bluewave Posted June 23, 2011 Share Posted June 23, 2011 Whatever, man. Believe whatever you want. The GFS is much better as both a tropical model and synoptic model than it was at this time last year. No doubt about that.I like the changes that I have seen so far.But you can see a slightly further south track off the 0 UTC than yesterday's 12 UTC GFS. 12 UTC 0 UTC Link to comment Share on other sites More sharing options...
am19psu Posted June 23, 2011 Share Posted June 23, 2011 Well, we'll see a mano a mano with the Euro in about a week. You're missing my point. What the GFS forecasted for Alex is completely irrelevant to what the GFS is forecasting for next week. I like the Euro solution better right now, as well, but not because the GFS busted too far north with a storm last year. Link to comment Share on other sites More sharing options...
OSUmetstud Posted June 23, 2011 Share Posted June 23, 2011 You're missing my point. What the GFS forecasted for Alex is completely irrelevant to what the GFS is forecasting for next week. I like the Euro solution better right now, as well, but not because the GFS busted too far north with a storm last year. There was a couple big busts by the EC last year...which model ended up with better skill scores...the GFS or the ECMWF? Link to comment Share on other sites More sharing options...
Ed Lizard Posted June 23, 2011 Share Posted June 23, 2011 You're missing my point. What the GFS forecasted for Alex is completely irrelevant to what the GFS is forecasting for next week. I like the Euro solution better right now, as well, but not because the GFS busted too far north with a storm last year. I didn't miss the point about the GFS change, its why I didn't post the GFS ensemble means image. By mano a mano, I meant hand to hand combat, as it were, between the models on a 2011 system. I wasn't really challenging your response to bluewave. I hope the GFS wins this fight for selfish reasons, but it will be an interesting test. Link to comment Share on other sites More sharing options...
bluewave Posted June 23, 2011 Share Posted June 23, 2011 warm/high height bias of the new GFS extends into at least Day 10. I would like to see storm by storm comparisons vs the Euro and how each model did in each particular region. The GFS may have improved but there still could be traces of old biases for surface tracks.Something like the GFS cold bias in the east which from time to time has shown up since the fix on a case by case basis vs the Euro. Link to comment Share on other sites More sharing options...
am19psu Posted June 23, 2011 Share Posted June 23, 2011 There was a couple big busts by the EC last year...which model ended up with better skill scores...the GFS or the ECMWF? The Euro. The GFS was #2 in the Atlantic last year. But that doesn't include situations for which no NHC forecast was available. Empirically, for me, a Euro/GFS blend has been the best way to go since last July. Link to comment Share on other sites More sharing options...
Ed Lizard Posted June 23, 2011 Share Posted June 23, 2011 BTW, I suspect where the system actually consolidates will be huge. I would still guess not much South of 20º leaving the peninsula to be a threat to Tampico which is a touch North of 22º. Anyway, a 6 day contest now before a center even forms, and lets see what the models do in 3 or 4 days... Link to comment Share on other sites More sharing options...
