Final Report Summary - INTRABASIS (Development and application of an individual trait-based simulator of Amazon forest dynamics)
Permanent Amazonian forest-plot censuses have revealed large-scale spatial patterns in forest structure and dynamism, with western forests having up to a three times greater rate of aboveground biomass production. The possible causes for this variation could be the spatial distribution of tree species composition or the climatic and edaphic gradients. Exploring these spatial patterns is important to both:
(a) understand the way diversity and carbon-cycle processes interact; and
(b) project the vulnerability of those ecosystems to climate change.
In INTRABASIS, our main aim was to develop a vegetation dynamic model to simulate the inherent heterogeneity found within Amazonian rain forests, following a data-constrained approach and using information from an extensive network of permanent forest plots (RAINFOR).
Trait-based tropical forest simulator (TFS) incorporates two innovative components. Firstly, it simulates the growth of each tree in a stand using an individual-based modelling architecture. Each tree measured in the RAINFOR plots, is fully represented in TFS and directly associated to field observations. Secondly, TFS avoids the use of plant functional types, i.e. it does not group together species with a 'supposed' common response to environmental resources. Functional diversity is represented by implementing an innovative trait randomisation algorithm highly constrained from observations.
We firstly evaluated the ability of the model to realistically simulate water fluxes at the canopy level. At this level, climate data from four eddy flux tower sites were used to force TFS, while observed latent heat fluxes served as the main diagnostic. The model performed well and had an adequate response to variations in climate and soil water availability. Subsequently we validated the model at the tree level. To do that we used data from a subset of 11 intensive measurement RAINFOR plots, where a comprehensive decomposition of carbon fluxes is available. The model was set up for 6 year-long simulations and a mean annual aboveground biomass growth was estimated and compared with observed data. A good agreement between modelled and observed growth rates was achieved. Finally, we used the model to simulate three important stand level parameters, i.e. gross primary productivity (GPP), net primary productivity (NPP) and carbon use efficiency (CUE) in 69 plots. Variations of these parameters were explored along climatic and soil gradients and were in agreement with published studies. GPP showed a positive association to annual precipitation (PA). NPP increased with mean annual temperature (TA), PA and soil fertility (FS). CUE increased with FS but decreased with TA and soil depth. However, apart from the examined environmental axes, these variations could also be driven by biotic factors and their interactions to environmental drivers.
In order to further explore the causes of GPP, NPP and CUE variations we conducted a randomisation exercise. We initially aimed to disentangle the effects of FS on stand level properties. We thus compared simulations along the 69 plots with a set of simulations were linkages between FS and foliage properties were 'bypassed'. These results indicated that accounting for between plot differences in functional diversity is important to accurately simulate GPP, NPP and CUE, and thus incorporating the effects of FS is necessary. Additional findings underline the caution of using PFTs that categorise tree species response to environmental gradients.
The model is currently enhanced to include regeneration and mortality dynamics, and it is linked to other large-scale modelling approaches. Additionally, it is implemented under scenarios of climate change, to project the vulnerability of Amazonian forests. These findings could help to better understand the potential atmosphere-biosphere feedbacks as well as to provide guidelines towards a sustainable land use management of the Amazon basin.
(a) understand the way diversity and carbon-cycle processes interact; and
(b) project the vulnerability of those ecosystems to climate change.
In INTRABASIS, our main aim was to develop a vegetation dynamic model to simulate the inherent heterogeneity found within Amazonian rain forests, following a data-constrained approach and using information from an extensive network of permanent forest plots (RAINFOR).
Trait-based tropical forest simulator (TFS) incorporates two innovative components. Firstly, it simulates the growth of each tree in a stand using an individual-based modelling architecture. Each tree measured in the RAINFOR plots, is fully represented in TFS and directly associated to field observations. Secondly, TFS avoids the use of plant functional types, i.e. it does not group together species with a 'supposed' common response to environmental resources. Functional diversity is represented by implementing an innovative trait randomisation algorithm highly constrained from observations.
We firstly evaluated the ability of the model to realistically simulate water fluxes at the canopy level. At this level, climate data from four eddy flux tower sites were used to force TFS, while observed latent heat fluxes served as the main diagnostic. The model performed well and had an adequate response to variations in climate and soil water availability. Subsequently we validated the model at the tree level. To do that we used data from a subset of 11 intensive measurement RAINFOR plots, where a comprehensive decomposition of carbon fluxes is available. The model was set up for 6 year-long simulations and a mean annual aboveground biomass growth was estimated and compared with observed data. A good agreement between modelled and observed growth rates was achieved. Finally, we used the model to simulate three important stand level parameters, i.e. gross primary productivity (GPP), net primary productivity (NPP) and carbon use efficiency (CUE) in 69 plots. Variations of these parameters were explored along climatic and soil gradients and were in agreement with published studies. GPP showed a positive association to annual precipitation (PA). NPP increased with mean annual temperature (TA), PA and soil fertility (FS). CUE increased with FS but decreased with TA and soil depth. However, apart from the examined environmental axes, these variations could also be driven by biotic factors and their interactions to environmental drivers.
In order to further explore the causes of GPP, NPP and CUE variations we conducted a randomisation exercise. We initially aimed to disentangle the effects of FS on stand level properties. We thus compared simulations along the 69 plots with a set of simulations were linkages between FS and foliage properties were 'bypassed'. These results indicated that accounting for between plot differences in functional diversity is important to accurately simulate GPP, NPP and CUE, and thus incorporating the effects of FS is necessary. Additional findings underline the caution of using PFTs that categorise tree species response to environmental gradients.
The model is currently enhanced to include regeneration and mortality dynamics, and it is linked to other large-scale modelling approaches. Additionally, it is implemented under scenarios of climate change, to project the vulnerability of Amazonian forests. These findings could help to better understand the potential atmosphere-biosphere feedbacks as well as to provide guidelines towards a sustainable land use management of the Amazon basin.