bluewave

Tom Skilling and Matt Lanza have a nice in depth discussion here on what the lack of balloon soundings may mean for the accuracy of weather model forecasts. https://www.facebook.com/TomSkilling/posts/weather-forecast-models-produce-forecasts-of-varying-accuracy-interestingly-some/1319935109491429/ With weather balloon observations regularly going off line at a whole series of key observation stations, we're moving into uncharted territory and hoping forecast models still behave. The concern is legitimate. Weather satellite and in flight aircraft data may help fill the void--but the data they provide is different than the more thorough balloon measurements. We may not see issues with model accuracy each day. But it's been established in some research work--and we know for a fact---that the upper air observations gathered by weather balloons carry disproportionate weight in influencing model accuracy. That's because their measurements of the atmosphere are in many ways more complete than other sources--such as satellite or in flight aircraft. NASA runs a global weather forecast model--the GEOS model. Among its roles is to look at the impacts of and distribution of gases and particulates in the atmosphere. It's also been useful in studying greenhouse gases and also the movement of dust and other particulates off the Sahara Desert westward out over the Atlantic and all the way to South and North America. It could also be used, were, God forbid, biological agents released into the atmosphere or a nuclear accident sent a huge amount of radioactivity into the atmosphere to track these agents and where they might go. So the NASA GEOS model isn't exclusively used for weather forecasting alone. But it IS an atmospheric forecast model and the data fed into it is comparable to the data fed into the full range of weather forecast models. You might have asked when you hear discussions of the fact different models come up with varying forecasts at times---"why is it they employ multiple models in the weather forecast process? Why not just one??" It's a reasonable question--and the answer is quite simple. It's because NO single model is perfect. Nor are our measurements of the atmosphere. Having multiple solutions from multiple models, each making slightly varied assumptions on how to handle the physics of the imperfectly measured atmosphere, offers human forecasters a range of possible forecasts scenarios--each valid yet some more likely to work out than others. When a common set of forecast solutions appears from varying models, it makes it easier to "cull" the less likely forecasts from the mix settling on the most accurate of the varied forecasts. NASA modelers have spent time looking at the observations fed into their model. By studying how the model behaves when certain data is flawed or missing, they've produced a ranking of which of the 210-million observations which go into many global forecasts based on their importance to the forecasts these models generate. A graphic I'm posting with this post lays out what they found. And interestingly, radiosonde data--in other words, weather balloon data--comes out as 2nd most influential of the different data sets which go into computer weather forecast models. This places weather balloon data in a uniquely powerful position. Eliminating balloon data from a set of stations, which has begun happening in greater numbers in recent months due to budget cuts, won't be something which messes up model predictions every day or in the same way. What it DOES suggest is that unexpected, hard to identify impacts may well occur. And who's to say that won't happen on a day when tornadoes or a snowstorm is a occurring and you need an accurate model read on how the extreme weather situation is evolving----or, equally scary, when a hurricane is approaching. That unpredictable and varying accuracy is introduced into the weather forecast environment is very concerning. Also, since many of the stations which have been pulled off line by budget cuts are in tornado and severe weather prone areas, the uniquely accurate 3-D measurements of the atmosphere weather balloons provide is worrisome. ----------------------------------------------- One other period provided meteorological forecast modelers a unique opportunity to check out the impact of losing upper air data from weather balloons--and it resulted in a measurable loss in computer model accuracy. The world lost Russian weather balloon soundings during an outage in 2015. European Centre researchers studied the impact on the loss of these balloon reports during that outage. This was a much broader loss of weather balloon data than what is happening now. But the outage led to a significant loss in model forecast accuracy. Because global weather is part of a vast, interconnected system, the loss of a critical data set produces forecast accuracy issues upstream and downstream of the regions where such data is usually taken. The 2015 outage provided a rare glimpse of what happens when weather balloon data is lost. That was a much different era in meteorological science so it's difficult to precisely anticipate the sort of problems such an out would produce today. But the loss of data as critical as that which is provided by our radiosonde (i.e. weather balloon) network can't be dismissed. ---------------------------------- https://theeyewall.com/weather-balloon-launch-cuts-an-honest-look-at-how-it-should-impact-forecasts/

12z Euro doesn’t quite connect all the way to the offshore hurricane moisture. But still has some spots going 2”+ with a decent looking squall line as the low cuts to our west. Maybe with some luck we can get a little more of a moisture connection in later runs to push the rainfall potential higher.

