A lie will travel halfway around the world while the truth is lacing her boots, so the adage goes. And so, the showmen will continue to misrepresent natural disasters that befalls northern Australia while finding something popular to say at the very moment everyone’s attention is focused on that event.
For sure, the rainfall associated with Cyclone Jasper was extraordinary, but not unprecedented for the Cairns catchment. For sure, it is difficult to forecast weather and climate, but the skill of new systems based on artificial intelligence (AI) show great improvement, while the Australian Bureau of Meteorology remains wedded to its General Circulation Models.
Contrary to various popular claims, including by my colleague Peter Ridd*, the Bureau uses the same supercomputer and the same general circulation model to forecast rainfall whether considering the next three-hours, the next three months or the next three decades. It uses a simulation developed by the UK Met Office known as ACCESS-S2, that is also one of the Intergovernmental Panel on Climate Change’s CMIP6 models.
All the general circulation models are underpinned by the assumption that carbon dioxide drives climate change, and all these same models are focused on large scale processes making it difficult to accurately forecasts what matters to real people – local climate, especially extreme rainfall known to be associated with cyclones, and also seasonal rainfall deficient associated with drought.
The Bureau claims it can accurately forecast temperature to within 2 degrees Celsius on any day. This may be of intense political interest, but it is of little real value to the Australian community. Being able to accurately forecast rainfall would be much more meaningful.
There are four types of precipitation-forming processes, including cyclonic rotation (low pressure), though the IPCC’s models are all based on surface-heating (convection). It is perhaps for this reason that these models while accurately simulating general global patterns of rainfall remain unable to capture high intensity events over small areas including rainfall associated with cyclones. Some of these problems can be overcome through downscaling, but even then, it is unclear why the elevation of mountains is mostly underestimated while their spatial extent is overestimated. Fundamentally, and at the core of the problem, is that clouds formation occurs at a scale much smaller than the resolvable grid scale used within all the general circulation models.
The Bureau was able to accurately forecast the trajectory of Cyclone Jasper as it presented as a large and slow-moving weather system, but once the structure of the cyclone began to break down, and this occurred from December 9, ACCESS-S2 struggled to accurately forecast both direction and intensity. Worst, when the heaviest rainfall began eight days later, on December 17 in Cairns, the Bureau was unable to capture the extent of the downpour because its automatic weather recording system at Cairns airport failed. The same problem was experience during the Lismore flooding of April 2022, meaning that the true intensity of the rainfall from these events is not even properly documented.
I first became interested in rainfall forecasting using artificial intelligence in January 2011, following the flooding of Brisbane. My colleague John Abbot used artificial neural networks, a form of AI, for share market trading, and so successfully he bought a red corvette with the winnings one day. That same sports car was drowned in the 2011 Brisbane flooding.
Over the next five years to 2017, John Abbot and I successfully published a dozen research papers on our new technique using AI for monthly rainfall forecasting. We published in the best international peer-reviewed journals and as book chapters following AI conferences.
Our very first paper about forecasting monthly rainfall – for 17 locations in Queensland 12 months in advance – was published in Advances in Atmospheric Sciences which is an Elsevier journal sponsored by the Chinese Academy of Sciences.
Back 12 years ago, when we were pioneering the technique, the Chinese were very interested, and prepared to publish us, but not our own Bureau of Meteorology who scoffed at the idea that AI could be used for weather or climate forecasting.
“But the climate is on a new and dangerous trajectory”, said Oscar Alves who then headed-up the long-range forecasting unit at the Bureau in Melbourne.
At the time the Bureau was using a statistical technique for its medium-term forecasts, while busy developing its own general circulation model known as POAMA. POAMA was subsequently used for operational forecasts from June 2013, before being replaced by ACCESS-S1 in August 2018. POAMA proved a disaster and was replaced without even a media release announcing the change over.
It took the Bureau 20 years to develop POAMA, and they pulled it after just five years and so many disastrous rainfall forecasts that were never acknowledged.
John Abbot and I spent an afternoon with Alves, back in August 2011. John Abbot and I wanted to collaborate. We were convinced back then that AI could significantly improve the skill of the Bureau's rainfall forecasts. Alves had no interest in learning anything new. Alves now heads up the Bureau's Earth System Modelling unit.
All these years later and the Bureau is still refusing to consider the value of AI for forecasting rainfall extremes whether the consequence of a cyclone or a drought.
