Constraining LAnd Responses by Integrating ObservatioNs

Climate-carbon feedbacks are a key unknown when projecting future climate change. Large ranges of feedbacks have been identified within the climate models used to make climate projections. Therefore, to make reliable and believable climate projections, there is an urgent need to reduce this uncertainty.

In CLARION, I will focus on land-surface models (LSMs), the terrestrial component of climate models.

I will investigate how observationally-constrained probability density functions (PDFs) of key model parameters (generated via Data Assimilation - DA) can be used to reduce the range of possible future climate-carbon cycle feedbacks.

This will be done in three key steps using the UK JULES (Joint UK Land Environment Simulator) LSM, though, critically, the methods developed throughout this project will be applicable to any LSM.

First, I will identify the key climate model parameters controlling the carbon, water, and energy cycles, and their relationship with future carbon climate projections. This will be done through sensitivity analysis experiments based on multi-ensemble runs.

Second, using sophisticated Bayesian techniques and the extensive amount of in situ and Earth Observations available, I will calibrate these parameters to create observationally-constrained PDFs.

Finally, these PDFs will be used to constrain the range of climate-carbon cycle projections by propagating the reduction in parameter uncertainty.

This project will build on my strong expertise in DA techniques and will complement the Host Institute's cutting-edge emergent constraint work, which considers narrowing the range of climate feedback across models.

This project will be a unique opportunity to take my demonstrable technical know-how and apply it to the climate change problem, generating high-impact results.

During CLARION, I will also launch a JULES DA working group which will bring together the different UK expertise, encouraging collaborations throughout the project and beyond.