FAQs - BOM BARRA2 (ob53)

I cannot see the BARRA2 local file paths from Gadi or ARE? 

Please check that you have registered for the relevant BARRA2 project code as local file paths are only visible to registered project members. You can check your registration status from MyNCI User Portal. 

If you are registered but having trouble from a Gadi job, please check that you have included the relevant data collection code(s) in the PBSPro storage directive of your job submission script, e.g.,

#PBS -l storage=gdata/ob53

What dataset/variables are available in the NCI collection?

Please refer to BARRA2 Parameter Descriptions for a list of available variables and descriptions.

The parameter descriptions are still being expanded and may be incomplete in the meantime. You can browse the full BARRA2 holdings in the NCI Data Catalogue.

Why is the dataset/variable I am after not available in the NCI collection?

Please check it is a modelling output of BARRA2. For BARRA-R2/RE2, please refer to Appendix of Su et al., (2022) Bureau Research Report 067. 

If it exists, the dataset/variable is stored on tape archive and a request can be made to the Bureau of Meteorology. Please email help@nci.org.au (Subject: BARRA2 data collection).

What if I find an error or issue with the data? 

If you notice an issue within the NCI replica collection, please email help@nci.org.au (Subject: BARRA2 data collection). NCI will check if it is a local issue or can assist with reporting the issue upstream to Bureau of Meteorology. Details of any issues affecting local usage will be added to this page.

Where can I find metadata information?

The BARRA2 data files are stored in netCDF format, containing global attributes that inform how the files are being produced and variable-level attributes that inform how to interpret a variable. Most of the BARRA2 modelling variables are mapped onto the CORDEX-CMIP6 variable names.

When will the BARRA2 data for last 3 months be available?

At this time, BARRA2 data is extended on a monthly basis, 3-4 months behind present day. This time delay is due to the fact that BARRA2 is nested in ERA5; ECMWF updates their quality-controlled ERA5 data 3 months behind present day.

For instance, at the start of May 2025, ERA5 releases its January 2025 reanalysis data. This enables BARRA2 to analyse January 2025 observations, using ERA5's January analyses as the boundary conditions. An added complication is that the BARRA2 analyses for the last day of January 2025 requires ERA5's February data to complete. As a result, we release BARRA2's January 2025 data early June 2025.

What is the time period of the BARRA2 reanalysis?

It covers the historical period from January 1979 to present day. We updates BARRA2 3-4 months behind present day. This time delay is due to the fact that BARRA2 is nested in ERA5; ECMWF updates their quality-controlled ERA5 data 3 months behind present day.

What time zone is used for BARRA2 data?

All time values given in the BARRA2 datasets are in the UTC timezone.

How many pressure levels are available?

Datasets for air temperature (ta*), zonal wind (ua*), meridional wind (va*), vertical wind (wa/wap*), geopotential heights (zg*), and specific humidity (hus*) are available for multiple pressure levels. The data for the pressure levels closer to the surface are provided as 1-hourly grids, while the data for pressure levels further up the atmosphere as 3-hourly grids.

Specifically, the pressure levels available for BARRA-R2/RE2 (AUS-11/AUS-22) are,

• Hourly (1hr): 1000, 925, 850, 700, 600, 500, 400, 300, 200 hPa  (9 levels). An extra ta950 is also available.
• 3-hourly (3hr): 250, 150, 100, 70, 50, 30, 20, 10 hPa (8 levels).

The pressure levels available for BARRA-C2 (AUST-04) are,

• Hourly (1hr): 1000, 975, 950, 925, 900, 850, 800, 750, 700, 600, 500, 400, 300, 200 (14 levels). 
• 3-hourly (3hr): 250, 150, 100, 70, 50, 30, 20, 10 hPa (8 levels).

Please refer to BARRA2 Parameter Descriptions for a list of available variables and descriptions.

How to distinguish between time-instantaneous and time-aggregated variables, and interpret 'cell_method' in the netCDF files?

To describe the characteristic of a field that is represented by cell values, we define the cell_methods attribute of the variable. This is a string attribute comprising a list of blank-separated words of the form "name: method".

Each "name: method" pair indicates that for an axis identified by name, the cell values representing the field have been determined or derived by the specified method. For example, if data values have been generated by computing time means, then this could be indicated with cell_methods="t: mean", assuming here that the name of the time dimension variable is "t". The token name can be a dimension of the variable, a scalar coordinate variable, or a valid standard name.

For more information and examples on cell_methods please refer to CF Conventions here: https://cfconventions.org/Data/cf-conventions/cf-conventions-1.7/build/ch07s03.html

The cell_methods used in BARRA2 reanalysis are either instantaneous in time (i.e., "point" for a subsampling and selecting instantaneous points in time), aggregated over time ("minimum"/"maximum"/"mean" for aggregating over a certain time period), or applied to an area (e.g., interpolated).

