Ocean-atmosphere influences on tropical Pacific cloud feedbacks

点击翻译按钮获取中文摘要

Cloud feedback, particularly resulting from tropical high cloud, is the most uncertain contribution to equilibrium climate sensitivity (i.e. global temperature rise from doubling of CO2)1,2. Clouds in the eastern tropical Pacific region exhibit some of the largest spread in cloud feedback across climate models. A focus on drivers of cloud radiative effects in the region is needed in order to better understand controls on cloud feedback.The El Nino Southern Oscillation typifies the large-scale coupling of ocean and atmosphere in the Pacific; a coupling that must be considered in relation to drivers of cloud feedback. However, radiative effects stem directly from the distribution and microphysical properties of clouds. The project proposed here will use Earth observations, existing in-situ campaign data and reanalysis data to characterise the small and large-scale ocean-atmosphere influences on the Pacific cloud radiative effect, and then relate this to global model variability in the region.An ongoing NERC programme, CloudSense, aims to reduce uncertainty in climate sensitivity. Two of the programme's projects, DCMEX and CIRCULATES, are between them studying the effects of microphysics, aerosol and large-scale circulation on cloud feedback. But neither has focused on the ocean-atmosphere interaction in the equatorial Pacific. These projects can guide, as well as learn from, the PhD research proposed here.Objectives / Methods / DatasetsCharacterise variability in clouds and their radiative effect in satellite (GOES/CERES) observations of tropical Pacific cloud, building on previous analysis3.Explore the potential importance to Pacific cloud variability of observed ocean-atmosphere interaction variables (e.g SST, ocean heat content variability, near surface air temperature and humidity, wind speed, strength of Walker/Hadley circulations, cloud base temperature, or aerosol concentration). The work will begin by decomposing "input" variables that influence cloud and "output" cloud variables into principal components and associating inputs and outputs using multiple linear regression. If it is necessary to address the scientific questions, we will progress to more sophisticated techniques, such as an autoencoder, that are better able to represent variance by using nonlinear relationships but retain a small number of features for ease of interpretability.This empirical approach will further develop the "cloud-controlling factors" approach4, which so far has not included a focus on ocean interactions and important cloud microphysical processes. Key datasets include MODIS/CALIPSO satellite data, HadISST, UK Met Office EN4, and TOGA-COARE/CEPEX Pacific field campaigns5,6,7.Apply identified important variables (including their interactions) to global model understanding. Assess the representation of variables in global climate models and whether the variables can explain the large model spread in Pacific cloud feedback, or whether different variables are important in the models.