NOAA however rose slightly, from 0.877°C to 0.899°C. TempLS grid also rose, from 0.704°C to 0.75. This is a pattern often observed in the past, where GISS follows TempLS mesh, and TempLS grid tracks NOAA. It is expected from the different ways they are constructed. I'll show the map comparisons below the fold. The updated comparison plots with 1998 are here
Here is the GISS map for the month
The Moyhu spherical harmonics map is here:
Update: I've mentioned the pairing of TempLS grid and NOAA before https://moyhu.blogspot.com.au/2014/08/templs-and-noaa-are-converging.html, and less markedly, GISS and TempLs mesh. Here's a current plot of recent years:
An interesting and quite plausible explanation (grid vs mesh) for the odd slight uptick for NCEI versus the moderate to large drops for all other global temperature anomaly estimates (except still waiting on HadCRUT4) from May to June. Here's what I get for June minus May differences in degrees Celsius:
ReplyDelete+0.02 NCEI
-0.06 RSS V3.3
-0.14 GISS
-0.15 WxBELL
-0.15 UM CCI
-0.21 UAH V6b5
-0.21 BEST ice air
The 0.23C range between these estimates suggest a fairly large uncertainty for the actual monthly change.
It's beginning to look like the reanalysis estimates will show a substantial uptick for July somewhere around +0.1C from June, barring a sudden plunge over the remainder of the month. If this uptick happens, it will be interesting to see how the other estimates compare. I noticed in the CFSV2 estimates from UM CCI that a large high spike in the Antarctic temperature anomaly appears to be a major factor in the reanalysis global July uptick so far and despite a slowly dropping tropic zone temperature anomaly that seems to be connected to hints of a La Nina getting started in the SSTA patterns.
Bryan,
DeleteYes, I think the basic difference is interpolation, which reflcts the weight given to polar temperatures. I first wrote about TempLS grid and NOAA here. I made a new comparison plot which I've added to the post.
I like the mesh approach versus the grid approach for climate work. A grid is fine for 2D work, but for a nearly spherical globe, the mesh approach seems better to me. Neither is going to work well for data sparse areas and that is why I like the weather model input reanalysis approach since it uses a much large measurement set. It would be interesting to see the global forecast modeling input measurements reanalyzed using the mesh approach. Unfortunately, that's way over my head and I'm getting too old to learn very many new tricks.
DeleteJMA also support the behaviour of gridded datasets, up 0.04 C from May to June.
ReplyDeleteThe Gistemp 12 month running mean is 1.01 right now. To stay above 1, the rest of 2016 has to be similar to 2015. I think the 12 month mean will continue to rise through August or September, but not in Oct-Dec, and it is not very likely that the year 2016 will be above 1 C.
However, a transition to GHCN v4, which is scheduled to be done in 2016, can change everything...
Olof, what will change with GHCN v4?
DeleteBryan, the GHCN v3 Arctic cooling bias will hopefully disappear.
DeleteSee e g http://www-users.york.ac.uk/~kdc3/papers/coverage2013/update.140404.pdf
More stations in the Arctic, supporting each other with similar high trends, may convince the GHCN algorithms that the fast warming in the Arctic is real...
Olof, I looked at the UM CCI CFSR/CFSV2 monthly estimates for the warmest and coldest months and for the annual averages for the Arctic zone (60-90N) and Antarctic zone (60-90S) and found the following trends:
DeleteArctic zone (60-90N)
July (warmest month on average) 1979-2015
Slope: +0.0229C/year [R2=0.3127]
Annual average 1979-2015
Slope: +0.0462C/year [R2=0.6205]
January (coldest month on average) 1979-2016
Slope: +0.596C/year [R2=0.2389]
Antarctic zone (60-90S)
January (warmest month on average) 1979-2016
Slope: +0.0025C/year [R2=0.0041 reflective of no trend]
Annual average 1979-2015
Slope: -0.0253C/year [R2=0.4042]
August (coldest month on average) 1979-2015
Slope: -0.0469C/year [R2=0.2319]
These data suggest winter temperatures in the Arctic are rising faster than summer temperatures, but both have trended upward significantly since 1979. Conversely, the Antarctic winter temperatures have trended downward significantly while the summer temperatures (which average well below 0C for the zone) show little trend since 1979. I am not sure how these trend estimates compare to GHCN-based and satellite-based trend estimates for the polar zones.
