When BOM / Weatherzone / Models get it wrong
- Thread starter Edge72
- Start date
-
Join the conversation with a better online community
To be fair, Everyone is entitled to their opinion.
I just find it a bit weird (ie . obvious) that Edge only ever appears on here and comments when the BoM get something slightly wrong. He/she never comments when they get it right. Also, never offers an alternative opinion or some educated reasoning and evidence why the BoM may have been wrong.
= troll.
I think BOM have done a great job in the Sydney area in recent weeks.
Not a whole lot has changed. Neg zonal pattern. Plenty juice upstairs which has lead to persistent periods of onshore, shower activity. Coupled with varying intensity of inland trough activity with MJO cycles.
NBD IMO.
Given what is going on the SEQ trough thread, i would appreciate what folk's expectations of weather modeling are.
Mine is that they " improve over time "which has been spectacularly demonstrable across a suite of models in the last 10 years.
The immense computational power that drives these models and the brain warping maths and physics behind them is not only an asset for weather forecasting but science in general.
That folk are quibbling over what is effectually a level of model precision that was unimaginable 20 years ago shows how far we have come.
Give me science over witch doctors anyday.
Well naturally models do get better over time. That being said though, they did very poorly for this particular scenario. It's no one's fault though. As good as the models usually are, they aren't perfect, and not every imaginable weather pattern has happened in our lifetime or been recorded into the data they utilise. I think for the most part, some of us just feel disheartened because we had too high an expectation for this week's rain. We rely on BOM and models to give us an idea what will happen so we can prepare, but it also excited some of us. I don't blame the BOM or models though. It's the weather, part of what makes it so interesting is that no one person has all the answers.
Improving over time is a good one. And I remember 80s when ECLs would pop up with basically no prior warning, and that cyclone paths were forecasted largely by extrapolating recent movement, and changes of direction were never predicted. So definitely improved over time.
I think another one is, are the models better than the alternative? If we get rid of all the models would we be better off relying on expert opinion on how current patterns are likely to change? No one is even really trying this.
Or perhaps, yes the models are helpful, but the current mob running them are incompetent and if someone else took over they would do better. Well there are a bunch of different agencies that run models, the European, the US, Australia, Canada, Japan and more I'm sure. The European is regarded the best according to statistics, and no one seems to seriously question this. One issue is all are government, so perhaps a private company would do better. There is nothing in principal stopping this, and there was some noise in 2016 when Panasonic claimed to have developed a better model than anyone else had, but I haven't heard anything further. An issue there is whether the company can make a profit out of it, and whether we would be better off with a commercially run model which may have more restrictions on obtaining data unless you pay the $.
A further issue out of the SEQ event is BOM's warning protocol. One thought I had was that the BOM should have cancelled their warnings much quicker. Ironically last night I thought it was clearly over, but now this morning I've changed my mind and think severe weather is still possible but the warning has been cancelled. BOM may know something I don't.
I can imagine that for last night's case that it is hard to retract a warning once given, especially if there is any chance of an error. Yesterday morning it seemed all evidence pointed towards severe weather being the most likely outcome and I think everyone in the SEQ thread was in agreement. There were a few posts pointing out that it was on a knife edge being so close to the coast and a small shift could see it out to sea. So issue a warning, with a small chance that the weather doesn't eventuate, then you point to all the models predicting severe weather and say 'yes I made a mistake, but the warning was justified'.
In contrast last night with the heavy rain being further east I made the prediction that nothing would happen. What if I was wrong? Well I'm a random armchair expert and if wrong I might apologize for my error, or maybe just pretend it never happened and continue to post my thoughts in this forum. Maybe someone might have a go at me if they have dislike me for some reason. What if the BOM forecaster decides nothing is going to happen and retracts the forecast? A quick retraction will raise eye brows and someone will jump on the BOM for confusing people and quickly changing their mind. If severe weather then does eventuate, the BOM would likely be crucified. It is likely such an event would attract the attention of the politicians, and no one in a government agency wants to do is make a mistake that attracts the attention of the politicians. A BOM forecaster needs to be 110% sure before quickly retracting a warning once issued if circumstances change.
And some users are trying to hold the BOM to a standard of accountability it has not planned its operational or data and media management for.
Be interesting to see the outcome of this court case against BOM.
Balloon company in Victoria trying to claim damages based on unfortunate consequences, legal expenses and insurance claims based, they claim, on reasonable actions consequent to reviewing BOM information.
https://www.theage.com.au/national/...u-to-court-over-forecast-20210215-p572jv.html
This kind of development suggests BOM would prefer to continue to err on the side of caution.
Which can only lead to a reduction in accuracy over the long term as BOM modify forecast data to protect themselves based on legal advice. BOM needs to be able to issue forecasts and warnings without fear or favour and be absolutely independent.
The emphasis in that case will not be the accuracy of a forecast. It will be in not having appropriate procedures to ensure that an accurate forecast was disseminated.
Disclaimers and qualifications to be added to all BOM data feeds no doubt.
I’d be surprised if the case succeeded given the broader public expectation and consequently existing common law position that a generic forecast may not apply locally in all instances.
This would see a significant shift in our existing understanding of public liability with insurance implications across travel, agriculture, transport.
To my knowledge those suffering damages due to flood, fire, cyclone damage and their insurers haven’t gone after the BOM wrt. the absence or presence of emergency warnings.
Insurers have relied on their Act of God out clause, and the underinsured and uninsured have always turned to government emergency funding in dealing with ‘unforeseen’ events.
Adventure companies engaged in risky sports may have less claims under our common law than pastoralists who have experienced stock losses - our existing insurance suggests a ballooning company is trying to pass off an inherently risky activity as normalised, palming off the inherent risk to a third party -
Might not fly....
It's not an action about the forecast. They are suing because, although the BoM had an accurate forecast, they delivered an outdated, inaccurate forecast.
That is understood.
However, the degree to which BOM is liable for the actions of people relying on BOM data, irrespective of the status of the data is moot.
Slightly OT here, But does anyone know if any of the large / popular weather models have any element of AI involved?
It seems like a perfect job for an NN system to me?
Big data sets, lots of inputs, you can generate results for +6 hrs, +12, etc.
and then you get a perfect result when you get to that time to train your previous inputs.
Surely with time, you could train an AI NN to be pretty damn good at weather prediction?Similar threads
