Why not use models 12 hours from event?

[Atlas]

I've noticed this happening over the years, and I'm kind of curious why, and if its reasonable or not. Yesterday once we got within 12 hours of the storm a lot of people say, "Its nowcast time stop looking at models." While I agree with this to a certain extent since this field is called meteorology and not modelology. I still disagree as well. As we get closer and closer to an event shouldn't the models get better and better. In fact I still find the NAM useful at 6 and 12 hours. It was obvious from the 18z NAM yesterday where the best bands would continue through the night. It comes down to the saying "garbage in garbage out" because once we are 6-12 hours out there should be less chaos on the models. Just wanted an opinion from you guys for future events. While I agree with the '.nowcast' sentiment. I also find it useful to look at models' 9-12 hour forecasts since there is a good chance it will help you see where the storm might head and which places will see the highest QPF Thoughts?

[Burghblizz]

Most people say "it's Nowcast time" when the a fore mentioned models don't show what they want.

But sure, I would think verification would be at their highest for the models, so a combination of that plus real weather would be a good approach.

[Ellinwood]

brb watching the radar...

Models and nowcasting don't exactly go hand-in-hand. Models are supposed to be used for getting the right idea, and when you hit that <12 hour window mets will resort to nowcasting via radar and mesoscale/current analyses from places like the SPC to track the storm. It is often easier and more accurate to nowcast, especially considering the mesoscale features of the system that simply won't be picked up on by the models. Models like the GFS/NAM/etc. give us very little useful information inside of 12 hours compared to what we can see from current analyses and radar information.

Yesterday the model watchers were still talking about the possibility of BWI-DCA and the surrounding areas getting 4-6"+ because of the QPF forecasts on the models, but you could see that it was a long shot and was very unlikely via the radar and nowcasting.

[JoMo]

I've noticed this happening over the years, and I'm kind of curious why, and if its reasonable or not. Yesterday once we got within 12 hours of the storm a lot of people say, "Its nowcast time stop looking at models." While I agree with this to a certain extent since this field is called meteorology and not modelology. I still disagree as well. As we get closer and closer to an event shouldn't the models get better and better. In fact I still find the NAM useful at 6 and 12 hours. It was obvious from the 18z NAM yesterday where the best bands would continue through the night. It comes down to the saying "garbage in garbage out" because once we are 6-12 hours out there should be less chaos on the models. Just wanted an opinion from you guys for future events. While I agree with the '.nowcast' sentiment. I also find it useful to look at models' 9-12 hour forecasts since there is a good chance it will help you see where the storm might head and which places will see the highest QPF Thoughts?

I agree, however, sometimes the models are still wrong. I remember a storm that didn't quite go as expected many years ago. Winter storm warnings were issued from both Springfield, MO and Wichita, KS NWS offices. A couple hours after they had issued the warning Wichita looked at the satellite and saw the storm was heading more NE instead of E and cancelled all warnings and advisories as they would just get rain. Springfield continued to ride the models with the winter storm warning and busted horribly due to the models being incorrect. This was within the 12 hour window as well.

[Analog96]

I've noticed this happening over the years, and I'm kind of curious why, and if its reasonable or not. Yesterday once we got within 12 hours of the storm a lot of people say, "Its nowcast time stop looking at models." While I agree with this to a certain extent since this field is called meteorology and not modelology. I still disagree as well. As we get closer and closer to an event shouldn't the models get better and better. In fact I still find the NAM useful at 6 and 12 hours. It was obvious from the 18z NAM yesterday where the best bands would continue through the night. It comes down to the saying "garbage in garbage out" because once we are 6-12 hours out there should be less chaos on the models. Just wanted an opinion from you guys for future events. While I agree with the '.nowcast' sentiment. I also find it useful to look at models' 9-12 hour forecasts since there is a good chance it will help you see where the storm might head and which places will see the highest QPF Thoughts?

That statement was indeed correct. The GFS, for example, from 12 hours out, only gave me 0.81 QPF. The earlier runs were much more correct.

[sauss06]

At some point, a forecasters experience of their local would have to come into play as well, correct?

[LongRanger]

one reason: by the time you see model run's output, several hours have elapsed since the taking of observations that serve as input for that run

those several hours are a significant % of the "12 hours from event" that you mention

radar, by comparison, is much more current... often just a few minutes old, and reflects what has actually happened

[Reale WX]

Models do not predict EXACTLY where the mesoscale bands are setting up and where the precip is cutting off... radar and satellite do. It is all about observations and current information (as was mentioned) from spc and other sites. The RUC is useful because it comes out every hour, however it consistantly placed the strongest bands too far west until they were already set up over Eastern NJ. MM5 is good too. I take one glance at the new 12,18 or whatever model, see if its still on board with whats expected and continue to forecast and make updates.

[Analog96]

one reason: by the time you see model run's output, several hours have elapsed since the taking of observations that serve as input for that run

those several hours are a significant % of the "12 hours from event" that you mention

radar, by comparison, is much more current... often just a few minutes old, and reflects what has actually happened

This, too. For example, the 00Z GFS usually starts running around 10:30 PM. 00Z is 7 PM, so the "initial" data is already 3.5 hours old by the time we can see it.

[baroclinic_instability]

NOWcasting is very important, but the question then becomes, "Ok, this is happening and this is how it has trended the last, say 6 hours, now what?". Under that scenario, a good forecaster can add input into the forecast, but models can be of help in situations like that. What I would do is glance at the latest guidance and see how well they are handling the latest trends. Sometimes models won't catch on to features until the very last second, and most of the time these are mesoscale/sub-synoptic features such as heavy bands, for instance. Heavy bands alone can form for a various number of reasons, but being able to identify the forcing mechanism is problem number one followed by making use of the guidance to determine how long said banding may last. This is one example, but I can guarantee, when actually forecasting, no meteorologist would simply stop looking at the guidance without giving it consideration.

[WeatherFox]

Yes, for sure - local geography, climatology, & modernization of nearby waterways, hills, valleys, urban heat island, climate data etc are certainly part of the weather forecasting equation.

At some point, a forecasters experience of their local would have to come into play as well, correct?