Meanwhile Google is now using AI for weather forecasting with their GraphCast, run on a desktop, outperforming all the GCMs run on supercomputers.
Google’s GraphCast works from the same principles John Abbot and I used: recurrent cycles that can be found in weather and climate data – as long as the data hasn’t been remodelled to fit the human-caused global warming theory. And the Chinese are now working on AI systems that can forecast both the intensity of rainfall during cyclonic events, as well as trajectory.
I have no doubt that the rainfall forecasts for North Queensland following landfall by Cyclone Jasper would have been far superior if ten years ago the Bureau had began to invest in AI technology, had began to develop some capacity in this very different technique.
There has been commentary suggesting Cyclone Jasper resulted in unprecedented rainfall in the headwaters of the Barron River causing flooding of Cairns, particularly of the Northern Beaches. There is a long continuous rainfall record for Kuranda indicating that while December 2023, and the rainfall associated with Jasper was extraordinary, there are higher totals for previous year going back to 1911.
The Australian emergency management minister, Murray Watt, has ordered a review of the weather warning systems used by the Bureau of Meteorology, while claiming it will become increasingly difficult to predict the weather because of climate change. He should be calling for much more than this, and not using the excuse of 'climate change' for both the failed 3-day rainfall forecast and also the failed warning system after the heavy rains began to fall.
For the last twenty years various international working groups associated with the IPCC and World Meteorological Organisation have been making submissions and predictions regarding the likely effects of global warming on tropical cyclones. These reports have indicated the maximum intensity of cyclones is unlikely to significantly increase, certainly not beyond 10-20 percent. A ‘Statement on Tropical cyclones and Climate Change’ in 2006 by Dr G. B. Love the Permanent Representative for Australia indicated that rainfall intensity could increase, due to the increasing water vapour content of the atmosphere. Twenty years later, the data shows that both the intensity and number of cyclones has been declining.
The extraordinary rainfall associated with the flooding of Lismore and surrounds in early 2022, may have been exacerbated by the increase in water vapour content from the explosion of the volcano Hunga Tunga in January of that year. It could be that the Hunga Tonga eruption has also caused a depletion of ozone in the stratosphere, after temporarily increasing the water vapour content. There has been no overall increase in water vapour content of the lower troposphere associated with increasing atmospheric levels of carbon dioxide.
The general circulation models have difficulty simulating the local impact of volcanic ash on rainfall intensity and global temperatures, and this is a problem because aerosols can supercharge the atmosphere making rainfall more intense.
Importantly, and contrary to recent popular commentary, there are no two separate parts to the Bureau: one making operational weather forecasts and one concerning itself with climate change. Since June 2013, forecasts for the next three-hours and next three-months have relied on a general circulation model and since 2018 specifically on ACCESS, with some adds-on to provide more resolution. There is an urgent need for the skill of this general circulation model to be properly assessed. This could be done as a matter of urgency and through a comparison of forecast versus predicted rainfall for the Cairns catchment as it fell through December 2023, particularly after cyclone Jasper made landfall.
There is also a need for the Bureau to quantify the skill more generally of ACCESS against the skill of the new AI weather and climate forecasting systems including Google's GraphCast and the Pangu-Weather AI model. Pangu AI can predict both the direction of cyclones and their likely impact, particularly their capacity to generate intense rainfall over a small area.
There has been much talk about the unprecedented.
What would be both unprecedented and welcome would be for the Bureau to start doing KPI's and measure predictions against actual rainfall totals. This needs to be done for the three-hour 'rain burst' events associated with low pressure systems and also their longer seasonal rainfall forecasts.
The Bureau forecast that this summer would be an exceptionally dry one for the same regions that have now flooded, using the same ACCESS general circulation model. The drought forecast has also caused unnecessary hardship, with farmers selling livestock at discounted prices as so many anticipated being unable to feed their stock.
Ends.
John Abbot and Jennifer Marohasy outside the Administrative Appeals Tribunal in Brisbane, February 2023.
And just filing this here, by Peter Ridd and republished from The Australian, click here.
* Recent criticism of the Bureau of Meteorology for failing to predict the recent spate of extreme weather is unfair, is ultimately counter-productive and misses far more serious failings of the BOM.