The Canadian is the only model with hurricane so far west. The Euro and GFS are further east. So a nice compromise would be a deepening low going to our west opening a moisture channel with the hurricane staying offshore like a PRE or IVT to enhance the rainfall potential.

We need the rains so hopefully the low tracking to the west now can tap some moisture from the hurricane. If we can actually verify a soaking rainfall, then maybe we can begin putting the drought behind us. All of our long range forecasts have verified drier than originally forecast. If the 12z Euro comes in really juicy, then it could be the beginning of a change to wetter. We will see…

Warmer storm track on the latest EPS with the low now tracking to our west instead of the east in earlier runs. We have seen this quite a bit in recent years. Have been getting warmer storm tracks even with -NAO and +PNA patterns. The WAR is a little stronger than earlier runs. New run stronger WAR linking up with the Greenland block Old run weaker WAR

The only way to determine whether a lack of weather ballon data is causing a specific model error is to do a data denial experiment which is impossible without the data. https://www.cnn.com/2025/10/14/weather/alaska-storm-weather-balloons-trump-cuts-nws-climate How big of a difference the missing balloon data made, though, may never be known. The best way to determine that would be to run computer models with weather balloon data fed into them and without it, in what is known as a data denial experiment — impossible to do without the data itself.

Nino 1.2 continues to run warmer than 3.4. This has been the recent theme with the EWBs not able to penetrate closer to the SA coast. It all began with the record WWB near the SA coast back in March 2023.

Pretty strong signal for a little cooler than average trough in the Northeast into early November.

Can’t really lean on any AI data unless you know the integrity of the data used to train the AI.

The last time a Greenland Block worked in our favor with a KU here along the 1-95 corridor was February 2021. The MJO 8 in January 2022 was more of a Pacific driven snowfall pattern. Most other Greenland blocks like December 2022, March 2023, and February 2025 were Southeast ridge link ups with Great Lakes cutters or I-78 to I-84 hugger tracks. Prior to 2019 there Southeast ridge link ups with Greenland blocks were rare. Like in December 2012 and the late 1990s. But those were usually followed up by more favorable Greenland blocking patterns and KU snowstorms in the next months or years which hasn’t happened yet.

If I had to draw up a set of maps now, would probably just use the EPS mean from days 3-7 to leave enough wiggle room for changes in later model runs.

Each of the models for next week have a different solution. The two upper lows and the hurricane are all in different positions at day 5. Plus the strong block to the north. So we are getting the windshield wiper effect with a new solution every run. There is even a 3rd upper low now showing up later in the forecast period dropping down from Canada to complicate matters.

We may never know for sure but the loss of weather balloon data is probably having some effect on the model performance. https://theconversation.com/typhoon-leaves-flooded-alaska-villages-facing-a-storm-recovery-far-tougher-than-most-americans-will-ever-experience-267423 But as the storm approached Alaska, everything went sideways. The weather model forecasts changed, reflecting a faster-moving storm, and Halong shifted to a very unusual track, moving between Saint Lawrence Island and the Yukon-Kuskokwim Delta coast. Unlike Merbok, which was very well forecast by the global models, this one’s final track and intensity weren’t clear until the storm was within 36 hours of crossing into Alaska waters. That’s too late for evacuations in many places. Did the loss of weather balloon data canceled in 2025 affect the forecast? That’s a question for future research, but here’s what we know for sure: There have not been any upper-air weather balloon observations at Saint Paul Island in the Bering Sea since late August or at Kotzebue since February. Bethel and Cold Bay are limited to one per day instead of two. At Nome, there were no weather balloons for two full days as the storm was moving toward the Bering Sea. Did any of this cause the forecast to be off? We don’t know because we don’t have the data, but it seems likely that that had some effect on the model performance.

