Guest post by Bob Tisdale
THE INDIAN AND ATLANTIC OCEANS
INTRODUCTION
This post is a continuation of Part 1 – Satellite-Era Sea Surface Temperature Versus IPCC Hindcast/Projections. It examines the differences between multi-model mean of the IPCC 20C3M (Hindcast)/SRES A1B (Projection) for Sea Surface Temperatures and the observations using Reynolds OI.v2 Sea Surface Temperature (SST) anomalies, a satellite-based SST dataset. Part 1 covered the global oceans and the Pacific. This post discusses the Indian and Atlantic Oceans.
Note that Part 1 was updated with additional comparisons that started and ended during ENSO-neutral years. The update was added April 11, a day after the post was originally published. Refer to the discussion after Figure 17 in Part 1. The update also appeared in my April 11, 2011 at 5:48 pm comment in the cross post at Watt’s Up With That (Satellite-Era Sea Surface Temperature Versus IPCC Hindcast/Projections – Part 1). Using ENSO-neutral years for the start and end years did not have a major impact the divergence between the models and observations, so they won’t be considered in this post.
This post includes a discussion of the Atlantic Multidecadal Oscillation (AMO). If this topic is new to you, refer to An Introduction To ENSO, AMO, and PDO — Part 2
We’ll begin the post with the comparisons for the Indian Ocean. The Atlantic will then be discussed, with an emphasis on the North Atlantic.
INDIAN OCEAN
Like the Global Sea Surface Temperatures and those of the North and South Pacific basins, the linear trend of the IPCC Hindcast/Projection (multi-model mean) for the Indian Ocean Sea Surface Temperatures is significantly higher than the satellite-era observations. Refer to Figure 1. The model mean has a linear trend of approximately 1.6 deg C per Century, while the observations show a linear trend of only 0.9 deg C per Century.
Figure 1
We can separate the Indian Ocean data into east and west subsets, using 80E as the dividing longitude, then compare the trends on a latitudinal (zonal mean) basis. As discussed in Part 1, the zonal mean data in these posts are based on the SST anomalies for 5-degree latitude bands (80S to 75S, then 75S-70S, etc.), from pole to pole. The graphs present the linear trends of the SST data for those latitude bands in Deg C/Decade, with the data starting in January 1982 and ending in February 2011. Figure 2 shows how the model hindcasts/projections for the East and West Indian Ocean subsets both follow somewhat similar patterns, with higher linear trends in the tropics than at the high latitudes of the Southern Hemisphere.
Figure 2
But the trends of the observations show little similarities between the east and west portions of the Indian Ocean, Figure 3. South of 40S there are portions of the Indian Ocean that have cooled over this period, but the cooling was not forecast by the models. And note the warming in the East Indian Ocean between the latitudes of 60S and 45S. The model mean projections also failed to forecast it.
Figure 3
Figures 4 and 5 compare the observations and the model data for the West Indian and the East Indian subsets. They show how poorly the models hindcast/project the rise in Sea Surface Temperatures on a zonal-mean basis.
Figure 4
#######################################
Figure 5
In short, the model hindcast/projections of the rise in Indian Ocean Sea Surface Temperatures during the satellite era show no similarities to the observed rise.
ATLANTIC OCEAN
For the South Atlantic, the linear trend of the multi-model mean, Figure 6, is twice the observed linear trend. Note how the South Atlantic SST anomalies were basically flat from the late 1980s to 2008, yet the models show a nearly continuous rise. (Of course, that excludes the dips and rebounds in the early 1990s resulting from the eruption of Mount Pinatubo.)
Figure 6
The model hindcasts/projections for all of the basin subsets we’ve examined so far–the North and South Pacific, the Indian Ocean, and the South Atlantic–have all been significantly higher than the observed rise in Sea Surface Temperatures. The North Atlantic is the exception to this. Refer to Figure 7. The North Atlantic SST observations rose faster than the model mean. The models show a linear trend for the North Atlantic of approximately 1.6 deg C per Century, a trend that is similar to the other ocean basins. But the observed rise in the SST anomalies of the North Atlantic is approximately 2.5 deg C per Century.
Figure 7
And if we compare the trends for the models and observations on a zonal mean basis, the high latitudes of the North Atlantic have the greatest divergence.
Figure 8
The additional variability in the North Atlantic is not unusual. During the cooling period from 1944 to 1976, the mid-to-high latitudes of the North Atlantic showed the greatest cooling as well, Figure 9.
Figure 9
WHY ARE THE MODEL TRENDS FOR NORTH ATLANTIC LOWER THAN THE OBSERVATIONS DURING THE SATELLITE ERA?