Presumably the CFSR/CSV2 data set includes measurements from many drifting buoys that are not included in GHCN data sets as well as more land stations and thus should provide much better spatial coverage in the Arctic zone. Likewise, in the Antarctic zone there appears to be many more synoptic weather stations reporting temperature than there are GHCN stations. I have not looked to see if the ERA/ERAI data are available for the same zones but it would be interesting to compare the trends.
Oops, typo on the Arctic January trend. Should be +0.0596C/year. Only off by a factor of ten.
DeleteOlof R on July 20, 2016 at 6:48 PM
DeleteYes indeed, Olof, but their gridded data has an annual cycle
http://ds.data.jma.go.jp/tcc/tcc/products/gwp/temp/map/download.html
There you find only the last year's data.
It would be so interesting to get their new data up to june, e.g. to bring such charts up to date:
http://fs5.directupload.net/images/160721/aijwvumk.pdf
A question.
When you compare Klaus Wolter's Multivariate ENSO Index with troposphere data, you see a clear correlation between MEI and TLT/TTT for the 1997/98 and 2015/16 events, but not for the 1982/83 Nino event, which peaked like in 1997/98 at a MEI value above 3:
http://fs5.directupload.net/images/160722/mdul58if.pdf
There are other places showing a lack of correlation, but this one is the most impressive.
Do you have an idea about this rather unexpected discrepancy?
Everybody talks about strong correlations between El Nino and UAH, but I guess they all mean 1997/98 or 2015/16 (few people seem to know about 1982/83).
Bryan, Knmi climate explorer has several reanalyses, and it is quite easy to mask different parts of the earth as well..
DeleteI am not convinced that the reanalyses do a good job in Antarctica, there is quite a spread among them, and Era-interim, for instance, doesn't agree with South Pole Amundsen Scott station, reporting a trend that is 0.2 C/decade lower than CRUTEM4, or 0.3 lower than BEST..
BEST uses all available data from Antarctica, also small automatic weatherstations, and show generally higher trends than the other observational and reanalysis datasets..
Bindidon, the el Chichon eruption "neutralised" the 1982-1983 el Nino.
Also, I believe that there is seasonality in all blended global surface datasets, caused mostly by NH SST. And it grows larger with older base period climatologies..
Olof, thanks for the info about Antarctica. I have dabbled with KNMI recently but have much to learn about how to better use it. I like the BEST approach of using more data, even when the period of record is shorter and that is why I also believe that even more data should be used.
DeleteThe concept of a global surface temperature average at about 2m above ground level is useful, but comes with substantial uncertainties, especially with older estimates (and that includes paleo proxies). As far as I know there are no long-term GHCN temperature measurements at 2m under forest or jungle canopy areas and these areas have decreased significantly over the last 100 years. Then there is urbanization. What do we use for 2m temperature measurements to represent the middle of a city with tall buildings and paved streets? The temperature environment there is certainly much different than it was before the city was built. Some GHCN measurements were likely taken from the tops of buildings in the past and I would not be surprised if a few may still be from the top of buildings. Of course there area also many other considerations on uncertainty like topography and ground cover, but my point is that the uncertainty associated with trying to estimate a global temperature or even a regional temperature is probably larger than most people realize. Using temperature anomalies probably reduces this uncertainty slightly, but I suspect not much.
Using more data is a double edged sword because large data sets tend to introduce more data quality problems. I see evidence of plenty of temperature data quality issues in reviewing daily weather maps of temperature observations from NOAA and OGIMET. The larger errors are relatively easy to spot, but smaller errors, such as from the lack of proper shielding and aspiration with direct sunlight, are more difficult to assess (I see hints of this with some of the drifting buoys in the Arctic recently for instance and this would introduce a small high bias). Hopefully the larger errors are identified and removed from reanalysis efforts, although I don't know how well this is done.
I agree with your assessment of the 1982-1983 El Nino. The El Chichon eruption effects caused a significant cooling of global surface temperatures and probably interfered somewhat with the calibration of satellite estimates of lower tropospheric temperatures and sea surface temperatures (increasing the uncertainty of those estimates). Large explosive volcanic eruptions are definitely a big wild card in climate forecasts and also in trying to interpret the paleo proxy evidence.
Bryan, GISS has a god piece on absolute temperatures vs anomalies
Deletehttp://data.giss.nasa.gov/gistemp/abs_temp.html
How useful are absolute temperatures from weather stations? In the best case they represent the temperature for a standardized environment; flat, open, short grass. How doesn't this compare with the actual environment? Is it representative for the average regional or global environment?