Weather prediction is difficult. At best one can hope only to improve probabilities. And the weather hardest to predict is extreme events associated with storms. These systems are extremely “nonlinear”, to use the parlance of meteorology.
When there are large quantities of moisture in the lower levels of the atmosphere, the air need be lifted only slightly to trigger a violent updraft.
It is a huge slow-motion explosion where the fuel is the invisible water vapour turning into cloud. The amounts of energy involved can be huge – think Hiroshima atom bomb – and a tiny perturbation can set them off. It is often stated that a butterfly flapping its wings could trigger the storm, at least theoretically.
This is one of the least predictable phenomena on Earth. At best, weather prediction can indicate only that such storms are likely at a rough time and place. Perhaps the BOM can get the final warnings out a little faster, but a storm can morph into a supercell in a few minutes.
BOM’s performance in predicting the ultimate landfall of Tropical Cyclone Jasper was nothing short of brilliant. For days before it crossed the coast, the bureau predicted it would end up near Cairns. And that is where it went. The cyclone did minimal damage, but the rain cell associated with it sat stationary around Cairns for days, causing flooding. If the cell had moved, even slowly, Cairns would have been just extremely wet rather than breaking records. But that detail is beyond prediction.
The result of unjustified expectations of prediction accuracy will result in the bureau being forced to cover itself and issue warnings whenever there is a minute possibility of extreme weather. The predictions will become meaningless.
The BOM has a truly superb observation network of rain radars, rain gauges and flood levels. Millions of people use these, especially in country areas, for everything from bringing in the washing to gauging when it will be possible to drive across a flooded creek. This network gives us remarkable ability to see what is happening. Thirty years ago, we were almost blind compared with today.
So give the BOM a break, at least on this matter. But there are two BOMs. There is the operational weather BOM, which does the daily forecasts and measurements, and then there is the climate change part of the BOM. And that is where the criticism should be levelled.
The climate models used by the BOM and many other groups regularly are used to predict, with certainty, the end of the world because of “global boiling”. But those models are little better than a guess. We have no idea what caused historical climate change such as the Little Ice Age of a few centuries ago and the hot climate of the Egyptian period. Climate models fail on this. The bureau’s failure to acknowledge model weaknesses is unscientific. Uncertainties must be stated. If the BOM proclaims its predictions for the year 2100 are excellent, it can hardly complain when people get upset when its forecast for this afternoon turns into a dud.
Another major problem within the bureau is the section dealing with long-term temperature measurements. Most long-term measurements have been modified (homogenised), almost always making past temperatures cooler.
The BOM does not dispute it has done this, but there is a huge argument about whether it has done it in a justifiable way, and BOM has failed to release all its data about these temperature adjustments. This is inexcusable and breeds concerns about the bureau’s scientific integrity.
There is also the habit of the BOM to associate every extreme, or record-breaking, weather event with climate change. In fact, record events are inevitable every year because of the huge scale of the observation network.
But the climate section of the BOM uses record events for political purposes.
Should we have an inquiry into the BOM? Yes. But the good guys of the BOM short-term weather forecasting department need to stand up against the anti-science catastrophists in their climate department. Otherwise they deserve to be tarred with the same brush.
Peter Ridd is a physicist, adjunct fellow with the Institute of Public Affairs and chairman of the Australian Environment Foundation.
Just filing this here:
by Peter Hannam, in The Guardian, Sunday 21st January 2024.
https://www.theguardian.com/technology/2024/jan/21/can-the-power-of-artificial-intelligence-be-harnessed-help-to-predict-australias-weather
"Kerry Plowright had his feet up and was watching TV one evening late last year when his phone warned of incoming hail.
“I was stunned when I walked out the door because there was just this roar,” he says, describing the sound of hailstones hitting roofs in the New South Wales town of Kingscliff. He had just enough time to move his cars under canvas sails, sparing them from damage.
Plowright isn’t alone in having little warning before wild weather during Australia’s seemingly relentless summer of extremes. This season may include a second tropical cyclone to strike Queensland.
The European Centre for Medium-Range Weather Forecasts’s AI model has the potential tropical cyclone Kirrily crossing the Queensland coast next Thursday night into Friday
The Albanese government has launched an inquiry into warnings issued by the Bureau of Meteorology and emergency authorities after complaints by councils and others that some alerts lacked accuracy and timeliness.
But Plowright’s case is a little different – his hail heads-up was triggered by data generated by his own firm, Early Warning Network.