If we could find a way to replicate the cooler high temperature patterns that we have been getting in May recently at other times of the year, then we would be onto something. Islip , NYPeriod of record: 1963-09-05 through 2025-10-23DateLowest maximum temperatures (degrees F)Top Record 2nd Record 3rd Record 5/1 49 in 2016 52 in 1978 53 in 2019 5/2 51 in 1995 53 in 1988 53 in 1975 5/3 49 in 1985 53 in 2016 53 in 1986 5/4 47 in 1987 51 in 2016 54 in 1978+ 5/5 44 in 1978 46 in 1987 53 in 2016 5/6 50 in 1978 52 in 2016 52 in 1967 5/7 44 in 1967 50 in 2022 52 in 1965 5/8 53 in 1971 54 in 1992 54 in 1989+ 5/9 49 in 2020 49 in 1972 50 in 1977 5/10 52 in 1989 54 in 2024 55 in 1995+ 5/11 54 in 1998 54 in 1989 55 in 1995+ 5/12 48 in 2010 51 in 2008 55 in 1966 5/13 51 in 2019 51 in 2017 51 in 2002 5/14 52 in 2019 57 in 2006 57 in 1984+ 5/15 53 in 1978 55 in 1973 56 in 1967 5/16 53 in 1984 53 in 1978 54 in 1996 5/17 56 in 1978 57 in 1994 58 in 1966 5/18 49 in 2007 53 in 2002 56 in 1968 5/19 51 in 1976 53 in 1994 55 in 1993 5/20 51 in 2000 54 in 2005 55 in 2008+ 5/21 49 in 1990 54 in 2000 56 in 2025 5/22 53 in 2025 54 in 2003 57 in 2000 5/23 54 in 2003 54 in 1982 57 in 1973 5/24 56 in 1982 57 in 2003 57 in 1969 5/25 49 in 1967 51 in 2013 51 in 2005 5/26 52 in 1967 53 in 2003 57 in 1973 5/27 56 in 1974 57 in 1973 58 in 1983+ 5/28 60 in 2014 60 in 2009 60 in 1996+ 5/29 51 in 2021 58 in 2017 60 in 1996+ 5/30 52 in 2021 60 in 2017 61 in 1965 5/31 55 in 1984 61 in 1967 63 in 1992

Most spots should finish warmer than average but the cooler spots like LGA, BDR, and NYC should finish below. October departures through the 23rd EWR….+2.5 NYC…..+1.4 JFK……+1.8 LGA……+0.8 HPN……+2.9 ISP……..+3.3 BDR…..+0.4 New Haven….+3.8 New Brunswick…+2.3

It’s probably a combination of factors with this being the first El Niño with the North Pacific so warm. Plus the lower aerosols got a boost after the new 2020 shipping regulations. So there were probably multiple factors which enhanced the warming both and before this event. So the El Niño was probably the trigger and the magnitude of the warming came from other places than directly the El Niño 3.4 ONI. There is another theory that the PCC could have also been involved since this El Niño started with so much 1.2 warming. But even past El Niños with strong 1.2 warming had a lag in global temperatures into the fall and not an immediate spring increase. https://www.nature.com/articles/s41467-024-52731-6 Instead, if the cyclic IPO dominates the recent cooling, we may expect a strong warming when it reverses. In support of the first possibility, we have identified an emerging climate change signal in the tropical Pacific across different observational datasets and we call it the PCC.

It’s always a good question to ask as we haven’t had a solid MJO 8 during the winter since January 2022. The standard MJO playbook has been a weakening before reaching 8 and then reloading back into the IO through the MC to WP. If we do eventually see another MJO 8 again, it probably won’t be forecast much more than a week or two in advance. But I am hoping we can see at least another weak reflection of January 2022 type event a some point in the coming winters.

It’s really the Arctic down to the mid-latitudes as the tropics have been a little cooler due to the La Niña. With Canada being so warm, it won’t really be that cold here for late October. Plus you can see the major drought across Canada with all the high pressure at 500 mb down to the surface.

