June 2019 observation and discussion

mwilli wrote:noaa is upgrading the gfs model,i would like to send the article here..if njstrong weather is on facebook i could send it there...let me know

rb924119 wrote:

billg315 wrote:Looks like the new GFS core (FV3) is up and running. We'll see how it performs in the coming months as Fall and Winter slooowly get closer.

Anybody else find it curious that they are releasing it as operational now that it’s summer, when a not insignificant portion of its problems were highlighted with its snowfall prediction? I’m really nervous about this model.

Quietace wrote:

rb924119 wrote:

billg315 wrote:Looks like the new GFS core (FV3) is up and running. We'll see how it performs in the coming months as Fall and Winter slooowly get closer.

Anybody else find it curious that they are releasing it as operational now that it’s summer, when a not insignificant portion of its problems were highlighted with its snowfall prediction? I’m really nervous about this model.

As we know it was delayed for that issue, however, if you read the implementation notice you will find that it states:

"EMC has conducted 3 years of retrospective experiments, including the real-time parallel, covering the past 3.5 years for a  comprehensive evaluation of the Q2FY19 GFS implementation. GFS V15 shows equal or improved forecast skills in many areas, especially for 500-hPa height anomaly correlations, precipitation diurnal cycle and ETS score over the CONUS, surface 2m temperature, stratospheric ozone and water vapor, and hurricane intensity over all basins. Several individual case studies illustrate the model occasionally produces excessive snow in the medium range. EMC also noted a persistent cold bias that increases with forecast time. EMC will continue to explore ways  to address these issues."
There were two updates that were utilized to fix this problem, with a power point link I have included below that describes the issues and fixes in more detail. However, it is noted that while the severity of the issues is moderated, it is not a complete fixed. On one hand, the summer release allows them additional time to provide updates to fix this issue while operating a slightly higher skill score model during a time of year where this bias has little impact. However, the counter-argument is the assumption any fixes can be implemented before the winter.

http://www.emc.ncep.noaa.gov/users/Alicia.Bentley/fv3gfs/updates/MEG_4-04-19_FV3GFS_COLD.pptx

rb924119 wrote:

Quietace wrote:

rb924119 wrote:

billg315 wrote:Looks like the new GFS core (FV3) is up and running. We'll see how it performs in the coming months as Fall and Winter slooowly get closer.

Anybody else find it curious that they are releasing it as operational now that it’s summer, when a not insignificant portion of its problems were highlighted with its snowfall prediction? I’m really nervous about this model.

As we know it was delayed for that issue, however, if you read the implementation notice you will find that it states:

"EMC has conducted 3 years of retrospective experiments, including the real-time parallel, covering the past 3.5 years for a  comprehensive evaluation of the Q2FY19 GFS implementation. GFS V15 shows equal or improved forecast skills in many areas, especially for 500-hPa height anomaly correlations, precipitation diurnal cycle and ETS score over the CONUS, surface 2m temperature, stratospheric ozone and water vapor, and hurricane intensity over all basins. Several individual case studies illustrate the model occasionally produces excessive snow in the medium range. EMC also noted a persistent cold bias that increases with forecast time. EMC will continue to explore ways  to address these issues."
There were two updates that were utilized to fix this problem, with a power point link I have included below that describes the issues and fixes in more detail. However, it is noted that while the severity of the issues is moderated, it is not a complete fixed. On one hand, the summer release allows them additional time to provide updates to fix this issue while operating a slightly higher skill score model during a time of year where this bias has little impact. However, the counter-argument is the assumption any fixes can be implemented before the winter.

http://www.emc.ncep.noaa.gov/users/Alicia.Bentley/fv3gfs/updates/MEG_4-04-19_FV3GFS_COLD.pptx

