GaWx Posted June 28, 2011 Share Posted June 28, 2011 The GFS has recently been modified (spring 2011), which seems to have warmed it substantially at day 5 in the NH as per the graph below. Day 6 is even warmer but the graph isn't shown here. This graph suggests that the GFS is now the warmest of the main medium range models by a good margin (at +6.08 meters at day 5). It is also warmest at day 6 with its +6.95 meters. In the ten years I've been following these graphs (including summers), I've never seen the GFS nearly as warm as this. It still has the Euro as being warm (+3.87 at day 5 and +4.17 at day 6), but not as warm as the GFS. The CDN is neutral while the UKMET and NOGAPS are cold. It remains to be seen how this may translate to the upcoming winter bias in the U.S. for the medium range. Keep in mind that the graph is for the NH as a whole. Will the GFS become as warm or warmer than the Euro in the U.S. this winter? Stay tuned. Link to comment Share on other sites More sharing options...
am19psu Posted June 28, 2011 Share Posted June 28, 2011 IIRC, we noticed this last summer with the GFS after the upgrade and then it lost a good deal of the warm bias over winter. EDIT: Yep. Here is the graph since 1996. You can pick out the upgrade in July and then the bias went away in winter. It is definitely higher right now than it was last summer, though. Link to comment Share on other sites More sharing options...
InstantWeatherMaps Posted June 28, 2011 Share Posted June 28, 2011 Ouch... What happened to the old cold-biased GFS? Link to comment Share on other sites More sharing options...
GaWx Posted July 2, 2011 Author Share Posted July 2, 2011 Ouch... What happened to the old cold-biased GFS? Some GFS runs a few days ago had KATL up to a whopping/near record high +25 C, with sfc highs in the low 100's! Now it has backed off to "only " +21 C, a cooling of 7F. Can you say warm bias? Link to comment Share on other sites More sharing options...
Ed Lizard Posted July 6, 2011 Share Posted July 6, 2011 Neighboring NWS office has issues with GFS MOS. GFS MOS STILL HAS NOT BEEN ABLE TO RESOLVE ITS COOL BIAS WITH HIGHTEMPERATURES AND HAVE THEREFORE GONE WITH A BLAND OF PERSISTENCE AND IN-HOUSE BIAS CORRECTION NUMBERS. && .PRELIMINARY POINT TEMPS/POPS... AUSTIN CAMP MABRY 76 102 75 102 76 / - - - 0 - Link to comment Share on other sites More sharing options...
baroclinic_instability Posted July 8, 2011 Share Posted July 8, 2011 Interesting data. Before reading this, I had noticed what had seemed to be a significant decrease in performance with the GFS this summer after what was a mainly stellar winter. I know dtk or other modelers may come and bash me, but it is just a subjective observation of overall forecast performance. All the great attributes of the upgraded GFS from this winter (bias, trends, overall "hit rate" of big storms) have decreased significantly this summer. I wondered quietly if changes to the convective scheme to account for increased spectral resolution have resulted in some negative changes in summer patterns? Just a thought as I honestly have no idea, but the GFS has not been particularly stellar regarding overall synoptic patterns this summer--especially the more significant upper level PV anomalies/baroclinic waves. Link to comment Share on other sites More sharing options...
