There is an urgent need to quantify how climate conditions that have no analogy in the past can modify water stress experienced by vegetation and how this can feedback on global carbon and water cycles. In order to do so, this project enhances how current terrestrial biosphere models simulate vegetation functioning and represent spatial heterogeneities.

The results of the project will be instrumental for modelling and understanding future carbon and water cycles in different ecosystems, addressing questions of ecosystem resilience under water stress, as well as to facilitate solutions in water and forestry management problems.

Plant water stress is an emerging issue of global significance because it has been observed to affect places and biomes previously not considered at risk as the tropical forest. It can induce ecosystem transformations and modify the provision of ecosystem services. Notwithstanding the importance of water limitations, there are fundamental gaps in our mechanistic understanding of plant response to water scarcity and on the role played by spatial heterogeneities in ecosystem responses. As a result, predictive capabilities in terms of carbon and water cycle remain unsatisfactory. This project addresses knowledge gaps that are critical for quantitative predictions of vegetation response to water stress across a range of spatial and temporal scales through targeted numerical experiments carried out with an enhanced mechanistic model of the terrestrial biosphere, Tethys-Chloris (T&C), and by means of detailed plant hydraulic observations in a few trees in a newly established NUS ecohydrological open-Lab.

We are currently studying if precipitation and evapotranspiration trends are leading to an acceleration of the water cycles over land. We are analysing by means of mechanistic ecohydrological modelling, reanalysis data, and outputs of General Circulation Models, how climate change modify the residence time of water in the soil, and how this is connected to vegetation water stress (Fig. 1).

Fig. 1. Annual time series of average residence time of water in the top 1 m of soil as computed from ERA-5 reanalysis data and corresponding global map of average residence time.

Then, we will combine a mechanistic terrestrial biosphere model tested in a representative number of locations and ecosystems worldwide with machine learning methods. The idea is to use machine learning to map model parameters anywhere in the globe and then simulating the response for each “grid cell” or large regions globally (Fig. 2). The novelty of this approach is to use machine learning techniques on model simulations (or derived parameters) carried out for a large number of representative sites rather than on observed data directly. Subsequently, we will systematically test the effect of different representation of plant hydraulic parameterizations, which can be fundamental to decide which model component is affecting more model performance and where to invest in terms of model developments and additional field campaigns/measurements.  Finally, we will instrument a few trees at NUS campus with a number of sensors to monitor their hydraulic functioning. This will serve the double purpose of creating a living laboratory where students can be practically introduced to ecohydrological variables and processes and it will provide continuous data on tree water status, which can guide further model developments of plant response to water stress. 

Fig. 2. Description of the alternative way to make global scale predictions of water, carbon and energy fluxes, based on extrapolating model soil and vegetation parameters anywhere globally by means of machine learning techniques informed by parameters calibrated in “control locations” and maps of available covariates.

The project is intended to address knowledge gaps that are critical for quantitative understanding of vegetation dynamics with direct implications for carbon dioxide (CO₂) uptake and storage, resilience of vegetated landscape to water stress, biome transitions, and ultimately the preservation of water resources.

For more details, please contact:
Assoc Prof Simone Fatichi
Email: ceesimo@nus.edu.sg