Early Warning Network analyses data from radars and remote sensors to detect and issue alerts on extreme heat, rainfall and flooding. It counts local councils and big insurers among its customers.
Private businesses have long offered services based on data from BoM or agencies such as the European Centre for Medium-Range Weather Forecasts (ECMWF). But Early Warning Network is starting to test artificial intelligence models that promise to make a lot more weather information available both rapidly and at low cost.
“You have to pay a bucket load for [ECMWF] data,” Plowright says. “We don’t need now a supercomputer to go and run a forecast that will be extremely accurate up to 10 days, especially for extreme weather.”
Artificial intelligence “is going to be absolutely phenomenal with weather and ultimately climate too, once it gets there”, he predicts.
Juliette Murphy, a water resources engineer, is similarly excited. She foundedFloodMapp to give communities more time to prepare after monitoring devastating floods in Queensland’s Lockyer region in 2011 and then in the Canadian city of Calgary two years later.
FloodMapp uses machines that learn from each model run as well as traditional physics-based hydrology and hydraulic models. Even relatively basic computers can comb through “really large datasets” quickly to identify likely effects of a flood, she says.
Her clients include Queensland’s fire and emergency services. Its results complement BoM’s, helping authorities decide which homes to evacuate and which roads to close. “That’s important not least because almost half of flood deaths involve people in cars,” Murphy says.
A BoM spokesperson says the bureau had been “proactively and safely engaging with artificial intelligence capabilities for several years”.
“This area of research is one of many initiatives the bureau actively pursues to improve its services to government, emergency management partners and the community,” she says.
Justin Freeman, a computer scientist, ran BoM’s research team which was working on machine learning before he left in in late 2022 to set up his own firm, Flowershift.
Flowershift is building a geospatial model trained on existing observational data. “We would be filling in gaps around what the current forecast products are”, such as providing forecasts in remote regions of Australia or beyond, Freeman says.
“There’s a lot more flexibility to be able to explore things [outside BoM] and use technologies which are very new,” says Freeman, who still does contract work for the bureau. “We’ve got this whole new different class of models which are completely different to what [the bureau had] been running for the last 50 years.”
There are many potential uses for models that can analyse data cheaply and then supply localised information. Farmers, for instance, could ask, “Should I spray my crops this week?” and be told why or why not, Freeman says.
“It hasn’t been that long that we’ve had access to something like ChatGPT,” he says. “Look forward like another two years, five years – it’s just going to accelerate and get better and better.”
Some BoM and climate researchers, though, caution how much AI-based models, such Google’s GraphCast or Nvidia’s FourCastNet, can improve on numerical models that churn out a range of probabilities.
‘Very scary’: Mark Zuckerberg’s pledge to build advanced AI alarms experts
“For ‘simple’ weather forecasting and for downscaling physical model data I reckon [there’s] massive potential,” one bureau scientist says. “For warning us of real dangers when the atmosphere gets violent, I’d be very cautious.
“And with climate change, we need to better understand things that are well outside the norm.”
Sanaa Hobeichi, a post-doctoral researcher at the ARC Centre of Excellence for Climate Extremes, says there are still benefits despite the limitations.
Existing climate models typically offer only “coarse” resolutions, such as estimating rainfall changes over areas 150km by 150km. In Sydney, for instance, a model that size would capture the city, mountains and a lot else and so be of limited use.
Google’s GraphCast forecast model has a resolution down to 28km by 28km, while Hobeichi says some AI can model just 5km by 5km.
A challenge, though, is that machine-learning techniques inherit and potentially extrapolate imperfections of the traditional models they train on.
Jyoteeshkumar Reddy Papari, a post-doctoral CSIRO researcher, notes that the ECMWF was initially sceptical of AI but has lately started its own experimental model. It is also displaying several others on its website, including Google’s.
“Countries that don’t have good meteorological organisations are relying on these machine learning models because they are super easy to learn and are publicly available,” he says. “So some of the African countries are using these forecasts.”
Google researchers last year claimed GraphCast “significantly outperforms the most accurate” operational systems in 90% of 1380 targets. Tropical cyclones, atmospheric rivers and extreme temperatures were predictions it made which were better than traditional models and improvements are ongoing.
