Nitrogen limitation in the Earth System: merging top-down and bottom-up approaches to modelling land surface processes

Nitrogen limitation in the Earth System: merging top-down and bottom-up approaches to modelling land surface processes

How does nitrogen availability limit carbon sequestration over long time scales, in real and model ecosystems, and for a changing global climate system?

Background

The terrestrial carbon sink has absorbed 1/3 of the anthropogenic carbon dioxide (CO2) emitted since the start of the Industrial Revolution, primarily through increased primary production. However, nitrogen (N) availability constrains primary productivity in many temperate and boreal ecosystems. The extent to which N might limit the ability of terrestrial ecosystems to sequester current, and in particular, future CO2 emissions remains highly uncertain. To predict future terrestrial carbon storage capacity, and to identify which ecosystems are most likely to store carbon without N limitation, we need to know where N limitation occurs, over what time scale and how this is affected by ecological and environmental contexts. Dendro-ecological records provide a largely untapped yet powerful resource for information on long-term plant-N interactions and how these have varied under different climatic and biotic contexts. This project proposal is to develop novel model-data assimilation techniques to entrain these  measurements, and critically improve the representation of nitrogen limitation within global land surface models with longterm observations from existing tree-ring records spanning the last few decades.

Aims of the Project

Use dendro-ecological data to constrain contemporary land surface models, and assess the implications of such constraint on future projections of terrestrial carbon stores.

Methods to be used

The main computer model, on to which we will map our findings, is JULES (Joint UK Land Environment Simulator). JULES is the primary UK land surface model, and a version of it is contained within the Met Office climate simulation framework (UKESM). Recent enhancements include an initial representation of the Nitrogen cycle, and it is these components that this project will inform as to their suitability, accuracy and parameterisation.

Novel statistical learning techniques will be employed to derive mechanistic equations from point-scale and employing available dendro-ecological data. These methods will be extended to capture ecological processes operating across multiple individual study sites –  to find emergent properties when aggregating to larger spatial scales. These data-led findings will be compared to the N-cycling part of the JULES model.

CEH (Chris Huntingford) maintains the IMOGEN global impacts model. This simulation framework emulates the full suite of climate models in the UN IPCC CMIP5 database. Within IMOGEN is JULES, plus a full global carbon cycle, and so this prediction framework will show the global implications of adjusting the Nitrogen cycle. Although IMOGEN operates globally, it is computationally fast, and is therefore used frequently to prototype new land surface configurations before eventual implementation in the Met Office climate model. This can include a range of alternative N cycle representations in JULES.

It is worth noting, therefore, that this proposal opens the remarkable opportunity to characterise a critical piece of the Earth System. Furthermore, it sets out a feasible pathway from measurements right through to performing global climate model simulations.

Project Description

The standard method of projecting how climate will change into the future is through the use of Earth System Models (ESMs). These estimate the statistics of meteorological change expected for different potential levels of future atmospheric greenhouse gas concentrations. ESMs have capitalised on nearly 60 years of research in weather forecasting, and so most features of atmospheric dynamics are calibrated well. However, on the decadal-centennial timescale, global environment attributes that can be considered fixed for weather forecasts of just days ahead (e.g. ice-cover), are instead evolving. Hence, more recently, these other parts of the Earth system have needed to be modelled, and this includes vegetation carbon stores. With less lead time of development, these components still need substantial advancement. Besides being an integral part of the coupled climate-carbon cycle system, there is much policy interest as to whether terrestrial carbon stores can continue to offset ~33% of emissions. Furthermore, policymakers ask whether deliberate land-use change could even increase current CO2 offset level, to raise the chances that climate change can be constrained to key thermal thresholds such as two degrees of global warming.

A fundamental uncertainty in modelling terrestrial ecosystem stores is the extent to which the N (nitrogen) availability impacts on vegetation carbon cycling. The JULES land surface model has only recently included a framework to describe this process, but it is only in provisional form, and many parameters are little more than initial estimates. Yet one form of data that has huge potential to inform the calibration of the N part of JULES are dendro-ecological measurements of plant biomass and N availability dynamics. Mechanistic modelling of these data have been used to infer the extent to which N availability constrains biomass dynamics and to identify the presence of plant-soil feedbacks, but to date, such research has only been operating on local scales. A fundamental question is “How generalizable, to global scales, are the insights gleaned from studies at smaller scales?”. The student will answer that question by using novel AI-based approaches to infer the equations that best describe plant-N relationships at the regional to continental-scale from existing dendro-ecological time-series data. They will also uncover the biological and physical covariates that best explain differences in plant-N relationships and will translate these insights into improved representation of biogeochemical cycling in the JULES land surface model.

This initiative will generate multiple outputs. However of particular interest is that it will contribute towards an overarching assessment of whether N-effects will downregulate future fertilisation as atmospheric CO2 rises. Hence it will aid improvement in models of global climate-carbon cycle interactions.

There is a strong training element to this project. The student will be exposed to the full end-to-end scientific procedure of using models to infer ecological processes from data, through to performing projections of large-scale geochemical cycle change under climate change. They will have access to existing dendroecological datasets and the opportunity – if they choose – to collect additional records where required. Coaching will be provided on the modelling and data analysis methods to be employed in the study. The student will join the Ecosystem Ecology group in the Department of Zoology and be a part of the Oxford Long Term Ecology Laboratory. They will also benefit from access to nearby Centre for Ecology and Hydrology, Wallingford, as well as visits to Exeter University and the Meteorological Office (also in Exeter).

Specialised skills required

The student will encounter quite substantial involvement with statistical and process modelling, with attendant calculus. Hence a minimum of an Alevel in mathematics is advisable. Ideally the student will also have some experience of – or be willing to learn – developing, testing,  managing and operating computer code in FORTRAN and/or Python.

If you are interested in this project please email elizabeth.jeffers@zoo.ox.ac.uk in the first instance