On Foster and Rahmstorf 2011 – Global temperature evolution 1979–2010
Guest post by Bob Tisdale
Note (May 23, 2013): Update 3 appeared in the cross post at my blog but not here at WUWT. I’ve added it here.
UPDATE 3 (January 14, 2012): I displayed my very limited understanding of statistics in this post. This was pointed out to me a great number times by many different people in numerous comments received in the WattsUpWithThat cross post.The errors in that initial portion of the post were so many and so great that they detracted from the bulk of the post, which was about the El Niño-Southern Oscillation. Please disregard this post and the WUWT cross post, and any other cross posts that may exist.
I have reissued the ENSO-related portion of the post herewith a number of additions. If Anthony Watts cross posts the new version at WattsUpWithThat, I’ll provide a link here.
Originally, when I wrote the post about Foster and Rahmstorf (2011), I had not included my error-filled discussion about their regression analysis. That was a last minute addition. Lesson learned.
HHHHHHHHHHHHHHHHHHHHHHHHHH
UPDATE 2 (January 5, 2012): PLEASE READ. Three things: First, I did not understand that a “linear time trend” used by Foster and Rahmstorf (2011) is different than a “linear trend”. My confusion also led to confusion for many bloggers who read my post and who commented on the WattsUpWithThat cross post. My apologies. For those interested, the “linear time trend” is discussed under the heading of “Data as trend plus noise” on the Wikipedia Trend Estimationwebpage.
Second, in addition to MEI, AOD, and TSI as independent variables, I mistakenly used the values of the linear trend, which EXCEL calculated with its LINEST function from the monthly GISS data, as the fourth independent variable. And this added to the confusion of those who were interpreting the equations. In retrospect, I should not have included the equations. I should have included a table that listed the coefficients instead.
Third, in my haste to publish this post, I failed to explain the steps I used to process the data, and it may have been confusing to those who were looking at equations and graphs. I performed the regression analyses with the “raw” monthly data; then using the resulting coefficients, I made the adjustments to the monthly data. (I had prepared a graph using monthly data, similar to F&R’s Figure 4, with 1979-2010 as base years. But I felt my version was an unintelligible spaghetti graph with little value, so I didn’t include it.) I then converted the adjusted data to annual data; and last, changed the base years to 1979-2010.
The bottom line: Although I mistook a linear trend for a linear time trend, and although I did not include all of the additional data refinements used by Foster and Rahmstorf (2011), it’s difficult to see any difference between my Figure 7 and their Figure 5. There were other bloggers commenting on the thread of the WUWT cross post who got similar results using different methods. Does this mean the results of Foster and Rahmstorf (2011) are robust as some comments on the WUWT tread claimed? No. ENSO is a process, not an index, and it can’t be account for using linear regression analysis. This was illustrated clearly and discussed in detail under the heading of ENSO IS NOT AN EXOGENOUS FACTOR.
UPDATE 1 (January 3, 2012): Under the heading of ENSO IS NOT AN EXOGENOUS FACTOR, I changed the wording of a sentence, crossing out “create” and replacing it with “recharge”.—Thanks, Steve Allen.
OVERVIEW
This post examines a curious aspect of the multiple linear regression analysis performed by Foster and Rahmstorf in their 2011 paper “Global Temperature Evolution 1979–2010”. I find it very odd that a factor upon which the paper appears to rest was not presented in detail in it. Please understand right from the start, for this portion of the post, I am not implying that there is something wrong with this specific aspect of the paper; but I’m also not agreeing with it. I’m presenting it for discussion.
The second part of this post is a discussion of one of the exogenous factors that Foster and Rahmstorf (2011) has attempted to remove. The problem: it is not an exogenous factor. And there is a third discussion about a dataset that’s present in the spreadsheet provided by the lead author Grant Foster (aka Tamino) but, curiously, not mentioned in the paper.
Not surprisingly, Foster and Rahmstorf (2011) made the rounds at the blogs of the proponents of anthropogenic global warming. Joe Romm praised it with the post Sorry, Deniers, Study of “True Global Warming Signal” Finds “Remarkably Steady” Rate of Manmade Warming Since 1979. SkepticalScience covered the paper in their post Foster and Rahmstorf Measure the Global Warming Signal. And RealClimate gave it an honorable mention by including it as one of the topics in its Global Temperature News post.
INTRODUCTION
Foster and Rahmstorf (2011) attempted to remove from 5 global temperature datasets the linear effects of 3 factors that are known to cause variations in global temperature.
They covered the period of 1979 to 2010. The obvious intent of the paper is to show that anthropogenic global warming continues unabated in all of those datasets. The independent variables listed in the abstract of Foster and Rahmstorf (2011) are El Niño-Southern Oscillation, volcanic aerosols, and solar variations. Foster and Rahmstorf (2011) appears to be a much clarified version of Tamino’s (Grant Foster’s) January 20, 2011 post How Fast is Earth Warming? After publication of the paper, Tamino discussed it in his post The Real Global Warming Signal and was kind enough to provide the source data and code in his post Data and Code for Foster & Rahmstorf 2011. The data Tamino provided is available here. It is a .zip file that Tamino has renamed a .xls file, as he explains, “in order to fool the wordpress software into believing that it’s an Excel file.” You will need to “Right Click and Save As” and then change the file name back to a .zip file in order to open it.
As noted above, in the abstract, Foster and Rahmstorf (2011) list the exogenous factors that are used as independent variables in the multiple regression analysis as “El Niño/southern oscillation, volcanic aerosols and solar variability.” Curiously, three paragraphs later, when they list the factors included in the multiple regression analysis again, Foster and Rahmstorf (2011) have added a fourth variable: linear trend. The last sentence of the third paragraph under the heading of “Introduction” reads:
“The influence of exogenous factors will be approximated by multiple regression of temperature against ENSO, volcanic influence, total solar irradiance (TSI) and a linear time trend to approximate the global warming that has occurred during the 32 years subject to analysis.”
But one of the bases for the paper is to illustrate how similar the trends are after the adjustments for ENSO, Total Solar Irradiance, and Volcanic Aerosols have been made, so including the linear trends of those datasets in the regression analysis seems odd. As a result, I went in search of another reason why Foster and Rahmstorf (2011) would have needed to include the linear trend in their regression analyses. As I note in the following, I’m using commercially available add-on software for EXCEL to perform the multiple regression analyses. Since I have no other means to verify the results, other than reproducing the results of one of their graphs, I’ll request that you confirm the following results if you have that capability.
WHY DID FOSTER & RAHMSTORF NEED TO INCLUDE A LINEAR TREND IN THE MULTIPLE REGRESSION ANALYSIS?
The only reason that I can see that Foster and Rahmstorf (2011) needed to include the trend in the multiple regression analysis is, the adjustment factor for the solar data is the wrong sign when the multiple regression analysis uses only ENSO, Solar, and Volcanic Aerosol data as independent variables. Let me explain in more detail. But again, please understand, for this portion of the post, I am not implying that there is something wrong with this specific aspect of the paper; and again, I’m also not agreeing with it. I found this interesting.
With the data provided by Tamino, I used Analyse-It for EXCEL software to perform a multiple regression analysis. (For those with EXCEL who have no means to perform a multiple linear regression analysis and want to verify my results, Analyse-It is available free on a 30-day trial basis.) My initial analysis included Tamino’s favorite global Surface Temperature dataset GISS as the dependent variable and the Multivariate ENSO Index (MEI), the Total Solar Irradiance (PMOD), and the Volcanic Aerosol Optical Depth data (AOD) as the independent variables. I lagged the MEI data by four months, the PMOD data by one month, and the AOD data by seven months, in agreement with Table 1 of Foster and Rahmstorf (2011), which is also Table 1 in this post. And in this analysis, I did not include the GISTEMP linear trend as an independent variable.
Table 1
The multiple regression analysis using only the ENSO (MEI), Solar (PMOD), and Volcanic Aerosol (AOD) data resulted in Equation 1:
EQUATION 1:
GISS = 123.6 + 0.06769MEI(4m lag) – 0.09025TSI.PMOD(1m lag)– 3.837AOD (7m lag)
I highlighted the solar variable scaling factor in boldface to emphasize the fact that the sign is negative. It would need to be positive to reproduce the results of Foster and Rahmstorf (2011). The signs of the ENSO and volcanic aerosol factors are what one would expect, Figure 1. It’s only the sign of the solar coefficient that is the opposite of what Foster and Rahmstorf (2011) present, Figure 2 (which is their Figure 7).
Figure 1
HHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHH
Figure 2
And that makes a monumental difference to the outcome of Foster and Rahmstorf (2011). If we adjust the GISS surface temperature data with the factors presented in Equation 1, then the rise is not continuous. Refer to Figure 3. The peak year for the adjusted GISS-based global Surface Temperature data is 2002.
Figure 3
To confirm the results of Foster and Rahmstorf (2011), I added the 0.167 deg C/Decade linear trend of the GISS global surface temperature anomaly data to the independent variables. The lags of the ENSO (MEI), Solar (PMOD), and Volcanic Aerosol (AOD) data remained the same as above.
The multiple regression analysis using the ENSO (MEI), Solar (PMOD), and Volcanic Aerosol (AOD) data and the linear trend resulted in Equation 2:
EQUATION 2:
GISS = -91.43 + 1.024Trend + 0.0761MEI(4m lag) + 0.06694TSI.PMOD(1m lag)– 2.334AOD (7m lag)
The sign of the Total Solar Irradiance coefficient now agrees with what Foster and Rahmstorf (2011) presented, as shown in Figure 4. Note that including the trend as an independent variable also influenced the scaling of the ENSO (MEI) and Volcanic Aerosol (AOD) data. It increased the scaling factor of the ENSO data a little, but decreased the scaling factor of Volcanic Aerosol significantly. Of course, the inclusion of the trend as an independent variable, with the change in sign of the Solar influence, also gives the adjusted GISS data results that Foster and Rahmstorf (2011) wanted, Figure 5, with the rise in temperature relatively steady over the 32 year period. And note that the trend of 0.172 deg C per decade is comparable to the findings of Foster and Rahmstorf (2011) shown in Table 1 for GISS data.
Figure 4
HHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHH
Figure 5
Of course, I did not include [and Foster and Rahmstorf (2011) could not have included] the trend adjustment from Equation 2 when the corrected data was presented in Figure 5. If the trend adjustment was included, the corrected data would have no trend. That means, it appears Foster and Rahmstorf (2011) needed to include the trend of the GISTEMP data in the regression analysis only to assure the sign of the solar influence they sought.
Foster and Rahmstorf (2011) would have gotten similar scaling factors for the ENSO (MEI), Solar (PMOD), and Volcanic Aerosol (AOD) data if they had simply detrended the GISS Global Surface Temperature data.
EQUATION 3:
Detrended GISS = -86.31 + 0.0759MEI(4m lag) + 0.0632TSI.PMOD(1m lag) – 2.37AOD (7m lag)
REVERSED SIGN OF SOLAR INFLUENCE IS COMMON TO ALL GLOBAL TEMPERATURE DATASETS
Someone is bound to ask whether the GISS Global Surface Temperature dataset is the only dataset with these results. The answer is no. If the linear trend is not included in the multiple linear regression analyses, the sign of the solar coefficient is the opposite of what Foster and Rahmstorf (2011) would had to have used for the NCDC and HADCRUT global land plus sea surface temperature datasets and for the RSS and UAH global Lower Troposphere Temperature data. The resulting equations from the linear regression analyses of the other datasets are presented in equations 4 through 7. The lags for the independent variables are as listed in Table 1 above:
EQUATION 4 (NCDC Land Plus Ocean Surface Temperature):
NCDC = 109.1 + 0.05495MEI(2m lag) – 0.0796TSI.PMOD(1m lag)– 3.113AOD (5m lag)
EQUATION 5 (Hadley Centre HADCRUT Global Surface Temperature Anomalies):
HadCRUT3v = 92.21 + 0.06421MEI(3m lag) – 0.0673TSI.PMOD(1m lag)– 3.293AOD (6m lag)
EQUATION 6 (RSS MSU Lower Troposphere Temperature Anomalies):
RSS33 = 61.44 + 0.1285MEI(5m lag) – 0.04489TSI.PMOD(0m lag)– 4.863AOD (5m lag)
EQUATION 7 (UAH MSU Lower Troposphere Temperature Anomalies):
UAH = 72.94 + 0.1332MEI(5m lag) – 0.05338TSI.PMOD(0m lag)– 5.139AOD (6m lag)
If we use those coefficients, the five datasets do not produce the nice continuous rise in Global Temperatures that Foster and Rahmstorf (2011) wanted to present, as shown in Figure 6. For the three Surface Temperature anomaly datasets (GISS, HADCRUT, NCDC) 2002 has the highest temperature. It’s only the two Lower Troposphere Temperature anomaly datasets that have 2010 as the warmest year.
Figure 6
And as one would expect, if the linear trends of the other global temperature datasets are included in the independent variables, the signs of the solar coefficients are positive. Refer to equations 8 through 11.
EQUATION 8 (NCDC Land Plus Ocean Surface Temperature, with trend):
NCDC = -106.7 + 1.085Trend + 0.06832MEI(2m lag) + 0.07813TSI.PMOD(1m lag)– 1.68AOD (5m lag)
EQUATION 9 (Hadley Centre HADCRUT Global Surface Temperature Anomalies, with trend):
HadCRUT3v = -119.2 + 1.093Trend + 0.07519MEI(3m lag) + 0.08723TSI.PMOD(1m lag)– 1.858AOD (6m lag)
EQUATION 10 (RSS MSU Lower Troposphere Temperature Anomalies, with trend):
RSS33 = -135.5 + 1.05Trend + 0.1342MEI(5m lag) + 0.09923TSI.PMOD(0m lag)– 3.479AOD (5m lag)
EQUATION 11 (UAH MSU Lower Troposphere Temperature Anomalies, with trend):
UAH = -105.7 + 0.9953Trend + 0.1381MEI(5m lag) + 0.07742TSI.PMOD(0m lag)– 3.871AOD (6m lag)
With the linear trend included in the multiple regression analyses, the coefficients in the equations above provide the adjustments that Foster and Rahmstorf (2011) presented, Figure 7. I’ve included their Figure 5 as my Figure 8 as a reference.
Figure 7
HHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHH
Figure 8
THE ASSUMPTION ABOUT THE LINEAR TREND
I’m sure some will attempt to argue that including the trend in the regression analyses is necessary since computer model-based studies have shown the rise in global surface temperature is caused by anthropogenic forcings during the period of 1979 to 2010. But of course, that argument assumes climate models can be used for determining the cause of the rise in Global Surface Temperatures during any period. We have recently illustrated and discussed that the climate models used by the IPCC in their 4th Assessment report have shown no skill at reproducing the global surface temperatures over any period during the 20th Century. Refer to the summary post ON THE IPCC’s UNDUE CONFIDENCE IN COUPLED OCEAN-ATMOSPHERE CLIMATE MODELS – A SUMMARY OF RECENT POSTS. The second problem with their assumption is that the global oceans, which cover about 70% of the surface area of the globe, show no signs of the influence of anthropogenic global warming during the satellite era. And that brings us to…
ENSO IS NOT AN EXOGENOUS FACTOR
Foster and Rahmstorf (2011) included ENSO as one of the exogenous factors they attempted to remove from the instrument temperature record. But ENSO is not an exogenous factor. ENSO is a coupled ocean-atmosphere process that periodically discharges heat to the atmosphere during an El Niño. The El Niño causes changes in atmospheric circulation patterns, which cause temperatures outside of the eastern tropical Pacific to vary, some warming, some cooling, but in total, the areas that warm exceed those that cool and global surface temperatures rise in response to an El Niño. The patterns of warming and cooling during a La Niña are similar to an El Niño, but the signs are reversed. And that’s really all that a paper such as Foster and Rahmstorf (2011) could hope to account for including ENSO in the regression analysis. But there is much more to ENSO.
