Project Proposal: Benchmarking of ACCESS-S2 precipitation against other international global seasonal forecast models available on C3S

Earth System Experiments
1

Experiment title :bell:: Benchmarking of ACCESS-S2 precipitation against other international global seasonal forecast models available on C3S

Summary :bell:: Short lead-time seasonal precipitation forecasts from the ACCESS-S2 seasonal forecasting model will be evaluated against AGCD and ERA5 and additionally bench-marked against other global seasonal forecast models (ECMWF UK Met Office, Météo-France, the German Weather Service (DWD), the Euro-Mediterranean Center on Climate Change (CMCC), the US National Weather Service’s NCEP (NCEP), Japan Meteorological Agency (JMA) and Environment and Climate Change Canada (ECCC).) currently publicly available on C3S portal.

Scientific motivation: To better understand the source of precipitation biases in the ACCESS-S2 seasonal forecast model, with a particular focus on southwest Western Australia, a region where agriculture is predominantly rain-fed with minimal to no irrigation. This will help directly inform biases emerging the from UM, which is directly relevant to ACCESS-ESM-1.X.

Experiment Name :bell:: ACCESS-S2 precipitation forecast skill in southwest WA
People :bell:: Jatin Kala (Murdoch University), Debbie Hudson (BOM), and Rebecca Firth (PhD student)
Model: ACCESS-S2
Configuration: N/A - we will use existing outputs
Initial conditions: N/A - we will use existing outputs
Run plan: N/A - we will use existing outputs
Simulation details: N/A - we will use existing outputs
Total KSUs required :bell:: approx 5KSU
Total storage required :bell:: approx 3 to 5 TB
Outputs: N/A - we will use existing outputs
Restarts: N/A - we will use existing outputs
Related articles: https://doi.org/10.1071/ES24004

Analysis: Analysis will use of Python and CDO, largely single CPU jobs. Some jobs may use DASK with up to 16 CPUS, but not much more. All analysis will use existing data-sets

Conclusion: This is a “low-hanging fruit” project, requiring minimal SUs, but will be very useful towards better understanding biases emerging from the UM in ACCESS-S2. It is acknowledged that this is not directly related to ACCESS-ESM-1.X, however, both models use the same atmospheric model and lessons learnt in this project will be very useful to the broader community.

2 Likes
2

Hi @jatin, thanks for submitting this project proposal! I just wanted to check whether the 3-5 TB storage required would be on lg87, and if so whether you had a rough idea of how long you would need the storage for.

3

Hi Spencer,Yes, we will need the storage on the relevant working group’s project at NCI. This is my first proposal, so I am not across all the project codes associated with ACCESS-NRI. I would say probably around 9 months to 1 year, upon completion, all data will be removed of course.

Regards

jatin