“One particular example we often mention was Hurricane Lee, because it was the first time that we observed in real time how GraphCast was predicting a hurricane trajectory that originally differed from the traditional systems, and eventually was shown to be the right trajectory,” said Alvaro Sanchez-Gonzalez, a Goggle researcher.
“It was detected in real time and it was verified by independent sources.”
Current tracking of the potential cyclone in the Coral Sea – to be named Kirrily if it forms as expected by Monday – will also be monitored to see how models compare.
ECMWF’s machine learning coordinator, Matthew Chantry, says AI models are “a very exciting avenue as a companion system for traditional forecasting” although the latter retains some advantages.
“Tropical cyclone intensity estimates are a good example,” he says. “It’s an open question whether these flaws are maintained as the technology matures – it is still very early days.”
Authorities act based on the probabilities calculated by traditional models but that needs a very large supercomputer. “With AI forecasts, this is dramatically reduced, with some estimates suggesting a 1000-times reduction in the energy to make a forecast. Cheaper systems could therefore be a force for equality.
“This reduced cost could also be invested into larger ensembles, meaning that we have a better idea of low-probability but extreme events that could occur.”
And as for predicting effects of a heating planet?
“The problem is significantly harder than weather forecasting, with less data,” says Chantry. “That said, in a changing climate, where evidence suggests an increase in extreme events, then any help with predicting these events has significant value.”
Competitive Enterprise Institute should dismiss him immediately for ad hominem.
"The litigation targets two writers: Rand Simberg, analyst at the rightwing thinktank Competitive Enterprise Institute, who published a piece comparing Mann to a convicted serial child molester, and the National Review blogger Mark Steyn, who in a blogpost favorably quoted Simberg and called Mann’s research “fraudulent”
Siliggy, did you contact the BoM for the information you are interested in? They clearly not that it is available on request.
Sliggy,
In 1893 the Great Flood of Brisbane left a path of destruction in its wake. The total rainfall in Brisbane over 8 days was about 20 inches (500 mm) and the Brisbane River rose 23 feet 9 inches (7.2 metres) above the mean spring tide and 10 feet over the previous highest flood mark (1890). Brisbane suffered approximately £2,000,000 worth of damages. The Victoria Bridge and the Indooroopilly Railway Bridge were swept away and in Queen Street, the businesses of Finney, Isles and Co, drapers, Perry Brothers, the goldsmiths, Hall Company, H. L. Davies, and Gordon and Gotch, all suffered major damage.
Brisbane wasn't the only area hit. The countryside for miles on either side of the Mary River was devastated and the loss of settlers enormous. The Mayor of Brisbane composed the following cable for the Lord Mayor of London: "Brisbane, Maryborough, Gympie, Ipswich, Bundaberg and Rockhampton inundated by floods. Destruction of property and loss of life enormous. Relief urgently required. " Against the wishes of the Queensland public, this message was never sent. (Western Mail, 18 February 1893 p.40)
This was a cyclone that mainly hit the Sunshine Coast. It was called the "Mooloolah Event". At Buderim, in the month of February 1893, 1819 mm of rain was recorded at the PO. Even higher rainfall was recorded at the observatory near Maleny by the most experienced meteorologist (Inigo Jones)at the time. The record there is still the highest rainfall in Australia for a 3 day period. The 1893 total rainfall in Buderim was 3848mm but the highest was in 1898 of 3997mm when Brisbane again flooded. The lowest rainfall in Buderim was in 1902 of 519mm in the Federation drought. The floods and drought had nothing to do with CO2 and there is nothing about the present that is exceptional.
Ooops. "back out again to the south west." That should be south east. After Townsville Sigma first went north then crossed it's own path to head off toward N.Z.
Cyclic based forecast. Not a meteorologist. Don't take this too seriously. Why not look at non BoM history for clues to what happened? Previous cycles point at Townsville where in 1896 cyclone Sigma hit. Newspapers record major destruction, "hardly a house escaped damage". An old track map shows the cyclone bouncing off the coast and heading back out again to the south west. Flooding rain was guided into central Queensland and all the way down the coast further than Maryborough. More than a week later newspapers compare the Mary river flood at three feet higher than a flood in 1887. As I type there is no warning for the Mary river here. http://www.bom.gov.au/qld/warnings/flood/index.shtml
There is no mention of either the 1887 flood or the 1896 one here. http://www.bom.gov.au/qld/flood/brochures/mary/mary.shtml#FloodWarningsandBulletins