PhiEaglesfan712 Posted 3 hours ago Share Posted 3 hours ago 14 hours ago, 40/70 Benchmark said: I'd kill for 2016-2017 right now...featured about normal snowfall and I haven't seem that since 2017-2018. Maybe in your place it was about normal snowfall, but in most places, especially Philly on southward, it was a well below normal snowfall winter. For Baltimore and DC, it was a Top 10 least snowy winter. Plus, it torched from November-February, with very little snowfall during those months. Only March was colder than average and had above average snowfall. Easily one of the most depressing winters ever. Besides, I'm pretty sure either 2020-21 or 2021-22, depending whether you're inland or on a coast, was better than 2016-17. Link to comment Share on other sites More sharing options...
bluewave Posted 3 hours ago Share Posted 3 hours ago 21 minutes ago, GaWx said: Chris, Very well said about model biases! I couldn’t agree with you more. When significant clearcut cold biases are identified, it isn’t cherry picking to point out a much too cold model verification. If a model is too cold and has too low H5 hts in the E 1/2 of the US, say, 75-80% of the time, it isn’t cherry picking. The same could be said for any identified clear cut biases for any model. Thanks. I have kept a running tally of past model forecasts and where their errors or biases have been. I use them in the various threads for my forecast discussions. While AI models on their own haven’t been very good with real world use compared to the numerical models, they may have some promise if they could figure out how to incorporate AI to bias correct the numerical models in real time. I think the ECMWF is trying to do this. But the computer power involved in that is pretty demanding. Link to comment Share on other sites More sharing options...
mitchnick Posted 3 hours ago Share Posted 3 hours ago 19 minutes ago, bluewave said: Models having inherent biases have nothing to do with cherry picking. They are a function of the way the models work. The CanSIPS is well known for showing too much cold where the troughs are forecast especially in the East. One of the few parts of the country that can verify colder than these seasonal forecasts is the Rockies and Upper Plains. But that doesn’t do the posters living in the I-95 corridor any good. Plus the forecasts beyond one month are a crap shoot anyway. So don’t get too attached to what any of the seasonal models are showing now for next winter. The CanSIPS and other models don’t have a ton of skill forecasting the 500mb pattern more than 30 days out. So we have to wait until the 1st of each month for the monthly forecast. Then adjust the temperatures warmer in the spots where a trough is indicated for models like the CanSIPS. With the exception of Western areas which can sometimes verify colder. But usually not the East. The one model truism over the years is that the Southeast Ridge or general 500 mb heights in the East verify higher than forecast. This is why the heavy snows which were indicated as far south as NY Metro last February got pushed up into Canada from Toronto to Montreal. So understanding these model biases is essential if you want to have a better handle on the medium to long range and seasonal forecasts. Cansips is not perfect, and I never said it was. But your statement it's biased in one direction doesn't always hold true. Ask the central, southern, and eastern areas of the conus about 1/25. Top map Cansips 500mb 1/25 forecast from 7/24 Middle map 2m 1/25 forecast from 7/24 Bottom map actual 1000mb map from 1/25 I'd like to know how many other seasonal models were showing any BN temps in January from their 7/24 forecast, other the Pac NW. 1 1 Link to comment Share on other sites More sharing options...
bluewave Posted 3 hours ago Share Posted 3 hours ago 11 minutes ago, mitchnick said: Cansips is not perfect, and I never said it was. But your statement it's biased in one direction doesn't always hold true. Ask the central, southern, and eastern areas of the conus about 1/25. Top map Cansips 500mb 1/25 forecast from 7/24 Middle map 2m 1/25 forecast from 7/24 Bottom map actual 1000mb map from 1/25 I'd like to know how many other seasonal models were showing any BN temps in January from their 7/24 forecast, other the Pac NW. You are using an example of the CanSIPS missing the 500mb forecast from multiple months out. My observation is based on the months when the 500mb forecast is reasonably close. The updated January 2025 forecast which came out on January 1st was too cold in the Great Lakes and Northeast even though the 500mb forecast did better once we got closer in. Link to comment Share on other sites More sharing options...
