long term 5 yrs+

[bdgwx]

21 hours ago, snowlover91 said:

Here's something else to consider. With urbanization and the well documented heat island effect, what's to say the rise in temperatures experienced in recent years isn't partially affected by the heat island effect? How do we get an accurate comparison when the temperatures in the past didn't have this type of influence as widespread as most official NWS stations do now?

Keep in mind that there are two broad classes of datasets that yield global mean temperatures. The first is surface station (land, buoy, ship) based datasets like NASA GISS, NOAAGlobalTemp, and HadCRUT. These datasets must make adjustments for the urban heat island, station relocations, etc. The UHI effect must (as is) removed because there is an overweighting of urban vs rural stations. The stations do not form a homogeneous network of measurement because of the clustering of stations. The second type is reanalysis like CFSR, ERA, etc. These datasets assimilate more than 10 million observations per day and compute a true global mean temperature from a homogeneous grid mesh. If a grid cell has a UHI component so be it. Since it's treated no different than any other grid cell (it's equally weighted) that adjustment does not need to take place. Afterall, UHI or not that IS the temperature of that grid cell and you must count it just like all of the others. Unsurprisingly the global mean temperature computed from the surface-only datasets matches quite well with the reanalysis datasets. It's a testament to the confidence in our measurement of the global mean temperature. That is, many different institutions using wildly different methodologies and inputs all come to the same basic answer.

[Vice-Regent]

In my estimation, ocean temperatures are the best way to measure global warming. For that we have the impressive distribution of surface buoys and Argo floats.

http://www.argo.ucsd.edu/

[Sophisticated Skeptic]

Unless aliens come down and make their presence known globally....to warn and get the attention of the public, about the numerous hazards the planet faces ...were doomed. 

There are sooo many things polluted right now (atmosphere, oceans, surface) , and since the world ‘climate change’ turns people’s brains slow, and makes it seem like a political crazy person term...the real hazards on our planet will never get the attention  they need..either covered up by fake news media, or laughed at by the global elite.  Anything clean or healthy for the planet is nerfed or run into the ground by the oil giants...keeping the planet toxic and unhealthy is their motto, as long as their pockets keep getting bigger.

[skierinvermont]

On 3/16/2018 at 11:28 AM, bdgwx said:

Keep in mind that there are two broad classes of datasets that yield global mean temperatures. The first is surface station (land, buoy, ship) based datasets like NASA GISS, NOAAGlobalTemp, and HadCRUT. These datasets must make adjustments for the urban heat island, station relocations, etc. The UHI effect must (as is) removed because there is an overweighting of urban vs rural stations. The stations do not form a homogeneous network of measurement because of the clustering of stations. The second type is reanalysis like CFSR, ERA, etc. These datasets assimilate more than 10 million observations per day and compute a true global mean temperature from a homogeneous grid mesh. If a grid cell has a UHI component so be it. Since it's treated no different than any other grid cell (it's equally weighted) that adjustment does not need to take place. Afterall, UHI or not that IS the temperature of that grid cell and you must count it just like all of the others. Unsurprisingly the global mean temperature computed from the surface-only datasets matches quite well with the reanalysis datasets. It's a testament to the confidence in our measurement of the global mean temperature. That is, many different institutions using wildly different methodologies and inputs all come to the same basic answer.

From what I understand, CFSR and ERA are not considered appropriate tools for assessing long-term temperature trends. They both utilize the same data found in surface and satellite data sets of global temperature (GISS, BEST, HadCRUT, RSS etc.) but without careful quality control to maintain continuity of the record (say when one satellite is replaced with another with slightly different calibration, satellite drift etc.). 

Overall, they show very similar results to these other observational sources because they are based on the same data. But they contain potential errors without incorporating any additional data or quality control. In other words, they can only be more error prone and in no way could the be less error prone. They can be useful to predict (model) surface temperature where observational data is not available. But that doesn't improve the long-term quality or reliability.

https://pdfs.semanticscholar.org/8cb9/de0babe76f4bbd498dbcf6ded7dfa3b597f0.pdf

See section 4 on long term trends in reanalysis products

"Because of the addition or removal of observational platforms as the assimilation progresses, a reanalysis product may contain spurious trends or discontinuities that are not physical in nature and are an artifact of the assimilation system alone."

 

 

[bdgwx]

3 hours ago, skierinvermont said:

From what I understand, CFSR and ERA are not considered appropriate tools for assessing long-term temperature trends. They both utilize the same data found in surface and satellite data sets of global temperature (GISS, BEST, HadCRUT, RSS etc.) but without careful quality control to maintain continuity of the record (say when one satellite is replaced with another with slightly different calibration, satellite drift etc.). 

Overall, they show very similar results to these other observational sources because they are based on the same data. But they contain potential errors without incorporating any additional data or quality control. In other words, they can only be more error prone and in no way could the be less error prone. They can be useful to predict (model) surface temperature where observational data is not available. But that doesn't improve the long-term quality or reliability.

https://pdfs.semanticscholar.org/8cb9/de0babe76f4bbd498dbcf6ded7dfa3b597f0.pdf

See section 4 on long term trends in reanalysis products

"Because of the addition or removal of observational platforms as the assimilation progresses, a reanalysis product may contain spurious trends or discontinuities that are not physical in nature and are an artifact of the assimilation system alone."

