The objective of this project is to evaluate the representation of vegetation in the UK’s flagship Earth System Model (UKESM) using state-of-the-art Machine Learning methods and observations of the Earth’s climate and vegetation. This interdisciplinary project will provide experience in applied machine learning for climate science.
Department: Meteorology
Supervised by: Ranjini Swaminathan
Gross Primary Production (GPP) is the flux of carbon into the land and oceans driven by the process of photosynthesis, and controlled by the presence and amount of vegetation in any given region. It is an indicator of how much atmospheric carbon dioxide, a major greenhouse gas, is used up by photosynthesis to produce new plant structures. It is thus an important component of terrestrial ecosystems and the global carbon cycle. Earth System Models (ESMs) are computer programs that simulate processes in the atmosphere, ocean and land, their interactions and how carbon flows through the earth system. ESMs simulate GPP and can be used to study spatial and seasonal patterns of GPP including how this influences the Earth’s climate in the future. However, there is considerable uncertainty and divergence in GPP estimates from ESMs making it challenging to infer accurate estimates of atmospheric carbon dioxide needed for carbon budgets. This study will directly contribute to ongoing research exploring machine learning approaches to evaluate the terrestrial carbon cycle in TerraFIRMA, a five year multi-million pound, multi-center project currently in its first year. The student will extend an existing ML framework developed by the PIs for evaluating GPP simulations in UKESM to answer specific questions on how well UKESM simulates GPP across different regions when compared to observations and the atmospheric factors most relevant for GPP. Answers to these research questions will help improve our capability to more accurately simulate GPP in ESMs for carbon budget estimates.
In the first week, the student will be required to get familiar with the ML algorithms and Python software to be used for analysis. Over the next few weeks, time will be spent equally on applying ML tools to climate data and in the analysis of results. In the last week of the program the student will be required to complete a short writeup on the work done including any barriers and challenges faced.
Python programming proficiency: Required. Basic understanding of the carbon cycle and atmospheric processes : Desirable but not essential. Familiarity with Linux systems : Desirable. Machine learning experience : Desirable but not essential. Familiarity with NetCDF data files : Desirable but not essential.
The student will get first hand interdisciplinary research experience and an opportunity to contribute to a large research project. The student will also benefit from interactions with the wider ESM community and will have potential co-authorship if any of their work/results are included in research publications. Additionally, the student will also develop skills in handling different kinds of climate data and in understanding and interpreting results from ML algorithms for climate data analysis.
Provisionally in the Department of Meteorology or at the Library.
9:00-17:00
Monday 05 June 2023 - Friday 14 July 2023
The deadline to apply for this opportunity is Monday 3rd April 2023. Students should submit their CV and Cover Letter directly to the Project Supervisor (click on supervisor name at the top of the page for email). Successful candidates will be invited for an interview.