Thanks for posting, Ryan! Mind if I borrow that link to share elsewhere? And in response to yours, I honestly don’t see how they can say it definitively improved, when the raw fractional increases were 1-4% (from what I could tell on mobile). That doesn’t scream “better model” to me. Now, will future fixes increase that, possibly significantly? I hope so! Secondly, there were other issues aside from the snowfall forecasts that led to its suspension. Even though they have started fixing them with results in the right direction, why they didn’t keep it on the shelf and run it in parallel to the base GFS in real time for a longer period baffles me. There’s no way you’re going to convince me that the changes they made were compartmentalized and will only affect cold weather output. Therefore, we really don’t know how/if any biases were truly fixed and/or if any new ones emerged. Sure, they ran hindcasts, but that doesn’t tell you much IMO when the answers to the equations are known and you reverse engineer the equation. It’s just like solving a math problem in grade school: How much easier was it to “show your work” if you knew the answer that you had to get because you happened to remember a homework problem verbatim that showed up on the test, versus if you had no idea and were shaky on the algorithm to solve the problem? No comparison. Same thing here. Why they aren’t continuing to run it in parallel and compare its raw outputs through the next ten months is counterintuitive to me. What happens if it shows the 1-4% change this trop season and this winter, which is for all intents and purposes, negligible, and there is no longer the base GFS to fall back on? Then what? This whole thing seems very much rushed to me in the name of ego. If my tax dollars are being used for this, idc if it takes us ten years to do it, as long as it’s done right. But this just seems knee-jerk, especially when this is going to be THE model used in our field, with lives and property on the line.

Quietace wrote:

rb924119 wrote:

Quietace wrote:

rb924119 wrote:

billg315 wrote:Looks like the new GFS core (FV3) is up and running. We'll see how it performs in the coming months as Fall and Winter slooowly get closer.

Anybody else find it curious that they are releasing it as operational now that it’s summer, when a not insignificant portion of its problems were highlighted with its snowfall prediction? I’m really nervous about this model.

As we know it was delayed for that issue, however, if you read the implementation notice you will find that it states:

"EMC has conducted 3 years of retrospective experiments, including the real-time parallel, covering the past 3.5 years for a  comprehensive evaluation of the Q2FY19 GFS implementation. GFS V15 shows equal or improved forecast skills in many areas, especially for 500-hPa height anomaly correlations, precipitation diurnal cycle and ETS score over the CONUS, surface 2m temperature, stratospheric ozone and water vapor, and hurricane intensity over all basins. Several individual case studies illustrate the model occasionally produces excessive snow in the medium range. EMC also noted a persistent cold bias that increases with forecast time. EMC will continue to explore ways  to address these issues."
There were two updates that were utilized to fix this problem, with a power point link I have included below that describes the issues and fixes in more detail. However, it is noted that while the severity of the issues is moderated, it is not a complete fixed. On one hand, the summer release allows them additional time to provide updates to fix this issue while operating a slightly higher skill score model during a time of year where this bias has little impact. However, the counter-argument is the assumption any fixes can be implemented before the winter.

http://www.emc.ncep.noaa.gov/users/Alicia.Bentley/fv3gfs/updates/MEG_4-04-19_FV3GFS_COLD.pptx

Thanks for posting, Ryan! Mind if I borrow that link to share elsewhere? And in response to yours, I honestly don’t see how they can say it definitively improved, when the raw fractional increases were 1-4% (from what I could tell on mobile). That doesn’t scream “better model” to me. Now, will future fixes increase that, possibly significantly? I hope so! Secondly, there were other issues aside from the snowfall forecasts that led to its suspension. Even though they have started fixing them with results in the right direction, why they didn’t keep it on the shelf and run it in parallel to the base GFS in real time for a longer period baffles me. There’s no way you’re going to convince me that the changes they made were compartmentalized and will only affect cold weather output. Therefore, we really don’t know how/if any biases were truly fixed and/or if any new ones emerged. Sure, they ran hindcasts, but that doesn’t tell you much IMO when the answers to the equations are known and you reverse engineer the equation. It’s just like solving a math problem in grade school: How much easier was it to “show your work” if you knew the answer that you had to get because you happened to remember a homework problem verbatim that showed up on the test, versus if you had no idea and were shaky on the algorithm to solve the problem? No comparison. Same thing here. Why they aren’t continuing to run it in parallel and compare its raw outputs through the next ten months is counterintuitive to me. What happens if it shows the 1-4% change this trop season and this winter, which is for all intents and purposes, negligible, and there is no longer the base GFS to fall back on? Then what? This whole thing seems very much rushed to me in the name of ego. If my tax dollars are being used for this, idc if it takes us ten years to do it, as long as it’s done right. But this just seems knee-jerk, especially when this is going to be THE model used in our field, with lives and property on the line.