 

 

 

Yes, they do utilize surface data, but that's only a subset of everything they can and do assimilate. Remember, they use very similar (3DVAR, 4DVAR, etc.) techniques that global numerical weather prediction models use for their analysis and so as a result they construct full 3D fields of the atmosphere from the surface up to the stratosphere. IMHO one the best places to start with reanalysis is the SPARC Reanalysis Intercomparsion Project. Refer to their first publication Introduction to the SPARC Reanalysis Intercomparison Project (S-RIP) and overview of the reanalysis systems for a brief primer. Also, this recent publication by the ECMWF group comparing reanalysis and surface-based datasets for climate research purposes is also interesting. But yeah, like any dataset, reanalysis isn't perfect. It's just another tool for climate research. 

[skierinvermont]

8 hours ago, bdgwx said:

 

Yes, they do utilize surface data, but that's only a subset of everything they can and do assimilate. Remember, they use very similar (3DVAR, 4DVAR, etc.) techniques that global numerical weather prediction models use for their analysis and so as a result they construct full 3D fields of the atmosphere from the surface up to the stratosphere. IMHO one the best places to start with reanalysis is the SPARC Reanalysis Intercomparsion Project. Refer to their first publication Introduction to the SPARC Reanalysis Intercomparison Project (S-RIP) and overview of the reanalysis systems for a brief primer. Also, this recent publication by the ECMWF group comparing reanalysis and surface-based datasets for climate research purposes is also interesting. But yeah, like any dataset, reanalysis isn't perfect. It's just another tool for climate research. 

You're not identifying what additional data is being assimilated. Without additional data the only improvement is to use the modelling capability to interpolate between observations. That can improve the accuracy of short term variability, but won't improve the accuracy of long-term trends relative to conventional datasets.

I skimmed through most of your link and didn't find anything to say it's more accurate.. just that it interpolates and is released faster.

[AfewUniversesBelowNormal]

The coldest map I've seen in a while

[bdgwx]

13 hours ago, skierinvermont said:

You're not identifying what additional data is being assimilated. Without additional data the only improvement is to use the modelling capability to interpolate between observations. That can improve the accuracy of short term variability, but won't improve the accuracy of long-term trends relative to conventional datasets.

I skimmed through most of your link and didn't find anything to say it's more accurate.. just that it interpolates and is released faster.

The conventional datasets like NASA GISS, NOAAGlobalTemp, etc. almost exclusively use surface station data. Satellite datasets like RSS and UAH almost exclusively use satellite data. Reanalysis assimilates all of that plus a bunch of other stuff like aircraft, RAOBs, wind profilers, and more.

Also, I didn't mean to imply that reanalysis is more accurate. Though it very well may be. I know that NASA GISS publishes a 0.05C margin error on their global mean temperature. Same with the Berkeley dataset. I'm sure NOAAGlobalTemp and HadCRUT are much of the same. The satellite datasets (RSS and UAH) likely have higher error margins according to Dr. Mears who heads the RSS team. While on the other end Spencer/Christy have claimed they are more accurate. So I don't know. But, if I had to guess either way I lean towards them being less accurate for several reasons. But, it's important to note that neither RSS nor UAH are attempting to measuring the global mean temperature near the surface. It's actually a pretty deep layer of the lower troposphere from what I understand.

Anyway, my point was that between conventional, satellite, and reanalysis there are wildly different techniques employed using drastically different subsets of available data and yet they all come to the same basic conclusion within a reasonable margin error (the exception may be v6 of UAH which seems to be low outlier in terms of the warming trend compared to all of the others). That's a pretty good sign that our confidence in the global mean temperature trends are grounded in reality.

[skierinvermont]

19 hours ago, bdgwx said:

The conventional datasets like NASA GISS, NOAAGlobalTemp, etc. almost exclusively use surface station data. Satellite datasets like RSS and UAH almost exclusively use satellite data. Reanalysis assimilates all of that plus a bunch of other stuff like aircraft, RAOBs, wind profilers, and more.

Also, I didn't mean to imply that reanalysis is more accurate. Though it very well may be. I know that NASA GISS publishes a 0.05C margin error on their global mean temperature. Same with the Berkeley dataset. I'm sure NOAAGlobalTemp and HadCRUT are much of the same. The satellite datasets (RSS and UAH) likely have higher error margins according to Dr. Mears who heads the RSS team. While on the other end Spencer/Christy have claimed they are more accurate. So I don't know. But, if I had to guess either way I lean towards them being less accurate for several reasons. But, it's important to note that neither RSS nor UAH are attempting to measuring the global mean temperature near the surface. It's actually a pretty deep layer of the lower troposphere from what I understand.

Anyway, my point was that between conventional, satellite, and reanalysis there are wildly different techniques employed using drastically different subsets of available data and yet they all come to the same basic conclusion within a reasonable margin error (the exception may be v6 of UAH which seems to be low outlier in terms of the warming trend compared to all of the others). That's a pretty good sign that our confidence in the global mean temperature trends are grounded in reality.

I can certainly agree with all that. I'm just wary of using reanalysis like ERA and CFSR to measure global temperature for a few reasons. I've never seen them used in that way in climate studies to measure long-term global temperature. I've seen them used to study regional temperature or short term variability, but not as a measure of long-term temp. From what I remember the CFSR trend is very different (warmer) but that may have been corrected or maybe my memory is mistaken. For another, it's not incorporating more data it's just using the same data sources and combining and interpolating which could lead to sources of error.

And finally, in the paper you posted there is a note that the old version of ERA showed less warming because it used a fixed level of CO2. If the model requires increasing CO2 level to "measure" the global temperature accurately it's not really measuring or observing anymore and some deniers could accuse the science of circular reasoning.