Srain Posted June 23, 2011 Share Posted June 23, 2011 MARINE WEATHER DISCUSSION NWS NATIONAL HURRICANE CENTER MIAMI FL 300 AM EDT THU JUN 23 2011 GULF OF MEXICO... A RIDGE OF HIGH PRESSURE WILL GENERALLY PREVAIL ACROSS THE AREA THROUGH THE FORECAST PERIOD. SE RETURN FLOW WILL PREVAIL OVER THE WESTERN GULF. THE TROPICAL WAVE IN THE EASTERN CARIBBEAN...MENTIONED BELOW IN THE CARIBBEAN SECTION...WILL MOVE INTO THE GULF OF MEXICO MONDAY. NE WINDS AHEAD OF THE TROPICAL WAVE WILL INCREASE OVER THE SOUTHERN PART OF THE GULF SUNDAY. MODELS CONTINUE TO BE BULLISH WITH THE POTENTIAL DEVELOPMENT OF THIS WAVE IN THE SW GULF. SW N ATLC... A RIDGE OF HIGH PRESSURE ALONG 27N WILL SHIFT N TO NEAR 30N FRI NIGHT. IN GENERAL...MODERATE WINDS WILL PREVAIL OVER THE ARE FOR MUCH OF THE FORECAST PERIOD. THE ONLY EXCEPTION WILL BE OVER THE FAR S PART OF THE AREA WHERE FRESH TRADE WINDS ARE EXPECTED DUE TO A LOCALLY TIGHTER PRESSURE GRADIENT BETWEEN THE RIDGE AND A TROPICAL WAVE MOVING ACROSS THE CARIBBEAN. CARIBBEAN AND TROPICAL N ATLC W OF 55W... A TROPICAL WAVE IN THE EASTERN CARIBBEAN WILL REACH THE CENTRAL CARIBBEAN TODAY AND ENTER THE WESTERN CARIBBEAN FRI. THE WESTWARD PROPAGATION WILL DECREASE WITH CONVECTIVE ACTIVITY EXPECTED TO INCREASE AS THE WAVE MOVES SLOWLY ACROSS THE WESTERN CARIBBEAN OVER THE WEEKEND. THE TROPICAL WAVE WILL REACH THE YUCATAN PENINSULA SUN NIGHT...THEN MOVE INTO THE GULF OF MEXICO MONDAY. FRESH TO STRONG TRADEWINDS WILL PREVAIL OVER THE CARIBBEAN THROUGH THE FORECAST PERIOD. A SECOND TROPICAL WAVE WILL REACH THE TROPICAL N ATLC EARLY SUN...THE ENTER THE E CARIBBEAN EARLY MON. Link to comment Share on other sites More sharing options...
OSUmetstud Posted June 23, 2011 Share Posted June 23, 2011 I would like to see storm by storm comparisons vs the Euro and how each model did in each particular region. The GFS may have improved but there still could be traces of old biases for surface tracks.Something like the GFS cold bias in the east which from time to time has shown up since the fix on a case by case basis vs the Euro. Has it? Even if the GFS happens to be colder a few times than the Euro it does not mean it still retains a bias. Link to comment Share on other sites More sharing options...
OSUmetstud Posted June 23, 2011 Share Posted June 23, 2011 Last year the five day NHC track forecast error averaged 187 nm...with the error vector pointed toward the east...meaning the NHC (and the models, likewise) were mostly slow. Unfortunately, the NHC doesn't produce biases in their yearly verification reports. Link to comment Share on other sites More sharing options...
LakeEffectKing Posted June 23, 2011 Share Posted June 23, 2011 At these lead times, any biases in models are not as important as those biases inside T+84. More importantly is the mean spread of the model consensus at the lead times we are talking about. I'm sure statistically, there is SOME bias that can be teased out..(a few percentage points to a "biased" side) at the longer leads, but more focus should be given to the current progs by various models, as to the overall conditions in the area, now, and how they are doing vs. Verification. Link to comment Share on other sites More sharing options...
bluewave Posted June 23, 2011 Share Posted June 23, 2011 Has it? Even if the GFS happens to be colder a few times than the Euro it does not mean it still retains a bias. We had an old thread citing the examples that had an impact on operational forecasting. Link to comment Share on other sites More sharing options...
OSUmetstud Posted June 23, 2011 Share Posted June 23, 2011 We had an old thread citing the examples that had an impact on operational forecasting. okay...I can't see that thread. Was it a few specific examples of cold or was it an actual bias? Link to comment Share on other sites More sharing options...
Ed Lizard Posted June 23, 2011 Share Posted June 23, 2011 At these lead times, any biases in models are not as important as those biases inside T+84. More importantly is the mean spread of the model consensus at the lead times we are talking about. I'm sure statistically, there is SOME bias that can be teased out..(a few percentage points to a "biased" side) at the longer leads, but more focus should be given to the current progs by various models, as to the overall conditions in the area, now, and how they are doing vs. Verification. You won't make the famous LEK map before its declared a TC by NHC? This inspires confidence (not) since there is nothing there yet... Link to comment Share on other sites More sharing options...
bluewave Posted June 23, 2011 Share Posted June 23, 2011 okay...I can't see that thread. Was it a few specific examples of cold or was it an actual bias? There were several after the fix but I think they were on the old BB. Link to comment Share on other sites More sharing options...
