BREATHE (Bridging Research on Environmental Attribution and Health Equity) is an ambitious international consortium, coming together to advance our knowledge of climate and health attribution.
This Wellcome trust funded project brings together researchers from the University of Bristol (UK), University of Oxford (UK), London School of Hygiene and Tropical Medicine (UK), Office for National Statistics (UK), CIDACS (Brazil), FFU (Brazil), PIK (Germany), Estación Biológica de Doñana - CSIC (Spain), and the University of Cape Town (South Africa), along with a wide range of international partners and stakeholders.
My role within the project is the development and analysis of a new detection and attribution large ensemble from the University of Bristol’s HadCM3B-ESM climate model. This large ensemble - which will feature 200+ initialisations, multiple experiments, and both concentration- and emission-driven simulations - will be based on the DAMIP2 protocol and submitted to CMIP7. With our large ensemble we aim to investigate the contribution of anthropogenic emissions to climate-related health outcomes, particularly in the UK and Brazil.
HadCM3B-ESM
The model
HadCM3B-ESM is an Earth System Model hosted at the University of Bristol. A new update to the Met Office HadCM3 model which was developed in the late 1990s, HadCM3B-ESM still performs well compared with contemporary climate models. The biggest strength of HadCM3B-ESM is its incredible speed, simulating approximately 100 model years per day on Bristol’s computing infrastructure. This is orders of magnitude faster than most other modern coupled models.
A consequence of HadCM3B-ESM’s speed is the ability to run large ensembles. Our HadCM3B-ESM submission to CMIP7 will be characterised by the use of both initial condition and perturbed parameter ensembles.
Over the past year I have contributed to the development of the model, implementation of forcings, and overseen the creation of a post-processing pipeline for our data. I have also run the long spinup simulations that are essential for ensuring the model is in equilibrium.
You can read about the previous iteration of HadCM3 in the 2017 paper by Paul Valdes et al.
We are also in the process of writing two new papers which will describe the updates to the model and its implementation for CMIP7.
DAMIP2 large ensemble
Within the CMIP7 submission I will be running the DAMIP2 simulations. Our goal is to produce a 200+ member initial condition ensemble, which will be many times larger than any other equivalent ensemble. We will also be running simulations both in ‘concentration-driven’ setup, in which atmospheric greenhouse gas concentrations are prescribed, and in ‘emission-driven’ setup, where we turn on the model’s interactive carbon cycle and instead prescibe the anthropogenic emission of carbon dioxide into the atmosphere. By running the same experiments in both setups, we can analyse the effect of carbon cycle feedbacks on the detection and attribution of human drivers of climate change and its impacts.
The size of our large DAMIP2 ensemble will have two key benefits:
- The ability to better constrain uncertainty in trend attribution Detection and attribution is, at its heart, a signal-to-noise problem. Our goal is to pick out a slowly varing forced signal from within a noisy and chaotic system. Whilst this noise, which is a result of internal variability, will be different in each run of our ensemble, the forced signal, which is a result of the boundary conditions we apply, will remain the same. Therefore, with a large enough ensemble we can ‘average out’ the noise across our ensemble, and better constrain the forced signal.
- The ability to consider extreme events Extreme weather events are, by definition, rare. Hence it can be difficult to generate a large enough sample of these events to perform any meaningful statistics. With a 200+ member ensemble, we will see enough rare events to perform probabilistic extreme event attribution. In this method we consider the likelihood of exceeding some threshold (this could be temperature, rainfall, wind, etc.) in both a historical and natural climate. We can then compare the two to quantify how much more likely a particular magnitude of event has become (or converseley, how much more severe a particular likelihood of event has become) due to anthropogenic climate change.
We are currenly putting together a paper which will describe the ensemble and give a summary of its outputs. Once the simulations have been run, they will also be made available through the ESGF CMIP7 platform.