Atmosphere WG Annual Meeting minutes 5/9/24
Science talks
-
Christian Jakob: Atmospheric modelling needs to build the weather-climate bridge
- 21st Century Weather Research Program - atmospheric modelling needs
- Weather systems science needs models and obs
- Higher resolution GCMs shown to better represent trends and SST warming patterns
- Need a 25km global model (resolution sits in between coupled NWP and seasonal)
- Need a large domain ~5km convection permitting model
- Discussed links between climate drivers and weather systems, and the need for higher res models to better represent the associated weather and forcings such as topography
- Need a 1km Australian model
- Need LES resolution relocatable small-domain model
- Qs: Need to move on from AUS2200, great start but have other needs and want more flexibility for running different domains and resolutions. Noted that the regional nesting suite can do what is needed, seems like education and training is the key thing missing rather than developing the technical infrastructure
-
Shaun Cooper: ACCESS-AE: a national, km-scale numerical weather prediction ensemble system
- Introduction to ensemble science and the motivations for running ensembles to quantify uncertainty
- ACCESS-AE: 12 members, 2.2km resolutions, RAL3.2, 48 hour forecasts
- Perturbations from IC, LBC, physics parameters
- Dec 10 - 17 2023 trial to test developments to the random parameter scheme, TC Jasper, fog, fire weather, tornado, heavy rainfall
- Fractions skill score to assess ensemble precipitation performance, ACCESS-AE precip compared to GPM shows good skill and the new random parameter scheme is marginally more skilful for small rain accumulations with an increase in skill at higher rainfall accumulations
- Spread/skill relationship for near surface fields, temp and winds, new random parameter scheme shows better spread/skill relationship with the increase in spread greater than the increase in RMSE
- Qs: ACCESS-TC model could be replaced with a larger domain ACCESS-AE, the current domain is not large enough. FSS is state of the art metric to assess spatial performance and spread/error shows that the ensemble is under spread, as are all convective scale ensembles, limit of running members is compute resources. Note that ACCESS-AE is a research system running at NCI using the regional ensemble nesting suite (same functionality as the regional nesting suite so you can run any domain at any resolution with multiple nests).
-
Mathew Lipson: A BARRA-R2 driven Regional Nesting Suite
- Suite to assess JULES changes on 12km resolution GAL9 regional climate simulations
- IC and LBC provided by BARRA-R2 with frames on disk, daily updating SST from ERA5
- Ran through the tasks in the suite and compute requirements
- 13 month simulation completed compared near surface temperature to AWS and BARRA-R2, not as accurate
- Specific humidity also not as accurate as BARRA-RS but clearly not expected to be
- Rainfall has a different pattern and biases compared to BARRA-R2
- Can set up a smaller nested region, e.g. 1.5km, since this is the nesting suite
- 12km model is cheap to run compared to convective scale national models
- Discussed outstanding technical issues and whether the community is interested
- Qs: Rainfall bias from tropical Australian variability, benefits of comparing driving data ERA5 and BARRA-R2 would be interesting particularly given the issues with land initialisation, no evaluation of land surface from 13 month simulation yet
-
Thi Lan Dao: Joint modulation of coastal rainfall in Northeast Australia by local and large-scale forcings: Observations versus AUS2200 simulations
- Coastal rainfall affected by both local variability and large scale climate, aim to understand the interactions between these
- Using AUS2200 as need km-scale model to represent local processes, ERA5 with BARRA-R2 soil moisture
- 3 MJO events, 180 days
- Compared simulations to radar derived rainfall, captures the coastal rainfall reasonably well so the model is a viable tool to understand the processes responsible for generating the rainfall patterns
- Analysed vertical cross-sections to understand off-shore/onshore rainfall propagation, noted the role of the sea breeze and density current in the lowest levels as well as the background wind
- For rainfall over land, convergence of sea breeze and westerly winds
- Limited samples to fully understand scale interactions
-
Claire Vincent/Andrew Brown: ACCESS modelling for coastal processes
- Planned activities: coastal winds, using high res models to understand BL structure
- Motivation: wind farms and the need for more energy when solar stops for the day
- Assessed diurnal cycle of wind from BARRA for 3 regions coast, off shore, inland
