ACS Hazard Team on Extratropical Storms

Last update: 01/08/2024 - now based on bias corrected data and improved regional averaging

Description

GitHub repository for ACS Hazard Team on Extratropical Storms to store, track and develop code.

Indices considered by the hazard team:

• Highest annual 24 hour total (RX1D) - this is the highest daily rainfall falling at each grid point in a year
• Highest annual 5-day total (RX5D) - this is the highest five-day rainfall total falling at each grid point in a year
• Highest annual hourly total (RX1H) - this is the highest rainfall falling over one hour at each grid point in a year, an indication of short-lived heavy rainfall relevant to flash flooding
• Proportion of observations influenced by a low (low_freq) - this indicates how often an extratropical cyclone and/or East Coast Low occurs

Index/metric	Summary of change for GWL2	Summary of change at GWL3	Additional information
RX1D	An increase is possible based on physical understanding, but may not be detectible relative to interannual variability.	Rainfall intensity will likely increase, with a best estimate of ~ +12%	Higher changes on the order of 15% per degree of warming are possible (+29% at GWL3). Regional variations in changes are not robust.
RX5D	An increase is possible based on physical understanding, but may not be detectible relative to interannual variability.	Rainfall intensity will likely increase, with a best estimate of ~ +8%	Higher changes on the order of 15% per degree of warming are possible or +29% at GWL3. Regional variations are not robust, but declines are possible in some regions, particularly where total rainfall declines, with declines projected for more regions in the Queensland ensemble.
RX1H	Rainfall intensity will likely increase, with a best estimate of ~ +12%	Rainfall intensity will very likely increase, with a best estimate of ~ +29%	Regional models are unable to simulate key processes for subdaily rainfall, and underestimate future increases. Based on multiple lines of evidence, the best estimate is a +15% (+7 to +28%) increase per degree of warming; this means increases of more than 50% are posible at GWL3
low_freq	The number of lows will likely decrease in southern Australia, with a best estimate of -8%	The number of lows will very likely decrease in southern Australia, with a best estimate of -13%	Larger declines of 30% or more are possible for GWL3, particularly in the cool season and in southeast Australia. Trends in the most intense and impactful lows may differ, noting that extreme rainfall and sea levels associated with lows will likely increase

Products:

Status of the NCRA deliverables.

The three dots (in order from first/top/left to last/bottom/right) represent the datasets used to compute indices:

• Dot 1: Pre-processed BARPA/CCAM – downscaled but NOT bias-corrected, 5 km spatial resolution (deliverable for 30 June)
• Dot 2: Bias-corrected BARPA/CCAM – downscaled AND bias-corrected, 5 km spatial resolution (deliverable for 31 July)

Note that bias correction is not available for pressure data

In terms of the colors:

• 🟢 The data is available in its final official form
• 🟡 The data creation is currently in progress and available soon
• 🔴 The data processing has not yet started
• ⚪ Not intended for delivery/not applicable

Index/metric	time series (ts)	GWLs 2D	MME 2D	MME 2D change	(Notes)
low_freq	🟢	🟢	🟢	🟢	bias correction not available
RX1D	🟢 🟢	🟢 🟢	🟢 🟢	🟢 🟢
RX5D	🟢 🟢	🟢 🟢	🟢 🟢	🟢 🟢
RX1H	🟢 🟢	🟢 🟢	🟢 🟢	🟢 🟢	partially bias corrected (from the bias-corrected daily data)

Data location

• Raw data: /g/data/ia39/ncra/extratropical_storms/5km/GWLs/
• Bias corrected data: /g/data/ia39/ncra/extratropical_storms/bias-corrected/ACS-QME-AGCD-1960-2022/GWLs/
• Data for the Queensland 15-member ensemble has also been processed, but has not yet been made available in ia39.
• Reported changes also rely heavily on the recent Wasko et al. (2024) sytematic review of changes in rainfall intensity, in support of updates to the Australian Rainfall and Runoff guidelines.

Summary statistics

This table gives the Australian average of the median % change between values at GWL1.2 and at GWLs 1.5, 2, and 3.