One example is mon/sfcWindmax, that has cell_methods given by "time: maximum (interval: 1 hour) area: interpolation (method: bilinear) time: maximum (interval: 1 day) time: mean (interval: 1 month)". Its calculation comprises of a series of steps, namely:

• Taking maximum over one-hour intervals
• Horizontal spatial interpolation to the final grid via bilinear method
• Taking maximum over one-day intervals
• Taking mean over one-month intervals

It's important to note that time values are different between instantaneous and time-aggregated variables. Accumulated variables use time values in the centre of their time-aggregation window, e.g., 00:30 for a 1h-averaged value between 00:00 to 01:00. The accumulated variables will have an additional coordinate, 'time_bnds' to indicate these time windows.

The variable names themselves does not indicate how the variable might have been aggregated over time or space. To find out about the variable aggregation you can check the header of the netcdf file (e.g., using ncdump -h [filename]) and looking for cell_methods. Because BARRA2 (and BARPA) follows the CORDEX-CMIP6 protocol as closely as possible, you can also check the CORDEX-CMIP6 variable list here: https://docs.google.com/spreadsheets/d/1qUauozwXkq7r1g-L4ALMIkCNINIhhCPx/edit#gid=1672965248. In this spreadsheet column c ("ag") will give you information about the aggregation of a variable. 

This also applies to BARPA projections.

How to interpret instantaneous wind components data?

Instantaneous fields provide the value a field valid at a model time-step with no additional time processing applied. The variables "uas" and "vas" are instantaneous estimates of the westerly and northerly components of horizontal wind at 10m. These have been calculated within BARRA2 reanalysis by integrating the similarity equations from the surface to 10m (U_0 = westerly component of surface current if over sea, 0.0 if over land).

BARRA2 (and BARPA) does not explicitly simulate processes at time-scales smaller than their own timesteps. These fields may be interpreted either as a true instantaneous value with sub-timestep processes unresolved, or as an average calculated over the timestep period. BARRA-R2 has a 5-minute timestep.

This also applies to BARPA projections, with BARPA-R has a 7.5 minute timestep.

How to interpret maximum wind gust data?

WMO define the wind gust strength as the maximum of the wind averaged over 3 second intervals. As 3 seconds is much shorter than the model timestep, a parameterisation scheme is used within BARRA2 reanalysis to derive the maximum wind gust within each timestep. The hourly variable "wsgsmax" then represents the maximum of this parameter computed over all timesteps within each hour. 

This also applies to BARPA projections.

How can I download the netcdf files?

Registered NCI users, see

• NCI User Guide on downloading data from NCI gdata (from /g/data/ob53/BARRA2)
• Example Juypter Notebooks, developed by Bureau of Meteorology and NCI, on accessing and visualising BARRA data **

Non-NCI registered users, see

• Example Juypter Notebooks, developed by NCI, on accessing or downloading data from NCI THREDDS server 
• Python script thredds_download.py, developed by Bureau of Meteorology, to download BARRA data in bulk from NCI THREDDS server **

** This also applies to BARPA projections, with BARPA-R on /g/data/py18/BARPA

What are the spatial extents of the data?

The BARRA2 data is available across a number of spatial domains, which are distinguished by different "domain_id", see table below.

The "domain_id" is constructed in two parts XXX-YY, where XXX indicates domain extent and YYY indicates the grid resolution.

What is the projection datum of the data?

The numerical model assumes a spherical Earth (radius 6371229 metre) for computational efficiency and numeric simplicity. This allows for a uniform latitude-longitude grids. While the model calculations were conducted on a sphere, the differences between a sphere and an ellipsoid are small compared to the scale of atmospheric processes.

However, the modelling uses observations such as satellites that inherently account for Earth's actual shape. For instance, the modelled land surface - the land surface cover, topography, and land sea mask - are described by satellite-based products (ESA CCI, IGBP, SRTM) that use WGS84 (World Geodetic System 1984) datum. 

For users who wish to load the data as a layer into GIS (Geographic Information System) software, or compare directly with a spatial data such as a shapefile, we recommend the users to use WGS84 datum.

** This also applies to BARPA projections.

What are the influences of lateral boundaries on BARRA-C2?

The BARRA-C2 domain focuses on Australian land regions and adjacent coastal ocean areas. Nested within BARRA-R2, it relies on boundary conditions by the larger-scale model. Due to the lower resolution of BARRA-R2 (12 km), the lateral flow entering BARRA-C2 requires time to develop convective-scale structures, a process known as model spin-up. As a result, lateral forcing can introduce spurious artefacts, such as large biases, in BARRA-C2 data under in-flow conditions. These artefacts can be avoided by using BARRA-C2 data from areas farther away from the domain lateral boundaries. 

The stronger the inflow from BARRA-R2, the further into the model domain it takes before clouds and precipitation form properly in BARRA-C2. Examples of conditions with strong inflows include tropical cyclones, east coast lows and cold-air outbreaks. The model spin-up affects the representation of meteorological conditions differently at various vertical levels. For instance, impact on low-level clouds is more pronounced than on high clouds.

Users are encouraged to examine cloud (e.g., cll) and precipitation (pr) fields to identify the most suitable regions of BARRA-C2 data for their specific applications. Lack of cloud and precipitation near the lateral boundaries or strong lines of precipitation coming into the domain from the boundaries, can be indications of lateral boundary spin up.  

