Breakout Session 1: High resolution modelling. Co-chairs: Clothilde Langlais and Chris Chapman (CSIRO)

ACCESS Workshop 2024 ACCESS Workshop Day 2
1

High resolution modelling. Co-chairs: Clothilde Langlais and Chris Chapman (CSIRO)

Link: S1 B4 - Google Docs

Description: There is a clear movement of the community towards simulations at higher and higher resolutions.

  1. can we identify synergies between the earth system components?
  2. Are small scales important everywhere or regionally ?

While exciting, high resolution simulations present challenges.

  1. What are the major technical roadblocks?

  2. What are the major scientific roadblocks?

  3. What tools do we have now, and what will we need to develop?

  4. What are the “low hanging fruit”

    Conclusions/Actions from Breakout Session

    Please use this thread for further discussion on this talk.

2

I am not sure if it was mentioned by Richard, but Kirsty Hanley and Humphrey Lean have run a 300m ensemble over London:
The performance of a variable‐resolution 300‐m ensemble for forecasting convection over London - Hanley - Quarterly Journal of the Royal Meteorological Society - Wiley Online Library(https://rmets.onlinelibrary.wiley.com/doi/10.1002/qj.4794?af=R)

3

Notes from the session:

High resolution modelling discussion

Some representation of : Atmosphere modelling community, ocean modelling community and coupled ocean-atm modelling community, including ocean BGC and atmospheric composition.

Why?

  1. Better understanding of small scale processes in their own right

Dynamical understanding (Better understand their drivers) and impact

Processes important for hazard: dry condition for fire risk, coastal rainfalls, intense localised rainfall and wind, tropical cyclone.

Localised effect -Processes that interact with and/or driven by complex orography, topography, coastlines: waves, tide and inundation ; Ice-shelf ocean interaction ; shelf-break current interaction with canyons , tidal mixing

Small scale Processes and events that are involved in downstream impact : nutrient supply, water quality, larval dispersal – usually the purpose of coastal modelling

Submesoscale processes in the ocean : coastal and open ocean

  1. understand small scale processes to fix biases in large scale

examples: Southern Ocean radiation biases , small scale convection in tropical band, submesoscale ocean processes

2ways to get benefit:

  • ¡ upscale effect : small scale impact large scale - improve biases in HR global run
  • ¡ better understanding of HR process to inform better parameterization in LR global run
  1. What is the community doing?

We contribute/invest in to 1. and 2.upscale, not so much 2.parameterization

But lots of unknown that need to be understood:

  • o Is there benefit? And Can we quantify it?
  • o Why high res give us improvement?
  • o How to assess performance / validate with obs?
  • o What is an optimal resolution? need km’ scale coupled : 1/4 atm 1/10 ocean , in 7-y 5km atm and ocean?

comment about value in translating the learning of HR into parameterization scheme for LR:

ML space could inform the parameterization.

Example: cloud formation module translated into ML-informed module (which learn from HR)

How?

Lots of different pathways to go to high resolution

Lots of questions, without clear answer – the how? is a scientific question in itself:

  • o Global high res or downscaling?
  • o Atm only, ocean only, coupling at high resolution ?
  • o For downscaling which boundary conditions are best?
  • o Does global model provide good enough boundary condition to the regional downscaling?
  • o Does global high res provide better climatology ?

Comments :

  • o Role to play for relocatable tool for downscaling
  • o Role of ~100m resolution model:

Benefit in doing some trial run at a few 100m resolution, for small domains (example for Paris Olympics)

Very HR in coastal modelling (down to ~50m), but big limitation is to have the right atm forcing (sea-breeze, diurnal cycle , island effect, cap effect….)

Across time-scale value:

event type study or/and climate study or/and decadal study?

Depend on the science Q? but high values across time scale:

For example: extremes rainfall (intensity and location)

high value for forecasting – high value for climate but harder for climate to get/understand the return periods/ frequency

Role of ACCESS-NRI and community

  1. benchmarking in regional modelling?

can we come up with a protocol to assess the value of HR.

what are the right metrics?

Can we assess benefits across spatial scales : local event/process, but also regional benefit.

The community needs a framework to assess the value of increasing resolution. K-scale project from the MetOffice provides a potential example.

  1. 1 big Challenge with HR is large database and analysing TB

Postprocessing discussion:

Do we want an equivalent of ESMval tool?

It is important to store the data in the right format - right chunking. in particular space vs time.

We need/want to share the postprocessing methodologies.

Compute vs storage discussion :

compute cheap, storage expensive.

Try to find the right balance between Diagnostic online vs offline

What to save? Discussion:

Finding the right balance between: “Please Don’t output 200 variables”, and “please output more”.

Governance discussion:

  • o Better tracking of what is going in : input, initial conditions, forcing…etc
  • o Sometimes outputs are “too easily” available – data not used in the right way. Misused of data.
  1. Coordination across communities.

More question than answers. But Clear need and will for coordination and knowledge sharing.

Easy hanging fruit?

A larger atm domain at HR could serve the ocean community.
In coastal ocean: do we need long term projection at the boundary or could a short-term or scenario type forcing be useful? time-slice or scenario are good.

1 Like