The best guess so far is the big reduction in sulphur dioxide emissions over the last decade from China and shipping lead to less clouds over the Pacific which was masking the CO2 emissions warming. https://e360.yale.edu/digest/asia-air-pollution-sulfates-warming

It could just be that mid-latitude Pacific warming is acting as a higher latitude El Niño. So this is why the warming occurred earlier in 2023 than during past El Niño events. Plus the warming lingered longer than previous El Niños.

https://theconversation.com/typhoon-leaves-flooded-alaska-villages-facing-a-storm-recovery-far-tougher-than-most-americans-will-ever-experience-267423 Remnants of a powerful typhoon swept into Western Alaska’s Yukon-Kuskokwim Delta on Oct. 12, 2025, producing a storm surge that flooded villages as far as 60 miles up the river. The water pushed homes off their foundations and set some afloat with people inside, officials said. More than 50 people had to be rescued in Kipnuk and Kwigillingok, hundreds were displaced in the region, and at least one person died. Typhoon Halong was an unusual storm, likely fueled by the Pacific’s near-record warm surface temperatures this fall. Its timing means recovery will be even more difficult than usual for these hard-hit communities, as Alaska meteorologist Rick Thoman of the University Alaska Fairbanks explains. Disasters in remote Alaska are not like disasters anywhere in the lower 48 states, he explains. While East Coast homeowners recovering from a nor’easter that flooded parts of New Jersey and other states the same weekend can run to Home Depot for supplies or drive to a hotel if their home floods, none of that exists in remote Native villages. account Rick Thoman ‪@alaskawx.bsky.social‬ Follow Track and intensity of Typhoon/ex-typhoon Halong in early October. The storm passed over ocean water significantly warmer than normal for virtually its entire track, from southeast of Japan to landfall on the northeast Bering Sea coast. #akwx #weather #Climate #Alaska @climatologist49.bsky.social ALT October 22, 2025 at 3:14 PM Everybody can reply

Just a little below average for this time of year since the Northern Hemisphere is so warm.

This gradient between the record SST warmth east of Japan and the Arctic cold over NE Asia is driving this very fast North Pacific Jet. So it has been very challenging to sustain the -EPO and -WPO intervals for long. Prior to 2019, we would get extended -EPO and -WPO intervals instead of these rapid shifts between positive and negative states. https://downloads.psl.noaa.gov/Public/map/teleconnections/epo.reanalysis.t10trunc.1948-present.txt

This just highlights the issue with the models beyond 120 hrs being really poor in general with storms along the East Coast. But we see some general trends since the weekend. Everything is further east than the other day with more high pressure over the Northeast. While there could certainly be coastal storm of some type during this period, how far west it comes is still in question. So we just have to be patient and see what the models show once we are under the 120hr range when details become more obvious. New run more high pressure to the north Old run less high pressure to the north

I realize that. But some of the very wet posts showing up on social media were based on the wet OP Euro runs and not the ensembles. I posted the OP since it’s what has been getting the most attention. There are a few issues with the ensembles also. First, they can smooth out important features which lead to lesser amounts like stronger high pressure to the north. On paper the late month period looks like events in the past which have produced decent rainfall events here. But as we have seen in recent times, as we approached the event time the rainfall amounts became lesser. The Euro originally forecast the coastal event around the 13th to be further west with a more consolidated low pressure. There were several runs which had the heavier rains further west than they wound up. So it was more of a coastal event with a sharp rainfall cutoff and two weaker lows. So now I don’t have any confidence yet in the wetter model runs for the last week of October. Since there may be more high pressure and 500 mb ridging pressing down from the north than the smoothed ensembles are showing. Would like to see wetter solutions hold to 120hrs and under to be more believable. The old saying is that the long range OP runs are just another ensemble member. I am hoping the end of the month period is close to some of the EPS forecasts. But will not have confidence in the details and if a sharp cutoff until under 120 hrs. We will need one of the wetter long range forecasts to verify in order to get the ball rolling on a reversal of the drought which began back in September 2024. Since most outcomes have been drier than the original day 6-10 forecasts over the last year.