The models do not consider the additional mode of natural variability in the North Atlantic known as the Atlantic Multidecadal Oscillation or AMO.
The AMO is typically illustrated with the North Atlantic SST anomalies detrended, but there are other ways to illustrate the additional variability. A simple way is to subtract Global SST anomalies from North Atlantic SST anomalies. Refer to Figure 10, which I’ve borrowed from the post An Introduction To ENSO, AMO, and PDO — Part 2″ href=”http://bobtisdale.wordpress.com/2010/08/16/an-introduction-to-enso-amo-and-pdo-part-2/”>An Introduction To ENSO, AMO, and PDO — Part 2. As illustrated, when smoothed with a 121-month running-average filter (commonly used for the AMO), the curve of the dataset that was created by subtracting the Global SST anomalies from the North Atlantic SST anomalies is very similar to the AMO data that’s calculated by detrending the North Atlantic SST anomalies. Note how both curves rose drastically from 1982 to present (the period illustrated in this post). The North Atlantic SST anomalies are responding to the additional natural mode of variability, and that additional variability is not included in the IPCC hindcasts/projections.
Figure 10
It is well known that the hindcasts/projections presented by the IPCC do not include the Atlantic Multidecadal Oscillation. Kevin Trenberth of the National Center for Atmospheric Research (NCAR) discussed this in his 2007 guest post at Nature.com, Climate Feedback: Predictions of climate. The third paragraph reads, “None of the models used by IPCC are initialized to the observed state and none of the climate states in the models correspond even remotely to the current observed climate. In particular, the state of the oceans, sea ice, and soil moisture has no relationship to the observed state at any recent time in any of the IPCC models. There is neither an El Niño sequence nor any Pacific Decadal Oscillation that replicates the recent past; yet these are critical modes of variability that affect Pacific rim countries and beyond. The Atlantic Multidecadal Oscillation, that may depend on the thermohaline circulation and thus ocean currents in the Atlantic, is not set up to match today’s state, but it is a critical component of the Atlantic hurricanes and it undoubtedly affects forecasts for the next decade from Brazil to Europe. Moreover, the starting climate state in several of the models may depart significantly from the real climate owing to model errors. I postulate that regional climate change is impossible to deal with properly unless the models are initialized.” [My caps]
The boldfaced sentence about the Atlantic Multidecadal Oscillation should also have included forecasts for Arctic Sea Ice, since Arctic Sea Ice is strongly impacted by the Sea Surface Temperatures of the North Atlantic.
LONG-TERM NORTH ATLANTIC SST OBSERVATIONS VERSUS INDIVIDUAL ENSEMBLE MEMBER HINDCAST/PROJECTIONS
The fact that the IPCC models are “not set up to match today’s state” of the Atlantic Multidecadal Oscillation has been known since 2007 when Nature published Kevin Trenberth’s guest post. The failure to properly model and present the AMO has impacted the differences between models and observations shown in Figures 7 and 8 and is likely the reason why the North Atlantic is the only ocean basin where the observed rise in SST anomalies exceeds the model projections. Refer to Table 1.
Table 1
I also wanted to see how the individual IPCC hindcasts/projections represented the AMO. Animation 1 is a .gif animation. It includes the HADISST-based Atlantic Multidecadal Oscillation, calculated as the North Atlantic SST anomalies minus Global SST anomalies. The 32 IPCC AR4 ensemble members that provided TOS (SST) data and that are available through the KNMI Climate Explorer are also presented, individually, in sequence. They are calculated the same way; that is, Global SST anomalies for each ensemble member are subtracted from their respective North Atlantic SST anomalies. (Note that I excluded ensemble member 18 since there was missing data and I did not want to track it down.) As you will note, some appear to create a multidecadal signal in the North Atlantic. Others do not. Some show the North Atlantic SST anomalies rising faster than Global SST anomalies, while in others it’s reversed, with Global SST anomalies rising faster than those of the North Atlantic. One thing is certain: There is little to no agreement among the climate models on the future state of the North Atlantic.
Animation 1
CLOSING
With the exception of the North Atlantic, the IPCC Sea Surface Temperature hindcasts/projections (multi-model mean) for the ocean basins have significantly higher linear trends than what has been observed during the satellite era. That is, since 1982, the Sea Surface Temperatures for the Indian, North and South Pacific, and South Atlantic have warmed much more slowly than projected by the models. The models have failed to capture the higher rise in North Atlantic SST anomalies because they do not include the additional natural variability that’s attributable to the Atlantic Multidecadal Oscillation.
