For example, in San Diego it looks ok for the next week or so for astronomy and craps out after that. I tend to feel that when the longer range forecast is poor for astronomy or other desired activity, they are accurate; but when the longer range forecast is good for the desired activity, the forecast is inaccurate.. Must be a conspiracy.
How accurate do you feel long range [ 7+] day forecasts are?
Posted 28 February 2021 - 01:49 AM
Here in the Northeast, most forecasts past three days are inaccurate.
- lsfinn and Ring_Singularity like this
Posted 28 February 2021 - 03:00 AM
I always wonder this myself. I think they are only really accurate for 72 hours; beyond that, unforeseen changes start playing a bigger and bigger role. Sometimes they can see rain coming a week ahead out here in Los Angeles, and that's impressive, but I think the temperature frequently deviates from the expectation. Two week forecasts are pretty useless. Notice how they generally seem to predict more of the same, regardless of what the current conditions are?
Speaking of conspiracies, the weather out here is always pretty good near the full Moon, and it starts to crap out a week before new Moon. I once heard a study say the Full Moon actually increases cloudiness, but I'm willing to bet most CNers will disagree . We all know nothing attracts inclement weather like a moonless night.
As for the forecasts, have there been any studies comparing the predicted temperature/precipitation with the actual values? What are the mean deviations at one day, three days, a week, etc.? How unpredictable are certain places vs. others? How has accuracy been improving over time? How do the different services/models compare in accuracy? Seems like a good grad school project. Any meteorologists out there with insight into this topic?
Posted 28 February 2021 - 03:12 AM
Here, UK, not a chance. Meteoblue do a new model run 4x a day and so every 6 hours it can and often does change. I have found that yesterdays forecast for today is very often wrong.
Does mean that planning is almost impossible, or meaningless.
One oddity I have found is that the Meteoblue site has a "rain forecast animation", the problem is they display the rain for the last 1.5 hours, and then for the next 0.75 hours. Why is it not the other way around? 0.75 hours ago and the next 1.5 hours. Would seem to make more sense.
Suppose if you said a 50% chance of being right and with 28 possible model runs the chance is (1/2)28 = 0.0000000037, or 0.00000037% chance of getting it right. Lets call it 0.
- deepwoods1 likes this
Posted 28 February 2021 - 03:54 AM
In California, they are pretty bad at forecasting when it comes to weather patterns where I observe. They can't predict clouds or precip in the mountains that well for even a day ahead or day of. I make a call based on what I see on the satellite the afternoon of, the patterns I see on webcam at a location near the observing site, and what I see looking at the mountains in the late afternoon.
There is a lot of "picket fence" effect to the forecasts where they predict overcast for a night, clear the next, overcast the next, clear, and so on...and manage to get it exactly backwards for a week straight, despite updating every day. They can get stuck out of phase, then the next month they can be on again.
They get some of the general precipitation right for maybe 2-3 days out, but since it rains so rarely here that isn't exactly difficult... I assume that much of the problem with multi day forecasting here is that there is less time between when weather moves in from off shore. That and the mountains cause a lot of north/south disruption that can be trickier at a local level. Honestly, we had much better long range forecasts in the midwest, but I attribute that to the bigger patterns being less disrupted in the short term.
Posted 28 February 2021 - 04:35 AM
- deepwoods1 and jcj380 like this
Posted 28 February 2021 - 04:45 AM
Here in the Northeast, most forecasts past three days are inaccurate.
Agreed and if a major winter storm is in sight, they are inaccurate beyond 48 hours for the most part.
Posted 28 February 2021 - 06:25 AM
Posted 28 February 2021 - 06:37 AM
Well, San Diego is an exception. The forecast for any time X days in the future, where X is smaller than 5,000, is partly sunny with moderate temperatures. But that's really a statement about climate, not a weather forecast.
Here in the U.S. Northeast, I find the forecast generally pretty reliable for 3 or 4 days out, in most circumstances. If you track the forecast daily, you get a pretty good sense for how reliable it is. Some kinds of weather patterns are very stable, in which case the 3-day forecast is rock solid. Sometimes the weather is unstable, and it's impossible to guess from one hour to the next.
Now that's all for general daytime activities, where thin cloud layer, or a few big puffy clouds, don't matter at all. Astronomers are a lot more picky. Although I usually have a pretty good hunch about what the next night's observing will be like on the morning of the same day, I never get my expectations too high (or low).
In general, I'd say that the forecast 6 or 7 days out is slightly better than a roll of the dice, and beyond that it's exactly as good as a roll of the dice.
- Jon Isaacs likes this
Posted 28 February 2021 - 06:44 AM
In VA. my long range forecast is about 2 hours even then it can change. So much precipitation since spring of last year here. I believe we have had 3 sunny days since the new year maybe 4.