The temperature at the nearby weather station can't tell me the temperature in my garden, but the the temperature change, or anomaly, will likely be similar to that of my garden..
Olof, thanks for the link. I am very familiar with all of the issues discussed there. I believe climate, like weather, can be viewed from different scales. In see a continuum that runs from microscale to mesoscale to regional and global scales, with no sharp boundaries to distinguish between each scale. Your garden and my backyard fall into the microscale. Lake, river, coastal, forest, grassland, hilly, and mountainous environments are commonly in the mesoscale range. I would prefer to define regions based on consistent geography, but they are often expanded to include subcontinents, continents, seas, or oceans.
DeleteMy backyard is covered by a canopy of trees and I have been measuring the temperature from a wireless remote sensor mounted on a fence post in heavy shade at about 1.5 meters above the ground for several years now. The heavy shade mutes the high and low temperatures compared to what is monitored at nearby airport weather stations with sensors open to the sky. On average my backyard is slightly cooler than the nearby airport weather stations, but they all show differences as well. The official Austin weather station is not at an airport but is currently at the Camp Mabry National Guard military base and is on a ridgetop surrounded by 10-15 m trees at about 10-20 m distance (where it was moved from the old Mueller Airport about 18 years ago). Needless to say this location surrounded by nearby trees has much lower wind speeds than the other National Weather Service automated weather station at Austin Begstrom Airport only about 17 km away located near a wide open runway. The wind sheltered environment at Mabry causes higher temperatures on sunny days with light winds, often as much as 1C to 2C. Also the ridgetop location at Mabry causes higher minimum temperatures at night in winter with light winds and clear skies compared to the flat low-lying Bergstrom measurements, with Mabry often showing minimums 4C to 6C higher. The anomalies from these stations are usually fairly close, but rarely match. In early 2008, a USCRN site was added near Austin to the northwest about 46 km northwest of Mabry and about 300 m higher. I have not yet compared the measurements, but it is on a high ridgetop with no nearby trees and will only represent similar locations in the area. I wish they could add a companion USCRN station at the bottom of the nearby Cow Creek valley about 150 m lower where I see much lower temperatures on clear light wind winter nights (viewing several temperature measurements in the valley from the Weather Underground WunderMap).
Nick's TempLS mesh map indicates lots of variation in the temperature anomalies between stations in relatively high density areas like the US. If we had even higher density and resolution I would expect even more variation. To me this a sign of the substantial uncertainty in trying to assemble the very spatially limited data into regional, global land only, or global land/ocean temperature anomaly estimates. The oceans of course have a whole different set of measurement and anomaly estimation problems.
This is the comparison to 1998 that I think could cause 2016 to finish over 1.00 ℃.
ReplyDeleteThanks Olof for your hint on El Chichon. I well remember Chichon's and Pinatubo's eruptions, but until now didn't put them in relation with a global El Niño weakening.
ReplyDeleteThe correctness of your opinion is best shown when comparing the MEI with temperature records restricted to the Tropics latitude stripe 25N-25S (I did with UAH TLT, RSS TTT, IGRA surface and GHCN unadjusted).
But your hint made me curious; I inspected the chart for other similar periods. And discovered one about 10 years later, where the temperature response to the MEI not only was far weaker, but even negative, as if the ENSO event at that time had been a La Niña.
It is the period 1991-1993: exactly that corresponding to the Pinatubo eruption, calssified with a volcanic explosivity index of 6 (compared to 5 for El Chichon and St Helens). Interesting.
Maybe we would find similar correlations at the time of other eruptions like Grímsvötn or Tambora if the historic MEI record was accurate enough.
If anyone thinks that a volcanic eruption has any effect on the El Nino event itself, they are mistaken. It can have a compensating side-effect on global temperature but that is an orthogonal behavior. To be charitable, I think that's what Olof meant when he used the term "neutralised". One only needs to look at the strength of the SOI signal during El Chichon to see that it didn't come close to derailing it. The SOI measure is of a pressure differential in the dipole region and that was one of the strongest ever seen.
DeleteSame thing for QBO. During Pinatubo, the QBO behavior proceeded like clockwork, but the stratospheric temperature showed a hump. See Figure 10 in this recent paper
http://iopscience.iop.org/article/10.1088/1755-1315/31/1/012032/pdf
That's evident whut. The error is mine: I should have written "temperature response weakening".
Delete