The rises in the model-mean Sea Surface Temperatures also do not capture the observed changes on latitudinal (zonal mean) bases. This could strongly impact any regional forecasts made by the IPCC in AR4.
The failure of the model-mean hindcasts/projections to capture the observed rise, or lack thereof, in regional sea surface temperatures was also discussed in the post How Can Things So Obvious Be Overlooked By The Climate Science Community? It includes links to numerous posts that discuss the natural causes of the rise in Sea Surface Temperatures since 1982.
We often read statements by alarmists and climate scientists that “x” part of the globe is warming faster, or that some regional climate indicator is responding quicker, than projected by climate models. Looking back at the time-series and zonal-mean graphs presented in this series of posts, the model-mean sea surface temperature data does not come close to representing reality, so the inability of the models to project regional warming, or to project the correct timing of certain regional indicators, is not surprising and should be considered model failings, not signs of impending doom. Some might even view the facts that the models were not initialized to the observed state, that the models failed to replicate some ocean processes, and that other known ocean processes were intentionally excluded from the models as the means for the models to underestimate the impacts of those ocean states and processes on many regional climate indicators.
SOURCE
The data presented in this post is available through the KNMI Climate Explorer:
http://climexp.knmi.nl/selectfield_obs.cgi?someone@somewhere
Kevin Trenberth is a master of the pithy comment. Interesting quote.
I wonder what you would get, if you calculated; not the US CO2 output emissions, but the emissions divided by the gross domestic product.
Easiest way to shut down the CO2 is to simply stop doing anything productive; including growing food. CO2 emissions will plummet.
Bob:
If the IPCC ocean temp projections were not initialized to observations and did not include AMO what exactly did they do? Was it just a proportionate assumption as to how surface temps would respond to a presumptively warming atmosphere?
What little I have read and understood about ocean temps, boundary layers and lag times leads me to believe there are bigger unknowns than a few giant squids lurking in the oceans, as was discussed thoughtfully here: http://noconsensus.wordpress.com/2011/04/05/234-5/
“Initialize” means to inscribe one’s initials. Plotting a time-series from its initial point means to “initiate” rather than “initialize” the data-stream. How this foolish error has attained such currency is a mystery to anyone respecting English usage. Absent authors’ basic literary competence, what confidence can readers have in any scientific study’s explication?
George Tobin says: “If the IPCC ocean temp projections were not initialized to observations and did not include AMO what exactly did they do?”
I can’t answer your question. I assume it varies from model to model and would depend on the complexity of the model and the assumptions made while creating each model.
initially, i thought i might reply to mr. blake’s monogrammatical error… but nah – better to do something unexpectorated.
Huh – comparing climate modeling projections with observation – did not the Great Gavin say that such should not be done?
I find the complexities of your chosen subject extraordinary difficult to follow, and regret that my age related lack of time and comprehensive abilities preclude my ever being able to apply the knowledge that you disseminate to an area of particular interest to me as a board member of a Scottish salmon river.
It seems to me that that the decline in Atlantic salmon and sea trout numbers might be related to the AMO and AO, with the warm part of the cycle causing a decline and cooling aiding recovery.
This would probably be more discernable in the case of Pacific salmonids, as the shorter cycle of the PDO and ENSO could be more revealing.
I would very much appreciate your thoughts on this matter and any related information that others in this community might wish to bring to bear.
Those of you predisposed to the /sarc tag need not reply; I have already rehearsed the carbon dioxygenated water theories!
Interesting post. However, you state that “the IPCC Sea Surface Temperature hindcasts/projections (multi-model mean) for the ocean basins have significantly higher linear trends than what has been observed during the satellite era”. Eyeballing the temperature time series and the fitted linear trends, I would guess that the uncertainty intervals for the fitted slopes for most, if not all, individual time series overlap — this would indicate non-significant differences between the time series/hindcast trends. It would be most helpful for a qantitative comparison if you would give the uncertainty estimates for your linear fits, I am sure that your analysis tool of choice supplies them readily.
It is worth noting, however, that even if the uncertainty intervals of the individual time series trends overlap, the uncertainty intervals on the collective trends estimated from the ensemble of time series and hindcasts likely will not.
I am a bit shocked and dismayed that the models could be wrong………. /sarc
Great post Bob.
ZT says: “Huh – comparing climate modeling projections with observation – did not the Great Gavin say that such should not be done?”