Notes:

• The first column gives our best estimate of the index in the current climate (GWL1.2, 2011-2030) based on the bias corrected datasets. In the case of lows, this has been replaced with the BARRA-R2 climatology during 1991-2020, as the raw ACS data underestimates observed low frequency.
• GWL1.2 represents the "current climate", around 2011-2030. The changes shown represent future changes, which are projected to occur on top of changes that have already occurred between a pre-industrial climate and the present.
• Regional average changes for each GWL are calculated for each ensemble member prior to calculating the ensemble statistics (median and 10th/90th percentiles)
• Ranges give the 10th-90th percentile of the ACS ensemble. Numbers are shown in bold where at least 65% of ensemble members agree on the sign of the change ("likely"), and italics indicates agreement across at least 90% of the ensemble ("very likely")
• For the case of low frequency, "Australian" averages are calculated only for latitudes south of 30S, to avoid contamination by trends in tropical lows

Index/metric	GWL1.2	GWL1.5	GWL2	GWL3	Notes
low_freq*	1.1% of the time ~16 times per year	-2% (-13% to +5%)	-8% (-17% to +6%)	-13% (-24% to -4%)	*calculated for locations south of 30S Current climate is from BARRA-R2 6-hourly data (1991-2020) Larger declines are possible based on the Queensland data (GWL3: -19% (-27% to -13%))
RX1D	58.8mm	+1% (-3% to +4%)	+6% (-4% to +14%)	+12% (-2% to +19%)	Best estimate based on the AR&R review is +8% (+2 to +15%) per degree of warming, implying possible increases of 29% or higher at GWL3
RX5D	101.4mm	-1% (-4% to +3%)	+4% (-8% to +11%)	+8% (-6% to +14%)	Consistent with the AR&R review, increases of 29% or higher are possible at GWL3
RX1H	18.1mm	+1% (-1% to +6%)	+5% (-2% to +12%)	+13% (+1% to +19%)	Regional models miss key processes likely to lead to intensification of hourly extremes. The best estimate based on the AR&R review is +15% (+7-28%) per degree of warming, implying increases of 13-56% at GWL3

RX1D	Low frequency
Multi-model median of the % change in RX1D between GWL1.2 and GWL3	Multi-model median of the % change in low frequency between GWL1.2 and GWL3

Details on extratropical lows

Lows are initially identified as individual cyclone centres at a single atmospheric level, and can have a range of intensities.
For the purposes of this intial dataset we include:

• Surface lows that have closed circulation and persist for at least 6 hours
• Which have a matching low at 500hPa within 500km at least once

Any grid point within a 5 degree (500km radius) of the low centre is considered to be influenced by the low.
This is used to calculate what proportion of all 6-hourly observations are influenced by a low.
Note that this tracking does not disinguish between extratropial lows or tropical lows (including but not limited to tropical cyclones)

This method identifies all lows, including tropical and subtropical lows that may affect northern regions of Australia. The raw ACS model data also does not fully replicate observed patterns, with a tendency to generate too many lows in northrn Australia during the warm half of the year, and too few lows in southern Australia.

Details of code

This is a directory of code developed by the NCRA Extratropical Storm Hazard Team

The early files are a dump of code used to create preliminary maps of extratropical storms and extreme rainfall as part of the NCRA Extratropical Storm Hazard Team 2-pagers

The files uploaded in June 2024 were used to develop the MVP grids stored in /g/data/ia39/ncra/extratropical_storms/

Shell scripts based on cdo are used to extract annual RX1D, RX5D, and RX1H datasets from the raw files.

• run_rx5d_5km.sh is the main file for the MVP RX1D/RX5D datasets, while run_rx5d_NHP.sh applies the same methods to the older NHP data and run_rx5d_5km_biasadj.sh applies the same methods to the bias corrected datasets (QME and MRNBC)
• As there is no 5km hourly data, the RX1H is first calculated on the native grid (e.g. run_rx1h_barpa_v3.sh, run_rx1h_ccam_v3.sh, run_rx1h_qld.sh) and then regridded to 5km using CDO (regrid_RX1H_5km.sh)
• Lows are tracked using a different set of code from https://github.com/apepler/cyclonetracking, and the raw track files at from https://github.com/apepler/cyclonetracking were converted to annual grids using R, extract_gridded_lows.R. This is performed on a 1 degree grid by default, and regridded to 5km using regrid_lows_5km.sh

In addition, there are preliminary code for calculating GWL grids and generating figures; these are likely to be replaced by python code in future versions.

• extract_grids_annmean_regs_tocsv.R takes annual gridded datasets and converts them to annual timeseries of means for each NCRA region
• generate_GWL_grids_5km.ncl generates netCDF grids for different GWLs using the 13-membr ACS ensemble
• analyse_GWL_plots_5km.ncl uses the same process but generates plots instead of aving the netCDF grids

Authors and acknowledgment

Hazard team:

• Acacia Pepler (BOM, lead)
• James Risbey (CSIRO, alternate lead)
• Carly Tozer (CSIRO)
• Tess Parker (CSIRO)
• Danielle Udy (BOM)