dtk Posted July 8, 2011 Share Posted July 8, 2011 Interesting data. Before reading this, I had noticed what had seemed to be a significant decrease in performance with the GFS this summer after what was a mainly stellar winter. I know dtk or other modelers may come and bash me, but it is just a subjective observation of overall forecast performance. All the great attributes of the upgraded GFS from this winter (bias, trends, overall "hit rate" of big storms) have decreased significantly this summer. I wondered quietly if changes to the convective scheme to account for increased spectral resolution have resulted in some negative changes in summer patterns? Just a thought as I honestly have no idea, but the GFS has not been particularly stellar regarding overall synoptic patterns this summer--especially the more significant upper level PV anomalies/baroclinic waves. Why would you be bashed for making a subjective observation regarding "overall" performance. Historically, the GFS has been much better in the winter hemisphere than summer (this is probably true in general for global NWP models, due to the much bigger reliance on parameterizations during the convectively active regions/seasons). The drop in skill in the NH (particularly for the past several weeks) is a bit troubling. We did not see this in August/September of last year immediately after the model change. Unfortunately, there really isn't anything in the works in terms of substantial GFS upgrades in the near future....but we are working on a package (that includes a HUGE change to how we initialize the model) for next spring. I'm sure there are people looking at both the increased warm bias as well as the recent drop in skill. Getting model changes into the operations is quite a difficult/long procedure, so I wouldn't expect that anything would be implemented soon. If anything is found, it would likely be included with the spring 2012 bundle. [ As an aside, constructive model criticism is actually helpful; particularly when it comes from mets looking at the products on a regular basis. I don't think that people giving their subjective input should be bashed (by me, nor anyone else). ] Link to comment Share on other sites More sharing options...
GaWx Posted July 8, 2011 Author Share Posted July 8, 2011 Some GFS runs a few days ago had KATL up to a whopping/near record high +25 C, with sfc highs in the low 100's! Now it has backed off to "only " +21 C, a cooling of 7F. Can you say warm bias? It turned out that KATL's hottest was 95 F. Granted that's still hot, but it was ~8 F cooler than what those GFS runs were predicting. Per today's (7/8) 6Z gfs (per MeteoStar), it is doing it again. It has KATL's highs up to 103 F for both 7/12 and 7/13. Alltime hottest is only two F hotter at 105 F. Keep in mind that KATL is at 1,000 foot elevation. 850's are predicted to get as hot as +26 C as of 8 PM on 7/13 (0Z on 7/14)! Looking back at past intense heatwaves there, I don't recall having ever seen anything hotter than 26 C at 850. (The hottest may be ~+25.5 C per my memory.) So, I'm educatedly guessing that the 850's will verify at least a couple of degrees C cooler than this. Let's see what actually verifies at both the surface and at 850. I'll say this: if KATL were to somehow achieve +26 at 850, then I would expect 103 F to verify. I just don't see it getting nearly as hot as +26 C at 850. MeteoStar output KATL: http://wxweb.meteost...shtml?text=KATL Link to comment Share on other sites More sharing options...
Ellinwood Posted July 8, 2011 Share Posted July 8, 2011 It turned out that KATL's hottest was 95 F. Granted that's still hot, but it was ~8 F cooler than what those GFS runs were predicting. Per today's (7/8) 6Z gfs (per MeteoStar), it is doing it again. It has KATL's highs up to 103 F for both 7/12 and 7/13. 850's are predicted to get as hot as +26 C as of 8 PM on 7/13 (0Z on 7/14)! Looking back at past intense heatwaves there, I don't recall having ever seen anything hotter than 26 C at 850. (The hottest may be ~+25.5 C per my memory.) So, I'm educatedly guessing that the 850's will verify at least a couple of degrees C cooler than this. Let's see what actually verifies at both the surface and at 850. MeteoStar output KATL: http://wxweb.meteost...shtml?text=KATL RE: That 95F high... for several runs the GFS ENSEMBLE had highs of 100-101 for KATL. The GFS has been overdoing the ridge in the eastern U.S. lately, most likely due to 1) poor representation of the NAO, which has consistently been verifying more negative than the medium range suggests, and 2) under-representation of widespread convection in the Midwest and Tennessee Valley. Link to comment Share on other sites More sharing options...
