Final Report Summary - GEM-TRAIT (GEM-TRAIT: The Global Ecosystems Monitoring and Trait Study: a novel approach to quantifying the role of biodiversity in the functioning and future of tropical forests.)
The GEM-TRAIT project directly addressed one of the greatest challenges in global change science: how will the terrestrial biosphere respond to global atmospheric change and, more specifically, how does the biodiversity of the biosphere moderate or affect that response? The project focused on tropical forests and savannas, home to over 60% of global biodiversity and global productivity.
It approached the challenge by focussing first on the lack of data on ecosystem function and functional diversity from tropical biomes, and then tackling some of the challenges of using these data to develop theory, modelling and remote sensing approaches.
To collect the data on ecosystem functioning, the first focus of the project was to collect and coordinate a global network of intensive ecosystem productivity and carbon cycle measurements, the GEM network (GEM: gem.tropicalforests.ox.ac.uk). This network collected data, at monthly resolution, from 84 plots in 17 locations across the tropics. This was a huge coordinated global effort: at each site this involved a team of local technicians or students spending one to two weeks every week in the forest collecting data on tree growth, litter sampling, carbon dioxide flux measurements, root growth studies and many other aspects. Sites ranged from the slopes of the Andes mountains, through the forests and savannas of Bolivia and Brazil, to the rarely studies forests of West and Central Africa, the fragmented forest landscapes of Borneo, the forests of NE Australia and on a coral atoll in the South Pacific. Overall this constitutes by far the largest dataset on tropical ecosystem and function, enhancing the total available data for the tropics by an order of magnitude.
This monitoring period also coincided with the El Niño event of 2015/16, the strongest such event for at least two decades. This event caused major disruption of the global tropical carbon cycle, and this project was in the fortuitous position of being able to track the detailed impacts of this event across all out sites, providing detailed and global insights into the processes involved in turning the tropics into a carbon source during such events. These insights are being used to inform the next generation of biosphere model development.
Our second major data collection effort was to describe the functional biodiversity of the trees in the sites across the network. The approach we took here was to conduct intensive field campaigns, typically 3-6 months in duration, involving up to around 30 people. Typically this involved tree climbers sampling leaves of the major trees from the upper canopy, and then the leaves and wood being sampled for structure, photosynthesis, chemistry and many other features. We first tried our methodologies in a six-month campaign in Peru in 2013, before extending them to Brazil in 2014, Australia and Ghana and Malaysia in 2016, and Gabon and Polynesia in 2017 In total we sampled ten locations across the tropics in five major biogeographical regions (Americas, Africa, SE Asia, Australia and Oceania), collecting leaves from 4014 trees and over 1231 species This is by far the largest leaf functional traits dataset from the tropics.
Beyond this major data collection and databasing effort, our next challenge was to integrate the function and diversity data to test and assist theory and model development. In this effort we initially focused our attention on our elevation gradient in Peru (our earliest and most advanced dataset). Our expectation has been that productivity would decline with elevation in Peru because of lower temperatures. However, our results showed a striking result that, because of the change in species composition, there was very little change in productivity with elevation, i.e. in the long term tropical ecosystems show much less sensitivity to temperature because the species composition of the community shifts to adapt to the local environment. We were able to demonstrate this both through theory and model development. We have now begun to apply a similar analysis to our other data, such as the wet-dry gradients in Ghana and Brazil, to understand whether a similar pattern can be seen in understand how ecosystems respond to drought.
The third key strand of this project was to explore the potential of multispectral and hyperspectral remote sensing approaches to be able to map and detect canopy leaf traits and look at shifts over time. We have demonstrated that such hyperspectral approaches do work (using aircraft-collected data in Peru and drone-collected data in Ghana), and how now exploring the potential of the multispectral data such as from the new Sentinel-2 satellite to map such traits, and changes in traits, at a much larger scale.
Overall, this project has been a success. Our data collection efforts have exceeded what we initially proposed, and theory and model development approaches have been successfully developed and tested. We are making the data collected widely available through an in-house database, and I expect that both my team and the wider scientific community will continue to work with this large dataset for many years to come.
It approached the challenge by focussing first on the lack of data on ecosystem function and functional diversity from tropical biomes, and then tackling some of the challenges of using these data to develop theory, modelling and remote sensing approaches.
To collect the data on ecosystem functioning, the first focus of the project was to collect and coordinate a global network of intensive ecosystem productivity and carbon cycle measurements, the GEM network (GEM: gem.tropicalforests.ox.ac.uk). This network collected data, at monthly resolution, from 84 plots in 17 locations across the tropics. This was a huge coordinated global effort: at each site this involved a team of local technicians or students spending one to two weeks every week in the forest collecting data on tree growth, litter sampling, carbon dioxide flux measurements, root growth studies and many other aspects. Sites ranged from the slopes of the Andes mountains, through the forests and savannas of Bolivia and Brazil, to the rarely studies forests of West and Central Africa, the fragmented forest landscapes of Borneo, the forests of NE Australia and on a coral atoll in the South Pacific. Overall this constitutes by far the largest dataset on tropical ecosystem and function, enhancing the total available data for the tropics by an order of magnitude.
This monitoring period also coincided with the El Niño event of 2015/16, the strongest such event for at least two decades. This event caused major disruption of the global tropical carbon cycle, and this project was in the fortuitous position of being able to track the detailed impacts of this event across all out sites, providing detailed and global insights into the processes involved in turning the tropics into a carbon source during such events. These insights are being used to inform the next generation of biosphere model development.
Our second major data collection effort was to describe the functional biodiversity of the trees in the sites across the network. The approach we took here was to conduct intensive field campaigns, typically 3-6 months in duration, involving up to around 30 people. Typically this involved tree climbers sampling leaves of the major trees from the upper canopy, and then the leaves and wood being sampled for structure, photosynthesis, chemistry and many other features. We first tried our methodologies in a six-month campaign in Peru in 2013, before extending them to Brazil in 2014, Australia and Ghana and Malaysia in 2016, and Gabon and Polynesia in 2017 In total we sampled ten locations across the tropics in five major biogeographical regions (Americas, Africa, SE Asia, Australia and Oceania), collecting leaves from 4014 trees and over 1231 species This is by far the largest leaf functional traits dataset from the tropics.
Beyond this major data collection and databasing effort, our next challenge was to integrate the function and diversity data to test and assist theory and model development. In this effort we initially focused our attention on our elevation gradient in Peru (our earliest and most advanced dataset). Our expectation has been that productivity would decline with elevation in Peru because of lower temperatures. However, our results showed a striking result that, because of the change in species composition, there was very little change in productivity with elevation, i.e. in the long term tropical ecosystems show much less sensitivity to temperature because the species composition of the community shifts to adapt to the local environment. We were able to demonstrate this both through theory and model development. We have now begun to apply a similar analysis to our other data, such as the wet-dry gradients in Ghana and Brazil, to understand whether a similar pattern can be seen in understand how ecosystems respond to drought.
The third key strand of this project was to explore the potential of multispectral and hyperspectral remote sensing approaches to be able to map and detect canopy leaf traits and look at shifts over time. We have demonstrated that such hyperspectral approaches do work (using aircraft-collected data in Peru and drone-collected data in Ghana), and how now exploring the potential of the multispectral data such as from the new Sentinel-2 satellite to map such traits, and changes in traits, at a much larger scale.
Overall, this project has been a success. Our data collection efforts have exceeded what we initially proposed, and theory and model development approaches have been successfully developed and tested. We are making the data collected widely available through an in-house database, and I expect that both my team and the wider scientific community will continue to work with this large dataset for many years to come.