ENSO is also a process that redistributes the warm water that was leftover from the El Niño itself and enhances the redistribution of the warm water that was created by the El Niño outside of the eastern tropical Pacific. The redistribution carries that warm water poleward and into adjoining ocean basins during the La Niña that follows an El Niño. La Niña events also recharge part of the warm water that was released during the El Niño. Sometimes La Niña events “overcharge” the tropical Pacific, inasmuch as they create recharge more tropical Pacific ocean heat than was discharged during the El Niño that came before it. That was the case during the 1973/74/75/76 and 1995/96 La Niña events. Refer to Figure 9. The 1973/94/75/76 La Niña provided the initial “fuel” for the 1982/83 Super El Niño and the multi-year 1986/87/88 El Niño. And the 1997/98 “El Niño of the Century” was fueled by the 1995/96 La Niña. The process of ENSO cannot be accounted for through linear regression on an index. This was illustrated and discussed at an introductory level in the post ENSO Indices Do Not Represent The Process Of ENSO Or Its Impact On Global Temperature.
Figure 9
Foster and Rahmstorf (2011) cited Trenberth et al (2002) Evolution of El Nino–Southern Oscillation and global atmospheric surface temperatures” as one of their ENSO references. But Trenberth et al (2002) include the following disclaimer in the second paragraph of their Conclusions, (their paragraph 52, my boldface):
The main tool used in this study is correlation and regression analysis that, through least squares fitting, tends to emphasize the larger events. This seems appropriate as it is in those events that the signal is clearly larger than the noise. Moreover, the method properly weights each event (unlike many composite analyses). Although it is possible to use regression to eliminate the linear portion of the global mean temperature signal associated with ENSO, the processes that contribute regionally to the global mean differ considerably, and the linear approach likely leaves an ENSO residual.
The ENSO “residuals” are a significant contributor to the rise in Global Sea Surface Temperatures during the satellite era as we shall see. Did Foster and Rahmstorf (2011) consider these residuals in their analysis? Nope. They assumed the rise was caused by anthropogenic forcing, and they assumed a linear trend represented it.
A more recent paper was overlooked by Foster and Rahmstorf (2011). Compo and Sardeshmukh (2010) “Removing ENSO-Related Variations from the Climate Record” seems to be a step in the right direction. They write (my boldface):
An important question in assessing twentieth-century climate is to what extent have ENSO-related variations contributed to the observed trends. Isolating such contributions is challenging for several reasons, including ambiguities arising from how ENSO is defined. In particular, defining ENSO in terms of a single index and ENSO-related variations in terms of regressions on that index, as done in many previous studies, can lead to wrong conclusions. This paper argues that ENSO is best viewed not as a number but as an evolving dynamical process for this purpose.
And as Compo and Sardeshmukh have suggested, Foster and Rahmstorf (2011) have reached the wrong conclusion.
Note: Compo and Sardeshmukh missed a very important aspect of ENSO. They overlooked the significance of the huge volume of warm water that is left over from El Niño events and they failed to account for its contribution to the rise in global Sea Surface Temperature anomalies since about 1976.
Let’s not forget the much-heralded Thompson et al (2008) paper “Identifying signatures of natural climate variability in time series of global-mean surface temperature: Methodology and Insights.” Thompson et al (2008) is the basis for the new and improved HADSST3 global sea surface temperature anomaly dataset from the Hadley Centre. Thompson et al (2008), like Foster and Rahmstorf (2011), is flawed because they attempt to remove the ENSO signal from the Global Surface Temperature record and claim the remainder of the rise in surface temperature is caused by anthropogenic forcings. In the Introduction, Thompson et al (2008) write (my boldface):
In this study we exploit a series of novel methodologies to identify and filter out of the unsmoothed monthly mean time series of global-mean land and ocean temperatures the variance associated with ENSO, dynamically induced atmospheric variability, and volcanic eruptions. The impacts of ENSO and volcanic eruptions on global-mean temperature are estimated using a simple thermodynamic model of the global atmospheric–oceanic mixed layer response to anomalous heating. In the case of ENSO, the heating is assumed to be proportional to the sea surface temperature anomalies over the eastern Pacific…”
That is a monumental assumption, and it’s the same flawed assumption made by Foster and Rahmstorf (2011).
But it was that specific language in Thompson et al (2008) that caused me to divide the Sea Surface Temperature anomalies of the Global Oceans into the two subsets, and those were the East Pacific from pole to pole (90S-90N, 180-80W) and of the Rest-Of-World (Atlantic-Indian-West Pacific) from pole to pole (90S-90N, 80W-180). And by coincidence, I used the Sea Surface Temperature dataset (Reynolds OI.v2) that’s used in the GISS Land-Ocean Temperature Index, which is Tamino’s favorite global Surface Temperature anomaly dataset. I first presented the Sea Surface Temperature for those two subsets in the March 3, 2011 post Sea Surface Temperature Anomalies – East Pacific Versus The Rest Of The World. (For those who are interested, there are about a dozen additional posts that discuss ENSO and the multiyear aftereffects of specific ENSO events linked at the end of that post.)
The East Pacific Sea Surface Temperature anomalies from pole to pole, Figure 10, are dominated by the variations in tropical Pacific caused by ENSO, and as a result, the variations in the East Pacific Sea Surface Temperature anomalies mimic ENSO, represented by the scaled NINO3.4 Sea Surface Temperature anomalies. The trend of the East Pacific Sea Surface Temperature anomalies is relatively flat at 0.011 deg C/Decade.
Figure 10
The reason the trend is so flat: warm water from the surface and below the surface of the western Pacific Warm Pool is carried eastward during an El Niño and spread across the surface of the eastern tropical Pacific, raising sea surface temperatures there. And during the La Niña events that follow El Niño events, the leftover warm water is returned to the western tropical Pacific. Due to the increased strength of the trade winds during the La Nina, there is an increase in upwelling of cool subsurface waters in the eastern equatorial Pacific, so the Sea Surface Temperatures there drop. In other words, the East Pacific is simply a temporary staging area for the warm water of an El Niño event. Warm water sloshes into this dataset from the western tropical Pacific and releases heat, and then the warm water sloshes back out.
But the warm waters released from below the surface of the West Pacific Warm Pool during the El Niño are not done impacting Sea Surface Temperatures throughout the global oceans, and they cannot be accounted for by an ENSO index. That leftover warm water is returned to the West Pacific during a La Niña event that follows an El Niño, much of it remaining on the surface. The Sea Surface Temperature in the western Pacific rises as a result. At approximately 10N latitude, a slow-moving Rossby wave also carries leftover warm water from the eastern tropical Pacific back to the western Pacific during the La Niña. Ocean currents carry the warm water poleward to the Kuroshio-Oyashio Extension (KOE) east of Japan and to the South Pacific Convergence Zone (SPCZ) east of Australia, and the Indonesian Throughflow (an ocean current) carries the warm water into the tropical Indian Ocean. And as noted above, due to the increased strength of the trade winds during the La Nina, there is an increase in upwelling of cool subsurface waters in the eastern equatorial Pacific, so the Sea Surface Temperatures there drop. But that cooler-than-normal water is quickly warmed during the La Niña as it is carried west by the stronger-than-normal ocean currents that are caused by the stronger-than-normal trade winds. And the reason that water warms so quickly as it is carried west is because the stronger-than-normal trade winds reduce cloud cover, and this allows more downward shortwave radiation (visible sunlight) to warm the ocean to depth. This additional warm water helps to maintain the Sea Surface Temperatures in the West Pacific and East Indian Oceans at elevated levels during the La Niña and it also recharges the West Pacific Warm Pool for the next El Niño event. Refer again to Figure 9, but keep in mind that it presents the Ocean Heat Content for the entire tropical Pacific, not just the Pacific Warm Pool.
And what happens when a major El Niño event is followed by a La Niña event? The Sea Surface Temperature anomalies for the Atlantic, Indian, and West Pacific Oceans (the Rest-Of-The-World outside of the East Pacific) first rise in response to the El Niño; the 1986/87/88 and 1997/98 El Niño events. Then the Sea Surface Temperatures of the Atlantic, Indian, and West Pacific Oceans are maintained at elevated levels by the La Niña; the 1988/89 and 1998/99/00/01 La Niña events. The results are the apparent upward shifts in the Sea Surface Temperature anomalies of the Atlantic, Indian, and West Pacific Oceans from pole to pole (90S-90N, 80W-180), as illustrated in Figure 11.
Figure 11
The dip and rebound starting in 1991 is caused by the volcanic aerosols emitted by the explosive volcanic eruption of Mount Pinatubo. And the reason the Rest-Of-The-World Sea Surface Temperature anomalies respond so little to the 1982/83 Super El Niño is because that El Niño was counteracted by the eruption of El Chichon in 1982.
To assure readers that the upward shifts in Rest-Of-The-World Sea Surface Temperature anomalies coincide with the 1986/87/88 and 1997/98 El Niño events, I’ve included an ENSO index, NINO3.4 Sea Surface Temperature anomalies, in Figure 12. The NINO3.4 Sea Surface Temperature anomalies have been scaled (multiplied by a factor of 0.12) to allow for a better visual comparison and shifted back in time by 6 months to account for the time lag between the variations in NINO3.4 Sea Surface Temperature anomalies and the response of the Rest-Of-The-World data.
Figure 12
But the ENSO Index data is visually noisy and it detracts from the upward shifts, so in Figure 13 I’ve isolated the data between the significant El Niño events. To accomplish this, I used the NOAA Oceanic Nino Index (ONI) to determine the official months of those El Niño events. There is a 6-month lag between NINO3.4 SST anomalies and the response of the Rest-Of-The-World SST anomalies during the evolution phase of the 1997/98 El Niño. So the ONI data was lagged by six months, and the Rest-Of-The-World SST data that corresponded to the 1982/83, 1986/87/88, 1998/98, and 2009/10 El Niño events was excluded—left as black dashed lines. All other months of data remain.
Figure 13
And to help further highlight the upward shifts, the average Sea Surface Temperature anomalies between the major El Niño events are added in Figure 14.
Figure 14
Based on past posts where I’ve presented the same dataset, some comments suggest the period average temperatures are misleading and request that I illustrate the linear trends. Figure 15 illustrates how flat the trends are between the 1986/87/88 and 1997/98 El Niño events and between the 1997/98 and 2009/10 El Niño events.
Figure 15
Back to the East Pacific data: If we adjust the East Pacific Sea Surface Temperature anomalies for the effects of volcanic aerosols, Figure 16, the linear trend is slightly negative. In other words, for approximately 33% of the surface area of the global oceans, Sea Surface Temperature anomalies have not risen in 30 years.
Figure 16
Note: The method used to adjust for the volcanic eruptions is described in the post Sea Surface Temperature Anomalies – East Pacific Versus The Rest Of The World, under the heading of ACCOUNTING FOR THE IMPACTS OF VOLCANIC ERUPTIONS.
And if we adjust the Rest-Of-The-World Sea Surface Temperature anomalies for volcanic aerosols, Figure 17, we reduce the effects of the dip and rebound caused by the 1991 eruption of Mount Pinatubo. And the trend of the Rest-Of-The-World data between the 1986/87/88 and 1997/98 El Niño drops slightly compared to the unadjusted data (Figure 15), making it even flatter and slightly negative.
Figure 17
In summary, ENSO is a coupled ocean-atmosphere process and its effects on Global Surface Temperatures cannot be accounted for with linear regression of an ENSO index as attempted by Foster and Rahmstorf (2011)–and others before them. We can simply add Foster and Rahmstorf (2011) to the list of numerous papers that make the same error. Examples:
Lean and Rind (2009) How Will Earth’s Surface Temperature Change in Future Decades?
And:
Lean and Rind (2008) How Natural and Anthropogenic Influences Alter Global and Regional Surface Temperatures: 1889 to 2006
And:
Santer et al (2001), Accounting for the effects of volcanoes and ENSO in comparisons of modeled and observed temperature trends
And:
Thompson et al (2008), Identifying signatures of natural climate variability in time series of global-mean surface temperature: Methodology and Insights
And:
Trenberth et al (2002) Evolution of El Nino–Southern Oscillation and global atmospheric surface temperatures
And:
Wigley, T. M. L. (2000), ENSO, volcanoes, and record-breaking temperatures
Additionally, Foster and Rahmstorf (2011) assumed that the global warming signal is linear and that it is caused by anthropogenic forcings, but those assumptions are not supported by the satellite-era Sea Surface Temperature record as shown above. The global warming signal is not linear, and the El Niño events of 1986/87/88 and 1997/98 are shown to be the cause of the rise in sea surface temperatures, not anthropogenic greenhouse gases.
THE ATLANTIC MULTIDECADAL OSCILLATION
Those who have downloaded Tamino’s allfit2.xls file here (and changed it back to a .zip file) will notice that the data in Column AA is identified as “AMO”. And yes, that is Atlantic Multidecadal Oscillation data from the NOAA Earth System Research Laboratory (ESRL) AMO website.
Note: The current AMO data and the data listed in Tamino’s file are slightly different. The reason: The ESRL AMO data is constantly evolving. Each month, when the new North Atlantic (0-70N, 80W-0) Sea Surface Temperature data are added, the data is detrended with the new data.
One could only speculate why Tamino included the AMO data in the spreadsheet–and why the data in the spreadsheet extends back to 1950, when the paper only deals with the period of 1979 to 2010. And one can also wonder why Tamino would include the ESRL AMO data, which is based on Kaplan North Atlantic Sea Surface Temperature anomaly data, when no surface temperature datasets (GISS, HADCRUT or NCDC) use Kaplan SST. It’s like subtracting the Hadley Centre’s CRUTEMP land surface temperature data from GISS LOTI data to determine the Sea Surface Temperature portion of GISTEMP LOTI data. The datasets are not the same. I’ve already pointed this error out to Tamino and his disciples in the post Comments On Tamino’s AMO Post.
But for example, let’s satisfy your curiosity. Let’s assume you were wondering what the results would be if you were to account for the impact of the AMO on Northern Hemisphere surface temperatures, using a linear regression analysis with the ESRL AMO data as the independent variable and with GISS Northern Hemisphere Surface Temperature data as the dependent variable. We’ll confine the example to the Foster and Rahmstorf (2011) time period of 1979 to 2010. Refer to Figure 18. The AMO-adjusted Northern Hemisphere surface Temperature has a linear trend that is only 41% of the unadjusted Northern Hemisphere data. Hmm. That would mean the AMO was responsible for 59% of the rise in Northern Hemisphere surface temperatures based on linear regression analysis.