40/70 Benchmark Posted 3 hours ago Author Share Posted 3 hours ago 5 hours ago, bluewave said: No problem. We need some levity in here after what has happened since 2018-2019. My only comment on the CanSIPS is that temperatures never match the forecasted 500 mb pattern. Just look at the one month forecast for June to see what I am talking about. The 500mb forecast for June wasn’t bad. But the cold pool to the south was way overdone. So this has also been happening with the La Niña winters showing so much cold to the north around the Great Lakes, Northeast, and Canada. If that Southeast Ridge verifies like it’s showing for the winter, there will be much less cold to the north and the warmth to the south will be much more expansive. 500 mb pattern forecast not bad Cold pool to the south never verified Yea, the CANSIPS is always too cold, but I thought it did best last season. 4 hours ago, snowman19 said: Ok JB lol I'm JB because I reported model data? 1 2 Link to comment Share on other sites More sharing options...
40/70 Benchmark Posted 3 hours ago Author Share Posted 3 hours ago 27 minutes ago, PhiEaglesfan712 said: Maybe in your place it was about normal snowfall, but in most places, especially Philly on southward, it was a well below normal snowfall winter. For Baltimore and DC, it was a Top 10 least snowy winter. Plus, it torched from November-February, with very little snowfall during those months. Only March was colder than average and had above average snowfall. Easily one of the most depressing winters ever. Besides, I'm pretty sure either 2020-21 or 2021-22, depending whether you're inland or on a coast, was better than 2016-17. Yes...and maybe for most places last season was good, but it was nothing short of atrocious where I am. And no.....I would take 2016-2017 over any of those seasons mentioned. Link to comment Share on other sites More sharing options...
40/70 Benchmark Posted 3 hours ago Author Share Posted 3 hours ago 46 minutes ago, bluewave said: Models having inherent biases have nothing to do with cherry picking. They are a function of the way the models work. The CanSIPS is well known for showing too much cold where the troughs are forecast especially in the East. One of the few parts of the country that can verify colder than these seasonal forecasts is the Rockies and Upper Plains. But that doesn’t do the posters living in the I-95 corridor any good. Plus the forecasts beyond one month are a crap shoot anyway. So don’t get too attached to what any of the seasonal models are showing now for next winter. The CanSIPS and other models don’t have a ton of skill forecasting the 500mb pattern more than 30 days out. So we have to wait until the 1st of each month for the monthly forecast. Then adjust the temperatures warmer in the spots where a trough is indicated for models like the CanSIPS. With the exception of Western areas which can sometimes verify colder. But usually not the East. The one model truism over the years is that the Southeast Ridge or general 500 mb heights in the East verify higher than forecast. This is why the heavy snows which were indicated as far south as NY Metro last February got pushed up into Canada from Toronto to Montreal. So understanding these model biases is essential if you want to have a better handle on the medium to long range and seasonal forecasts. Just to be clear, I was just thought that the current consensus was noteworthy.....I'm not expecting a cold winter. 1 Link to comment Share on other sites More sharing options...
mitchnick Posted 3 hours ago Share Posted 3 hours ago 3 minutes ago, bluewave said: You are using an example of the CanSIPS missing the 500mb forecast. My observation is based on the months when the 500mb forecast is reasonably close. Go back to December 2024 to see what I am talking about. Ok, we're back to the 1 particular month issue again, so you'd have to do some kind of compilation of comparing 1 month forecasts verifications. But I've always considered the Cansips a seasonal model even though it forecast period begins at 1 month. For purposes of this discussion, I'm looking for forecast trends in modeling for 6 months out. And as I said, Cfs2 and Euro seasonal show similar trends in areas of at or BN as the Cansips. Link to comment Share on other sites More sharing options...
40/70 Benchmark Posted 3 hours ago Author Share Posted 3 hours ago How did I know posting the seasonal consensus would spark a debate because it isn't a blow torch. 1 Link to comment Share on other sites More sharing options...
mitchnick Posted 3 hours ago Share Posted 3 hours ago Just now, 40/70 Benchmark said: How did I know posting the season consensus would start debate because it isn't a blow torch. I didn't want to say it, but thought the same thing. 2 Link to comment Share on other sites More sharing options...