RealClimate compares observations and model forecasts on an annual basis:
http://www.realclimate.org/index.php/archives/2009/12/updates-to-model-data-comparisons/
And:
http://www.realclimate.org/index.php/archives/2011/01/2010-updates-to-model-data-comparisons/
Kristoffer Haldrup says: “Eyeballing the temperature time series and the fitted linear trends, I would guess that the uncertainty intervals for the fitted slopes for most, if not all, individual time series overlap — this would indicate non-significant differences between the time series/hindcast trends.”
The linear trends of the models for most ocean basins are twice the observed trend. Those are not non-significant differences. On zonal mean bases, the models do not represent observations in any way.
John Blake says: “’Initialize’ means to inscribe one’s initials.”
Merriam-Webster disagrees with you:
“to set (as a computer program counter) to a starting position, value, or configuration”
http://www.merriam-webster.com/dictionary/initialize
Bob, if the one-sigma uncertainty intervals on the time series/hindcast slope estimates overlap, then the difference between the two slopes is insignificant, even if one slope is twice the other. Example: observed slope = 0.1 degr/dec +/- 0.1 and hindcast slope = 0.2 degr/dec +/- 0.05 degr/dec would not be a significant difference between the two if the uncertainty intervals represent one-sigma bounds.
This is why uncertainties on estimated parameters should be reported in any analysis.
John Blake says:
“Initialize” means to inscribe one’s initials
My Concise Oxford Dictionary uses: initial (vt) to inscribe one’s initials.
No need for the -ize suffix.
#
#
John Blake
April 19, 2011 at 12:28 pm
“Initialize” means to inscribe one’s initials. Plotting a time-series from its initial point means to “initiate” rather than “initialize” the data-stream. How this foolish error has attained such currency is a mystery to anyone respecting English usage. Absent authors’ basic literary competence, what confidence can readers have in any scientific study’s explication?
###
BZZZT! Wrong answer.
BTW, every time I see graphs like these, indicating temperature in 100th of a degree C, I wonder how reliable they could be. I have spent my entire working life involved with test and measurement, yet I would be hard pressed to give a real number representing the average temperature of my 500 L jungle stream Aquarium to within .01 degree C accuracy.
Bob, I don’t know if you have covered this before; but can you say something about the basic Sea surface Temperature sensing methodology. I gather this is Satellite based; I’m curious as to exactly what is being sensed ?
George
roger says April 19, 2011 at 1:38 pm
Climate Change and Long-Term Fluctuations of Commercial Catches – The Possibility of Forecasting
“”””” Jantar says:
April 19, 2011 at 3:04 pm
John Blake says:
“Initialize” means to inscribe one’s initials
My Concise Oxford Dictionary uses: initial (vt) to inscribe one’s initials.
No need for the -ize suffix. “””””
Well I don’t think that the OED would say “initialize” anyway; that’s American spelling; it would be “initialise” (but zed would be optional)
Figure 10 is a clincher. A sinusoidal curve would be a better fit than a straight line. It is clearer too if you go back further using HadCRUT back to 1850 – this gives two full wavelengths.
It immediately takes about 0.27 C away from the roughly 0.82 C rise between 1900 and 2000, since in 1900 the sine curve was at bottom and in 2000 it was at peak.
No hint of any similar sinusoidal signals in the IPCC model output in Animation 1, with possible exception of Ensembles 20 and 30.
George E. Smith says: “Bob, I don’t know if you have covered this before; but can you say something about the basic Sea surface Temperature sensing methodology. I gather this is Satellite based; I’m curious as to exactly what is being sensed ?”
George, in “A Real-Time Global Sea Surface Temperature Analysis”, Reynolds (1988) provided a basic description of the Reynolds OI.v2 data.
ftp://ftp.emc.ncep.noaa.gov/cmb/sst/papers/reynolds_1988.pdf
Of course, it’s been updated since then. All of the pertinent papers are listed (linked) at the bottom of the OI webpage:
http://www.emc.ncep.noaa.gov/cmb/sst_analysis/
The huge weakness – or strength, if they were right – in the IPCC AGW meme is that increases in atmospheric CO2 has a simple, linear relationship (at current levels) to global temperatures. The disconnect between the models and observations for the SSTs is about 58%, i.e. the models predict temperature rises that are about 58% greater than observed. The conclusion, based on the simple ppmv/C* viewpoint, must mean that the radiative forcing for a doubling of CO2 is about 2.38 W/m2 at this time, not 3.75 W/m2. That would be a reasonable correction is the science was not “settled” to a “97%” consensus level with a “>95%” certainty.