Figure 18
And that’s in line with generalization made by Tamino’s associates at RealClimate in their Atlantic Multidecadal Oscillation (“AMO”)webpage. There they write that the AMO is:
A multidecadal (50-80 year timescale) pattern of North Atlantic ocean-atmosphere variability whose existence has been argued for based on statistical analyses of observational and proxy climate data, and coupled Atmosphere-Ocean General Circulation Model (“AOGCM”) simulations. This pattern is believed to describe some of the observed early 20th century (1920s-1930s) high-latitude Northern Hemisphere warming and some, but not all, of the high-latitude warming observed in the late 20th century. The term was introduced in a summary by Kerr (2000) of a study by Delworth and Mann (2000).
59% is definitely “some, but not all”.
Tamino continues to complain that one can’t make adjustments for the AMO because it includes a global warming component. For example, in a response to a December 22, 2011 at 6:11 pm comment by Colin Aldridge,Tamino writes:
As for AMO, unlike ENSO (or PDO for that matter) it IS temperature. Pure and simple, nothing more nothing less. Attributing temperature change to temperature change seems kinda stupid.
Hmm. I believe Tamino misses the point that the AMO is a mode of additionalvariability and that it is detrended over the entire term of the data.
Further to this end, I discussed and illustrated for Tamino that we can subtract the “warming signal” of the Global Sea Surface Temperature anomalies excluding the North Atlantic from the North Atlantic Sea Surface Temperature anomalies. That way we’re left with only the additional variability of North Atlantic Sea Surface Temperature anomalies caused by the AMO. And that’s really how the AMO should be expressed. We’ll call the difference the North Atlantic Residual. The North Atlantic Residual has, approximately, the same trend as the AMO for the 1979 to 2010 period of Foster and Rahmstorf (2011), as shown in Figure 19.
Figure 19
Note: The North Atlantic Residual data presented in Figure 16 is based on the combination of HADISST data for the years 1979 to November 1981 and Reynolds OI.v2 SST data from December 1981 to present in agreement with the GISS recipe listed on their GISS Surface Temperature Analysis webpage. To remove the North Atlantic Sea Surface Temperature data from the Global data, the North Atlantic surface area for the coordinates of 0-70N, 80W-0 was determined to represent 11% of the surface area of the global oceans.
And as an additional check of the sign of the solar correction, I performed multiple linear regression analyses with GISS Northern Hemisphere Surface Temperature data as the dependent variable and using the AMO data as an independent variable in one instance and the North Atlantic Residual data in a second. The analyses also included the ENSO (MEI), Solar (PMOD), and Volcanic Aerosols (AOD.NH) as independent variables with the same lags as the global data. In both instances, the sign of the solar correction was the opposite of what Foster and Rahmstorf (2011) were looking for, as shown in equations 12 and 13:
EQUATION 12:
GISS.NH = 62.25 + 0.001696MEI (4m lag) – 0.04528TSI.PMOD(1m lag)– 1.683AOD.NH (7m lag) + 0.866AMO (0m lag)
EQUATION 13:
GISS.NH = 72.12 + 0.04751MEI (4m lag) – 0.05258TSI.PMOD(1m lag)– 2.413AOD.NH (7m lag) + 0.72N. Atl. Residual (0m lag)
A closing AMO note: For an additional discussion on how the North Atlantic impacts the Sea Surface Temperatures of the periods between the upward shifts caused by the 1986/87/88 and 1997/98 El Niño events, refer to the post Supplement To “ENSO Indices Do Not Represent The Process Of ENSO Or Its Impact On Global Temperature”.
CLOSING COMMENTS
I found the inclusion of a linear trend in the regression analyses performed by Foster and Rahmstorf (2011) to be very interesting. It appears the linear trends were included simply to cause a solar correction that was the sign the authors wanted for their adjustments. One might think, if the basic results of the paper were dependent on whether a linear trend was included in the multiple regression analyses, this would have been discussed in the paper. And again, if you have the capability, and if you’re not satisfied with the similarities between my results and the Foster and Rahmstorf (2011) results (Figures 7 and 8), please confirm the multiple regression analyses results presented above with and without the linear trend.
This post also illustrated and discussed the error in their assumption that regression analysis can be used to remove the impacts of ENSO on Global Surface Temperature. ENSO is a process that is not fully represented by ENSO Indices. In other words, the ENSO indices only represent a small portion of the impacts of ENSO on Global Surface Temperatures. Attempting to use an ENSO index as Foster and Rahmstorf (2011) have done is like trying to provide the play-by-play for a baseball game solely from an overhead view of home plate.
The assumption made by Foster and Rahmstorf (2011) that a linear trend provides an approximate “global warming” signal was shown to be erroneous using Sea Surface Temperature data. When broken down into two logical subsets of the East Pacific and the Atlantic-Indian-West Pacific Oceans, Satellite-era Sea Surface Temperature data shows no evidence of an anthropogenic global warming signal. It only shows upward shifts associated with strong ENSO events. This would seem to complicate any attempt to justify the inclusion of the linear trend to reverse the sign of the solar adjustment.
And thanks to Tamino for including the Atlantic Multidecadal Oscillation data in his spreadsheet. It allowed me to illustrate the significant impact the AMO can have on Northern Hemisphere surface temperatures.
Happy New Year to all.
ABOUT: Bob Tisdale – Climate Observations
SOURCES
The spreadsheet that served as the source of the data for the regression analyses was linked to Tamino’s (Grant Foster’s) post Data and Code for Foster & Rahmstorf 2011.
To save you some time, here’s a copy of the file that contains the spreadsheet from Tamino’s blog that I’ve uploaded to mine, allfit2 as of 12-21-11. Again, you’ll have to download the file and change it to a .zip file in order to open it.
The Reynolds OI.v2 Sea Surface Temperature data used in the ENSO discussion is available through the NOAA NOMADS website here.
The Aerosol Optical Thickness data used in the volcano adjustments of the Sea Surface Temperature data in Figures 13 and 14 is available from GISS the Stratospheric Aerosol Optical Thickness webpage here.
Typo?:
“Total Solar Irradiance (PMOD), and the Volcanic Aerosol Optical Depth data (AOD) as the independent variables. I lagged the MEI data by four months, the PMOD data by one month”
Did you mean TSI?
OK, you had to repeatedly disclaim any direct implication that FR2011 deliberately fudged the “model” they used to force a positive trend line. But we get it anyway. Thanks!
Bob, I think you might have big fun with this:
http://www.sciencenews.org/view/feature/id/337207/title/Software_Scientist
http://creativemachines.cornell.edu/eureqa_download
Sort of like Analyse-It on steroids.
What is it with “linear trends” wrt climate !!! DOH !!!
The only linear trends that exist in climate are those created by the person doing the analysis.
In other words.. they don’t exist !!!
By including a linear trend for warming in their analysis as an independent variable, Foster & Rahmstorf 2011 have demonstrated that global warming is well correlated with global warming.
Have F&R cited a single reference where this technique is recognized as mathematically valid statistical technique for linear regression?
Look at the equations:
GISS = -91.43 + 1.024Trend + 0.0761MEI(4m lag) + 0.06694TSI.PMOD(1m lag)- 2.334AOD (7m lag)
What they are saying is:
Trend (GISS) = 1.024 Trend(GISS) + “other factors”
therefore:
– 0.024 Trend(GISS) = “other factors”
This has the effect of burying the “other factors” as -0.024 * GISS, and simply fitting GISS to itself.
It is mathematical nonsense. You might as well fit the other factor to zero.
I see. So you remove all natural cooling variables, but allow all warming variables.
“Attempting to use an ENSO index as Foster and Rahmstorf (2011) have done is like trying to provide the play-by-play for a baseball game solely from an overhead view of home plate.”
Thanks, Bob Tisdale. I’ll have to re-read this post a time or two, and hope that all your readers will have perspectives expanded, and trued.
Take a series of data points on a line with slope 1, intercept 0,0). Call these GISS. Call the slope Trend.
GISS = (0,0)(1,1)(2,2)(3,3)(4,4)
By linear regression fit this to
GISS = 1.0 * Trend + 0.0 * anything
subtract and you get
0 = 0 * anything
Therefore you have proven that you have removed “anything” from GISS and whatever is left is a valid Trend.
note: To obscure what you are doing, use 1.024 instead of 1.0 as the weight for Trend.
“By including a linear trend for warming in their analysis as an independent variable, Foster & Rahmstorf 2011 have demonstrated that global warming is well correlated with global warming.”
I like it !! the old AGW self perpetuating fraud strikes again.
I wouldn’t waste any of my valuable time on the Tamino’s nonsense.
I get the overwhelming sense that these guys are trying to pick fly sh*t out of the black pepper in order to “illustrate” an assumed AGW signal. Lardy, fellers, give it up.
Werner Brozek says: “Did you mean TSI?”
I could have used TSI. PMOD is the supplier of the TSI composite.
Brian H says: “OK, you had to repeatedly disclaim any direct implication that FR2011 deliberately fudged the “model” they used to force a positive trend line.”
You need to go back and read the post again because you missed it by a mile. They appear to have included the trend as an independent variable in the regression analysis to invert the solar adjustment, giving them the results they wanted.
Mr. Tisdale has once again shown that ENSO cannot be treated as an index but must be recognized as the natural process that it is. I am astonished that he has done such a good job in describing this natural process.
Warmists will fight to the death to overturn Mr. Tisdale’s results. They must because they cannot recognize ENSO as a natural process. To do so would ruin their “radiation only” theory of climate. Once natural processes are taken into account they interfere with the incremental progress that must be shown for the Warmists’ “radiation only” account to make sense.
Hats off to you again, Mr. Tisdale. Somebody give this man something like a Genius Grant.
“…SkepticalScience covered the paper in their post Foster and Rahmstorf Measure the Global Warming Signal…”
Isn’t this the same SkS that has problems with “…(1) deletion, extension and amending of user comments, and (2) undated post-publication revisions of article contents after significant user commenting…”?
And we’re expecting a reliable response on the article?
You know we’ll NEVER see any replies critical of the article there (or on “open” mind).
I used the included Analysis ToolPak Add-In in Excel 2010 and got the same MR results as Equation 1 and Equation 2 above.
It doesn’t make any sense to me to include a linear trend of your “dependent” variable as an independent variable in a regression analysis.
Gees, If they had weighted that linear trend much more in their “adjustment” they would have ended up with a perfectly straight line for all the plots in Figure 5.
Were the reviewers on this paper Mann et al ??
Fred Berple – spot on! That was my feeling as well.
Bob – why don’t you publish this stuff as a paper – the only way you tear down the edifice of BS is to play in the same boxing ring. (Mixing my metaphors).
So all we need to do is understand what drives the size and frequency of oceanic oscillations and that would get us a long way. Easier said than done. I really enjoyed this and it helped me to understand a little more. Thanks.
I actually understood every word and concept and don’t need tylenol!! Where was this tisdale guy when I needed a good physics prof? well done. well presented. thx
Bob Tisdale says:
January 2, 2012 at 2:24 pm
I could have used TSI. PMOD is the supplier of the TSI composite.
As recently admitted [SORCE, Sedona, 2011] as I have pointed out years ago, PMOD has uncompensated degradation with the result that the is a false ~0.2 W/m2 difference between the minima in 1996 and 2008. http://www.leif.org/research/PMOD%20TSI-SOHO%20keyhole%20effect-degradation%20over%20time.pdf
@Bob: This seems like an important point you are making. However, I’m a little short on time to digest it. Is it possible to get an executive summary that includes the personalities involved? What I can get out of this article is that the AGW trend discovered was a linear trend artificially introduced.
Declaring anything about climate with 30 years worth of data is like looking out the window when it’s raining, and 5 minutes later when it’s still raining declaring a global flood is occurring.
I have a gut feel that there’s a chicken/egg thing going on here. While I am not as expert in the workings of ENSO as Tisdale, I might say something like:
The atmospheric circulation patterns that cause El Niño also cause temperature changes outside the eastern tropical Pacific …
The difference being subtle but important in my mind. I see the ENSO cycle as a response to atmospheric circulation changes and not a driver of them. The slacking of the trade winds causes El Niño and the return of strong trades result in La Niña. That being said, however, I do find myself in agreement with the notion that ENSO is misunderstood and misapplied in many cases.
Goldie says:
January 2, 2012 at 2:56 pm
“So all we need to do is understand what drives the size and frequency of oceanic oscillations and that would get us a long way. Easier said than done.”
That’s science. Well, once you engage in the long slog that is the empirical research. Nothing that Warmists would deign to do.
Considered a response comment or a full article in the journal that published the Foster paper? This kind of deconstruction has only that weight given it in the journals, at least with respect to the policy makers. Gonna do it?
Leif Svalgaard says: “As recently admitted [SORCE, Sedona, 2011] as I have pointed out years ago, PMOD has uncompensated degradation with the result that the is a false ~0.2 W/m2 difference between the minima in 1996 and 2008. http://www.leif.org/research/PMOD%20TSI-SOHO%20keyhole%20effect-degradation%20over%20time.pdf”
Happy New Year, Leif. I was hoping you’d stop by to add a note about Foster and Rahmstorf’s use of PMOD data. I used it to duplicate their results.
Regards
JDN says: “What I can get out of this article is that the AGW trend discovered was a linear trend artificially introduced”
Nope. Foster and Rahmstorf included the trend as an independent variable in the regression analysis to invert the solar adjustment, giving them the results they wanted.
Regards
The correlation to the solar cycle TSI is negative because the temperature record is much more variable than the solar cycle. TSI is not significant variable.
TSI, by itself, yields a non-viable regression formula. It just trying to get as close to the average temperature anomaly over the period ie. a straight line. If you lag it by 5.5 years, getting the sign reversed back to positive, it still results in a straight line. ie no correlation.
Taking TSI out of the equation and putting the AMO back in results in a linear warming trend of 0.066C per decade from 1979 to 2010 for GISS and 0.056C per decade for RSS.
crosspatch: It’s the relocation of the warm water during the El Nino and the increased surface area of that warm water, along with the relocation of and increases in convection and precipitation, etc., that cause the changes in atmospheric circulation. Keep in mind that the equatorial Pacific stretches almost halfway around the globe. When all of that warm water and the accompanying convection, etc., shifts that far (or even half way for a central Pacific El Nino), atmospheric circulation also shifts and everything changes globally.
D. W. Schnare says: “Considered a response comment or a full article in the journal that published the Foster paper? This kind of deconstruction has only that weight given it in the journals, at least with respect to the policy makers. Gonna do it?”
I could be a co-author.
crosspatch says:
January 2, 2012 at 3:27 pm
Yes, crosspatch. Some serious empirical research will be necessary before the causes and effects receive a final sorting.
As regards publishing this article, Mr. Schnare, I cannot imagine that a Warmist journal would publish an article that does not treat ENSO as an index. Maybe Mr. Tisdale could attempt to publish just his remarks on the use of the trend.
Thanks for the post Mr Tisdale. Any chance you could write a book taking we lay readers v-e-r-y s-l-o-w-l-y through all of your research? Sometimes it feels as if I’ve started a book on string theory at chapter 11!
Nonetheless, I get the gist. If the authors remove most aspects of climate from the climate, they get what they want. Along the lines of “it was necessary to destroy the [village – self-snip] climate in order to save it”.
Or these ‘real’ rates of inflation beloved by economists and central bankers, which remove unpredictable elements like food and fuel. Statistically perfect – unless you’re human and need to eat or keep warm.
My immediate thought on reading Tamino’s paper was that they had selected three surface temperature records that have much greater length than the satellite records, and yet looked at a small subset of those records, obstensibly to be compatible with a much shorter sattelite record. What does the surface record show over it’s full length?
