Tornado prediction or forecasting possible using Supercomputers?

[akerkarprashant]

TL;DR

Can Supercomputers predict or forecast Tornadoes?

Can Tornado be predicted or forecasted using Supercomputers ?

Thanks & Regards,
Prashant S Akerkar

 

[Baluncore]

Can Tornado be predicted or forecasted using Supercomputers ?

Yes. That is how weather forecasting is done now.
One hour predictions are accurate. One day predictions are less accurate.
Radar will show the current position. How much accuracy do you require?

 

[akerkarprashant]

More the accuracy more the better. Once predicted accurately by Supercomputers, warnings can be issued by Television broadcasting channels, radio & other electronic media to alert the people as where a Tornado is going to take place.
Adequate measures can be taken by people once alerted forehand of possible occurrence of Tornado.

https://www.bbc.com/news/world-us-canada-59667108

Example : Recent Tornado in Kentucky.

 

[Baluncore]

Prediction of a tornado will not reduce the physical damage. There may be no time for evacuation, so houses will still need to be built with safe tornado shelters.

 

[akerkarprashant]

As you mentioned, there may be no time for evacuation, if the prediction was done well in advanced with 100% accuracy by Supercomputers, this could have been possible?

Buying House/Property (General- Non life insurance) policy can assist in claiming money from the General insurance company for the damage caused by the Tornado.

 

[Baluncore]

As you mentioned, there may be no time for evacuation, if the prediction was done well in advanced with 100% accuracy by Supercomputers, this could have been possible?

The error in the prediction will probably be more than the distance you can evacuate in the time available. There will be traffic jams and deaths on the road during evacuation. It is lower cost and safer, to have part of the house built as a secure shelter for the residents and for storage of valuables.

 

[anorlunda]

if the prediction was done well in advanced with 100% accuracy

With today's technology, 100% accuracy may identify the likelihood of a tornado. Still, the location where it touches down, and where it travels after touchdown may be 50 miles wide. It sounds like you're thinking a forecast pinpointing it to a single house or neighborhood.

 

[hutchphd]

Also one should consider the predictions of Hurricanes. The programs have become pretty good at determining the track of these storms but it is apparently (I have heard the "experts" say this) much harder to determine the strength of the storm with time. For tornados one needs to decrease the granularity and probably other factors get more difficult, too. This is akin to calculating the onset of turbulent flow and so the butterfly effect is doubtless operating in overdrive.

 

[Rive]

Can Tornado be predicted or forecasted using Supercomputers ?

The possibility of a stormy weather which may lead to a tornado can be predicted by days ahead with good accuracy, but the exact locations and tracks cannot. They are fairly random within an area (this area may be really big).

Once a storm is formed and starts moving the area it might affect can be predicted fairly well, just as the chance of tornadoes it may bring, but it'll not be absolutely accurate. Storms still can jump, shift and turn and do unexpected things. This kind of prediction/warning (in case of tornado-relevant storms) can be a few hours ahead.

Once an actual tornado formed/spotted an even more accurate warning can be issued, but this kind of warning relevant only for smaller areas and may come less than hour before the problem arrives. This really does not gives enough time for anything else but to seek shelter (evacuation attempt likely would make things far worse - a car is NOT a shelter).

 

[russ_watters]

A tornado watch is issued hours before, indicating conditions favor tornado formation. A tornado warning is issued when a probable tornado is detected. These alerts are pushed to cell phone pop-ups (overriding screen locks and volume controls). A warning indicates you may have only seconds and at most a few minutes to find shelter before the tornado hits. They don't guarantee a hit, but cover an area of a few square miles where the spinning top is likely to wander.

 

[russ_watters]

Also one should consider the predictions of Hurricanes. The programs have become pretty good at determining the track of these storms but it is apparently (I have heard the "experts" say this) much harder to determine the strength of the storm with time.