You are preaching to the choir. As someone who runs WRF and MPAS daily, I still believe they chose the wrong core and I won't get over it for a while

Quietace wrote:

rb924119 wrote:

Quietace wrote:

rb924119 wrote:

billg315 wrote:Looks like the new GFS core (FV3) is up and running. We'll see how it performs in the coming months as Fall and Winter slooowly get closer.

Anybody else find it curious that they are releasing it as operational now that it’s summer, when a not insignificant portion of its problems were highlighted with its snowfall prediction? I’m really nervous about this model.

As we know it was delayed for that issue, however, if you read the implementation notice you will find that it states:

"EMC has conducted 3 years of retrospective experiments, including the real-time parallel, covering the past 3.5 years for a  comprehensive evaluation of the Q2FY19 GFS implementation. GFS V15 shows equal or improved forecast skills in many areas, especially for 500-hPa height anomaly correlations, precipitation diurnal cycle and ETS score over the CONUS, surface 2m temperature, stratospheric ozone and water vapor, and hurricane intensity over all basins. Several individual case studies illustrate the model occasionally produces excessive snow in the medium range. EMC also noted a persistent cold bias that increases with forecast time. EMC will continue to explore ways  to address these issues."
There were two updates that were utilized to fix this problem, with a power point link I have included below that describes the issues and fixes in more detail. However, it is noted that while the severity of the issues is moderated, it is not a complete fixed. On one hand, the summer release allows them additional time to provide updates to fix this issue while operating a slightly higher skill score model during a time of year where this bias has little impact. However, the counter-argument is the assumption any fixes can be implemented before the winter.

http://www.emc.ncep.noaa.gov/users/Alicia.Bentley/fv3gfs/updates/MEG_4-04-19_FV3GFS_COLD.pptx

Thanks for posting, Ryan! Mind if I borrow that link to share elsewhere? And in response to yours, I honestly don’t see how they can say it definitively improved, when the raw fractional increases were 1-4% (from what I could tell on mobile). That doesn’t scream “better model” to me. Now, will future fixes increase that, possibly significantly? I hope so! Secondly, there were other issues aside from the snowfall forecasts that led to its suspension. Even though they have started fixing them with results in the right direction, why they didn’t keep it on the shelf and run it in parallel to the base GFS in real time for a longer period baffles me. There’s no way you’re going to convince me that the changes they made were compartmentalized and will only affect cold weather output. Therefore, we really don’t know how/if any biases were truly fixed and/or if any new ones emerged. Sure, they ran hindcasts, but that doesn’t tell you much IMO when the answers to the equations are known and you reverse engineer the equation. It’s just like solving a math problem in grade school: How much easier was it to “show your work” if you knew the answer that you had to get because you happened to remember a homework problem verbatim that showed up on the test, versus if you had no idea and were shaky on the algorithm to solve the problem? No comparison. Same thing here. Why they aren’t continuing to run it in parallel and compare its raw outputs through the next ten months is counterintuitive to me. What happens if it shows the 1-4% change this trop season and this winter, which is for all intents and purposes, negligible, and there is no longer the base GFS to fall back on? Then what? This whole thing seems very much rushed to me in the name of ego. If my tax dollars are being used for this, idc if it takes us ten years to do it, as long as it’s done right. But this just seems knee-jerk, especially when this is going to be THE model used in our field, with lives and property on the line.