This is typical of Tamino. Grant is an excellent mathematician who has no idea about the the data he is looking at. Its all just numbers to him and he doesn’t understand what they mean. That and the fact he’s (self confessed) biased to finding anthropogenic warming in those numbers. Rhamstorf is equally biased.
Their paper is worthless because of this.
F&R 2011 – What it actually shows.
This should be fairly simple to follow if you understand algebra and substitution.
EQUATION 2:
GISS = -91.43 + 1.024Trend + 0.0761MEI(4m lag) + 0.06694TSI.PMOD(1m lag)- 2.334AOD (7m lag)
(1) GISS = 1.024Trend + bx + c
(2) GISS = 1.0Trend + 0.024Trend + bx + c
(3) GISS = (GISS + d) + 0.024Trend + bx + c
(because y = mx + d, where m=slope=trend, d=y intercept)
(4) 0 = 0.024Trend + bx + e
(5) Trend = -(bx + e)/0.024
F&R have not solved for GISS. By including Trend(GISS) as an independent variable they have eliminated GISS. What they have shown is that the Trend in GISS can be fully explained as a linear result of MEI, TSI, and AOD, without any reference to CO2.
In other words, F&R have proven that Climate Change is fully explained by the Multivariate ENSO Index (MEI), the Total Solar Irradiance (PMOD), and the Volcanic Aerosol Optical Depth data (AOD).
In other words, F&R have proven that CO2 has no role in climate change.
crosspatch says: “…I see the ENSO cycle as a response to atmospheric circulation changes and not a driver of them. The slacking of the trade winds causes El Niño and the return of strong trades result in La Niña….”
Yabbut what causes the trade winds to slacken and unslacken? Hmmm? 🙂
ferd berple says: “By including a linear trend for warming in their analysis as an independent variable, Foster & Rahmstorf 2011 have demonstrated that global warming is well correlated with global warming….”
It struck me when I read the post as one of the most bizarre bits of Warmist
mathdigital manipulation I’ve ever seen. It makes no sense to me at all. But I realized after Climategate I that Warmist spewings must get increasingly detached from reality. The actual science is so far at odds with AGW science fiction that Warmists must go further and further into the Twilight Zone every month to keep the hoax going. It’s a Trenberthsty.Bob, This is an excellent, probing piece of scholarship by you.
What were the reviewers thinking? The first error, if it is an error, of reversing the sign of TSI.PMOD(1m lag), should have been picked up. It is so fundamental to the analysis that even if it is not an error, it deserved more explanation in the original paper.
Unless peer review is to be allowed to take another degrading hit, the reviewers should be named, shamed and sacked if your analysis proves to be more correct than the version they approved.
I’d be writing to the publisher demanding that until this is sorted, the paper should lose its “peer reviewed” status.
crosspatch says:
January 2, 2012 at 3:27 pm
…………..
I have a gut feel that there’s a chicken/egg thing going on here.
=======================================================
lol, I’ve been lurking, waiting on the inevitable forcing/feedback argument.
ENSO is certainly a natural process, and must be seen as a separate temperature driver but there is more to the process than a simple El Nino/La Nina shuffle in my opinion. We are now witnessing back to back La Nina’s that occur during negative PDO phases, the current La Nina is not driven by a preceding El Nino. The warm pool above New Guinea is required to form to drive La NIna which this time around looks to have been fed by the warm pool of water that forms off Japan during negative PDO events. PDO/ENSO comparison graphs suggest the La Nina has two sources.
A negative PDO gives the option for more La Nina events, what creates the warm pool in the north western pacific being the key to the 60 year cycle. I suspect local cloud conditions brought about by magnetic changes in the atmosphere could be responsible, but further research is required.
“jorgekafkazar says:
It struck me when I read the post as one of the most bizarre bits of Warmist math digital manipulation I’ve ever seen. It makes no sense to me at all.”
Math has a couple of well known “problems”. They can either be used by the knowledgeable to fool the gullible, or by the clumsy to hang themselves. Anytime a term reduces to zero as is the case here, you have to suspect a problem. Which suggests there was no mathematician reviewing this paper.
(5*0) = (3*0)
1 = (3*0) / (5*0)
1 = 0/0
therefore
(5*0) = (3*0)
(5*0/0) = (3*0/0)
(5*1) = (3*1)
5 = 3
Thank you Mr. Tisdale. Your posts are always a great read, and even someone like me who sometimes fumbles with the technical aspect of the debate can understand it.
To me, it seems that the Foster&Ramstorf made an error by adding the ” linear trend” variable.
As many have pointed out, the weather and climate system is one of the most complex systems out there. It is completely NON-LINEAR.
This is what makes it so hard to predict, and something that the GCM’s will never be fully able to account for.
More smoke and mirrors from Tamino and Company…
this is one thing i hated about certain professors who made it their signature to FOOL people rather than allow them tho learn….
Climategate Continues with Tamino at the Helm….
Bob Tisdale says:
January 2, 2012 at 2:25 pm
You need to go back and read the post again because you missed it by a mile. They appear to have included the trend as an independent variable in the regression analysis to invert the solar adjustment, giving them the results they wanted.
————————————————————————————–
Where I come from this is called INTENTIONAL fraud.. Kinda like gluing together differing proxies to obtain a HOCKEY STICK..
You asked for a check on your regressions for Equations 1 and 2. Using R and the data from Tamino’s files, I get, for Equation 1:
Call:
lm(formula = giss ~ mei + volc + solar, data = bfeed)
Coefficients:
(Intercept) mei volc solar
123.64297 0.06769 -3.83664 -0.09025
and for Equation 2:
Call:
lm(formula = giss ~ mei + volc + solar + tau, data = bfeed)
Coefficients:
(Intercept) mei volc solar tau
-91.17277 0.07610 -2.33400 0.06694 0.01711
The tau is quite different from your trend, but is perhaps specific to how tau was calculated in the R code. (January 2009, for example, is 19.042, which is the year and fractional month minus 1990.) Everything else is pretty much the same.
I also felt there was an issue with including the trend in the regression. To me the trend should be what pops out at the end of the analysis when all other known factors are removed if it is a true artifact of the environment. This looks like introducing a trend and then weighting it until a straight line is attained that we can then call a trend. It is a little like making station adjustments to account for UHI by increasing the temperature at the station. If there is UHI then the adjustment to correct for it would be to decrease the station reading, but that is not what is done. Einstein’s greatest error was introducing the cosmologic constant to general relativity in order to freeze it into place. It was a fudge factor introduced to fit the universe to his ideal rather than to match the theory to the observed phenomena. I think this linear regression analysis is similar in scale.
Bob,
I used Statgraphics and reproduced your first regression exactly. Will check the others when I get a chance.
I am back to computer sized images and electricity instead of phone screens and hurricane lamps! And this post is just eye-delicious. Now on to the obvious. That any climate researcher would ignore how warm water created in one area then moves to another area and makes things warm there after a time apparently does not believe in the recent debris warning. Apparently the coastal villages on the Pacific side of the island of Japan are heading our way. How does all that debris do that? Don’t ask a climate scientists. Especially of the Tamino variety. They don’t believe in such things moving from one location in the ocean to another. To their way of thinking, the original debris sloshing about in Japan right after the tsunami was natural but the debris heading our way a year later is anthropogenic. Now that is unimpeachable logic, donchya think?
Following up on my previous post: the residuals from the Equation 1 regression have a trend, and the ACF looks bad, while the residuals from Equation 2 have no trend and the ACF looks pretty good. An ANOVA test also prefers Equation 2. I believe that these and a couple of other tests indicate that Equation 2 fits GISS better.
In both Equations, the parameters’ 95% CI don’t include zero, so they appear to be significant and their signs are not in question.
Perhaps using tau is a standard choice when using linear regression with time series, to avoid autocorrelation problems? Tau could have been statistically insignificant, in which case no trend would be indicated.
I think what Foster and Rahmstorf have done is very similar to what Mann did in a way to get his hockey stick. I would think that if you took the whole of Mann’s Hockey stick graph data you would once again get a very good fit of with a straight line representing the warming signal but with a much gentler slope than only taking the last 30 years or so of data. Probably the only requirement is the data shows a warming trend with time which clearly picking 1970’s to 2010 does. They allowed the ENSO (MEI), Solar (PMOD), and Volcanic Aerosol (AOD) variable to vary to what ever value they needed to be in order to get the predetermined straight line they wanted. Unfortunately, without including the linear trend for warming in the regression the sign for solar was wrong, so they had to add another term.
Another test of the method is to take other temperature./time regimes in the data sets and see what values result for the variables. Pick 1930-1970 for example and see what results you get. If their method is valid it should not only give results that look good for one particular point in time, but should be equally valid for any other set of points in time.
Sorry to add one more post, but Bob, I’d highly suggest that you get and learn R. It’s free and runs on any platform. Excel has historically had various issues with accuracy, its graphics are primitive, its use will give a bad first impression, and it doesn’t have all of the tools you need. (Some, you can purchase, but…)
For example, your Equation 1 and 2 discussion doesn’t get into how well either fits the data, and what problems may arise from using linear models on time series data without some kind of compensation. I’m sorry to say that I do believe Equation 2 actually is more accurate than Equation 1, and the sign flip may be a result of Equation 1 actually being pretty wrong.
Bob Tisdale
The models you use are different from the model in F&R, the latter include 4 more coefficients relating to a 2.order Fourier series. The lags are estimated from the model used in the multiple regression so you cannot use the lags estimated by F&R when fitting your models. And you can not use the same lags for your equations 1 and 2, in each case you need to calculate the appropiate lags using the particula model.
F&R models GISS as
GISS= c0 + c1*mei(tau-lag1)+c2*volc(tau-lag2)+c3*solar(tau-lag3)+c4*tau+c4*cos(2*pi*tau)+c5*sin(2*pi*tau)+c6*cos(4*pi*tau)+c7*sin(4*pi*tau)
c0-c8 are the coefficients are the lags to be estimated and lag1-lag3 are the lags to be estimated and tau is the time variable.
Your model 1 is:
GISS= c0 + c1*mei(tau-4)+c2*volc(tau-7)+c3*solar(tau-1)
and your model 2 (used in equation 2, and 4-11) is
GISS= c0 + c1*mei(tau-4)+c2*volc(tau-7)+c3*solar(tau-1)+c4*tau
You cannot just plug in the lags from F&R in your models.
I do also agree with Waynes argument that Equation 1 is a bad model since there is a trend in the residuals (and the trend is significant, even after correcting for autocorrelation).
Bob Tisdale writes,
“But one of the bases for the paper is to illustrate how similar the trends are after the adjustments for ENSO, Total Solar Irradiance, and Volcanic Aerosols have been made, so including the linear trends of those datasets in the regression analysis seems odd. As a result, I went in search of another reason why Foster and Rahmstorf (2011) would have needed to include the linear trend in their regression analyses.”
The reason F&R included a linear trend in their analysis is quite simple: in the 1979-2010 temperature series there is a linear trend. The whole point of the F&R analysis was to compare estimates of how steep this trend is, using different datasets after adjustment for several other factors that affect temperature.
Let’s keep it simple, setting aside for a moment the AR(1) errors and lag structure F&R specified. If you just regress, say, 1979-2010 GISS on MEI, AOD, and TSI as Tisdale tries to do, those three factors explain only 19% of the variance.
If you then add YEAR as a predictor (the linear trend), explained variance climbs to 62%. That’s a huge difference, more than tripling the proportion of variance explained, because the linear trend is a dominant feature of the data. MEI, ADO, and TSI do not predict this trend, so they give a much weaker fit by themselves.
Moreover, after adjustment for other predictors the trends turn out to be quite similar across four of the five datasets, with UAH being the exception. F&R Table 1 compares these trends.
Leif Svalgaard writes,
“PMOD has uncompensated degradation with the result that the is a false ~0.2 W/m2 difference between the minima in 1996 and 2008.”
Leif, what solar indicator would you recommend for this purpose?
On an OT,but related topic, a bit of news from Napa Wine Country
http://napavalleyregister.com/lifestyles/food-and-cooking/wine/a-look-back-the-year-in-wine/article_61480f34-3280-11e1-b5f0-0019bb2963f4.html
A look back the year in wine
• The cool summer led to “the latest start to harvest that any vintner or grower can remember,” according to the Napa Valley Vintners (NVV). The first grapes for the sparkling harvest were brought in on Aug. 29.
As the valley experienced another cooler than average growing season — a decade-long trend — according to the Napa Valley Vintners — the NVV released of a Napa Valley-specific climate study titled “Climate and Phenology in Napa Valley: A Compilation and Analysis of Historical Data by Dr. Daniel R. Cayan, Dr. Kimberly Nicholas, Mary Tyree, and Dr. Michael Dettinger.”
The Napa-specific study scrutinized weather and phenology (the growing cycle of grapevines) records from geographically diverse sites stations within Napa Valley over four years of the study, along with historic records.
“In brief, it finds that the region has experienced some warming, approximately 1 to 2 degrees Fahrenheit over the past several decades, but considerably less warming than would be inferred from the standard cooperative observer weather stations in Napa Valley,” it reported. “The warming has been primarily in winter, spring and summer, and it has concentrated during nighttime rather than daytime. Over the last several decades in growing season temperatures, there has been little warming in the daytime, and the available observations provide little evidence that the growing cycle of the grapevines has changed substantially.
“The results, overall, provide good short-term news that consumers are not “tasting” climate change in Napa Valley wines. It reinforces the firmly held belief among growers and winemakers that the taste profile of Napa Valley’s wines is driven by its place of origin, as well as by the solid direction of the in-field practices related to viticulture (clonal and rootstock selection, canopy management, irrigation, crop load and hang time, among others) along with stylistic preferences in winemaking.
Gneiss says:
January 2, 2012 at 7:55 pm
“The reason F&R included a linear trend in their analysis is quite simple: in the 1979-2010 temperature series there is a linear trend. The whole point of the F&R analysis was to compare estimates of how steep this trend is, using different datasets after adjustment for several other factors that affect temperature.”
There are no trends in data sets. Data sets contain only data. Trends are calculated by human beings. Reasonable human beings can disagree over the trends found in a data set.
You must have meant something else. Maybe try again.
“In other words, the East Pacific is simply a temporary staging area for the warm water of an El Niño event. Warm water sloshes into this dataset from the western tropical Pacific and releases heat, and then the warm water sloshes back out.
“But the warm waters released from below the surface of the West Pacific Warm Pool during the El Niño are not done impacting Sea Surface Temperatures throughout the global oceans, and they cannot be accounted for by an ENSO index.” [Bob Tisdale]
The above should be done in a nice calligraphy font and framed!
Well crafted.
Using TSI as a solar proxy is one mistake made in this process and one that the warmist brigade are happy to use. Solar influence on climate (isolating PDO) is more likely a result of the large fluctuations in solar EUV that effect cloud cover and atmospheric pressure patterns, this coupled with the PDO (ENSO) is enough to explain the temperature record.
Gneiss says:
January 2, 2012 at 8:09 pm
“PMOD has uncompensated degradation with the result that the is a false ~0.2 W/m2 difference between the minima in 1996 and 2008.”
Leif, what solar indicator would you recommend for this purpose?