40/70 Benchmark Posted 3 hours ago Author Share Posted 3 hours ago 2 minutes ago, mitchnick said: I didn't want to say it, but thought the same thing. If it were a wall-to-wall inferno and I posted it, do you think snowman would have posted a giggling emoji and called me Bluewave? He simply would have liked the post and perhaps made a passing comment about the IOD coupling with La Niña as added support. 1 Link to comment Share on other sites More sharing options...
mitchnick Posted 3 hours ago Share Posted 3 hours ago 1 minute ago, 40/70 Benchmark said: If it were a wall-to-wall inferno and I posted it, do you think snowman would have posted a giggling emoji and called me Bluewave? He simply would have liked the post and perhaps made a passing comment about the IOD coupling with La Niña as added support. Agreed, but you know as well as I do you'd never would have posted first! Lol Link to comment Share on other sites More sharing options...
bluewave Posted 3 hours ago Share Posted 3 hours ago 10 minutes ago, mitchnick said: Ok, we're back to the 1 particular month issue again, so you'd have to do some kind of compilation of comparing 1 month forecasts verifications. But I've always considered the Cansips a seasonal model even though it forecast period begins at 1 month. For purposes of this discussion, I'm looking for forecast trends in modeling for 6 months out. And as I said, Cfs2 and Euro seasonal show similar trends in areas of at or BN as the Cansips. The one particular month shows the greater bias of being too cold in the Northeast which is the same as the collection of all the months averaged out we have been discussing. Seasonal models beyond a month out are usually a crapshoot even when looking at the 500mb forecasts and not taking into account the general temperature biases for said patterns. The last time one of these seasonal models had a great winter forecast in October and verified close to having a great actual winter was the JMA winter 13-14 forecast with the wall of Pacific blocking which verified. Link to comment Share on other sites More sharing options...
40/70 Benchmark Posted 2 hours ago Author Share Posted 2 hours ago 10 minutes ago, bluewave said: The one particular month shows the greater bias of being too cold in the Northeast which is the same as the collection of all the months averaged out we have been discussing. Seasonal models beyond a month out are usually a crapshoot even when looking at the 500mb forecasts and not taking into account the general temperature biases for said patterns. The last time one of these seasonal models had a great winter forecast in October and verified close to having a great actual winter was the JMA winter 13-14 forecast with the wall of Pacific blocking which verified. I seem to remember the CANSIPS being pretty consistent with showing a colder look last season...maybe I'm wrong. Anyway, I don't argue the inaccuracy of climate models at 6 month leads....I just posted it because I thought that the consensus was interesting. Link to comment Share on other sites More sharing options...
mitchnick Posted 2 hours ago Share Posted 2 hours ago 1 minute ago, bluewave said: The one particular month shows the greater bias of being too cold in the Northeast the same as the collection of all the months averaged out we have been discussing. Seasonal models beyond a month out are usually a crapshoot even when looking at the 500mb forecasts and not taking into account the general temperature biases for said patterns. The last time one of these seasonal models had a great winter forecast in October and verified close was the JMA winter 13-14 forecast with the wall of Pacific blocking which verified. Seasonal models by definition are intended for periods beyond 1 month, which is why I consider weeklies as being more accurate for 1 month, though flawed too as we know. But that's why I look at forecast trends in the seasonal models as a whole as a better method than taking the forecasts at face value. Link to comment Share on other sites More sharing options...
GaWx Posted 2 hours ago Share Posted 2 hours ago 42 minutes ago, mitchnick said: Cansips is not perfect, and I never said it was. But your statement it's biased in one direction doesn't always hold true. Ask the central, southern, and eastern areas of the conus about 1/25. Top map Cansips 500mb 1/25 forecast from 7/24 Middle map 2m 1/25 forecast from 7/24 Bottom map actual 1000mb map from 1/25 I'd like to know how many other seasonal models were showing any BN temps in January from their 7/24 forecast, other the Pac NW. Hey Mitch, I’m confident that you realize that a model bias in whatever direction doesn’t mean it’s almost never going to miss in the opposite direction. A typical model bias could be something like verifying consistent with the bias 65% of the time while verifying the opposing way 20% of the time and being close 15% of the time. For example, the April 30th CANSIPS forecast for May was actually too warm, especially in Midwest/OH Valley/NE (though a bit too cool E VA S to FL): 4/30/25 CANSIPS for May ‘25 (in C vs 1981-2010): Actual May ‘25 (in F) vs 1981-2010: But the 2/28/25 CANSIPS verified as being significantly too cold for Mar ‘25 vs the reality of a very warm month. It was mainly warmer than normal but not nearly warm enough in most of the E US. Whereas CANSIPS was much too warm in Jan and too warm in Feb, it was much too cold in Dec. Link to comment Share on other sites More sharing options...