This is completely in the reverse, however, of the heat content of the oceans measurements, however. Between 1983 and 2011 the OHC increased about 0.34 GJ/m2, or 1.22E23J. This 28 year period, or 8.84E8 seconds, means that the oceans have accepted a W/m2 of 0.39 W/m2 AVERAGED over 28 years. Taking the beginning period at 1983 as having an additional Nil W/m2, then the difference is twice the average in 2011, or 0.77 W/m2 over a period in which the pCO2 has increased about 56 ppmv. If CO2 were to be, directly and indirectly through increased water vapour, this would mean a radiative doubling for CO2 of about 4.52 W/m2 (relative to 1983). And that is not considering the Land Heat Content or Atmospheric Heat Content. Which means that by Heat Content considerations the IPCC “settled” impact of CO2 is UNDERESTIMATED.
The IPCC models for CO2 radiative heating on one hand indicate more than observed warming of the SST or atmospheric temperature , but do not contribute enough heat for the heat content of the oceans as measured. If the IPCC instead considered an insolation increase through cloud cover either on a gross level or on a time-and-location dependent basis, then the warming of the oceans, the warming of the land and then the warming of the atmosphere would make sense.
I don’t understand why the disconnect between what the IPCC says and what is observed, and how predictions failing to come true doesn’t make the “settled” science and “consensus” opinion whither while they whine. Still, we accept politicians who do not do as they say within days of being elected, so I suppose I shouldn’t be surprised even as I fail to understand.
Interesting stuff. Can any correlation be gleaned on the Arctic (and Antarctic) sea ice extents from those big peaks and valleys on the Atlantic graphs over the past 10 years ??
Also:
Initialize definition, by Googling, first hit:
1. Set to the value or put in the condition appropriate to the start of an operation.
2. Format (a computer disk).
Maybe it’s an Americanization, maybe not, but less scary than when I first came to live here and, while traveling by air a lot, was told on many occasions that the plane I was on was going to take off momentarily. I’m OK with that now, since the definition’s been expanded.
mo·men·tar·i·ly/ˌmōmənˈte(ə)rəlē/Adverb
1. For a very short time.
2. At any moment; very soon.
The animation was fantastic. It was amazing to see the variations roll by. It seemed as if, were they all printed together, that you’d wind up with a nice solid red ribbon across the graph.
George E. Smith.
“Well I don’t think that the OED would say “initialize” anyway; that’s American spelling; it would be “initialise” (but zed would be optional)”
The “ize” ending is the older English form, and is used by the OED for (e.g.) “realize”. On this, the Americans are (for once) correct in using the letter, even if they get its name wrong.
The nasty “ise” ending is a modern British and Australian innovation. I suspect that it stems from the British export drive at the end of WW2. Europe was in ruins, and had lost most of its zeds. Europeans (and especially Eastern Europeans) are heavy users of zeds, so Britain exported almost its entire stock to Europe. This resulted in a zed shortage in Britain, and spare esses were used instead.
No excuse for it these days, though. Plenty of cheap zeds are available from China.
Bob Tisdale says:
April 19, 2011 at 2:18 pm
Re: Initialization
This refers to the process of determining the initial conditions for the differential equations you are solving in your climate model (well, for those groups which actually know the equations they’re solving, that is). And, of course, climate IS solved as an initial value problem.
Still, it’s amazing that they do so poorly with their hindcasting…nothing like having the answer to work towards…
O.K. So in a pub argument can I say “The Warmists said the seas would warm up, but in fact they are cooling down”?
It is like predicting hourly temperature and not considering the diurnal cycle. And assigning the observed temperature rise in the morning to something else than rising Sun.
Compare reality
http://i51.tinypic.com/k8epd.jpg
with the model
http://climexp.knmi.nl/data/isst_mpi_echam5_sresa1b_280-360E_0-70N_n_++_1850:2010a.png
I popped over to the links at RealClimate that Bob Tisdale provided, and saw that the HadCRUT3v, NCDC and GISTEMP temperature records show essentially the same flat or gently rising trend since 1998 that the UAH satellite records show. Yet the conclusion there is that “global warming continues. Did you really think it wouldn’t?”
If one looks at the trend from around 2001 to 2009, it’s obvious that the model completely missed the downturn — the divergence is tremendous, and only the El Nino of last year allowed reality to “catch up” to the model. Does it not matter to them that the warming has nothing to do with CO2, just as long as the numbers are close?
George E. Smith says:
Well I don’t think that the OED would say “initialize” anyway; that’s American spelling; it would be “initialise” (but zed would be optional)
Actually, OED has this:
initialize, v.
[f. initial n. + -ize.]
1. rare. a. intr. To use an initial or initials instead of the full name. b. trans. To designate by an initial or initials.
2. trans. (Computers.) To set to the value, or put in the condition, appropriate to the start of an operation. Const. to.