The best we have is SORCE TIM TSI. Unfortunately it only goes back to about a bit before 2004. What I would do is to use PMOD until 1996 and then calculate the quantity D = – 002836 t + 0.00093266 t^2 – 0.00010134 t^3 W/m2, which is the degradation of PMOD where t is the time in years since 1996, then calculate PMOD(t) + D(t). At some point in time there will be new PMOD, but for now, I would use the above. The problem with PMOD was the [wrong] assumption that a cavity not exposed to sunlight would not degrade.
I am starting to be amazed by what I see as complete idiocy in an increasing number of scientific papers. These are people who are supposed to know what they are doing. For example, I cam across one that uses O-18 temperature proxy data to create an atmospheric CO2 reconstruction going back 20 million years. The implication being that the “scientist” believes that there is a direct correlation between atmospheric temperature and CO2 and this “belief” is so strong that he can make a CO2 reconstruction simply from O-18 records.
This is absolute fairy story stuff. I suppose anyone can get a research grant these days.
Bob,
Haven’t finished reading your latest post yet, but I your comment below:
“…inasmuch as they create more tropical Pacific ocean heat than was discharged during the El Niño that came before it. ”
Don’t you mean recharge more tropical Pacific ocean heat to depths, instead of “create more…”?
Thanx.
The East Pacific is also a staging area for the cold upwelling produced from La NIna. This upwelling is a product of warm water over New Guinea.
Geoff Sharp says:
January 2, 2012 at 9:01 pm
Using TSI as a solar proxy is one mistake made in this process and one that the warmist brigade are happy to use. Solar influence on climate (isolating PDO) is more likely a result of the large fluctuations in solar EUV
The fluctuations in EUV are so minute that they have almost no effect on surface tropospheric temperature.
Leif Svalgaard says:
January 2, 2012 at 9:21 pm
which is the degradation of PMOD where t is the time in years since 1996, then calculate PMOD(t) – D(t).
Should be a minus sign.
SRJ says:
F&R models GISS as
GISS= c0 + c1*mei(tau-lag1)+c2*volc(tau-lag2)+c3*solar(tau-lag3)+c4*tau+c4*cos(2*pi*tau)+c5*sin(2*pi*tau)+c6*cos(4*pi*tau)+c7*sin(4*pi*tau)
——————————————–
What is the point of such a model ?
mei, volc and solar modelled this way do not allow an accumulation of energy as temperatures are only allowed to follow their ups an downs with a lag. As we know that there have been decadal and multicentury natural trends in the past, this model is not only oversimplistic, it is false for the purpose.
That deficiancy easily explains, why the contribution is only 19%. Cosmic rays and perhaps other contributions are missing as well.
I would really urge scientific institutions to assure that their researchers quickly retract poor science.
Gneiss says:
January 2, 2012 at 7:55 pm
“The reason F&R included a linear trend in their analysis is quite simple”
No, if Bob has Equation 2 correct then including the trend in the dependent variable as an independent variable with a weighting close to 1.0 removes the dependent variable from both sides of the equation. It is a common mathematical sleight of hand that delivers a nonsense result. The equivalent of 0/0, which the layman assumes is equal to one, but every mathematician knows is undefined, because 0/0 = any value. It is one of the oldest mathematical parlor tricks in the book, used to deliver any result you want.
Leif Svalgaard says:
January 2, 2012 at 9:33 pm
The fluctuations in EUV are so minute that they have almost no effect on surface tropospheric temperature.
Strange how a scientist after being shown by many that solar effects can be chemical and radiative that he continues with the same line. This can only mean an agenda is involved.
Bob, I recommend you have Steve at Climate Audit take a look at equation 2. If you have it correct then I believe he will find it interesting.
Bob.
Learn R
It is standard in undergraduate econometric courses to hammer home the implications of excluding relevant explanatory variables – in this case CO2 surely.
For this reason alone it is a revelation that this paper was allowed publication.
These climate “scientists” have done untold damage to the discipline of statistics at the same time as they have trashed their own field.
My concern with Equation 2 is not in how to calculate the result. Rather with the form of the regression. My concern is that 1.024 is close enough to 1 to suggest we might get a misleading result.
The point of the regression is to solve
(1) y = mx + b, where m = trend
So, by including the trend on the right as in equation 2 we have something like:
(2) y = 1.0mx + f(x) + g (f(x) is the lagged terms, 1.0 is approx equal to 1.024)
we know from (1) m = (y – b)/x,
so for (2) by substitution we have
y = 1.0((y-b)/x)x + f(x) + g
simplify:
y = y-b + f(x) + g
0 = f(x) + g – b
f(x) = b – g
so in the end we have not solved to y at all, we have solved for f(x)
y, the temperature series, appears nowhere in our solution.
A few thoughts on all this.
First, while the regression model that excludes time in predicting GISS (e.g.) isn’t robust I would have thought the issue of multicollinearity between time and TSI would have been worth some further discussion beyond the assertion that it didn’t matter. The fact that signs started changing in the model etc should have been worth a comment.
Probably at the moment the authours were saying:
“In addition, we did regression experiments with each of the exogenous factors omitted one at a time, then tested whether or not its influence was still present in the residuals. All three exogenous factors showed about the same influence when fit to these residuals, as when fit in a multiple regression using all variables.”
would have been a good time to mention this. This could have included a discussion of the extent to which TSI adds to the model (interestingly not much).
SRJ @ January 2, 2012 at 7:55 pm
Do you have some inside knowledge here? I read the paper to say they followed an iterative approach to determining lags:
“The influence of exogenous factors can have a delayed effect on global temperature. Therefore for each of the three factors we tested all lag values from 0 to 24 months, then selected the lag values which gave the best fit to the data.”
See also earlier:
“Therefore the multiple regression includes a linear time trend, MEI, AOD, TSI and a second-order Fourier series with period 1 yr.”
This doesn’t seem to me to imply that the lag coefficient estimates were included in the estimation of the regression coefficients or for that matter that the loss of degrees of freedom included in the calculation of errors etc was taken into account.
Gneiss @ January 2, 2012 at 7:55 pm
“If you just regress, say, 1979-2010 GISS on MEI, AOD, and TSI as Tisdale tries to do, those three factors explain only 19% of the variance.
“If you then add YEAR as a predictor (the linear trend), explained variance climbs to 62%.”
However if you include the ARMA(1,1) structure claimed by the authors and look instead at what deltaGISS does regressed against the balance of the vbles but with added GISS t-1 , the explained variance without YEAR rises to 45%. Nothing wrong with including previous states as a predictor in these kinds of systems?
All just goes to prove that it is physical models and hypotheses of causality that should count in the end.
Wayne says: “…Bob, I’d highly suggest that you get and learn R. It’s free and runs on any platform. Excel has historically had various issues with accuracy, its graphics are primitive, its use will give a bad first impression, and it doesn’t have all of the tools you need.”
The good thing about Excel is it’s very easy to learn. Virtually anybody who isn’t a total doofus can use it to do trends. Still, for serious scientific work, I’d learn R as both Mosh and Wayne suggest. Good advice.
Foster and Rahmstorf (2011) included ENSO as one of the exogenous factors they attempted to remove from the instrument temperature record. But ENSO is not an exogenous factor. ENSO is a coupled ocean-atmosphere process that periodically discharges heat to the atmosphere during an El Niño.
More to the point, ENSO is the name that we give to surface temperature measurements from within a particular area of the globe. “Correcting” a global surface temperature measurement to eliminate the influence of a surface temperature measurement from part of the globe is assinine.
crosspatch says:
January 2, 2012 at 9:24 pm
—————————————
Crosspatch, for paleogeochemical studies of atmosphere, the Geocarb III model is the current standard. Among other items, it uses an entire range of known carbon uptake and output processes as part of the model. It seems to be rarely cited by AGW but that may be due to the fact that it outright falsifies the idea of “tipping points.” Also, when plotted against estimated planetary temperatures there is no correlation between modeled CO2 levels and modeled temperatures. Search for phanerozoic_co2.txt for a small data at 10 my intervals. For the correlation of temperature and CO2 over the Phanerozoic see Veizer and Shaviv (2003) in GSA Today. Geocarb III models indicate that the last time CO2 was at the levels we presently see was during the Permian. For the remainder of the geological record, levels apparently were higher to very much higher – possibly as much as 26 times greater 500 mya.
Rahmstorf is a key panicmaker, he is up to his neck in funding connections via PIK and Schellnhuber.
I’m completely ignorant of statistics, of course haven’t read the original paper, and if I’m honest, didn’t take the time to really follow the post that’s been presented here fully. I’m commenting mainly in response to comments, particularly ones saying that including a linear trend makes no sense and winds up producing a sort of hidden mathematical identity. I think it does make sense, if I’m understanding it right.
Seems to me that they make the hypothesis that the linear trend from, say, 1979 to 2000 is caused by CO2 — the trend is just the CO2 signal (or whatever signal caused an upward trend back when temps actually were trending upward). So when they sum up the 3 other influences along with the linear trend, what they’re really doing is summing up known influences with the hypothetical CO2 signal. Now you can correlate those data with the actual measured temperature data, and it’s not going to be a perfect match, but *if* it correlates as strongly from 2000 to present when temps are flat as it does back when temps were rising, that’s some indication that the linear trend is still there in recent times, it’s just obscured by the other influences.
It seems like a pretty simple and unconvincing model, but if I’m understanding the idea right, including the linear trend is really the key part of the experiment.
alcheson says: “Another test of the method is to take other temperature./time regimes in the data sets and see what values result for the variables.”
That will be included in part 2.
Should have stopped playing with the data but for those of you who like this stuff I fitted a simpler deltaGISS, GISS-1, GISS-2, GISS-12 model and then had a look to see what adding YEAR did (the Fourier terms.were getting marginal). (Of course this is all getting shot with problems in the interdependence of the dependent vbles, but it makes a point)
Adding in YEAR now only attracts a coefficient of 0.055 degrees/decade – a third of that found by the authors of this paper. And if only the Excel F significance could be believed in this situation (it can’t) this is a better than the model with all that fancy stuff like volcanos etc in it.
I’m sure I can do better. GDP of Europe next. And I don’t even need to learn R.
SRJ says: “The models you use are different from the model in F&R, the latter include 4 more coefficients relating to a 2.order Fourier series.”
Does it make any difference to the outcome of this post? My Figure 7 and F&R’s Figure 5 are the same. If you were to exclude the trend from F&Rs analysis is the solar component inverted?
@ HAS at http://wattsupwiththat.com/2012/01/02/tisdale-takes-on-taminos-foster-rahmstorf-2011/#comment-851376
“SRJ @ January 2, 2012 at 7:55 pm
Do you have some inside knowledge here? I read the paper to say they followed an iterative approach to determining lags:”
I have no inside knowledge, I have just read the paper and played around with the R code.
I don’t think that my previous description excludes an iterative approach to determine the lags – if so it was unintended.
This is the code that is used to calculate the lags:
zfeed = data.frame(giss,mei,volc,solar,tau,f1.cos,f1.sin,f2.cos,f2.sin)
######################
# test a range of lags
######################
bestaic = 999999
for (lag1 in 0:24){
for (lag2 in 0:24){
for (lag3 in 0:24){
lagvec = c(lag1,lag2,lag3)
zfit = lagfit(zfeed,lagvec,ndx.beg)
if (AIC(zfit$model) < bestaic){
bestaic = AIC(zfit$model)
bestlag1 = lag1
bestlag2 = lag2
bestlag3 = lag3
}
plot(0,0,main=paste(lag1,lag2,lag3))
}
}
}
lagfit is a function written by Tamino that does multiple regression with lagged variables.
zfeed is the input dataframe, that defines which variables to include in the regression.
If one changes zfeed to match Bobs equation it is obvious that the lags should be recalculated since the estimation of the lags is dependent on zfeed.
Gneiss @ January 2, 2012 at 7:55 pm and Wayne @ January 2, 2012 at 7:33 pm:
A trend is not an exogenous factor. The first sentence of the F&R Conclusions reads,
“This analysis confirms the strong influence of known factors on short-term variations in global temperature, including ENSO, volcanic aerosols and to a lesser degree solar variation.”
I don’t see trend mentioned there at all. Therefore, including the trend has to be based on something else. The bottom line is, including the trend inverts the solar component, and that gives the adjusted data the continuous rise. That was the primary goal of F&R (2011). The next paragraph of the F&R conclusion reads:
“Perhaps most important, it enables us to remove an estimate of their influence, thereby isolating the global warming signal. The resultant adjusted data show clearly, both visually and when subjected to statistical analysis, that the rate of global warming due to other factors (most likely these are exclusively anthropogenic) has been remarkably steady during the 32 years from 1979 through 2010. There is no indication of any slowdown or acceleration of global warming, beyond the variability induced by these known natural factors.”
I don’t see the trend mentioned there as an exogenous factor either.
Regards
Leif Svalgaard
Bob Tisdale
Bill Illis
Geoff Sharp
and other contributors
The TSI and the sunspot number (SSN) may not be as good indicators as the geomagnetic activity’s effect on the Earth’s magnetic field
http://www.vukcevic.talktalk.net/Tromso.htm
and hence its effect on the AMO, and via it on the global temperatures
http://www.vukcevic.talktalk.net/GT-AMO.htm
Wayne says: “I’m sorry to say that I do believe Equation 2 actually is more accurate than Equation 1, and the sign flip may be a result of Equation 1 actually being pretty wrong.”
Or it could just as easily indicate the solar lag time is wrong. For years, studies have looked into the response time of surface temperatures to solar variability. If memory serves me well, the response times for land surface temperatures are typically measured in months, but for sea surface temperature, the lags are measured in terms of years to decades. But F&R come up with 0- & 1-month lags.
Steve Allen says: “Don’t you mean recharge more tropical Pacific ocean heat to depths, instead of ‘create more…’?”
Thanks. I’ll correct that on the cross post at my blog.
Geoff Sharp says:
January 2, 2012 at 10:28 pm
“The fluctuations in EUV are so minute that they have almost no effect on surface tropospheric temperature.”
Strange how a scientist after being shown by many that solar effects can be chemical and radiative that he continues with the same line. This can only mean an agenda is involved.
EUV varies on the order of milliWatts/m2 and is absorbed above 150 km altitude [ http://www.leif.org/research/Atmospheric-Structure.png ] and has not been shown by ‘many’ to have any effect on surface temperatures. You might provide some references if you believe that the climate response is caused by variations of EUV.
OK, so I’m as usual a bit confused. I see absolutely no problem with including ENSO, despite the use of impressive words like “exogenous” by people who don’t like Tamino.
Now to the use of the linear trend. Once you’ve got rid of volcanoes, ENSO and TSI, what is left is what we want to see – the trend due to increasing CO2. But this is due to another variable, CO2. So I’m not sure why Tamino didn’t use Log[CO2] with an associated climate sensitivity factor as a 4th variable along with ENSO, TSI & Volcanoes. There is nothing in the first 3 variables to create the observed linear trend, but there is in the Log[CO2] term.
Anyway, I think Bob is a bit out of his depth here. But what would I know.
MODERATORS — NAME COLLISION:
@ Wayne
‘Wayne’, sorry to have picked such a simple name two years ago as ‘wayne’ but could I ask you to pick something a bit more unique than the one capitalization so the other commenters here do not get the two of us mixed?
jorgekafkazar says:
January 3, 2012 at 12:00 am
“The good thing about Excel is it’s very easy to learn. Virtually anybody who isn’t a total doofus can use it to do trends. ”
What does this say about professors who work for CRU?
