All the Yoshino cherries and magnolias here are toast. Noticed that driving this afternoon.
My plums were brown even with covering. They were in full bloom. A few tight buds prevailed so might still get a few. Covered sauzee swirl and even the open blooms look good. One nectarine will still have some later blooms as well as a few others.
This past night saw temps around 19 from midnight until daylight. Gusts up to 40 mph midday which reeked havoc on tree covers
1 Like
Temperatures are all over the place here. Kansas was 80 degrees yesterday and it rained hard in the evening and showers continue. We will cool down soon as well. Let’s see what happens it’s anyone’s guess.
I’m no meteorologist, but over the years, I have not seen this to be the case in my region. In my memory the lows are as likely to sink below as ride above predictions- but predictions tend to be pretty accurate-at least in day preceding them. Also, it only takes one night. Two years ago we had an early spring, but didn’t have a single hard frost in April. We got though about the first week of May when we finally got the cold blast that destroyed my stone-fruit crop, which had fruits with brown embryos- one fricking night near the end of any possibility of such cold. Every fruit loss in my region that I can remember has occured because of one exceptionally cold night- as far as hard spring frosts. 24 or below. It tends to coincide with early springs, of course.
For this latest event, the ones that seemed to do more of a worst-case were calling for 19F a day or two before. The ones that were lacking a worst-case were calling for 22F. It was 21F for me, but the average around me was more like 23F.
Further away from the freeze, there was a bigger gap as the worst-case forecasts were calling for 19F or lower, and the non worst case places were calling for 24-26F. It is further from the event that there is often more of a difference as the worst case places are thinking “what is the lowest 20% range of the spread of what could happen” and the non worst case places are thinking “what is the midpoint”, and because there is more uncertainty in the forecast the spread is more. In this particular event it ended up closer to the early worst-case, but it doesn’t seem to usually go that way.
At least this is my read of these forecasts after watching them over the years. What happens most often is the worst-case numbers don’t come to pass and their numbers slowly bump up as we approach the event. To be concrete, I consider most forecasts to be worst-case and the two that are not are the Euro model (which is available in the WeatherPro app) and the DarkSky model (which is available in their app). Apparently-worst-case models include NWS (many apps use), Foreca, and AccuWeather.
… Back on the “Oh crap” topic, so far I have found all of one flower that looked damaged, one fully open peach blossom was browned. I am expecting more damage to show up still but am getting a bit more optimistic. Fortunately it is all done now, it was only 28F this morning and this should (should!!!) be the last temps below 25F.
I am wondering a bit if the wind didn’t actually help me avoid the freeze. Usually my worst damage is on windless nights when there is an inversion (the cold settles to the ground).
3 Likes
But for my region, the forecasts fluctuated both ways leading up to the event with the lowest holding true. Perhaps you are more of an optimist than I am so your memory works differently- I have to admit I am not referring to any written record. I do, however, obsess on weather forecasts and often look at them twice a day, sometimes from multiple sources.
There was variable damage here. Some pawpaw flower buds look droopy. Some Asian pear buds were black inside even though most were only barely exposed. Korean Giant and Shin Li had the most damage with maybe 5-10% of flowers still looking good. I guess I don’t have to worry about thinning those! My single peach tree had no visible damage to any buds I popped open. Many J plum flowers still look good despite most showing white, and some started blooming.
1 Like
Flower can look fine on the outside and even be attractive to bees when ovary’s are dead. You have to cut them open to know what is happening. Last time it happened to me, I kept hoping but brown equaled dead and if most seem brown you are unlikely to get any crop at all- or at least I didn’t.
1 Like
Scott- Doesn’t your slope buy you a few degrees?
It gets warmer in the day but when it is night and the wind is blowing hard it doesn’t really buy anything.
I still don’t see any damage, I took apart a few more fruitlets and they all looked great. I am starting to get a bit optimistic.
4 Likes
You seem to be suggesting that meteorologists deliberately exaggerate potential cold events while I assume they make predictions based purely on a scientific method as accurate as current technology allows. If I was a professional meteorologist, my sense of competence would be purely based on the accuracy of my forecasts, not on how well I prepare people for forecasts that end up wrong towards the cold side.
Or are you suggesting that the methodology unintentionally and consistently exaggerates lows and has not been observed and corrected by the scientists who obsess on this.
https://www.accessscience.com/content/weather-forecasting-and-prediction/742600
“exaggerate” is not the correct word. The models spit out a bell curve of potential temperatures, not one temperature. The question is what to tell the public: the middle of the bell curve, which is the average, or something a bit off the middle as that is still possible and could help people who need to protect against an adverse event. Another way to think about it is a forecast is for a geographical region of some size; do you tell the public the average low expected for that region, perhaps misleading some people in the parts slightly colder, or do you bump it down a bit so those people in the spots a bit colder will not be shocked and upset at how the forecast was off?
One thing I should make clear is this is just my own speculation. It is only by watching predictions on many forecasts over many years that it seems like there is more wost-case going on in some. This is not just for lows, it is also for rain, snow, etc forecasts.
11 Likes
@scottfsmith One thing I have noticed in this area in particular is there are certain common weather forecasting challenges that everyone knows about, but few meteorologists properly adjust for in their forecasts (even if they actually talk about it!)
One is the stubbornness of cold air off the ocean from an east wind in spring. We will get that damn east wind and it’ll be 53 degrees and cloudy at noon. The forecast is for a high of 75, and they keep saying “the models show the cold air eroding by afternoon”, but those of us who have lived here a while know that 75% of the time, the cold air never really erodes that day. They talk about it but fail to ever properly account for it in their forecasts.