Edited by jjbag, 28 February 2021 - 06:44 AM.
- wrnchhead likes this
Posted 28 February 2021 - 08:25 AM
Not accurate here, in fact we have a saying "If you don't like the Weather in middle Tennessee, wait a few minutes. It'll change!"
- jcj380 likes this
Posted 28 February 2021 - 08:38 AM
Posted 28 February 2021 - 09:59 AM
At T-3 days, you might start to see some agreement between some of the models
If you want to play around with the various weather models and their forecasts, see: https://weather.us/m...0228-1500z.html
- jcj380 likes this
Posted 28 February 2021 - 10:11 AM
Two days out, maybe 70% accurate. At 7 days, little better than a historical average, unless something like a polar vortex is involved.
Posted 28 February 2021 - 10:28 AM
Hah, ha, ha. Living in southeast Texas, I know they cannot predict reliably more than a few hours ahead. For example, two weeks ago we have temps remain below freezing for 3 days straight, even getting down to 10° one night (this was unusual for the area however). Then the following week temps during the day are upper 70s to low 80s with nightime lows in the upper 50s to low 60s. And it can go from exceptional drought to flash flooding in a matter of hours.
Posted 28 February 2021 - 11:04 AM
Up here in the San Juan's we have serious micro climates. From low lying Islands in the Salish Sea, to the foothills, and finally the Cascades to the east. Long range forecasts are only an indicator of potential cloud breaks. I only use them to charge my batteries and otherwise prepare to get ready. On observation night I have 2 choices; the front yard facing east or the back deck facing west.
- wrnchhead and Alan D. Whitman like this
Posted 28 February 2021 - 11:16 AM
Here in Colorado, anything past 7 hours is a crap shoot....but then, we can go from 70F to 8" of snow and a high in the teens by morning....
Posted 28 February 2021 - 11:38 AM
There are times of the year were cloudy weather is more likely than others. I usually don't plan be astronomy trips until late June. Our club's spring star parties are canceled more often than not.
July - October weather is more reliable in Reno. The only problem with July-September weather is California wildfire smoke.
One of the keys to enjoying trips with an astronomy focus is to have a backup plan for when you can't observe because of weather.
I think weather forecasts more than two days out are subject to change.
Posted 28 February 2021 - 01:29 PM
Biker Dib wrote: "Living in southeast Texas, I know they cannot predict reliably more than a few hours ahead. For example, two weeks ago we have temps remain below freezing for 3 days straight, even getting down to 10° one night (this was unusual for the area however). Then the following week temps during the day are upper 70s to low 80s with nightime lows in the upper 50s to low 60s."
Fact is, your exceptional damaging Arctic air mass was accurately predicted two weeks in advance by the models put out by the super-computers, but the state officials did not warn people to store water, and to turn off their water supply so the pipes would not burst and cause major damage to their home.
In the rare case that you actually need a reliable long-range forecast, such as for eclipse-chasing, compare all of the available models. If the models agree, you can probably rely on them. If the models disagree, or if your favourite model keeps changing, then you cannot have confidence. An experienced eclipse-chaser will drive to somewhere with a simple, reliable, upper ridge on the weather charts.
A sophisticated user of weather forecasts will get value from them just as a sophisticated user of forecasts of who will win the football game or the election or what the stock market will do will get value from the forecast while always ground-truthing them by what they see evolving in the sky.
Weather forecasts began as something to save mariners and serve farmers. For decades not a single unforecast hurricane has hit any coastline in the world.
A hay farmer who cuts his crop needs about five dry days for it to cure before he can take it up. Once he has cut, he is committed. If it rains heavily he can lose his crop. In my experience hay farmers are usually very sophisticated users of weather forecasts. If there has been a long dry spell and he has gotten part of his hay crop in, he wants to know whether he can continue cutting. If he hears a forecast of significant rain on day four he knows that the weather pattern is changing and a significant system is due through in the medium term. If it rains on day three instead of day four he understands that he was warned that the dry weather pattern was changing and the rain did indeed come, but it came a day earlier than forecast. He is grateful that he was warned.
But the unsophisticated weather forecast user who back on Wednesday planned a barbecue for the weekend says: "Those fools blew the forecast again! They forecast sunshine on Saturday and rain on Sunday, but it rained on Saturday and was sunny on Sunday!"
Well the barbecue guy got to blow off steam and show his ignorance. But the hay farmer was darn grateful that he was warned that a significant rainfall was coming in the medium term, a significant rainfall did occur, and he did not lose tens of thousands of dollars. The farmer is sophisticated enough to realize that although the timing of the forecast was a little off, the significant predicted change in the weather did occur.