GaWx Posted July 8, 2011 Author Share Posted July 8, 2011 RE: That 95F high... for several runs the GFS ENSEMBLE had highs of 100-101 for KATL. The GFS has been overdoing the ridge in the eastern U.S. lately, most likely due to 1) poor representation of the NAO, which has consistently been verifying more negative than the medium range suggests, and 2) under-representation of widespread convection in the Midwest and Tennessee Valley. 1) Thanks. Do you think that this too strong an eastern US ridge is at least loosely tied in with the overall NH strong warm GFS bias of this summer? 2) Update of the 30 day running mean of the hemispheric model bias for the GFS based on 500 mb height verifications (in meters): - Day 5 GFS has risen from +6.08 to +6.81. The 2nd warmest, the Euro, has fallen from +3.87 to +3.12. - Day 6 GFS has risen from +6.95 to +8.15. The 2nd warmest, the Euro, has fallen from +4.17 to +3.43. - I'm wondering if the GFS will peak around mid-summer (say mid July to early August) based on the past pattern of peaking in summer. http://www.emc.ncep....html/mnz56.html Link to comment Share on other sites More sharing options...
Ellinwood Posted July 8, 2011 Share Posted July 8, 2011 1) Thanks. Do you think that this too strong an eastern US ridge is at least loosely tied in with the overall NH strong warm GFS bias of this summer? From what I've seen this summer, the GFS has been over-amplifying the eastern US ridge as the trough in central/eastern Canada has been rather successful at suppressing it. Also, the GFS has been struggling a lot with the troughing along the West Coast... with the exception of the past several days, the trough has been digging and lingering longer than the GFS has anticipated, resulting in a large warm bias over the western US. Link to comment Share on other sites More sharing options...
okie333 Posted July 8, 2011 Share Posted July 8, 2011 Why would you be bashed for making a subjective observation regarding "overall" performance. Historically, the GFS has been much better in the winter hemisphere than summer (this is probably true in general for global NWP models, due to the much bigger reliance on parameterizations during the convectively active regions/seasons). The drop in skill in the NH (particularly for the past several weeks) is a bit troubling. We did not see this in August/September of last year immediately after the model change. Unfortunately, there really isn't anything in the works in terms of substantial GFS upgrades in the near future....but we are working on a package (that includes a HUGE change to how we initialize the model) for next spring. I'm sure there are people looking at both the increased warm bias as well as the recent drop in skill. Getting model changes into the operations is quite a difficult/long procedure, so I wouldn't expect that anything would be implemented soon. If anything is found, it would likely be included with the spring 2012 bundle. [ As an aside, constructive model criticism is actually helpful; particularly when it comes from mets looking at the products on a regular basis. I don't think that people giving their subjective input should be bashed (by me, nor anyone else). ] Does this "HUGE change" involve 4DVAR by any chance? EDIT: Nevermind. I see that it is a combination of EnKF and traditional 3DVAR. I also notice that you will be doubling the resolution in Fall 2012 and changing to a Semi-Lagrangian scheme at the same time. Link to comment Share on other sites More sharing options...
dtk Posted July 8, 2011 Share Posted July 8, 2011 Does this "HUGE change" involve 4DVAR by any chance? No, the 4DVAR development (for use with GFS) is not even close to ready for prime time [this is a development that has been ongoing for some time and will continue to be worked on in parallel]. We actually plan to implement a hybrid variational-ensemble (3DVAR-EnKF) algorithm. There are plans to continue to work on the 4DVAR, hybrid 4DVAR, as well as "ensemble 4DVAR" moving forward. Link to comment Share on other sites More sharing options...