Ed Lizard Posted June 23, 2011 Share Posted June 23, 2011 If this was last year, would there be planes in the Western Caribbean already dropping sondes? I would be curious as to what denser data sampling would do for long term (6 days or more) model forecasts. Link to comment Share on other sites More sharing options...
Srain Posted June 23, 2011 Share Posted June 23, 2011 In other NATL news, the Gulf of Guinea continues to cool and the monsoonal trough is lifting N in the Eastern Atlantic. Waves are becoming a bit more active and MIMIC TPW clearly shows the trough slowly lifting. Also of note is a vigorous wave in Central Atlantic that will be near the Windward Islands early next week. The wave in question re: SW Gulf development (currently in the Central Caribbean) is very moist and a hint of a roll or fold is showing up as well… Link to comment Share on other sites More sharing options...
dtk Posted June 23, 2011 Share Posted June 23, 2011 Last year the five day NHC track forecast error averaged 187 nm...with the error vector pointed toward the east...meaning the NHC (and the models, likewise) were mostly slow. Unfortunately, the NHC doesn't produce biases in their yearly verification reports. I can generate stats for any of the operational models (by season, basin, or individual storm) if there is something you're interested in seeing....(it's fairly trivial for me to do this for historical cases....but I can dig stuff up on storms for this year as they come in as well, it just takes a bit more work on my part to prep the files myself). Link to comment Share on other sites More sharing options...
am19psu Posted June 23, 2011 Share Posted June 23, 2011 I can generate stats for any of the operational models (by season, basin, or individual storm) if there is something you're interested in seeing....(it's fairly trivial for me to do this for historical cases....but I can dig stuff up on storms for this year as they come in as well, it just takes a bit more work on my part to prep the files myself). What'd be more interesting, I think, are the error stats for pre-TC systems. I don't think anyone is collecting them, though. Actually, that sounds like a Masters project or maybe even a PhD dissertation. Link to comment Share on other sites More sharing options...
OSUmetstud Posted June 23, 2011 Share Posted June 23, 2011 I can generate stats for any of the operational models (by season, basin, or individual storm) if there is something you're interested in seeing....(it's fairly trivial for me to do this for historical cases....but I can dig stuff up on storms for this year as they come in as well, it just takes a bit more work on my part to prep the files myself). I'd be interested in seeing model comparisons for storms in the atlantic basin last year...but if it's too much work for you don't worry about it Link to comment Share on other sites More sharing options...
dtk Posted June 23, 2011 Share Posted June 23, 2011 What'd be more interesting, I think, are the error stats for pre-TC systems. I don't think anyone is collecting them, though. Actually, that sounds like a Masters project or maybe even a PhD dissertation. Actually, I might be able to do verification on anything that has been at least identified as an invest...but we do not yet have the tools to do any quantitative verification of genesis. There are some folks (in and out of house.....particularly from the Navy and NHC) looking at things like this in experimental mode (looking at FAR, etc. for genesis); but it does sound like a good Masters project. Link to comment Share on other sites More sharing options...
dtk Posted June 23, 2011 Share Posted June 23, 2011 I'd be interested in seeing model comparisons for storms in the atlantic basin last year...but if it's too much work for you don't worry about it I actually already have the files to do the verification on disk (it turns out I'm doing some 2010 TC verification for an experimental version of our DA system, hopefully to be implemented next spring). I just need to to specify the models/storms and re-run the script. I can at least post a table with some raw numbers in a matter of minutes (statistical significance and error bars are another matter, and take a bit longer). Link to comment Share on other sites More sharing options...