- Night time peak in wind inland, daytime complex orography, night off shore
- Assessed winds at Gippsland where Australia’s first off shore wind farm will be located
- Variability of winds translates into large changes in wind power
- Sea breezes in the UM have been assessed in various studies in various regions, as turbines are at ~300m really need to understand the vertical structure of the sea breeze, BARRA at 12km looks good
- Also looking at diurnal cycle of convective gusts to understand wind variability
- BARPA-R vs BARPA-C demonstrated the need for km-scale models to capture high winds
- Science questions: role of sea breezes in modulating off shore winds, are there strategic locations for wind energy that would help balance power demand, role of SST gradients and coastal land, can those processes be resolved in high- resolution coupled models
- Qs: Will look at more than Victorian coastline but around the country, higher density of vertical levels in the model BL would be useful - note that this is what the UM vertical levels sets do, how well are downdraft gusts represented - we know that 4km model looks much better than 12km
Discussion
Reference datasets
-
Update from ACCESS-NRI (Clare Richards)
-
ACCESS-NRI has NCRIS funding for storage of community reference data sets
-
Datasets that are accessible, trustworthy and (re)useable
-
High level process described includes review, prioritise, propose, approve, release, maintain
-
Atmosphere WG requests grouped into different data sources
-
Scott noted that any data sets from the Bureau at NCI that are useful for the community should be flagged, as discussions about NCI maintaining data sets are occurring
-
Total storage available to ACCESS-NRI 600TB, looking to expand
-
Takes a lot of effort to document and curate data so want to understand what the community highest priorities are
-
WG needs and priorities
- Discussion on Group 2 data
- Matt on ERA5 processed data for nudging, needed for composition modelling, 3 people currently using this plus 2 more soon. Want the full time series, question about the temporal resolution required. Can’t use the NCI ERA5 data. Bec has scripts, they want ACCESS-NRI to develop their scripts and take over the data curation and management.
- Sonya on SST ancillary file data, needed for AMIP runs using CABLE, uses Met Office data but processed onto grid for ACCESS AMIP.
- Chermelle on OSTIA SSTs, OSTIA is high resolution so would be useful for AUS2200 and km-scale models rather than using ERA5.
- Yi on GPM, AGCD, Bureau AWS. GPM is possibly supported already by Paola. AGCD useful for land WG too. Up to date data required (3 months ideally), but that means working with the Bureau who generate the data and updating takes a lot of effort. Bureau station data needed for high temporal resolution model assessment - think the community needs to be clear about what they want e.g. temporal resolution as there is 1 min data available for some stations, what time period, etc.
Atmosphere Cookbook (Heidi)
- Navid has helped setup a GitHub repository with some Atmosphere ‘recipes’ to analyse model output
- Collated on the Atmosphere-Cookbook Github repo to share and document resources with community
- The cookbook uses ACCESS-NRI Intake Catalogue to load data (e.g., currently from Aus2200 and CMIP6 data)
- Anyone can contribute (or get help) by opening a new Github issue
- Went through the repository, showed Chris’s recipes to produce plots, went through an example to show how to modify to read in different data, variables, etc
- Yi and Chris note that they think this will be useful to help new users get started
- Emma asked whether this will be compatible with ESMValTool or other existing community tools, which was noted would be good but ESMValTool uses its own ecosystem but could be some similarity with some recipes
- Paul asked about the scope, COSIMA does produce publication quality figures
- Clare noted that this could provide information about what datasets are being used
- Matt asked about data ingestion. Heidi noted that you can only use this through the ACCESS-NRI Intake Data Catalogue, but as Matt pointed out there is a need for quick looks of lots of experiments that won’t go into the intake catalogue. Needs a data ingestion script that doesn’t use intake catalogue - what data format, UM file or netcdf, most users only look at netcdf model data.
- Discussion about data conversion, many existing tools.
- Should we have an Atmosphere Cookbook Hackathon?
- How is attribution of contributions acknowledged?
- You can add your model data to ACCESS-NRI Intake Catalogue, which is used to load data for the Cookbook recipes. Sonya Fiddes is keen to contact ACCESS-NRI to do this.