baroclinic_instability Posted July 9, 2011 Share Posted July 9, 2011 Why would you be bashed for making a subjective observation regarding "overall" performance. Historically, the GFS has been much better in the winter hemisphere than summer (this is probably true in general for global NWP models, due to the much bigger reliance on parameterizations during the convectively active regions/seasons). The drop in skill in the NH (particularly for the past several weeks) is a bit troubling. We did not see this in August/September of last year immediately after the model change. Unfortunately, there really isn't anything in the works in terms of substantial GFS upgrades in the near future....but we are working on a package (that includes a HUGE change to how we initialize the model) for next spring. I'm sure there are people looking at both the increased warm bias as well as the recent drop in skill. Getting model changes into the operations is quite a difficult/long procedure, so I wouldn't expect that anything would be implemented soon. If anything is found, it would likely be included with the spring 2012 bundle. [ As an aside, constructive model criticism is actually helpful; particularly when it comes from mets looking at the products on a regular basis. I don't think that people giving their subjective input should be bashed (by me, nor anyone else). ] I worded it poorly. I meant I may get eaten alive by you with your historically spot on retorts with great data to reinforce your points. I meant it in a nice way, lol. Link to comment Share on other sites More sharing options...
baroclinic_instability Posted July 9, 2011 Share Posted July 9, 2011 Why would you be bashed for making a subjective observation regarding "overall" performance. Historically, the GFS has been much better in the winter hemisphere than summer (this is probably true in general for global NWP models, due to the much bigger reliance on parameterizations during the convectively active regions/seasons). The drop in skill in the NH (particularly for the past several weeks) is a bit troubling. We did not see this in August/September of last year immediately after the model change. Unfortunately, there really isn't anything in the works in terms of substantial GFS upgrades in the near future....but we are working on a package (that includes a HUGE change to how we initialize the model) for next spring. I'm sure there are people looking at both the increased warm bias as well as the recent drop in skill. Getting model changes into the operations is quite a difficult/long procedure, so I wouldn't expect that anything would be implemented soon. If anything is found, it would likely be included with the spring 2012 bundle. [ As an aside, constructive model criticism is actually helpful; particularly when it comes from mets looking at the products on a regular basis. I don't think that people giving their subjective input should be bashed (by me, nor anyone else). ] I will send you a message with what I have seen including a good example from the current system bombing out in Alberta. Link to comment Share on other sites More sharing options...
cmichweather Posted July 9, 2011 Share Posted July 9, 2011 No, the 4DVAR development (for use with GFS) is not even close to ready for prime time [this is a development that has been ongoing for some time and will continue to be worked on in parallel]. We actually plan to implement a hybrid variational-ensemble (3DVAR-EnKF) algorithm. There are plans to continue to work on the 4DVAR, hybrid 4DVAR, as well as "ensemble 4DVAR" moving forward. Have you guys gotten 4DVAR running internally for testing yet or still working on it? I've noticed that usually a testing period precedes the para page like you said so i'm just wondering how close you guys are to para testing? It sure is a tough task to conquer and I understand why the ecmwf crowd are so proud of their model. Link to comment Share on other sites More sharing options...
GaWx Posted July 9, 2011 Author Share Posted July 9, 2011 It turned out that KATL's hottest was 95 F. Granted that's still hot, but it was ~8 F cooler than what those GFS runs were predicting. Per today's (7/8) 6Z gfs (per MeteoStar), it is doing it again. It has KATL's highs up to 103 F for both 7/12 and 7/13. Alltime hottest is only two F hotter at 105 F. Keep in mind that KATL is at 1,000 foot elevation. 850's are predicted to get as hot as +26 C as of 8 PM on 7/13 (0Z on 7/14)! Looking back at past intense heatwaves there, I don't recall having ever seen anything hotter than 26 C at 850. (The hottest may be ~+25.5 C per my memory.) So, I'm educatedly guessing that the 850's will verify at least a couple of degrees C cooler than this. Let's see what actually verifies at both the surface and at 850. I'll say this: if KATL were to somehow achieve +26 at 850, then I would expect 103 F to verify. I just don't see it getting nearly as hot as +26 C at 850. MeteoStar output KATL: http://wxweb.meteost...shtml?text=KATL Fwiw, the 6Z 7/9 gfs has backed off on the very intense heat with "only" +22 C at 850 and 98 F at the sfc for 7/12-3 for KATL. Compare this to just 24 hours earlier when that 6Z gfs had 26C as the hottest at 850 and 103 F as the hottest at the sfc. The 0Z 7/9 Euro also has its hottest at 850 at ~+22C. Even the Euro has cooled about 1 C in the last 24 hours. http://wxweb.meteost...shtml?text=KATL Link to comment Share on other sites More sharing options...