dtk Posted June 23, 2011 Share Posted June 23, 2011 First, my apologies since I cannot include the ECMWF forecasts in this comparison (we have very strict guidelines regarding use of their data, and we don't include the TC forecasts in the database we use for TC verification). I can probably dig it up somewhere but it will take time (and it would be comparable to / better than the GFS stats, generally speaking). A comparison of the GFS, UKM, Nogaps, and Canadian TC track verification for the Atlantic 2010 reveals pretty much what we'd expect: average track errors (NM) FOR HOMOGENEOUS SAMPLE 00 12 24 36 48 72 96 120 GFSO 12.3 33.8 49.3 67.7 81.7 122.3 175.7 232.9 NGPS 21.2 41.8 63.5 88.1 117.4 173.4 231.5 319.0 UKMO 20.4 39.5 55.8 75.6 87.1 113.6 154.4 203.8 CMC 33.2 44.7 59.0 73.9 95.3 136.8 216.2 274.4 #CASES 98 90 79 67 58 43 35 26 The GFS and UKMet are better for track forecasting than Nogaps and Canadian. I was surprised to see the UKMet was better than the GFS for the day 4-5 lead time. Some of this has to be taken with a grain of salt since the sample size is so small (especially for the day 4/5 lead times). This is because special care is given to ensure that the sample is homogenous (meaning that the storm has to exist and be verified for all models used in the comparison for it to be counted).....which can lead to a bias (for example by throwing out weak storms, if one of the models struggles to keep them in existence). At some point I'll re-run the stats to see how the inhomogeneous stats compare (this is a bit more time consuming). The other thing to keep in mind is that the day 4-5 track error tends to be dominated by huge errors from a select few cases (for example, in looking through the database, the GFS had errors ranging from 350-550 NM for a few runs for Colin and Lisa).....whereas the track errors were comparable for the more well developed, long lived cyclones (Earl/Igor for example). I'm just starting to look at the stats more closely, but the UKMet had a pretty significant bias in both the lat/lon directions, whereas the GFS didn't have much of a longitudinal bias at all (but did have a latitudinal bias of the opposite sign). Link to comment Share on other sites More sharing options...
Ed Lizard Posted June 23, 2011 Share Posted June 23, 2011 Oh, forgot the UK Met. Seems weak and down South at 120 hours. Link to comment Share on other sites More sharing options...
wxmx Posted June 23, 2011 Share Posted June 23, 2011 Last year, the current GFS, during Alex, was running as a parallel run. The old GFS was way north, while the para GFS (current GFS) was further south, but still a touch north of where LF ocurred. Link to comment Share on other sites More sharing options...
dtk Posted June 23, 2011 Share Posted June 23, 2011 Last year, the current GFS was running as a parallel run. The old GFS was way north, while the para GFS (current GFS) was further south, but still a touch north of where LF ocurred. Well, it was implemented on July 28, so these stats are a bit "dirty" in that sense (containing verification both before [Alex/Bonnie only] and after the upgrade). I could easily remove those two storms from this comparison, but I don't think it change the stats very much. Link to comment Share on other sites More sharing options...
Ed Lizard Posted June 23, 2011 Share Posted June 23, 2011 Well, the wave that will trigger the preexisting vorticity of the monsoon trough (judging all those working backwards from 850 mb vorts on the models) is approaching what I assume is an extension of the MT or ITCZ (SW winds yesterday at someplace in Panama called David), and while the wave is sheared to heck, it is becoming convectively active, and the feature in the SW Caribbean seems to be lighting up a bit as well.. Link to comment Share on other sites More sharing options...
VAwxman Posted June 23, 2011 Share Posted June 23, 2011 Bluewave's point earlier is right. We can't say definitively whether or not the GFS' poleward bias is gone. Just because the GFS doesn't have a bias to lower heights too much hemispherically really doesn't tell you much. Maybe it won't have the poleward bias, but there's no way to know yet. Link to comment Share on other sites More sharing options...
am19psu Posted June 23, 2011 Share Posted June 23, 2011 Bluewave's point earlier is right. We can't say definitively whether or not the GFS' poleward bias is gone. Just because the GFS doesn't have a bias to lower heights too much hemispherically really doesn't tell you much. Maybe it won't have the poleward bias, but there's no way to know yet. Given all of the upgrades last summer and its obviously better performance around the globe, why is your null hypothesis that the bias still exists? Link to comment Share on other sites More sharing options...
VAwxman Posted June 23, 2011 Share Posted June 23, 2011 Given all of the upgrades last summer and its obviously better performance around the globe, why is your null hypothesis that the bias still exists? Didn't say it still exists. I only said we don't yet know, which is true. We can't just assume the poleward bias is gone anymore than we can assume it will still be there like before. Link to comment Share on other sites More sharing options...
Recommended Posts
Archived
This topic is now archived and is closed to further replies.