bluewave Posted 2 hours ago Share Posted 2 hours ago 25 minutes ago, 40/70 Benchmark said: How did I know posting the seasonal consensus would spark a debate because it isn't a blow torch. To be fair, our really warm winters like 22-23 and 23-24 weren’t forecast to be that warm. The seasonal models had a pretty extensive cold bias for both those winters. The seasonal forecasts for years like 15-16, 16-17, and 19-20 were also much cooler than the record warmth which verified. I believe from memory that the 01-02 and 11-12 seasonal model forecasts missed the warmth also. So most of the time as the winters have been warming these seasonal forecasts have been too cold. Last October I was pointing out the colder La Niña mismatch potential that some of the seasonal models could have been missing. So I wasn’t surprised at the colder pattern last January than some of the seasonals had many months in advance. Link to comment Share on other sites More sharing options...
mitchnick Posted 2 hours ago Share Posted 2 hours ago 12 minutes ago, GaWx said: Hey Mitch, I’m confident that you realize that a model bias in whatever direction doesn’t mean it’s almost never going to miss in the opposite direction. A typical model bias could be something like verifying consistent with the bias 65% of the time while verifying the opposing way 20% of the time and being close 15% of the time. For example, the April 30th CANSIPS forecast for May was actually too warm, especially in Midwest/OH Valley/NE (though a bit too cool E VA S to FL): 4/30/25 CANSIPS for May ‘25 (in C vs 1981-2010): Actual May ‘25 (in F) vs 1981-2010: But the 2/28/25 CANSIPS verified as being significantly too cold for Mar ‘25 vs the reality of a very warm month. It was mainly warmer than normal but not nearly warm enough in most of the E US. Whereas CANSIPS was much too warm in Jan and too warm in Feb, it was much too cold in Dec. That's why I like the Cansips...it can be used to argue both sides pretty convincingly! But seriously, as I said in my last post, I think seasonal model trends for a future season, and not a 1 month forecast, are worth following. 1 2 Link to comment Share on other sites More sharing options...
40/70 Benchmark Posted 2 hours ago Author Share Posted 2 hours ago 12 minutes ago, bluewave said: To be fair, our really warm winters like 22-23 and 23-24 weren’t forecast to be that warm. The seasonal models had a pretty extensive cold bias for both those winters. The seasonal forecasts for years like 15-16, 16-17, and 19-20 were also much cooler than the record warmth which verified. I believe from memory that the 01-02 and 11-12 seasonal model forecasts missed the warmth also. So most of the time as the winters have been warming these seasonal forecasts have been too cold. Last October I was pointing out the colder La Niña mismatch potential that some of the seasonal models could have been missing. So I wasn’t surprised at the colder pattern last January than some of the seasonals had many months in advance. Seasonal models are never going to accurately depict any anomaly of that magnitude, though....and obviously all of the higher magnitude anomalies have been warm over the past decade, so that is going to cause a cold bias. Go back to the fall of 2014 and show me a seasonal that nailed that anomaly..... Link to comment Share on other sites More sharing options...
GaWx Posted 2 hours ago Share Posted 2 hours ago Just now, 40/70 Benchmark said: Seasonal models are never going to accurately depict any anomaly of that magnitude, though....and obviously all of the higher magnitude anomalies have been warm over the past decade, so that is going to cause a cold bias. Go back to the fall of 2014 and show me a seasonal that nailed that anomaly..... If someone had the time, they could go through AmericanWx threads for each winter and find CANSIPS forecast maps (however far back they go) and compare to what verified. Any volunteers? Link to comment Share on other sites More sharing options...
40/70 Benchmark Posted 2 hours ago Author Share Posted 2 hours ago 1 minute ago, GaWx said: If someone had the time, they could go through AmericanWx threads for each winter and find CANSIPS forecast maps (however far back they go) and compare to what verified. Any volunteers? I'll bet anyone $1000 that the CANASIPS seasonal was too warm in the fall of 2014 for the subsequent winter of 2014-2015. I mean in the seasonal mean...probably not for December. 1 Link to comment Share on other sites More sharing options...
Stormchaserchuck1 Posted 2 hours ago Share Posted 2 hours ago CPC doesn't have a super warm Winter forecast right now.. seasonal models going a little cooler is probably why. Link to comment Share on other sites More sharing options...
40/70 Benchmark Posted 2 hours ago Author Share Posted 2 hours ago Any long range ensemble suite is going to have a smoothed mean....and I know most of the climate guidance is an average of a number of runs smoothed out. This is why you never see 2 feet of snowfall predicated from an ensemble mean at day 7....it doesn't mean it can't happen, nor does it mean that the data doesn't have value. I agree the trend is somewhat important. Link to comment Share on other sites More sharing options...