You are preaching to the choir. As someone who runs WRF and MPAS daily, I still believe they chose the wrong core and I won't get over it for a while

rb924119 wrote:

Quietace wrote:

rb924119 wrote:

Quietace wrote:

rb924119 wrote:

billg315 wrote:Looks like the new GFS core (FV3) is up and running. We'll see how it performs in the coming months as Fall and Winter slooowly get closer.

Anybody else find it curious that they are releasing it as operational now that it’s summer, when a not insignificant portion of its problems were highlighted with its snowfall prediction? I’m really nervous about this model.

As we know it was delayed for that issue, however, if you read the implementation notice you will find that it states:

"EMC has conducted 3 years of retrospective experiments, including the real-time parallel, covering the past 3.5 years for a  comprehensive evaluation of the Q2FY19 GFS implementation. GFS V15 shows equal or improved forecast skills in many areas, especially for 500-hPa height anomaly correlations, precipitation diurnal cycle and ETS score over the CONUS, surface 2m temperature, stratospheric ozone and water vapor, and hurricane intensity over all basins. Several individual case studies illustrate the model occasionally produces excessive snow in the medium range. EMC also noted a persistent cold bias that increases with forecast time. EMC will continue to explore ways  to address these issues."
There were two updates that were utilized to fix this problem, with a power point link I have included below that describes the issues and fixes in more detail. However, it is noted that while the severity of the issues is moderated, it is not a complete fixed. On one hand, the summer release allows them additional time to provide updates to fix this issue while operating a slightly higher skill score model during a time of year where this bias has little impact. However, the counter-argument is the assumption any fixes can be implemented before the winter.

http://www.emc.ncep.noaa.gov/users/Alicia.Bentley/fv3gfs/updates/MEG_4-04-19_FV3GFS_COLD.pptx

Thanks for posting, Ryan! Mind if I borrow that link to share elsewhere? And in response to yours, I honestly don’t see how they can say it definitively improved, when the raw fractional increases were 1-4% (from what I could tell on mobile). That doesn’t scream “better model” to me. Now, will future fixes increase that, possibly significantly? I hope so! Secondly, there were other issues aside from the snowfall forecasts that led to its suspension. Even though they have started fixing them with results in the right direction, why they didn’t keep it on the shelf and run it in parallel to the base GFS in real time for a longer period baffles me. There’s no way you’re going to convince me that the changes they made were compartmentalized and will only affect cold weather output. Therefore, we really don’t know how/if any biases were truly fixed and/or if any new ones emerged. Sure, they ran hindcasts, but that doesn’t tell you much IMO when the answers to the equations are known and you reverse engineer the equation. It’s just like solving a math problem in grade school: How much easier was it to “show your work” if you knew the answer that you had to get because you happened to remember a homework problem verbatim that showed up on the test, versus if you had no idea and were shaky on the algorithm to solve the problem? No comparison. Same thing here. Why they aren’t continuing to run it in parallel and compare its raw outputs through the next ten months is counterintuitive to me. What happens if it shows the 1-4% change this trop season and this winter, which is for all intents and purposes, negligible, and there is no longer the base GFS to fall back on? Then what? This whole thing seems very much rushed to me in the name of ego. If my tax dollars are being used for this, idc if it takes us ten years to do it, as long as it’s done right. But this just seems knee-jerk, especially when this is going to be THE model used in our field, with lives and property on the line.

You are preaching to the choir. As someone who runs WRF and MPAS daily, I still believe they chose the wrong core and I won't get over it for a while

What would you have wanted them to go with?

Quietace wrote:

rb924119 wrote:

Quietace wrote:

rb924119 wrote:

billg315 wrote:Looks like the new GFS core (FV3) is up and running. We'll see how it performs in the coming months as Fall and Winter slooowly get closer.

Anybody else find it curious that they are releasing it as operational now that it’s summer, when a not insignificant portion of its problems were highlighted with its snowfall prediction? I’m really nervous about this model.