It is also strange that Foster put the AMO in his spreadsheet, but then did not use it. It is clearly a significant variable (while TSI is not). As long as it is detrended, it provides a very good explanation for the long-term cycles in the climate (albeit on top of some rising trend). One can go back to the 1850s and still see the AMO cycle in the climate (and who knows how far back this goes).
Obviously, Foster had tried it out (and he has written several articles about it before). He does not have a good explanation for not using it. He discarded it because it leaves a very low warming residual afterward.
Bob Tisdale at http://wattsupwiththat.com/2012/01/02/tisdale-takes-on-taminos-foster-rahmstorf-2011/#comment-851455
“Does it make any difference to the outcome of this post? My Figure 7 and F&R’s Figure 5 are the same. If you were to exclude the trend from F&Rs analysis is the solar component inverted?”
If I fit a model similar to your equation 1 I find the lags to be (MEI,AOD,Solar) = (8,10,4) and the coefficients of this model are:
Model1:
(Intercept) mei volc solar
158.48692653 0.06640975 -3.97792290 -0.11575444
AIC = -239.1566
Using a model similar to your model 2 I find the lags to be used as (4,8,1) and get the coefficients:
Model2:
(Intercept) mei volc solar tau
-85.83179633 0.07403464 -2.31884852 0.06303096 0.01702787
AIC = -508.7721
The number under tau is the trend estimate.
For model that also includes the 2. order Fourier coefficients the estimated lags are (28,11,4) and the coefficients and AIC:
Model3:
(Intercept) mei volc solar f1.cos
180.383579384 -0.071785962 -2.855324729 -0.131762917 0.017162310
f1.sin f2.cos f2.sin
0.018471758 -0.008205048 0.024409417
AIC= -256.1461
F&R’s original model, Model4:
(Intercept) mei volc solar tau
-83.498399079 0.079103123 -2.369367887 0.061321729 0.017092008
f1.cos f1.sin f2.cos f2.sin
0.018359550 0.038623288 -0.001051748 0.024893570
AIC= -532.3597
The AIC suggest that F&R’s model is the most likely. It also suggest that model that model2 is more likely than the 2 models without a trend (i.e. model 1 and 3).
And yes, the solar coefficient changes sign when the trend is included. But If a trend is not included in the model then there will be a trend in residuals (that is the case for model1 and model3) and that suggest that model should be changed.
I think that Wayne had a very good point earlier when he wrote:
“I’m sorry to say that I do believe Equation 2 actually is more accurate than Equation 1, and the sign flip may be a result of Equation 1 actually being pretty wrong.”
Citing from this comment: http://wattsupwiththat.com/2012/01/02/tisdale-takes-on-taminos-foster-rahmstorf-2011/#comment-851237
John Brookes says: “OK, so I’m as usual a bit confused. I see absolutely no problem with including ENSO…”
Then you didn’t read the discussion of the process of ENSO in the post above.
You continued, “Anyway, I think Bob is a bit out of his depth here.”
I’m not out of my depth on ENSO. And I presented a curiosity about the multiple regression analyses of F&R. I’m not out of my depth there either.
@Bob: You really need to show a plot of the residuals of Equation 1. Leaving out a discussion of the residuals from Equations 1 and 2 is misleading. Once you show that graph, it’s obvious, without any fancy statistics, that Equation 1 has left behind a linear trend in the data and not accounted for it.
Several people have commented that F&R put a linear trend into the equation and lo-and-behold out came a linear trend: they got out what they put in. This is incorrect. You could say the same thing about MEI or any other variable in the equation, right? The fact is, you can put any variables into a linear regression, and the way that it tells you which ones make sense is by the coefficients that come out the other side. If the linear trend didn’t belong, its coefficient would be essentially zero.
Don’t get me wrong: I think your physical argument about ENSO, etc, is valid. The problem is that you’re wasting a lot of correctness points on a statistical argument (Equations 1 and 2) that is incomplete and incorrect. Please, please correct/retract the whole “sign flip” thing. It sounds a lot like the hockey stick sediment issue, but it’s not.
Bob Tisdale writes,
“A trend is not an exogenous factor.”
A trend is caused by exogenous factors. What F&R succeed in showing is that the warming trend in temperatures cannot be explained by El Nino, volcanoes or solar irradiance. And moreover when you account for those three factors, the different temperature datasets roughly agree on how steep that trend is, although UAH stands a bit apart from the others.
SRJ, regarding you January 3, 2012 at 5:56 am comment. Thanks for the lags. I’ll look at the results later. And thanks for confirming that the solar signs are inverted if the trends are not included in the anaylsis. But we seem to have a difference of opinion on the use of the trend. You’re saying that including the trend in the analysis provides the best fit. I’m not disagreeing with you. But the trend is not an independent variable. It’s a function of the dependent variable. There’s no reason to include it in the regression analysis. Also the fact that the solar coefficient is opposite of what is expected if the trend is not included could simply mean the actual solar lag falls outside of the 24-month window of the F&R analysis. And last, ENSO should not be included as an independent variable.
I’m no expert on statistical analysis, but I did have to learn the subject in some depth back when I was studying to become a financial analyst.
I don’t remember the majority of the details, but I did commit to memory a few critical issues. One of which could be a factor here… when you’re performing multiple linear regression analysis (as was done in F&R2011), any significant correlations between your independent variables (a.k.a. multicollinearity) can lead to erroneous estimates for individual regression coefficients (possibly resulting in sign inversions). A quick check of the four independent variable data sets in a correlation matrix could help clear this up.
As for the inclusion of a linear trend as an independant variable, I’m rather surprised F&R wouldn’t just use a log of the actual atmospheric concentration of CO2 (and equivalents), since the data should be of good quality, and that’s the key relationship they’re actually trying to isolate and quantify.
HAS writes,
“However if you include the ARMA(1,1) structure claimed by the authors and look instead at what deltaGISS does regressed against the balance of the vbles but with added GISS t-1 , the explained variance without YEAR rises to 45%.”
Converting to first differences (deltaGISS) is a standard time series technique to remove a linear trend, and if we do remove the trend, then of course the three exogenous variables explain a larger fraction of what variance remains. That is one of F&R’s points too, that El Nino, volcanoes, and solar irradiance explain some of the short-term variation in temperature, after you adjust for the longer-term upward trend. But the upward trend is large, real, and of considerable interest here. F&R show that El Nino, volcanoes, and solar irradiance cannot explain the upward trend, it must be due to something else.
Working with first differences does not really hide the incline anyway, it just sweeps it into the y intercept.
Russ R at
January 3, 2012 at 7:43 am
“A quick check of the four independent variable data sets in a correlation matrix could help clear this up.”
F&R mentions this collinearity issue in their section 2. (Data).
Here is the correlation matrix for the 4 variables used in Bobs equation 2:
mei volc solar tau
mei 1.00000000 0.4396667 0.04446291 -0.2172003
volc 0.43966667 1.0000000 0.16409459 -0.3519571
solar 0.04446291 0.1640946 1.00000000 -0.4745288
tau -0.21720025 -0.3519571 -0.47452881 1.0000000
According to F&R these correlations are sufficiently small to say that the variables are certainly not collinear.
Bob Tisdale said
January 3, 2012 at 7:28 am
“You’re saying that including the trend in the analysis provides the best fit. I’m not disagreeing with you. But the trend is not an independent variable. It’s a function of the dependent variable. There’s no reason to include it in the regression analysis”
Yes there is. Models without a trend gives a trend in the residuals. Such models are bad. One way to improve the model to remove the residual trend is to use a model with a trend. If you are going to use a model without a trend you need a model that does not give a trend in the residuals. So how will you alter your model to accomplish that?
Bob Tisdale said:
“Also the fact that the solar coefficient is opposite of what is expected if the trend is not included could simply mean the actual solar lag falls outside of the 24-month window of the F&R analysis.”
I tested this by allowing solar lags up to 35. The best fit lags for F&R’s model were unchanged as (4,7,1).
Once again, Wayne2 makes a good point on this at
January 3, 2012 at 7:06 am
Leif Svalgaard says:
January 3, 2012 at 3:40 am
EUV varies on the order of milliWatts/m2 and is absorbed above 150 km altitude
That in no way proves the change is too small to significantly affect climate. The change in EUV from the quick check of articles I looked at was 100% of more. To prove this is not significant you would need to rule out any amplification via ozone, atmospheric height, etc.
The best that can be said about EUV variability is that at present it is unproven if this affects climate or not. To claim there is no effect is unscientific.
Wayne2, I just tried to post this but I got an error message in return. So just in case it shows up, that’s the reason for the duplicated reply.
You said, “You really need to show a plot of the residuals of Equation 1. Leaving out a discussion of the residuals from Equations 1 and 2 is misleading.”
It’s not misleading. You’re reading too much into this discussion. The intents of this part of the post were:
1) to show that the sign of the solar component is inverted if the trend is not included in the regression analysis, and
2) to show that it impacts the results.
That’s all.
And everyone who has checked the results with different regression analysis tools has found the sign of the solar component is dependent on whether the trend is included in the analysis. The question is, since the trend is not an independent variable, should the impact of including it in the linear regression analysis have been discussed in F&R 2011?
It’s like detrending the Surface Temperature or TLT data before performing the regression analysis. You’d get different results from “un-detrended” data and you would need to explain why you detrended the data.
Gneiss says: “A trend is caused by exogenous factors.”
Agreed. For the global oceans, which represent 70% of the surface area of the globe, the trend is caused by ENSO.
You continued, “What F&R succeed in showing is that the warming trend in temperatures cannot be explained by El Nino, volcanoes or solar irradiance. ”
F&R only succeeded in showing that they misunderstand ENSO or have elected to misrepresent it. ENSO is a couple ocean-atmosphere process that cannot be represented by an index. I’ve clearly illustrated and discussed that above. Did you miss that part of the post?
Russ R. “I’m rather surprised F&R wouldn’t just use a log of the actual atmospheric concentration of CO2 (and equivalents)”
Don’t for get that CO2 isn’t the only anthropogenic influence – there’s also aerosol influences. Over the last 30 years or so they add up to what we would expect (from the physics) to be roughly a linear trend influence. But it’s not just CO2.
F&R point toward Lean and Rind (2008, 2009) as interesting papers taking sum forcings and working forward toward component analysis – this in addition supports taking these analyses further back than the 1970’s, over periods where a linear trend is not supportable due to differing changes in total forcing and the observed temperature record.
—
I have to agree with Wayne2, incidentally – the residuals left after the multiple regressions are extremely important in judging how much the various factors contribute. And running the regression without a linear component shows a distinct linear trend remaining in the residuals – meaning (a) not everything is being accounted for, and (b) that ENSO, TSI, volcanic forcings (and annual signals, also in the regression) do not correlate to an ongoing increase in temperature in this analysis.
The residuals indicate that the linear trend is present. Running the regression with a linear component gives residuals without a linear, or polynomial, or other trend – indicating that major components have been fairly well accounted for. There are likely other components – but the small residuals from this regression do not point at major contributions over that 30 year period.
And – very importantly – removing a linear trend (or any other component, for that matter) independently will distort the contributions of all components. The various contributions all act together – removing them separately creates order-dependent distortions of their attribution strengths, and very likely changes in best-fit lag times as well.
Regarding ENSO – Tisdale seems to be claiming (here and in previous posts) that the ENSO variation is releasing more heat to the surface than absorbing it over the last 30 years, causing the observed rise in temperatures by itself. That’s quite a claim, considering that ocean heat content (OHC) data (although rather spotty so far) indicates increasing deep ocean energy, not decreasing, that the ENSO has not as far as I know been identified to have 30+ year effects in the past (published papers supporting this claim would be welcome, if they exist), and that our understanding of the physics of the ENSO don’t support such a long-term effect.
Given the current understanding of the ENSO, treating the ENSO as an exogenous variation is quite reasonable.
ferd berple says:
January 3, 2012 at 8:58 am
That in no way proves the change is too small to significantly affect climate. The change in EUV from the quick check of articles I looked at was 100% of more. To prove this is not significant you would need to rule out any amplification via ozone, atmospheric height, etc.
The percentage change is irrelevant as it is the total amount of energy that counts. You are like one who judges the variation of a billionaire’s wealth by monitoring the loose change in his pockets. The density of the atmosphere where EUV is absorbed is less than a billionth of that at sea level. Ozone is not produced by EUV in the thermosphere.
The best that can be said about EUV variability is that at present it is unproven if this affects climate or not. To claim there is no effect is unscientific.
The null-hypothesis is that there is no effect from such an insignificant source. To claim that there is an effect even though unproven is unscientific.
SRJ, regarding your January 3, 2012 at 8:48 am comment, you appear to be rewording your best fit argument. As I noted for Wayne2 above, you’re reading too much into this discussion. The intents of this part of the post were:
1) to show that the sign of the solar component is inverted if the trend is not included in the regression analysis, and
2) to show that it impacts the results.
That’s all.
And everyone who has checked the results with different regression analysis tools, including you, has found the sign of the solar component is dependent on whether the trend is included in the analysis. The question is, since the trend is not an independent variable, should the impact of including it in the linear regression analysis have been discussed in F&R 2011?
KR says: “…and (b) that ENSO, TSI, volcanic forcings (and annual signals, also in the regression) do not correlate to an ongoing increase in temperature in this analysis.”
And the reason is, as discussed in great detail in the post, ENSO is a process that is not represented by an index.
You wrote, “Regarding ENSO – Tisdale seems to be claiming (here and in previous posts) that the ENSO variation is releasing more heat to the surface than absorbing it over the last 30 years…”
Reread the post above. You’ve misunderstood the discussion.
Now KR has joined Wayne2, and Russ Rm and Gneissm and John Brookes, in all working to help Bob Tisdale to a better appreciation of why a linear trend is properly encompassed in the Foster & Rahmstorf regressions.
Now, Bob, it seems to me that the principles of rational skepticism require that you thank these WUWT posters, and amend your own WUWT guest-post to acknowledge that this aspect of the Foster & Rahmstorf 2011 analysis is correct.
Or is my appreciation of WUWT as a forum for rational skepticism incorrect?
I was a puzzled when Tamino insisted AMO was just temperature in response to my multiple posts. It is clearly an oscillation even if its underlying causes are a mystery. There are two AMO indices, one raw and another detrended from a linear increase. Is your 59% increase in NH temperature based on the detrended or raw version? My view is that using the detrended one is more approriate for the purposes of evaluating F and R on its own terms. What do you think?
My view btw is that there is reasonable evidence that there is an underlying increase in temperature over the last 50 odd years that can be attributed to increases in CO2 but it is almost certainly overestimated because natural oscillations mask it ( as indeed F and R shows.. whatever the flaws in their paper) and the high forcings often used in calculations make little sense when evaluated against long term ( >1k years) trends of temperature vs CO2 values.
The problem with all these adjustments papers is that you can fit any graph if you use enough variables and you hit the correlation is not causation problem however hard you try to avoid it
So I don’t object to assuming a linear temperature response to increased CO2 but I am very suspicious when two long term impacts are ignored that would reduce the CO2 signal, notably
1. AMO as you describe and
probably of lesser significance
2. Non volcanic AOD impacts ( particulates etc from coal) for which Fand R would need to use a different AOD index
I am however mildly encouraged by any paper that attempts to remove non CO2 factors (especially from AGM enthusiats when they show 1.6deg per century as the current trend)
Theo Goodwin writes,
“There are no trends in data sets. Data sets contain only data. Trends are calculated by human beings. Reasonable human beings can disagree over the trends found in a data set.”