40/70 Benchmark Posted 2 hours ago Author Share Posted 2 hours ago 1 minute ago, Stormchaserchuck1 said: CPC doesn't have a super warm Winter forecast right now.. seasonal models going a little cooler is probably why. Right....this is a tough pill for some to swallow, apparently. 1 Link to comment Share on other sites More sharing options...
Stormchaserchuck1 Posted 2 hours ago Share Posted 2 hours ago Just now, 40/70 Benchmark said: Any long range ensemble suite is going to have a smoothed mean....and I know most of the climate guidance is an average of a number of runs smoothed out. This is why you never see 2 feet predicated from an ensemble mean at day 7....it doesn't mean it can't happen, nor does it mean that the data doesn't have value. I agree the trend is somewhat important. Seasonal models were actually really bad over the PNA region last Winter.. they maybe over-use the recent trend and ENSO. Link to comment Share on other sites More sharing options...
40/70 Benchmark Posted 2 hours ago Author Share Posted 2 hours ago 1 hour ago, Stormchaserchuck1 said: Seasonal models were actually really bad over the PNA region last Winter.. they maybe over-use the recent trend and ENSO. Yes, they are heavily biased towards stock ENSO composites......and I know Chris called for the mismatch period, as did I, but I don't think anyone forecasted a +PNA and near normal temps in the seasonal mean....as impure as the PNA was at times being biased west and what not.... A mismatch period is one thing, but that just wasn't a warm winter per modern 1991-2020 climo I do think this trend is noteworthy despite the continued cool ENSO favor to the hemisphere...which is why I posted it. Not to claim its going to necessarily be cold. Link to comment Share on other sites More sharing options...
Stormchaserchuck1 Posted 2 hours ago Share Posted 2 hours ago 3 minutes ago, 40/70 Benchmark said: Yes, they are heavily biased towards stock ENSO composites......and I know Chris called for the mismatch period, as did I, but I don't think anyone forecasted a +PNA in the seasonal. mean....as impure as it was at times being biased west and what not.... Phoenix being so hot for a long time last Summer was an accurate predictor... they broke their record streak of most consecutive 100+ days by 40! The other years that had the streak broken came out with a +PNA and -4f to -5f in the Midwest and East for December and January. I don't think they are as warm so far this Summer, but it is warming there always. Also, the AAO last August was record negative.. that rolled forward at 0.5 correlation to a -AO Dec and Jan.. again, something that hit. The PDO was strongly warm forecast, and that got blown out... Link to comment Share on other sites More sharing options...
40/70 Benchmark Posted 2 hours ago Author Share Posted 2 hours ago 3 minutes ago, Stormchaserchuck1 said: Phoenix being so hot for a long time last Summer was an accurate predictor... they broke their record streak of most consecutive 100+ days by 40! The other years that had the streak broken came out with a +PNA and -4f to -5f in the Midwest and East for December and January. I don't think they are as warm so far this Summer, but it is warming there always. Also, the AAO last August was record negative.. that rolled forward at 0.5 correlation to a -AO Dec and Jan.. again, something that hit. The PDO was strongly warm forecast, and that got blown out... I expect it to be warmer than last year, but not as warm as many of these seasons over the past decade...I don't think it will be prohibitively warm for a lot of us...but I also don't expect a predominate east coast storm track, either. I agree with Chris there. Link to comment Share on other sites More sharing options...
GaWx Posted 2 hours ago Share Posted 2 hours ago 19 minutes ago, 40/70 Benchmark said: I'll bet anyone $1000 that the CANASIPS seasonal was too warm in the fall of 2014 for the subsequent winter of 2014-2015. I mean in the seasonal mean...probably not for December. Does anyone know if archived Cansips seasonal forecasts are available somewhere so we can easily see how they’ve done overall? Link to comment Share on other sites More sharing options...
GaWx Posted 1 hour ago Share Posted 1 hour ago NWS DJF 2014-15 probabilities released in 11/2014: best cold chances centered in south Actual 2014-5 DJF vs 1981-2010: coldest in NE not south, which was only slightly BN Link to comment Share on other sites More sharing options...
Recommended Posts
Create an account or sign in to comment
You need to be a member in order to leave a comment
Create an account
Sign up for a new account in our community. It's easy!
Register a new accountSign in
Already have an account? Sign in here.
Sign In Now