It would be a better exercise to compare satellite data with ARGO data. A comparison between remote sensed and directly measured would be better than comparing satellite data with a model’s rubbish output.
DesertYote , each of the “measurements” of a given year’s temperature of course comes with an uncertainty, even if the mean is “known” with good accuracy – see e.g. Figure 3.6 in the IPCC AR4 WG1 report. It is approximately 0.2 degrees C, or about the same as the year-to-year difference. This is part of the reason why you need some fairly long time series (15y+) to say anything quantitative about trends that are on the same order of magnitude (per decade) as the uncertainty in the individual measurements. -The same thing goes for the modelled temperatures for each year, each can be assigned an uncertainty estimated from the ensemble of models run.
All of these uncertainties should enter into any calculation of trends and will significantly affect the “error bars” on the trend lines, as will the actual scatter in the data. The error bars on the individual zonal models and data sets should also be reported/shown in order to perform a meaningful, quantitative comparison between models and data.
This was the point I tried to make to Bob Tisdale above, sorry if I failed:)
John Marshall says: “It would be a better exercise to compare satellite data with ARGO data.”
ARGO surface data is not readily available for comparison. Also, ARGO floats are only at the surface every 10 days to transmit data, if memory serves me well.
RoHa says: “O.K. So in a pub argument can I say ‘The Warmists said the seas would warm up, but in fact they are cooling down’?”
No, but you could say that it’s the oceans are not warming as quickly as forecast. The linear trend for the models is about 50% higher than the observed trend for Global SST anomalies:
http://i55.tinypic.com/x3y4xs.jpg
But that also varies depending on the portion of the oceans we’re looking at. The East Pacific Ocean for example…
http://i53.tinypic.com/2h4xrwh.jpg
…is warming at a rate that’s only 14% of what was forecast.
Those graphs are from Part 1:
http://bobtisdale.wordpress.com/2011/04/10/part-1-%e2%80%93-satellite-era-sea-surface-temperature-versus-ipcc-hindcastprojections/
The other side of the coin is the North Atlantic discussed above, where the models left out a major mode of natural variability. In that case, the model prediction was less than the observed rise.
philincalifornia says: “Interesting stuff. Can any correlation be gleaned on the Arctic (and Antarctic) sea ice extents from those big peaks and valleys on the Atlantic graphs over the past 10 years ??”
Not really. Arctic sea ice extent is a function of many factors, including Sea Surface Temperature, ENSO mode, Sea Level Pressure (the Arctic Oscillation), the direction of the Beaufort Gyre, etc.
@ Kristoffer Haldrup ,
Hi Kristoffer,
Here is another way to test if there is a statistical difference between the actual data nad the models.
Create a new series by subtracting the model observations from the actual observations. If the trends in the 2 series are the same, then the mean of the new series should be 0. A test of the null that the mean of the new series does not differ statistically from 0 will give you the answer.
Otherwise, you could fuzzy up the models to create larger and larger error bars so that you could never refute the claim that the model trend and the actual trend are the same.
Typo above, ‘nad’ = and. Also, the statistical test would be a one-tailed test because we are testing if the mean is postivive, i.e. model anomaly > than actual anomaly.
” postivive”? Can’t type or proofread, I give up : (
Reference says:
April 19, 2011 at 3:31 pm
Climate Change and Long-Term Fluctuations of Commercial Catches – The Possibility of Forecasting
Thanks for your response, Reference, that really is quite helpful. I had hoped for input from other knowledgeable contributors, but it seems from the number of responders that this thread has not generated the usual interest.
However, there is enough in your link to enable my drawing attention to the possibility that “it was the oceans what dunnit!”
“philincalifornia says: “Interesting stuff. Can any correlation be gleaned on the Arctic (and Antarctic) sea ice extents from those big peaks and valleys on the Atlantic graphs over the past 10 years ??”
I also thought about this and I plotted the SST’s of the Arctic (60…90N) with the data from here: http://nomad3.ncep.noaa.gov/cgi-bin/pdisp_sst.sh?ctlfile=oiv2.ctl&varlist=on&new_window=on&lite=&ptype=ts&dir= and the result is this plot: http://www.dh7fb.de/noaice/sstark/image001.gif .
We see some correlation between the seaice-minimum and the max. SST-Anomaly.
When we take the data for the september seaice extent from NSIDC and make a scatter-diagramm
http://www.dh7fb.de/noaice/sstark/image002.gif with ice-extent
we find a good correlation between maximum sst-anomaly and ice-extent. There is still a question: “egg or chicken”? So we can discover the time-scale:
http://www.dh7fb.de/noaice/sstark/image003.gif and we find, that the maximum in sst-anomaly happens during week 33..35, about 3 weeks before the sea-ice minimum.