Data contain patterns, including (in this case) an obvious trend. Tisdale’s equation 1, omitting the trend, is misspecified as many commenters are pointing out. The mistake would have been caught by a statistician or data analyst if his paper went out for journal review.
This model implies, that after a hundred years of El Ninos of the same size, the warming would still be the same as after the first event. and after a giant volcanic eruption blocking sunlight for 30 years, all cooling would have taken place in the first year.
The model treats oceans as memoryless with regards to natural influences and implies oceans with zero or very low heat capacity.
I would say very unlikely, and not supported by climate trends in the past.
This model does not answer the question, it is flawed right form the start, because it distributes the unaccounted natural contributions into the trend variable, while it was supposed to separate them.
Bob Tisdale
Is it not surprising that using a different model gives different estimates for the coefficients.
I am not sure that I understand what you mean when you write “the trend is not an independent variable”. What I prefer to call the trend is the coefficient on the time term.
F&R write explicitly that the time trend is among the independent variables:
“Hence we computed the correlation between the independent variables used during the time span under study. The strongest correlation was between TSI and the linear time trend, with correlation coefficient − 0.47.”
Your notion of an independent variable seems to differ from F&R.
Bob Tisdale – “You wrote, “Regarding ENSO – Tisdale seems to be claiming (here and in previous posts) that the ENSO variation is releasing more heat to the surface than absorbing it over the last 30 years…”
Reread the post above. You’ve misunderstood the discussion.”
Then please, explain what you meant in the post above by:
“Sometimes La Niña events “overcharge” the tropical Pacific, inasmuch as they create more tropical Pacific ocean heat than was discharged during the El Niño that came before it. That was the case during the 1973/74/75/76 and 1995/96 La Niña events.”
You appear to be claiming a long term imbalance in ENSO heat exchanges with the deep ocean. And I was asking if you have any support (published papers) to point to in that regard.
Furthermore, WRT your statement “ENSO is a process that is not represented by an index.”:
If that is indeed the case, then the index won’t encompass all of the ENSO process effects. They will have to show up somewhere (some-when) else…
The multivariate el Nino index (MEI), as well as the southern oscillation index (SOI) which was tested with equivalent results, are indexes. The regressions performed by F&R used those indexes, along with TSI, volcanic data, annual signal, and a linear trend. The residuals from the regression are both quite small and statistically trendless – indicating that additional effects not attributed to those five components are also quite small.
Hence you are in essence claiming that those ill-defined ENSO effects outside the index are mistakenly mixed in with the other components?
TSI seems unlikely as ENSO does not cross-correlate with TSI, volcanic influences are likewise extremely unlikely candidates, as is the annual signal – perhaps the linear trend? If so, you are asserting that (a) the ENSO has a linear trend component not predicted by our understanding of the ENSO, and (b) the linear trend fully expected by the spectroscopy of greenhouse gases somehow doesn’t add up to the predicted linear trend. That’s a lot of assertions…
Colin Aldridge says: January 3, 2012 at 10:26 am
I was a puzzled when Tamino insisted AMO was just temperature in response to my multiple posts. It is clearly an oscillation even if its underlying causes are a mystery.
There is no mystery about the AMO, and it is same for the PDO and the ENSO. They are all caused by the change in the heat transport by the major currents; in case of the AMO it is circulation within subpolar gyre, balance between the Labrador vs. the N. Atlantic drift currents, for the PDO Kuroshio vs Oyashio currents, and finally for the ENSO it is the South Equatorial and the Counter Equatorial currents:
http://www.vukcevic.talktalk.net/A&P.htm
Data is clear,
http://www.vukcevic.talktalk.net/ENSO.htm
but the AGW science is preoccupied with the CO2, and the sceptic camp is determined to squeeze the extra energy out of the TSI, where there is very little that matters. One day when both sides of the argument will hit the buffers.
It is just a matter of the available energy distribution, either is reradiated back into the space in the equatorial region (global cooling), or transported by the ocean currents further towards the poles (global warming).
SRJ @ January 3, 2012 at 1:59 am
Yes, and because of the use of this iterative procedure you can replicate the results by assuming these lags and putting them into a model with the trig function (the latter being what Tisdale neglected I think). I do wonder why just one lag per vble – but that’s an empirical question related to Manfred’s comment above.
Gneiss @ January 3, 2012 at 8:26 am
“Converting to first differences (deltaGISS) is a standard time series technique to remove a linear trend, and if we do remove the trend, then of course the three exogenous variables explain a larger fraction of what variance remains. …. the upward trend is large, real, and of considerable interest here. F&R show that El Nino, volcanoes, and solar irradiance cannot explain the upward trend, it must be due to something else.”
Just getting a good model fit is a necessary but not sufficient condition to explain what is happening in a physical system – in the end time causes nothing it is just a surrogate for something else (hence my comment about using the GDP of Europe). F&R themselves characterise this process as an ARMA(1,1) process and there are a wide variety of ways to create this behaviour out of a complex system without resorting to exogenous forcings (e.g. hydrological processes). So simply for the sake of argument I thought it fun to extract this component before seeking the “unexplained”. Not surprisingly less falls into this category – the moment you give a system memory linear trends become less inexplicable.
The real point is that this is just number crunching without strong underpinnings in the physical world. F&R should postulate a physical model for what is happening and test it. This should hypothesis how the exogenous vbles impact on global temp and test in that context (are these linear relationships etc etc).
Being able to find a residue monotonically increasing process in a data set having fitted other vbles without this feature tells us nothing of much value unless we can identify the processes driving it. F&R’s analysis doesn’t seem to contribute much on this score.
Now, Bob, it seems to me that the principles of rational skepticism require that you thank these WUWT posters, and amend your own WUWT guest-post to acknowledge that this aspect of the Foster & Rahmstorf 2011 analysis is correct.
##################################
that would be the right thing to do
Now all you insisting on a linear trend, did you check Ln(YEAR)
Just asking.
If the trend is a continuation of the natural uptrend in temperature following the LIA, then don’t the authors need to separate the natural part of the trend from the CO2 part of the trend, and finally seperate the hypothesized man-made component of CO2 from the natural CO2?
Russ R. writes,
“I don’t remember the majority of the details, but I did commit to memory a few critical issues. One of which could be a factor here… when you’re performing multiple linear regression analysis (as was done in F&R2011), any significant correlations between your independent variables (a.k.a. multicollinearity) can lead to erroneous estimates for individual regression coefficients (possibly resulting in sign inversions). A quick check of the four independent variable data sets in a correlation matrix could help clear this up.”
As collinearity increases the standard errors become larger — it’s a matter of degree, and not true that “any significant correlation” leads to erroneous coefficients. If there were no correlations between the independent variables, there would be no need for multiple regression and the coefficients would be identical to simple regression coefficients.
Incidentally, the correlation matrix by itself can diagnose collinearity but not, by definition, multicollinearity (for which tolerance or other multivariate statistics are needed).
“As for the inclusion of a linear trend as an independant variable, I’m rather surprised F&R wouldn’t just use a log of the actual atmospheric concentration of CO2 (and equivalents), since the data should be of good quality, and that’s the key relationship they’re actually trying to isolate and quantify.”
CO2 and year, on the other hand, really are collinear over these years. Which means it would make little difference to include CO2 but not year as a predictor, or vice versa. Put another way, year serves well enough as a proxy for CO2, over 1979-2010.
Steven Mosher says: January 3, 2012 at 11:29 am
……
Hi Steve
I hope you are well and had good holidays.
Dr. Hathaway of NASA January’s SC24 sunspot max prediction to around 100.
See how he does here:
http://www.vukcevic.talktalk.net/NFC7a.htm
and compare with the planetary pseudo-science astrology practiced by Ptolemy of Alexandria.
I haven’t had time to read the whole posting and all comments, but I do have an idea about the issue, and unusually I may be inclined to side with F&R rather than Tisdale on this one.
Since global temperatures have increased over the chosen period, and TSI has on the whole decreased, it is natural that a regression should give a negative weight to TSI to get the best fit unless something else is included. By including a linear trend the TSI can get a proper positive weight. As explanation, F&R would have done much better, instead of a straight line against time, to use a linear function of something which has been increasing during that time. Rates of piracy and concentrations of CO2 are the obvious candidates 🙂 .
By using CO2 F&R could even have deduced a sensitivity for doubling CO2, except it would have been an overestimate because of the short time period and the phase of AMO, as others have noted.
Does not the simple understanding above strike a chord with you, gentle reader?
Rich.
HAS writes,
“Now all you insisting on a linear trend, did you check Ln(YEAR)
Just asking.”
Yes. That takes about 2 seconds. It makes no difference.
Gneiss @ January 3, 2012 at 12:14 pm
So which one is it? (In the real world it makes a difference)
For those interested in these regression discussions, I would recommend Lean and Rind 2008, 2009:
How natural and anthropogenic influences alter global and regional surface temperatures: 1889 to 2006 (http://climate.envsci.rutgers.edu/physclim/LeanRind2008GL034864.pdf),
How will Earth’s surface temperature change in future decades? (http://www.unity.edu/facultypages/womersley/2009_Lean_Rind-5.pdf)
They perform a multiple regression against ENSO, TSI, volcanic influences, and the net changes in anthropogenic forcings for the last 125 years, rather than a simple 30 year linear trend as in F&R 2011 (F&R 2011 reference both of these papers). While they don’t spend as much time as F&R do on residual analysis, they find quite similar results both for attribution strength and for lag times for these components.
Again, in order to properly perform a multiple regression of this kind you need to calculate the attributions simultaneously, not individually, or you will overattribute to the initial components and underattribute to the later ones in an order-dependent fashion.
Gneiss says:
January 3, 2012 at 12:06 pm
“If so, you are asserting that (a) the ENSO has a linear trend component not predicted by our understanding of the ENSO”.
—————————————-
I don’t know who is meant by “our”, but not asserting a linear trend component would make it very hard or impossible to exlain any multiyear trends before 1750.
Memoryless oceans are a pretty bold assumption, and Tisdale had produced multiple charts in the past showing the opposite and that there is a trend component.
The issue is here with Foster and Rahmstorf. THEY made the assumption that there is no linear trend component, and they have presented nothing to substantiate it. It would have been up to THEM to do so in a proper paper.
Because this model is already the main reason for the result, Foster and Rahmstorf failed to produce anything useful right from the start.
Manfred, you have attributed to me a statement that was actually made by KR, although I happen to agree with it. I don’t follow the rest of your post.
“The issue is here with Foster and Rahmstorf. THEY made the assumption that there is no linear trend component, and they have presented nothing to substantiate it.”
Very confusing. F&R is all about a linear trend component, which they do not assume but present much to substantiate.
M.A.Vukcevic says:
January 3, 2012 at 12:13 pm
Dr. Hathaway of NASA January’s SC24 sunspot max prediction to around 100.
Vuk, for the gazillionth time: Hathaway’s ‘prediction’ is not a real prediction, but simply an assimilation of the current data so far into the old standard NOAA panel prediction, with a variable weight factor that gives more and more weight to current observations the deeper into the cycle we are. Now, about 50-50.
Leif Svalgaard says:
January 3, 2012 at 9:23 am
The percentage change is irrelevant as it is the total amount of energy that counts.
And yet you still continue. I and others have shown you the chemical reaction along with the change in atmospheric heights are the components of EUV that science papers and bodies are now strongly suggesting as a viable climate driver. Get off the TSI band wagon.
You can check my article from June 2010 which has many links to articles and papers related to EUV and climate.
http://tinyurl.com/2dg9u22/?q=node/128
KR says: “For those interested in these regression discussions, I would recommend Lean and Rind 2008, 2009:”
Playing your troll role again, I see, KR. Your comment clearly shows you didn’t bother to read the post. I linked both Lean & Rind papers in my post, KR.
Lean & Rind, both papers, make the same error as Foster and Rahmstorf (2011). They assume the effects of ENSO can be removed from the instrument temperature record through regression analysis. The majority of the post above illustrates and explains why this is an error, KR. The error with their assumption is very visible in the Sea Surface Temperature graphs I’ve presented, and it should be relatively easy to understand with my descriptions of ENSO, unless, of course, someone like yourself elects not to read the post or elects not to understand what is written in the post. Either way, KR, you’ve made your choice.
Geoff Sharp says:
January 3, 2012 at 2:06 pm
And yet you still continue. I and others have shown you the chemical reaction along with the change in atmospheric heights are the components of EUV that science papers and bodies are now strongly suggesting as a viable climate driver. Get off the TSI band wagon.
You can check my article from June 2010 which has many links to articles
These papers are concerned with the stratosphere and UV, not with the thermosphere and EUV. EUV does not penetrate deeper than ~150 km [and BTW as you show varies in step with TSI within the uncertainty]. What chemical reactions do you think take place above 150 km?
Leif, easy question: where does the decimal belong, in your coefficient on t?
“What I would do is to use PMOD until 1996 and then calculate the quantity D = – 002836 t + 0.00093266 t^2 – 0.00010134 t^3 W/m2, which is the degradation of PMOD where t is the time in years since 1996, then calculate PMOD(t) + D(t).”
Steven Mosher says: “Now, Bob, it seems to me that the principles of rational skepticism require that you thank these WUWT posters, and amend your own WUWT guest-post to acknowledge that this aspect of the Foster & Rahmstorf 2011 analysis is correct.”
You continued, “that would be the right thing to do”
It is very obvious, Steven, that you, like the others who are arguing, didn’t read the post. The Overview of the post reads, starting with the second sentence: I find it very odd that a factor upon which the paper appears to rest was not presented in detail in it. Please understand right from the start, for this portion of the post, I am not implying that there is something wrong with this specific aspect of the paper; but I’m also not agreeing with it. I’m presenting it for discussion.
Further to that end, I have written something to the effect of the following more than once on this thread:
You’re reading too much into this discussion. The intents of this part of the post were:
1) to show that the sign of the solar component is inverted if the trend is not included in the regression analysis, and
2) to show that it impacts the results.
That’s all.
And everyone who has checked the results with different regression analysis tools has found the sign of the solar component is dependent on whether the trend is included in the analysis. The question is, since the trend is not an independent variable, should the impact of including it in the linear regression analysis have been discussed in F&R 2011?
Steven Mosher: I just discovered that I had attributed the following to you when it was actually written by A Physicist: “Now, Bob, it seems to me that the principles of rational skepticism require that you thank these WUWT posters, and amend your own WUWT guest-post to acknowledge that this aspect of the Foster & Rahmstorf 2011 analysis is correct.”
But since you followed it with, “that would be the right thing to do”, my reply above at January 3, 2012 at 2:28 pm is appropriate.
Gneiss says:
January 3, 2012 at 2:23 pm
Leif, easy question: where does the decimal belong, in your coefficient on t?
“What I would do is to use PMOD until 1996 and then calculate the quantity D = – 0.002836 t + 0.00093266 t^2 – 0.00010134 t^3 W/m2, which is the degradation of PMOD where t is the time in years since 1996, then calculate PMOD(t) – D(t).”
Gneiss says:
January 3, 2012 at 2:23 pm
Leif, easy question: where does the decimal belong, in your coefficient on t?