dtk Posted July 11, 2011 Share Posted July 11, 2011 Have you guys gotten 4DVAR running internally for testing yet or still working on it? I've noticed that usually a testing period precedes the para page like you said so i'm just wondering how close you guys are to para testing? It sure is a tough task to conquer and I understand why the ecmwf crowd are so proud of their model. It's still a serious work-in-progress, and something in the long term plans (not anytime in the foreseeable future). We do have a 4DVAR prototype, but I wouldn't even call it an alpha-version (it still has too many restrictions, is too inefficient, and can only be run at very low resolution). The hybrid system we plan to implement next spring is a pretty significant upgrade (even though it will still be 3D), and the nice thing is that it will also be applicable to 4DVAR if/when ready (either in full ensemble or hybrid mode....something that no operational center is doing...yet.). Link to comment Share on other sites More sharing options...
icebreaker5221 Posted July 12, 2011 Share Posted July 12, 2011 IIRC, we noticed this last summer with the GFS after the upgrade and then it lost a good deal of the warm bias over winter. EDIT: Yep. Here is the graph since 1996. You can pick out the upgrade in July and then the bias went away in winter. It is definitely higher right now than it was last summer, though. Where do you get verification back to 1996? If it's somewhere on the verification site we all know and love (http://www.emc.ncep.noaa.gov/gmb/STATS_vsdb/), I seem to be missing it! Link to comment Share on other sites More sharing options...
am19psu Posted July 12, 2011 Share Posted July 12, 2011 Where do you get verification back to 1996? If it's somewhere on the verification site we all know and love (http://www.emc.ncep.noaa.gov/gmb/STATS_vsdb/), I seem to be missing it! At the very top left where it says "Longterm" Link to comment Share on other sites More sharing options...
icebreaker5221 Posted July 12, 2011 Share Posted July 12, 2011 At the very top left where it says "Longterm" Hidden in plain sight! Thanks! Link to comment Share on other sites More sharing options...
GaWx Posted July 14, 2011 Author Share Posted July 14, 2011 It turned out that KATL's hottest was 95 F. Granted that's still hot, but it was ~8 F cooler than what those GFS runs were predicting. Per today's (7/8) 6Z gfs (per MeteoStar), it is doing it again. It has KATL's highs up to 103 F for both 7/12 and 7/13. Alltime hottest is only two F hotter at 105 F. Keep in mind that KATL is at 1,000 foot elevation. 850's are predicted to get as hot as +26 C as of 8 PM on 7/13 (0Z on 7/14)! Looking back at past intense heatwaves there, I don't recall having ever seen anything hotter than 26 C at 850. (The hottest may be ~+25.5 C per my memory.) So, I'm educatedly guessing that the 850's will verify at least a couple of degrees C cooler than this. Let's see what actually verifies at both the surface and at 850. I'll say this: if KATL were to somehow achieve +26 at 850, then I would expect 103 F to verify. I just don't see it getting nearly as hot as +26 C at 850. MeteoStar output KATL: http://wxweb.meteost...shtml?text=KATL Once again, despite a very hot and very uncomfy week with unbearable heat indices, the GFS verified quite a bit too hot. A couple of runs even had 106-108 F vs. the alltime high of 105! KATL's hottest verified to be "only" 97 F. Part of the problem appeared to be it assuming TD's in the 60's instead of 70's. Also, 850's at KATL verified no warmer than ~+22 C vs. the ~25.5-26 C of some runs. Link to comment Share on other sites More sharing options...
Recommended Posts
Archived
This topic is now archived and is closed to further replies.