When I worked at weather offices I delighted in working with hay farmers. I didn't just say that it would rain or not rain. I explained what weather systems were in play and what they might do, with some idea of the chance of a drenching rain versus a light sprinkle that would be no worse than the dew that they would have gotten on a clear night. The decision to cut or not remained the farmer's -- the optimists might cut and the pessimists might not. Some summers we didn't get the wanted large blocking upper ridges. Then you had to choose the periods when it was likely that whatever rain occurred wouldn't be too heavy or prolonged.
Any long range forecast can only deal with organized large-scale weather systems. Over and over again I have read posts on various astronomy discussion lists saying "The Clear Sky Clock blew it again, they promised clear sky and I got cloud." It would be obvious from their description that they had local low cloud. Guess what: the fine print on the Clear Sky Clock clearly says that it cannot predict local low cloud. How could anybody with any understanding of physics possibly think that the CSC could predict local stratus cloud, fog (just stratus cloud that is on the ground), or small convective clouds?
When you do use the CSC, do not just look at the bar graph for your site. Click on the squares for the hours of interest and you will see a chart that looks like a satellite chart. If you are located in the midst of a large area of blue, then it will probably be clear unless you get local low cloud. But if the chart shows just a narrow area of blue, obviously timing may change. If you just look at the CSC bar graph you will not know whether your blue square is in a large forecast clear area or a tiny one.
A retired meteorologist who is 12/12 for eclipse-chasing (10 totalities and 2 annulars) by using the available forecast products in advance to know whether to drive all night to get to a more favourable weather pattern, looped satellite images in the last 12 hours or so before the eclipse, and what I see evolving in the sky in the last hours
- Jon Isaacs, BKBrown, Tyson M and 2 others like this
Posted 28 February 2021 - 03:00 PM
Wait, so now instead of expecting some accuracy in forecasting, we should all become meteorologists? I’m sorry, that’s unfair. We default to the “experts”. I don’t need to be a Doctor or a Structural Engineer for society to advance. I need them to do their jobs accurately so I can do what I do. Having worked on a dairy farm in my younger days, the most accurate forecasting came from the cows. If they’re hanging together by the trees, it’s probably going to rain. Having kids in school shows how poorly forecasting is. It’s a family joke. So many times school has been canceled for the potential of snow, and we get rain, or a couple of times nothing. Just last week my region was supposed to get 16” of snow. Snow all day. Potential blizzard conditions. Doom and gloom. We got four, maybe five. Some wind. I wish my work allowed me to be wrong so much. I’d have less stress. I get it, the Earths systems are far more complex than computing power. Waaaaayyyyy too many variables. All weather is local. As a Scout leader, final decisions regarding the day many times happen within an hour of the event. Bottom line, as consumers of whatever it is, weather forecasting or telescopes or whatever, we expect that what we are told is the truth. Clear and steady skies!
- jcj380 likes this
Posted 28 February 2021 - 06:20 PM
Less than 25% and I'm being very generous. Around here even a 3 day forecast is questionable. I have more trust in a 24-48 hour forecast
- jcj380 likes this
Posted 28 February 2021 - 07:14 PM
In my experience flipping a coin would beat a 7-day forecast where I have lived.
- deepwoods1 and jcj380 like this
Posted 28 February 2021 - 08:49 PM
Zero confidence in them. They can’t even get current conditions right. Been sitting under solid gray skies while seeing the conditions are reported as “clear”
But what was the forecast 7 days ago? That's the question.
Here is SE Texas they are pretty spot on as far as temperatures go. Not so much for cloud cover and rain. However they are usually pretty accurate on humidity. With temps and humidity I can pretty much deduce what the observing conditions will be like up to a few days out.
With all that I still make go-nogo call around noon on the day I plan to observe.
Posted 28 February 2021 - 10:38 PM
PNW wrote: "Up here in the San Juan's we have serious micro climates. From low lying Islands in the Salish Sea, to the foothills, and finally the Cascades to the east."
Yes, knowing how your local topography weakens (in your case, in the rain shadow of the Olympic Mountains) or enhances weather systems is critical here in the west. The topography will have differing effects depending on what direction the atmosphere's steering flow is from (taken as the flow at 500 mb, or 18,000 feet).
The Seattle TV stations are going to focus on the expected weather in greater Seattle, but as you undoubtedly know, the San Juan Islands will frequently have less precipitation, less intense precipitation, and more breaks in the clouds than Seattle. And since they have less significant weather than many other parts of western Washington, the San Juans will receive less attention from the weatherman who will naturally focus his remarks on where the problems are, and also on where the populous areas are.
Local knowledge is important in understanding what to expect from a given weather pattern at your site. And, no, you don't need to be a meteorologist -- you just need to be an observant person who attempts to understand what they see happening in the sky and on satellite images, as most successful amateur astronomers are (just like long-lived commercial fisherman are!).
- Jon Isaacs and blue like this