Dr Sihan Li presents at AGU Fall Meeting
The world’s largest Earth and space science meeting – the AGU Fall Meeting – took place in New Orleans, Louisiana, from 11-15 December. Post-doctoral Research Associate Dr Sihan Li (Meredith) attended the meeting to give several presentations on behalf of the climateprediction.net project.
Her presentation Changing frequency of flooding in Bangladesh: Is the wettest place on Earth getting wetter? (Karsten Haustein, Peter Uhe, Ruksana Rimi, Akm Saiful Islam, Friederike Otto) in the session entitled ‘Improving Our Mechanistic Understanding of the Regional Climate Response to Anthropogenic Aerosols’, presented results from an analysis of extreme precipitation that led to the Bangladesh floods in summer 2016 (see also the REBuILD project).
Human influence on the Asian monsoon is exerted by two counteracting forces, anthropogenic warming due to the influence of increasing Greenhouse Gas (GHG) emissions, and radiative cooling due to increased amounts of anthropogenic aerosols. GHG emissions tend to intensify the water cycle and increase monsoon precipitation, whereas aerosols are considered to have the opposite effect.
In reality we are essentially committed to more rainfall extremes already as aerosol pollution will eventually be reduced regardless of future GHG emissions. Therefore it is crucial to assess the risk related to removing anthropogenic aerosols from the current world as opposed to standard experiments that use projected climate scenarios.
In the Session on ‘Novel Methods for Combing Physical Simulation, Machine Learning, and Data-Driven Analysis in Climate Studies and Geophysical Sciences’, Dr Li presented Using Perturbed Physics Ensembles and Machine Learning to Select Parameters for Reducing Regional Biases in a Global Climate Model (Sihan Li, David E Rupp, Linnia Hawkins, Philip Mote, Doug J McNeall, Sarah Sparrow, David Wallom, Richard Betts).
The study investigates the potential to reduce known summer hot/dry biases over Pacific Northwest in the UK Met Office’s atmospheric model (HadAM3P) by simultaneously varying multiple model parameters. Results illustrate the potential of using machine learning to train cheap and fast statistical emulators of climate models.
Finally, Changing Drought Risk In a Warming World- using event attribution methods to explore changing likelihoods of drought in east Africa in the past, present and future (Sarah O’Keefe, Sihan Li, Friederike Otto) in the session ‘Climate Extremes: Patterns, Mechanisms, and Attribution’, estimated current and future changes in the probability of drought in different East African regions, making use of the HAPPI project in which large ensembles of atmosphere-only models are run under historic, 1.5 and 2 degrees C conditions (Mitchell et al, 2017).
East Africa is particularly vulnerable to potential impacts of anthropogenic climate change, due to the particular climatic forces at play in the region and the population’s dependence on rain fed agriculture. However large natural inter-annual variability in the region has made the detection and attribution of anthropogenic forcing a challenge. The large ensemble multi-model framework in the HAPPI design allows for a more robust estimation of extremes than ever before.