A recent paper on this topic by Warner et al. (2025).

What are the convective parameters and how do they differ from other BARRA2 variables?

Convective parameters are variables that are used to characterise environmental conditions favourable for thunderstorms and their associated hazards such as lightning, large hail, damaging wind gusts, heavy rainfall, and tornadoes. Most were developed to support forecasting of these hazards, specifically in the United States; however, they are now widely used in both forecasting and climate research worldwide. Examples include convective available potential energy (CAPE), convective inhibition (CIN), 700–500 hPa lapse rates (LR75), 0–6 km bulk wind difference (BWD06), and 0–3 km storm-relative helicity (SRH03). 

The convective parameters for BARRA2 were derived using a system called ConvParams, which was developed at the Bureau of Meteorology to support operational thunderstorm forecasting. Calculations are performed using the native model-level data, providing a much more accurate representation of vertically integrated quantities such as CAPE, CIN, and SRH than would be possible using the published pressure-level variables. 

For BARRA-R2, around 100 convective parameters are available at hourly temporal resolution on the AUST-11 domain. For BARRA-C2, a subset of around 25 parameters are available at hourly resolution on the AUST-04 domain. A smaller subset of just 15 parameters is available for BARRA-RE2 on the AUST-22 domain. Descriptions of all the convective parameters, together with technical details of the calculation methods, can be found in the BARRA2 Parameter Descriptions. 

What are the different versions of CAPE and CIN, and which one should I use? 

For BARRA-R2, four different versions of "total" CAPE are available on the AUST-11 domain: surface-based CAPE (usually designated SBCAPE but simply named CAPE in BARRA2 to match the CORDEX naming convention), mixed-layer CAPE (MLCAPE), most-unstable CAPE (MUCAPE), and effective CAPE (EFFCAPE). Corresponding variants of "total" CIN are also available. Each of these variants uses different assumptions about the hypothetical air parcel that is lifted to perform the calculation (see BARRA2 Parameter Descriptions for details). 

The choice of which variant of CAPE (or CIN) to use depends on your application. SBCAPE and MLCAPE are suitable for diagnosing surface-based convection, while MUCAPE and EFFCAPE are capable of also diagnosing elevated convection. While SBCAPE remains widely used, the ML parcel is generally considered to provide a more accurate representation of surface-based convection (e.g., Craven et al., 2002). EFFCAPE has not seen widespread use in the literature and is therefore considered somewhat experimental. For most applications, MUCAPE is the preferred variant; however, SBCAPE and MLCAPE are favoured for analysing the environments of tornadoes, which exclusively occur with surface-based storms (e.g., Thompson et al. 2003). 

SBCAPE, MLCAPE, and MUCAPE (and corresponding CIN variants) are all also available for BARRA-C2 on the AUST-04 domain. For BARRA-R2, there are four additional "partial" CAPEs (MLCAPE03, MLCAPEx, MUCAPE0m20, and MUCAPEm10m30) and one additional "partial" CIN (MLCIN1). Details of these can be found in the BARRA2 Parameter Descriptions. 

What is the difference between the CAPE and CIN variables on the AUS-11/AUS-22 and AUST-04/AUST-11/AUST-22 domains and which should I use? 

The CAPE and CIN variables on the AUS-11 and AUS-22 domains are output directly from the ACCESS model  (i.e. Unified Model for the atmosphere component), whereas those on the AUST-04, AUST-11, and AUST-22 domains are calculated using the Bureau's ConvParams system. 

Due to inaccuracies in the method used to calculate CAPE and CIN within the ACCESS model, users are advised to favour the AUST-04, AUST-11, and AUST-22 variants where possible (see Known Issues). This will also ensure consistency with the other convective parameters. Be aware that the ACCESS version of CIN is defined as negative whereas the ConvParams version is positive.

Note that BARRA-C2 (AUST-04) only has the ConvParam versions of CAPE & CIN, unlike BARRA-R2 (AUS-11/AUST-11) and BARRA-RE2 (AUS-22/AUST-22) which have both versions available.

Are convective parameters from BARRA-C2 better than those from BARRA-R2? 

Due to its higher horizontal resolution, BARRA-C2 is expected to provide a better representation of small-scale atmospheric processes compared to BARRA-R2. In particular, BARRA-C2 explicitly simulates convective storms, providing information on their formation, intensity, and evolution, which is not available from BARRA-R2. However, convective parameters were primarily designed to characterise the "undisturbed" environment before storms form or move into a particular area. The presence of explicitly simulated storms can strongly modify the environment (a process known as convective contamination), resulting in convective parameter values that counterintuitively appear unfavourable for convection (e.g., zero CAPE or large CIN).

As such, caution is required when analysing and interpreting these variables from BARRA-C2. Comparison with other variables such as precipitation (pr), simulated radar reflectivity (radrefl1km or maxcolrefl), and screen-level temperature (tas) may be useful in identifying where convective contamination may be occurring. 

In general, users are advised to focus on the BARRA-R2 convective parameters and only consider the BARRA-C2 versions when investigating local variations, storm-induced modifications of the convective environment, or reanalysis sensitivity to resolution.