Constraining Estimates of Tropical Climate Variability.

Project Details

Key Questions

How will tropical climate extreme weather events change as atmospheric GHGs rise? 


As background climate changes, any additional variation in extreme events could compound the impact of increasing atmospheric greenhouse gases (GHGs). This concern is especially true for the tropics, with both high temperatures and rainfall amounts. It is therefore necessary to constrain Earth System Model (ESM) projections of changes in  interannual variations of weather, in addition to long-term trends, to support climate adaptation planning. Multiple new statistical (e.g. advanced “EOF” algorithms) and physically-based methods now exist (e.g. climate “Emergent Constraints”) to achieve this. 

Aims of the Project

- Provide highly refined estimates of the likelihood of extreme years or seasons, as a function of GHG level. 

- To understand the atmospheric drivers of projected changes. - Provide data in a format of use by impacts or land surface modellers

Project Description

A recent rapid-response project under the “Climate Science for Service Partnership (CSSP) (Brazil)” was awarded to CEH and Exeter University, to remove biases in projections of interannual variability of weather patterns across South America. The initiative was to provide quick “first look” estimates of change, but as the project continued, it became clear it was also prompting a range of especially exciting further questions. Despite the fact that South America has one of the strongest signatures of interannual variability – i.e. the El Nino Southern Oscillation (ENSO) – there remain surprisingly large differences between their projections by ESMs. In the CSSP project, we developed a technique of spatial “morphing” of ENSO, to normalise ESM projections for the historical period. We then assumed that morphing to remain valid into the future, to allow revised projections combining future climate change with corrected ENSO variability. 

The proposed DPhil project will first perform a detailed analysis of recorded weather patterns across the tropics over the last few decades. Four techniques to do this are available, and we anticipate all to be utilised to some extent. Implicit in the standard statistical approach of EOF analysis is the separation of time-evolving drivers (e.g. ENSO strength) and spatial pattern. An assessment will be made of the statistical literature to see if newer algorithms exist that still retain a form of separation and that account for the presence of time-evolving drivers. To support this, there will be an opportunity to use new AI methods to also scan across measurement data and the full ensemble of available climate models. AI methods, applied carefully, offer the opportunity to bring rigour to the statistical testing of whether nonlinear changes exist in either data or model timeseries. The third approach is to undertake a more applied mathematics approach to the meteorological equations that are solved in ESMs for tropical regions. To perform this task will require very detailed climate model diagnostics to constrain or search for common conservation properties implicit in the partial differential equations describing tropical weather. Our strong links to the UK Meteorological Office are likely to make available additional information that is not held in the standard ESM data repositories. The fourth search will be for system “teleconnections”, and AI methods may again be appropriate. Teleconnections are where one part of the climate system may be affected by another, often across different seasons and locations. Such discovered connections, if well-founded, provide opportunities to warn of approaching extreme weather events. 

Dependent on progress, there will also be opportunities to translate findings into impacts. For example, the JULES land surface model has a vegetation competition component, capable of responding to altered climate forcings. Providing the JULES land surface model with constrained estimates of future atmospheric variability will allow investigation into the iconic potential ‘tipping point’ of Amazon dieback.

Methods to be used

Emergent Constraints, statistical EOFs and also dynamical systems to model atmospheric teleconnections. 

Specialised skills required

This project is likely to be quite mathematical and statistical. 

Please contact Chris Huntingford on and Hannah Christensen  if you are interested in this project