Welcome! This is the landing page for my PhD research. On it you will find an overview of my PhD, outlines for my (proposed) projects and presentations about my work. Additionally you will find links to other repositories used for this PhD.
I am doing this PhD as part of the Environmental Intelligence CDT* at the University of Exeter, UK. This PhD is financially supported by the Met Office, for whom I work part-time as a research software engineer and to whom I am very grateful. I started this PhD in September 2020 and, due to the part-time nature of it, have a finish date of September 2026.
My supervisors are Prof. Brett Day
*CDT = Centre for Doctoral Training
As part of the UK Government’s Net Zero targets, tree cover of a significant proportion of the UK’s land area needs to increase over the coming decades. Given that 70% of the UK land area is used for agriculture these trees will replace some agricultural production. This has the potential to impact Britain’s food security, particularly under a changing climate. Therefore is it crucial that we have evidence to support the right land use decisions being made now. Current strategies focuses on a “land sparing” approach - planting large areas of woodland while intensifying agriculture on the remaining land (Lamb et al, 2016). Agroforestry, integrating trees into agricultural production systems, represents a “land sharing” approach and has the potential to be a more resilient system of planting trees while retaining food production capacity. The focus of my research, therefore, is to understand how these two approaches can be implemented to maximise carbon sequestration, food production and other ecosystem services.
The challenges for this research are threefold. Firstly, there is little empirical data on agroforestry in the UK. There are a very small number of long running experiments but not enough to be representative of the whole UK and thus too little to extrapolate from. This is where agroforestry modelling is useful. Agroforestry models can simulate multiple configurations of land use in a much shorter time than experimenting in the field. However, given the sparse number of agroforestry observations to fit model parameterisations to, simulating agforestry growth will introduce substantial uncertainty which must be quantified.
The second challenge is factoring climate change into an assessment of agroforestry potential. The UK has good climate projection data, courtesy of organisaitons like the Met Office, but by its nature is uncertain and comes as an ensemble of model runs. This uncertainty must also be quantified and propogated through any agroforestry modelling. This uncertainty could be useful in determining how resilient simulated land use configurations are to future climates.
The third challenge explores how agroforestry will be adopted and how uncertainty under climate change can be translated into the language of risk. For instance, under a changing climate it might be lower yielding but lower overall risk to plant short rotation coppice on a 7-year rotation rather than rely on the successful growth and harvest of a 50 year timber crop, even if the timber crop yields higher financial return and carbon sequestration in the full term of the simulation. Uncertainty under different climate scenarios can be analyised in this risk framing.
- How can agroforestry be modelled at scale in a UK context?
- How does a land-sparing (agriculture+forestry) approach to tree planting compare to a land-sharing (agroforestry) approach, when analysed at scale?
- How does uncertainty propogate from input data (e.g. climate variability) and other sources (e.g. model descrepency) through a scaled-up agroforestry model?
- How to use results, with uncertainty quantification, to assess tree planting strategies for different risk appetites?
Collated from literature and field studies. Collaborating with Prof. Paul Burgess on this. Extent, quality and completeness of data still to be assessed.
UK Climate Projection 2018 (UKCP18) data are a collection of gridded ensemble simulation outputs for the UK climate up to the year 2080. Data is available for the RCP8.5* climate scenario at 60km, 12km and 2.2km spatial scales, and daily, monthly and annual timescales.
CHESS-SCAPE is a downscaling of UKCP18 to 1km spatial scale and additional RCP scenarios, performed by the UK Centre for Ecology and Hydrology (UKCEH) as part of the Climate Hydrology and Ecology Research Support System (CHESS). It was designed to drive higher resolution land-surface models, such as the Joind UK Land Environment Simulator (JULES).
The daily, coarse 12km resolution regional climate version of UKCP18 is being used for development purposes with the agroforestry model in this PhD, but the aim is to use the 1km high resolution CHESS-SCAPE data once the model performance and experimental design has been finalised. However, there are now only 4 ensemble members rather than the 12 of the original UKCP18.
CHESS-SCAPE data is now available on CEDA's data archive:
https://data.ceda.ac.uk/badc/deposited2021/chess-scape/data
*RCP = Representative Concentration Pathway
Soil data acquired from a collaborator in the Net Zero Plus project, source unkown.
Other potentially useful data could be land cover data to exclude areas of the UK which are unsuitable for agriculture e.g. urban areas, existing forestry, nature reserves, etc.
After a literature review on agroforestry modelling, I chose Yield-SAFE for this investigation. It is a parameter-sparse, biophysical, mechanistic model which simulates biomass accumulation of co-located trees and crops. It requires only daily meteorological data and parameters describing the soil, tree and crop properties being simulated.
The plan is to follow an approach of based on Kennedy & O'Hagan, 2002, Bayesian calibration of computer models with guidance from Lambert, 2018, A Student's Guide to Bayesian Statistics. This is likely to require many thousands of model executions with a Monte Carlo based method, so the model should either be fast or emulated.
Other repsitories used in this PhD are:
- yield-safe-py
A Python implementation of the Yield-SAFE agroforestry model - yield-safe-ukcp18
Repository of code to prepare UKCP18 data for ingestion into Yield-SAFE model