s phenological behavior to our environmental and null variables. We applied this analysis to pollen trap data collected from 1996 to 2006 from two lowland Panamanian forests, Barro Colorado Island and Parque National San Lorenzo. We also tested the performance of our predictive model using published data of annual flowering and fruiting from BCI to corroborate that our approach could reproduce previously published results on tropical phenology.

Our results indicate that although the overall variation in temperature was 3.28 °C over the ten year period, pollen productivity at both sites was most consistently affected by changes in temperature. This result was replicated by the published BCI flower and fruit data, which also increased with increased temperatures, highlighting the significant influence of even subtle changes in temperature for tropical forest communities. We also observed that both pollen and fruit production were negatively correlated with precipitation, suggesting a mechanism for how climate may interfere with pollination success.