The other is how thunderstorms often die crossing the mountains. This seems to be a function of wind direction in part (S to SE winds feed and even intensify storms, NW to W winds often kill them, SW is kind of a wash). They even say “the storms will weaken crossing the mountains” then, in the same breath, say “70% chance of storms” that often don’t pan out.
I actually think TODAY will end up that way.
I’m confused. The forecasts I’m talking about predict a specific low and high for any given day, and I assume they simply try to be as accurate as possible. Where did you get the info that meteorologists routinely predict a lower number than the most likely one- it just makes no sense to me. There simply is only one number that will be the most likely one. The average gardener doesn’t even worry about lows until they put out their tender annuals. The predictions I get are for a specific weather station in the town they are predicting for, and it is usually a couple degrees higher that my own lows because I’m in a colder spot than that station.
Weather Underground lets me see results of several “weather stations” in the area where I live so I can see the range that is occurring at the moment, which helps to show where the colder areas are.
1 Like
The big thing is each weather site or app can use one of several forecasts from different entities. If they are using different models, they have different numbers, so accuweather for instance will often predict a lower low than weatherunderground for my zip code, but not always. So many of us check the different sites/apps to see what he various predictions are so that we can plan for the worst case scenario, just in case. For the last cold snap, I had a range from 20 to 24 as the predicted low.
1 Like
I get that- what stumps me is that anyone that puts a specific number up is judged by the accuracy of that number and over time the site that is most accurate would seem to have a big marketing advantage. I see no incentive in being cautious, because people are going to get mad if you are wrong either way. If you covered up your tomatoes and it wasn’t necessary because of one service’s inaccurately low prediction and you noticed that another service was more accurate, where are you likely to turn next time?
Yet Scott is speaking with the assurance of someone with inside knowledge of the industry. If they deliberately go with a number that is lower than the one they think of as most likely, than I assume in the summer they would predict higher temps than what they thought was most likely, at least when things reached a dangerous heat index.
I can’t help it. I obsess on weather predictions so I want to know how they work and I want a logical explanation, if there is one.
2 Likes
No – as I said above,
2 Likes
I have no affiliation, but I’ll give a plug to my preferred weather site: https://windy.com
They have a nice interface, and give easy access to satellite and radar, but their best feature (in my opinion) is that instead of trying to offer a single forecast they let you compare several. Currently, for my location in Vermont, they offer GFS (US National Oceanic and Atmospheric Administration), ECMWF (European Center for Medium Range Weather Forecasts), ICON (German DWD), Meteoblue (Swiss private), NAM (US National Center for Environmental Prediction Mesoscale), and HRRR (US NCEP High Resolution Rapid Refresh).
Over time, I’m starting to learn which forecast is more likely correct in which circumstances. But the ability to compare them all side-by-side is great. If they all agree, it’s probably accurate. If they disagree, it will usually be within the extremes. To try it, enter you location as exactly as you can, and then choose “Compare” at the bottom right.
7 Likes
Thanks. Looks very interesting. I like the easy to read 10 day on weatherunderground with the visual display of % chance of rain, cloud cover, etc. so I’ll have to study this one a bit to see everything it helps you see and compare.
1 Like
The issue here is that we’re talking statistical models, not arithmetic. There’s a couple of things you need to take into account. First, stats don’t really spit out a definitive answer so much as a range of possibilities or a predicted value with an expected error range +/- around it. Second, as others noted there are a bunch of different weather prediction models out there that all mostly use the same data, but handle it differently. None of these models is unilaterally better than the others. A given might predict certain events better than the other models, but fall short on other events.
The biggest tradeoff is between accuracy and precision. Some models will give very precise predictions (in terms of temp, precipitation, geography, etc, or some combination), but they might be less accurate than a less precise model. The other thing that happens is what’s known as overfitting the model. When you’re trying to build a statistical model, you can keep adding parameters until your equation perfectly fits every data point in your training data. Unfortunately, you’ll find that once you start throwing additional data at the model, it usually performs worse than a simpler model would have. Now, there’s no clear line where overfitting begins, and all the major weather models fall short of it. But some might come closer than others. This is actually quite closely linked to the accuracy/precision tradeoff.
image from Machine Learning: How to Prevent Overfitting | by Ken Hoffman | The Startup | Medium
On top of all of this, once the model spits out the results, they need to be interpreted by a human. The major weather sites mostly give you the model outputs as is and leave interpretation up to you. Don’t ever take the numbers at face value. You need incorporate your experience of how well a particular weather service has done predicting a particular condition at your particular location. The job of a TV meteorologist is to interpret the results for the audience, and they might notice some things that us laypersons don’t. But, they still have to cover some pretty broad strokes.
Also, here’s a great quote from Model charts - ECMWF, ICON, GFS, UKMO, GEM, etc. | Weather.us
Why are there so many models and how are they different?
Many different national weather centers have supercomputers that run weather models. Each of these is slightly different, using different equations to solve for various physical processes that shape our weather patterns. Many of them also have slightly different resolutions, and use slightly different combinations of initial data sources.
These slight differences multiply out through time because the atmosphere is a chaotic system. This also means any errors that the models make in the near term become exponentially larger with time. This is why the forecast for a week from now is far less accurate than the forecast for tomorrow.
Weather modelling centers attempt to control for the influence of chaos by running ensemble systems that each use slightly different initial conditions. Each ensemble “member” then produces a forecast as if its set of initial conditions were correct. This provides some way of quantifying how likely a given forecast outcome is, helping to show forecast uncertainty.
3 Likes