The project centralized multiple sources of forest inventories covering decades of tropical forest observations across four continents (part of the Global Ecosystem Monitoring – GEM – network), multiple datasets of anatomical, chemical, and physiological leaf and wood traits collected in the inventoried plots, following a single standardized protocol, as well as detailed climatic data covering the period 1970 – 2019. This resulted in reproducible commented R scripts as well as a final dataset that formed the basis of the whole project, a deliverable of the project that will ease future collaborations within and outside the GEM plot network within and across different tropical regions. The resulting database integrates 50 years of biological and environmental data in 102 tropical forest plots (71,291 trees, 3892 species – 649 with functional traits –, 253,159 tree growth values).
The project developed a data analysis framework aimed to deal with the complexity of the questions and data, statistically, while ensuring a transparent set of ecological assumptions were defined to allow causal inference based on observational data, that is, using a formal framework to approach cause-effect relations. This analytical framework is an important feature of the project, as it greatly reduces risks of otherwise frequent problematic statistical biases arising when not differentiating cause-effect relations from other non-causal associations. TropDemTrait therefore developed a theoretical and causal framework responding to the need of a formal and reproducible approach to transparently derive statistical models from a set of interdependent causal assumptions about the studied system, to justify a causal interpretation of model outputs, linking the statistical model to the biological/ecological question of the work. A series of advanced Bayesian growth and survival models were developed, used and compared, to integrate the causal modelling framework and questions of the project into statistical analyses to respond to the project’s questions. This workflow is a product of the project and will hopefully contribute to pushing Global Change Ecology and the timely questions it must address towards increasing reproducibility and more theoretically grounded advances. Simulated increases in climatic anomalies were then combined to the fitted model outputs to derive causal predictions of species and forests demographic responses for varying climate scenarios, average climates, and species phenotypes.