As we know it was delayed for that issue, however, if you read the implementation notice you will find that it states:

"EMC has conducted 3 years of retrospective experiments, including the real-time parallel, covering the past 3.5 years for a  comprehensive evaluation of the Q2FY19 GFS implementation. GFS V15 shows equal or improved forecast skills in many areas, especially for 500-hPa height anomaly correlations, precipitation diurnal cycle and ETS score over the CONUS, surface 2m temperature, stratospheric ozone and water vapor, and hurricane intensity over all basins. Several individual case studies illustrate the model occasionally produces excessive snow in the medium range. EMC also noted a persistent cold bias that increases with forecast time. EMC will continue to explore ways  to address these issues."
There were two updates that were utilized to fix this problem, with a power point link I have included below that describes the issues and fixes in more detail. However, it is noted that while the severity of the issues is moderated, it is not a complete fixed. On one hand, the summer release allows them additional time to provide updates to fix this issue while operating a slightly higher skill score model during a time of year where this bias has little impact. However, the counter-argument is the assumption any fixes can be implemented before the winter.

http://www.emc.ncep.noaa.gov/users/Alicia.Bentley/fv3gfs/updates/MEG_4-04-19_FV3GFS_COLD.pptx

Thanks for posting, Ryan! Mind if I borrow that link to share elsewhere? And in response to yours, I honestly don’t see how they can say it definitively improved, when the raw fractional increases were 1-4% (from what I could tell on mobile). That doesn’t scream “better model” to me. Now, will future fixes increase that, possibly significantly? I hope so! Secondly, there were other issues aside from the snowfall forecasts that led to its suspension. Even though they have started fixing them with results in the right direction, why they didn’t keep it on the shelf and run it in parallel to the base GFS in real time for a longer period baffles me. There’s no way you’re going to convince me that the changes they made were compartmentalized and will only affect cold weather output. Therefore, we really don’t know how/if any biases were truly fixed and/or if any new ones emerged. Sure, they ran hindcasts, but that doesn’t tell you much IMO when the answers to the equations are known and you reverse engineer the equation. It’s just like solving a math problem in grade school: How much easier was it to “show your work” if you knew the answer that you had to get because you happened to remember a homework problem verbatim that showed up on the test, versus if you had no idea and were shaky on the algorithm to solve the problem? No comparison. Same thing here. Why they aren’t continuing to run it in parallel and compare its raw outputs through the next ten months is counterintuitive to me. What happens if it shows the 1-4% change this trop season and this winter, which is for all intents and purposes, negligible, and there is no longer the base GFS to fall back on? Then what? This whole thing seems very much rushed to me in the name of ego. If my tax dollars are being used for this, idc if it takes us ten years to do it, as long as it’s done right. But this just seems knee-jerk, especially when this is going to be THE model used in our field, with lives and property on the line.

You are preaching to the choir. As someone who runs WRF and MPAS daily, I still believe they chose the wrong core and I won't get over it for a while

frank 638 wrote:A gorgeous Saturday to be at the beach 80 * and not humid.i can't believe in the long range there is no heat wave coming up not that I am complaining .I wonder how long will it last

billg315 wrote:The southern tail of that large area of heavy rain appears to be sinking south across the central part of the state and is likely to affect Middlesex, Monmouth and Ocean counties in the next hour or so.

algae888 wrote:If long range ensembles are accurate, july looks to start off cool with anomalous ridge out west and trough on east. No heat in sight. What a difference from last year. Three more days and sunlight starts getting shorter.

algae888 wrote:If long range ensembles are accurate, july looks to start off cool with anomalous ridge out west and trough on east. No heat in sight. What a difference from last year. Three more days and sunlight starts getting shorter.

skinsfan1177 wrote:Keep a eye on the sky strong line of storms in eastern PA. Depoints are up in the 70s and sun is out here. We may be in for a great storm this evening.