True, but the intensity forecasts have gotten good enough that decision makers need to start trusting them more. Right now they are willing to make decisions of forecast tracks but not forecast intensity. Louisiana wasted a day when they should have been evacuating before Ida this summer and then essentially blamed the prediction for not being convincing enough. (governor? Mayor of New Orleans? I don't remember who exactly said it)

 

[akerkarprashant]

 

[anorlunda]

The whole idea of large scale evacuations is a mess. A few years back on PF I posted about the lunacy of evacuating 6 million people from Miami/Fort Lauderdale. They not only clogged the highways, but many of them had to go as far as Atlanta to find accommodations. That means driving 10 hours at 70 mph, or 70 hours at 10 mph searching for food and fuel the whole way.

IMO, it would be wiser to invest thought and money into shelters closer to people's homes. Settlers in the Midwest long ago figured out that the needed a storm shelter for each house. IMO for cities, the idea of shelter should replace the idea of evacuate.

On the weather map this week, the tornado watch area was on the order of 25% of the area of the entire lower 48 states. Comparing to Europe, it could be like evacuating France.

 

[Astronuc]

I was reading the following article about the recent extreme tornadic activity.

A tornado expert explains why last week’s twisters were so devastating
The science behind tornadoes is still evolving
https://www.theverge.com/2021/12/13/22832647/tornadoes-science-behind-climate-change

There are a lot of different mechanisms that we’re aware of that potentially could produce a tornado. Generally speaking, you need a supercell thunderstorm. That supercell thunderstorm begins to rotate at the mid levels and that process produces a downdraft that then causes rotation at the surface to become more intense, and eventually we refer to it as a tornado when it exceeds 50 miles per hour. But the exact mechanisms of tornado formation are still an open topic of scientific research.

 

[anorlunda]

A tornado hit my RV park in 2017. It happened 5 hours before the arrival of Hurricane Irma. It is common for tornadoes to spawn in the outer fringes of the front right quadrant of a hurricane.

Interestingly, the real time NWS report based on radar, got the location wrong by 40 miles.

This link talks about spawning tornadoes from hurricanes. The ingredients are different than Great Plains tornadoes.

https://www.livescience.com/37235-how-hurricanes-spawn-tornadoes.html

 

[Astronuc]

The ingredients are different than Great Plains tornadoes.

'Tornado Alley' has traditionally been associated with the region from W. Texas (mainly panhandle) up through Minnesota into Wisconsin. However, it appears that the region for tornadoes has expanded or shifted slightly eastward, or both (into the Mississippi and Ohio River Valleys). We now see extreme weather events like last week in places that have not experienced such activity in last couple of hundred years or so. The Gulf and Atlantic coasts experince strong thunderstorms and an occasional hurricane, but that weather (and now climate) can be experienced in the NE US on a more frequent basis.

 

[jsgruszynski]

Yes. That is how weather forecasting is done now.
One hour predictions are accurate. One day predictions are less accurate.
Radar will show the current position. How much accuracy do you require?

However such simulation are still highly susceptible to "garbage-in-garbage-out" - can you actually determine the REQUIRED initial conditions and boundary conditions for the simulation to have any fidelity.

This is a problem for Navier-Stokes (differential equations for fluid dynamics used in weather simulations) which has chaos baked into the equations thus the Butterfly Effect applies if you do NOT have the correct initial and boundary conditions. This is why weather forecasts only go out 10 days: beyond that the odds of being completely wrong are higher than the odds of being a little bit right. Because of chaos in Navier-Stokes.

So the great hope is with a IoT mesh of sensors, you can get some better values for initial and boundary conditions but the improvement probably requires an exponential increase in sensor data for a linear improvement in accuracy.

Rather than simulation, I'd put money on remote sensing such as more advanced radars - at least that's empirical rather than the often wishful thinking of simulation.

 

[akerkarprashant]

Thanks.

https://www.google.com/search?q=fas...X&ved=0ahUKEwj4hcHtuan2AhWXxjgGHabsCyMQ_AUIBg

Thanks & Regards,
Prashant S Akerkar

 

[Frabjous]

High performance computing is way more than raw computing power. There are such a large number of processers, that the probability of one failing during a long calculation is approaching one. Memory and interconnects are also important hardware considerations. The numerics that drive these ratings can actually be a surprisingly small percentage of the compute time. It is also well-written codes that can deal with these and other large machine issues. It is having the tools to process the data. From my experience, “production run“ size does scale with the machine. Many times “number of runs” absorbs the capacity.