So it could be, that the sst-anomaly of the arctic ocean has a great influence (R=0,86 for 1990…2010) on the seaice-minimum (extent).
Kristoffer Haldrup says: “Bob, if the one-sigma uncertainty intervals on the time series/hindcast slope estimates overlap, then the difference between the two slopes is insignificant, even if one slope is twice the other.”
Point taken. For a post such as this where I’m comparing observations to the model-mean data, it would not be difficult to add 95% uncertainties to the graphs for those who are interested. But I would not go to the efforts that Lucia does at her blog The Blackboard…
http://rankexploits.com/musings/wp-content/uploads/2011/04/HadCrutMarch2011.jpg
…because I’m interested in also showing the differences in the year-to-year variability, and all of the additional trend lines obscure the actual data.
The graph is from Lucia’s recent post:
http://rankexploits.com/musings/2011/hadley-march-anomaly-0-318c-up/
If I were to present uncertainties, it would be more in line with how RealClimate presents the trend lines in their model vs observation graphs, with simple uncertainty notations. Example:
http://i55.tinypic.com/15nkcoj.jpg
And that graph is from this RealClimate post:
http://www.realclimate.org/index.php/archives/2011/01/2010-updates-to-model-data-comparisons/
A few notes:
The first thing to keep in mind is that the many of the visitors to WUWT and to my blog do not have science backgrounds. They’re looking for simple discussions and presentations. In fact, the posts that are visited most often, week after week, at my website are the three-part series titled “An Introduction To ENSO, AMO, and PDO”. With this in mind, adding uncertainty levels to the graphs and discussions would add another level of complexity to the posts.
Second, I personally view many of my posts as introductions to specific topics. I look for information on a topic (such as this discussion on model forecasts versus SST observations), and if I can’t find anything (no one had prepared a post comparing models to observations for the individual ocean basins), I investigate it myself. When I believe my visitors would be interested, I clean up the graphs and animations, and write a post. If someone wants to take the analysis further (and determine uncertainties, etc.), they are more than welcome to do so. That’s why I provide the link or links to the data source at the end of my posts.
You later wrote, “This was the point I tried to make to Bob Tisdale above, sorry if I failed:)”
Sorry I didn’t reply earlier, but I needed more than a few minutes to prepare this reply.
Regards
Glaciations can occur in three ways. The first, which is intuitive, is that both summers and winters get cooler, so that more snow falls and less snow melts. The snow builds up, compresses and becomes ice, and at the 120m thickness level flows of its own volition. This would be the “global cooling” scenario. The second is that greater amounts of snow falling during winters than are melted during the following standard-temperature summers. That would be the “harsher winters” scenario. The third, the non-intuitive way, is by summers being cooler, so that standard winter snowfalls aren’t melted – the “cooler summers” scenarios.
In terms of result, they are the same; in terms of cause, they are not.
I look at the rise and fall of the sealevel over the year and ask a question equivalent to the reversal of a glacial-start: which of the three is the cause? Are we seeing both increased melting and reduced freezing, increased melting but the same freezing, or the same melting but decreased freezing? Again, same result, but different causes.
The CO2 theory should be increased melting and reduced freezing. A prelim look makes me wonder, though, if the sea-level rise isn’t “standard” melting and decreased freezing. If the Earth’s axis were vertical and its weather consistent year-to-year and within regions, this would not be an issue. But the axial tilt and regional drought-floods, heatwave-cold that typify our planet and give us Russian, non-CO2 summers of fire, demonstrate that we should not ignore the time and location aspect of either “planetary” warming or warming causes. Between Jan 4 and July 4, i.e. between apogee and perigee positions of the Earth, there is a 20.8 W/m2 difference in insolation, a 6.1% difference. Of this, roughly 4.2 W/m2 is absorbed in January that is not absorbed during July. It is only the greater proportion of land in the northern hemisphere that caused the global temperature to be about 2.2C warmer during the July-summer period; by insolation, the world should be warmer during the Jan-summer period. If the Earth’s albedo were to stay the same all year (which it intuitively does not, given that same land/sea disproportion in the hemispheres), 11.7 W/m2 more strikes and is absorbed by the sea and not-sea during the January-summer than during the July-summer. Both the absorption difference of the atmosphere and that of the sea/not-sea are significantly greater than the alleged CO2 radiative difference over the last 30 years. Yet I fail to see where time and location differences have been demonstrated to be not material to our rise and fall of what are really quite marginal temperature differences even over the last several thousand years. The NOAA/NASA desire to average throughtout the year, get rid of any but the longer trends is not necessarily the way to truth. As any student of aviation or oceanic accidents will tell you, it is not one big error that leads to disasters, but the co-incidence of a number of, in themselves minor, errors that bring airplanes out of the sky and ships to wreck on rocky shores.
