A workshop of machine learning with climate and carbon modelling.
Process-based climate and carbon cycle models have been extremely successful at simulating observed trends and variability in the Earth system. However, many uncertainties and limitations remain due to an incomplete understanding of nature, as well as the dependence on empirical formulations which are often not well grounded in theory (particularly in modelling dynamics of the biosphere). Focussing on the biosphere, there is a promising opportunity to develop a near-real time system of carbon stocks and fluxes. This is particularly relevant as there is a growing need to monitor and report on the role of natural and managed ecosystems in our efforts to reach ‘net zero’ and limit global warming. However, the best approach to take is far from realised.
In recent years, a wealth of Earth observation data has become available, in particular novel remote sensing products now cover the globe with extremely high spatial resolutions of a few metres, enjoy high revisit frequency, and are capable of multi- or hyper-spectral observation. However, how we best exploit these huge volumes of data to increase knowledge and understanding of the earth system, in a timely fashion, is yet to be determined. A promising approach is to use machine learning techniques, developed and optimised for earth observation tasks, to extract important spatio-temporal features automatically and obtain robust results despite noisy data or limited annotation. These relatively new data processing methods therefore have the potential to further our understanding and modelling of the Earth system. Machine learning can also be used in many other ways to improve carbon and climate models.
The aim of this workshop is to make connections between machine learning and carbon/climate modelling and explore opportunities for collaboration. Please come and join for what promises to be an interesting and fruitful discussion between colleagues from the new ESE, EI network, Exeter climate research, GSI, IDSAI, Met Office etc.
A brief agenda:
1:00 – 1:10 Introduction
1:10 – 2:10 Sharing relevant research with mini talks (please bring one slide or one printed page of your research)
2:10 – 2:20 Break
2:20 – 2:50 Break-out groups for themed discussion focusing on strengths and opportunities
Theme one: Environment and carbon modelling challenges
Theme two: Machine learning techniques
2:50 – 3:20 Integration session to share the themed group discoveries
3:20 – 3:30 Workshop event remark and further actions
3:30 – 4:00 Networking with the researchers from multiple disciplines
To contact the organisers: Mike O’Sullivan (M.OSullivan@exeter.ac.uk), Chunbo Luo (c.luo@exeter.ac.uk).