A de-long-term-trended sea-level rise showing just the annual signal, or a rise-fall graph vs time would demonstrate what portion of the melt-freeze was (mostly) responsible for the increase in sealevel. The CO2 in our atmosphere has a signficant variation as shown by (again) data with the seasonal variations left in. IPCC CO2 theory says that this should make a difference, as air temperatures are controlled by the retention of heat by CO2 increases these days. The retention would be immediate, though at the sea-air and land-air interface there would be moderation due to the substrate giving up heat to the lower air. Still, when the rise of 2 ppmv/year is supposed to give rise to hurricanes galore, a drop of 4 ppmv in a year should show up.
This data manipulation is outside my current skills. Is there such an up vs down sealevel comparison? Is there a seasonal trend of SST vs CO2 fluctuations? I’d imagine that it would require sharp statistics to see the trends within the longer term, but that is what computer algorithms are for, are they not?
dh7fb, regarding your April 20, 2011 at 7:28 am comment, great analysis, especially the Arctic SST data leading Sea Ice Extent. Thanks.
A quick note. Sorry I didn’t include this earlier in my reply to philincalifornia. Reynolds OI.v2 SST anomalies in the Arctic should agree in part with Sea Ice data. The other variable that’s deternined as part of the Reynolds OI.v2 analysis is sea ice cover. If you scroll down to the “Sea ice cover” datasets at the KNMI Climate Explorer…
http://climexp.knmi.nl/selectfield_obs.cgi?someone@somewhere
…you’ll find the option for Reynolds OI.v2 Sea Ice cover data.
I had never checked to see how well the Reynolds Sea Ice cover data correlates with their SST data or the SST data with the NSIDC Sea Ice Extent data.
Thanks again.
Regards
DesertYote says: “BTW, every time I see graphs like these, indicating temperature in 100th of a degree C, I wonder how reliable they could be.”
And I often wonder, when I download the Reynolds OI.v2 SST data from the NOAA NOMADS website…
http://nomad3.ncep.noaa.gov/cgi-bin/pdisp_sst.sh?lite=
…whether they really need to present it in six significant figures. But they do.
Thanks, Bob Tinsdale. A lot of these articles are pretty tough for non-specialists with limited time. One-line summaries are really helpful.
“dh7fb, regarding your April 20, 2011 at 7:28 am comment, great analysis, especially the Arctic SST data leading Sea Ice Extent. Thanks.”
Thank you for your freindly statement! Maybe ( before the melt-saison) you find some time and interest for a post about the arctic. In response to an actual paper ( http://www.agu.org/pubs/crossref/2011/2010GL045698.shtml ) I made some thoughts about the relation between OHC of the arctic ( 60N…90N) of the month (JAS) with much ice-free water and the Ice-extent. At first I made a plot of the OHC of augusts – data and the ice-extent of septembers : http://www.dh7fb.de/noaice/ohceis.gif , Data from here: ftp://sidads.colorado.edu/DATASETS/NOAA/G02135/Sep/N_09_area.txt and KNMI for OHC. We see a correlation but with a time-lag of one year. If we print it in a scatter-plot: http://www.dh7fb.de/noaice/ohcext.gif (without 1995 wich could be an outlier) we find a tight correaltion (R=0,87) between the OHC of the year x and the ice-extent of the year x+1. Maybe the sun heats the upper 700m of the ocean during JAS and in November (still almost ice-free) the arctic ocean has a big OLR and the rest of the retained heat defines the melting of the next year? In the OHC-plot we see, that in 2010 the OHC of JAS was much lower than in the years before, so we could see a much more greater ice-extent in 2011, when we use the recursion-formula of the scatter-plot: 6,35 Mio qkm. If this would happen… it could be the rule of the test 🙂
regards DH7FB
RoHa, your sense of humour and your ability to cut the crap out of waffle is refreshing and made me smile. For some reason, many posters here remind me of the incredibly boring and pedantic Education lecturer who shouted at me ” The problem you have is that you are a divergent thinker who distills everything I say down to the essentials!” I thought the guy was paying me a compliment but he was quite angry and was attempting to demean me in front of the class. He also realised I am inclined to take the piss.
And, like you, I am a non-specialist who loves one-liners as they contribute immediately to my understanding.