“What I would do is to use PMOD until 1996 and then calculate the quantity D = – 0.002836 t + 0.00093266 t^2 – 0.00010134 t^3 W/m2, which is the degradation of PMOD where t is the time in years since 1996, then calculate PMOD(t) – D(t).”
A physicist says: “Now KR has joined Wayne2, and Russ Rm and Gneissm and John Brookes, in all working to help Bob Tisdale to a better appreciation of why a linear trend is properly encompassed in the Foster & Rahmstorf regressions.”
And you continued, “Now, Bob, it seems to me that the principles of rational skepticism require that you thank these WUWT posters, and amend your own WUWT guest-post to acknowledge that this aspect of the Foster & Rahmstorf 2011 analysis is correct.”
Sorry, A physicist, I was replying to comments in reverse order and I mistakenly attributed your second paragraph to Steven Mosher. So I’ll amend my reply to Steven for you. I wouldn’t want you feel left out.
It is very obvious, A physicist, that you, like the others who are arguing, didn’t read the post. The Overview of the post reads, starting with the second sentence: I find it very odd that a factor upon which the paper appears to rest was not presented in detail in it. Please understand right from the start, for this portion of the post, I am not implying that there is something wrong with this specific aspect of the paper; but I’m also not agreeing with it. I’m presenting it for discussion.
Further to that end, I have written something to the effect of the following more than once on this thread: you’re reading too much into this discussion. The intents of this part of the post were:
1) to show that the sign of the solar component is inverted if the trend is not included in the regression analysis, and
2) to show that it impacts the results.
That’s all.
And everyone who has checked the results with different regression analysis tools has found the sign of the solar component is dependent on whether the trend is included in the analysis. The question is, since the trend is not an independent variable, should the impact of including it in the linear regression analysis have been discussed in F&R 2011?
That may be “all,” but it is not enough, for the common-sense reason that KR gave:
Thanks are owed to KR for stating this point concisely and clearly (and for citing references).
All in all, it appears that the methods and conclusions of Foster and Rahmstorf (2011) have stood up reasonably well to WUWT’s skeptical review.
Leif Svalgaard says:
January 3, 2012 at 2:19 pm
These papers are concerned with the stratosphere and UV, not with the thermosphere and EUV.
Some studies deal with UV others with EUV. What is important is that you recognize the chemical reaction not necessarily the range of the spectrum. Variations in UV and EUV are many times greater than TSI.
Bob Tisdale
As per my post at 11:08 – if (as you assert) the ENSO cannot be represented by an index (MEI or SOI), then the multiple regression will attribute those non-index qualities to other components that better match them. In this case, that would probably have to be the linear trend component, as TSI and volcanic forcings don’t correlate well. The “step” progression you espouse, not correlated to those variation indices, should show up in the residuals if it exists – there is no sign of it.
F&R 2011 is supported by (and extends upon) those papers you list, among a great many others. You’ve yet to point out a paper that supports your hypothesis. So either there is some massive ENSO conspiracy (yeah, right, pull the other one, it’s got bells on), or your hypothesis just isn’t supported by the data. Unless you consider the preponderance of work and evidence to be “trolling…”
Again – what peer-reviewed work (not blog postings, mind you, but something that passes muster upon examination by folks with expertise in statistics and physics, something others might cite) can you point to that support a long term linear trend component to the ENSO? What physics supports such an integrating component to ENSO? And yes, I’ve read numerous posts of yours on the subject – you leave the mechanism(s) behind your claims vague and unsupported, and have not once replied directly to my queries.
At this point I’ve asked that question of you repeatedly, and and your reply consists of “read my stuff – it’s in there”, without explanation.
—
Enough said.
* Your comments on independent components (as per cross-correlation and orthogonality, mind you), multiple regression, and how F&R posed their questions have been shown incorrect by a number of posters.
* The linear trend residual in a regression _without_ a linear component shows that such a component needs to be included in the multiple regression.
* The ‘step’ progression you espouse, as uncorrelated to the indices used by F&R, would show in the residuals if it existed – it doesn’t.
* You have not presented anything that supports your ENSO/trend hypothesis, let alone shown any errors in the multiple works that do not find long term ENSO related temperature trends, etc.
Adieu
Leif Svalgaard says:
January 3, 2012 at 1:26 pm
Vuk, for the gazillionth time: Hathaway’s ‘prediction’ is not a real prediction, but simply an assimilation of the current data
Hi doc
Hope you had good holiday.
I suppose he just forgot to change caption on his graph
http://solarscience.msfc.nasa.gov/images/ssn_predict_l.gif
Leif Svalgaard says:
January 3, 2012 at 1:26 pm
we are. Now, about 50-50.
If your instrumentation on the hillside is working OK then the magnetic field is on the move:
http://www.vukcevic.talktalk.net/LFC6.htm
I haven’t updated my formula graph for a while, but if the MF is about to reverse, than formula is absolutely spot on for the start of 2012
http://www.vukcevic.talktalk.net/LFC2.htm
If so, even I would be surprised, honestly !
A number of things wrong here, but let’s take them one at a time. First, the reason you include a linear term in multiple regression is to remove lingering effects of autocorrelation. This is in fact exactly what happens in your version with no linear trend in place. It just so happens that the satellite record starts very near solar maximum, and ends very near solar minimum. So the TSI record shows a much steeper downward slope than it would have if we had chosen start and end years that were in phase with the Wolf cycle. Since the temp trend is up during the period 1979-2010, and the TSI trend is down during the same period, this false correlation makes it seem as though the Sun’s influence on the climate is backward. Clearly this is not the case, which should have flagged you right from the start that your statistics are flawed.
The way you get around this is to de-trend all the variables (including the dependent variable) before running the multiple regression, and that means adding or subtracting a linear term from each.
Thank you Bob for the best explanation of the ENSO effect so far. It’s obvious that including an ENSO index, trivilialises the ENSO effect to the point of meaningless. It’s also obvious that the heat exchanges in the oceans and the teleconnections involved are way more complex than have been described in the peer-reviewed literature so far. It’s been a while since I did stats at uni, so I’m not entering into the detail of the regression arguments, but I reckon you’re on to something. Whether these regressions are correct or not, its not that important. It’s a paradigm thing.
A physicist, regarding your January 3, 2012 at 2:57 pm comment, in which you quoted KR:
“KR reminds us: Again, in order to properly perform a multiple regression of this kind you need to calculate the attributions simultaneously, not individually[emphasis added], or you will over-attribute to the initial components and under-attribute to the later ones in an order-dependent fashion.”
A physicist, it really looks as though you, like KR, didn’t bother to read the post. I only used multiple regression analysis. If you had read the post, I’m not sure how you could have missed that.
And, in an attempt to head off your possible next comment, I’ll add the following. I also showed Foster and Rahmstorf (2011) would also have gotten similar results to what was presented in their paper if they had simply detrended the data. That was the discussion of Equation 3. I well understand that detrending the dependent variable before multiple regression is done, for example, because the trend can distort features of the dependent variable. The point is, Foster and Rahmstorf (2011) chose to include a trend as an independent variable and it’s not an independent variable. And the reason they gave for including the trend is based on an assumption that is contradicted by the Sea Surface Temperature record of the past 30 years.
BTW: A physicist, you remind me of another blogger who used to comment here. Did you recently change your name here at WUWT?
For me, F&R represents a powerful predictor of short term temperature evolution. Whatever forcings predominate in the identified trend (CO2, aerosol change, ocean cycles, cloud dynamics, add your own…) they may be reasonably approximated by a linear response over short time frames. The consistency of the +0.16C/decade trend dictates new global temperature highs whenever El Nino’s of sufficient strength arise.
Socratic, regarding your January 3, 2012 at 4:04 pm comment: I did not say in the post that excluding the trend provided right or wrong results. Nor did I say that including the trend provided right or wrong results. I noted that they provided different results. I also presented the fact that Foster and Rahmstorf would have gotten similar results to what they had presented in their paper if they had simply detrended the data.
I have been asking the following question in the thread that I should have included in the post:
since the trend is not an independent variable, should the impact of including it in the linear regression analysis have been discussed in F&R 2011?
They simply could have detrended the data and explained why they detrended it. But they chose to give a reason that is contradicted by the Sea Surface Temperature record of the past 30 years, which I explained and illustrated later in the post.
Bob Tisdale – “I only used multiple regression analysis… I also showed Foster and Rahmstorf (2011) would also have gotten similar results to what was presented in their paper if they had simply detrended the data.”
Minor, last comment: de-trending the data means that you are individually removing the linear trend, rather than accounting for it’s attribution in the multiple regression. And hence getting different results than (correctly) including a linear trend in the multiple regression.
The linear trend is an independent variable if it does not correlate with the other components used in the linear regression. This comes from the very definition of orthogonal (or near orthogonal) components. If you disagree with the linear trend (perhaps you feel the slope is incorrect), run a series of multiple regressions with different slopes including zero and look for the minimization of the residuals, as F&R 2011 did for the lags.
Geoff Sharp says:
January 3, 2012 at 2:59 pm
Some studies deal with UV others with EUV.
None that you reference deal with the effect of EUV on the climate.
What is important is that you recognize the chemical reaction not necessarily the range of the spectrum.
Again: what chemical reactions do you think EUV are responsible for, and how do they change the climate?
Variations in UV and EUV are many times greater than TSI.
The variation of energy involved [Watt/m2] is much much smaller than those of TSI.
Repeat: Changes in climate are not caused by changes in EUV.
M.A.Vukcevic says:
January 3, 2012 at 3:16 pm
I suppose he just forgot to change caption on his graph
More likely that you ignore or deliberately forget what I have told you many times.
I did not … “A physicist” is the sole name under which I have ever posted. So if there was a prior, similar poster, perhaps she was a physicist too?
Excellent points by Socratic, for which appreciation is given. Perhaps you are a physicist / statistician / mathematician / computer scientist, Socratic? 🙂
[Moderator’s Note: “A Physicist” is the only handle he has used here and has consistently used on other sites as well. Anyone with a half ounce of curiousity can learn a great deal more about him if they choose. -REP]
“Whatever forcings” should read “Whatever longer term forcings” in my previous post.
Bob Tisdale says:
January 3, 2012 at 5:10 pm
since the trend is not an independent variable, should the impact of including it in the linear regression analysis have been discussed in F&R 2011?
ABSOLUTELY because it is the elephant in the china shop in Equation 2. As already shown:
GISS = Trend * year + (y intercept)
Therefore, all the other terms reduce to a constant, the (0-y intercept). In other words, the other terms are assumed to have zero contribution to the increase in temperature, as a condition of equation 2.
Equation 2 assumes the other terms simply contribute to the wiggle in temperature around the trend, and then tries to approximate this wiggle with a multivariate linear fit. How is it not cooking the books on the order of hiding the decline?
Leif, thanks for the clarification. I calculated an adjusted TSI by your method. As a first rough cut, I fed it to a 4-predictor OLS model without F&R’s error or lag structure (but that obtains similar trends and general conclusions). Coefficients on other variables, including year, are the same to two decimal places with the old TSI or adjusted. Coefficients on TSI itself are similar but slightly weaker using the adjusted version.
Is that what you’d expect?
KR says: “As per my post at 11:08 – if (as you assert) the ENSO cannot be represented by an index (MEI or SOI), then the multiple regression will attribute those non-index qualities to other components that better match them. In this case, that would probably have to be the linear trend component, as TSI and volcanic forcings don’t correlate well. The “step” progression you espouse, not correlated to those variation indices, should show up in the residuals if it exists – there is no sign of it. “
I not only espouse a Step progression, I have shown it in the post above and in other posts that preceded it, which you have commented on. It’s tough to miss:
http://bobtisdale.files.wordpress.com/2012/01/figure-14.png
It therefore is not a matter of “if it exists”. I have shown that it exists in the Sea Surface Temperature records for the past 30 years for 67% of the surface area of the global oceans. For the other 33% of the global oceans there has been little rise in Sea Surface Temperature:
http://bobtisdale.files.wordpress.com/2012/01/figure-10.png
You continued, “And yes, I’ve read numerous posts of yours on the subject – you leave the mechanism(s) behind your claims vague and unsupported, and have not once replied directly to my queries.”
Now your comments have reached the level of nonsense. I actually discussed the mechanisms in the post above. In posts that I’ve linked specifically for YOU on previous threads, I have presented Sea Surface Temperature, Ocean Heat Content, Sea Level, Downward Shortwave Radiation, Cloud Amount, etc., data that support my descriptions of those mechanisms. In those posts and other linked to them, I have presented animated maps that show the response of numerous coupled variables, further clarifying my descriptions of mechanism. In the post above, I provided a link to an earlier post that further linked “about a dozen additional posts that discuss ENSO and the multiyear aftereffects of specific ENSO events linked at the end of that post.” I had called then to the attention of readers who were interested. If you’re not interested, so be it.
With respect to your request for peer-reviewed papers, I was the first to illustrate the “step progression” as you call them. I first noted them in two posts about three years ago. I have published dozens of posts that further clarify and explain them since then. In March of this year, I was the first to divide the global oceans into the two logical subsets that were illustrated in this post.
http://bobtisdale.wordpress.com/2011/03/03/sea-surface-temperature-anomalies-%e2%80%93-east-pacific-versus-the-rest-of-the-world/
A month later, I discussed them in a post titled, “How Can Things So Obvious Be Overlooked By The Climate Science Community?”
http://bobtisdale.wordpress.com/2011/04/05/how-can-things-so-obvious-be-overlooked-by-the-climate-science-community/
With that title and the content of the post, one might conclude that there are no peer-reviewed papers that support what I’ve presented.
You concluded your comment with, “Adieu.”
And I’ll reply, Good-bye.
HB says: “Thank you Bob for the best explanation of the ENSO effect so far.”
Thanks for the feedback. How could improve the explanation?
Ammonite, your post admirably summarizes the main thrust of several other WUWT posters, and for that matter, the main thrust of the Foster / Rahmstorf article too.
Like the navigator (James Earl Jones) says in Kubrick’s movie Dr. Strangelove:
KR says with respect to my comment that Foster and Rahmstorf (2011) could have detrended the data: “Minor, last comment: de-trending the data means that you are individually removing the linear trend, rather than accounting for it’s attribution in the multiple regression. And hence getting different results than (correctly) including a linear trend in the multiple regression.”
Finally, we can agree on something. It IS minor. The following graph compares the GISS data adjusted with the coefficient determined through the multiple regression with the linear trend (equation 2), and with the GISS data detrended (Equation 3). The differences are so small when they exist that they would not have impacted the outcome of Foster and Rahmstorf (2011):
http://i42.tinypic.com/hve8e9.jpg
Gneiss says:
January 3, 2012 at 6:26 pm
Is that what you’d expect?
Pretty much as the adjustments are small. My only point was that there is no reason to use a composite that is suspect. Might as well use the ‘best’ possible, corrected for known errors.
That the coefficients are so small just shows that solar activity is not a major player. This holds for TSI and for all the other indices that to first order follow TSI: sunspots, magnetic field, [inverse] cosmic rays, EUV, solar wind, number of CMEs, flares, etc.
M.A.Vukcevic says:
January 3, 2012 at 4:01 pm
If your instrumentation on the hillside is working OK then the magnetic field is on the move
You better check your arithmetic on that last point…
A physicist writes,
“Now, Bob, it seems to me that the principles of rational skepticism require that you thank these WUWT posters, and amend your own WUWT guest-post to acknowledge that this aspect of the Foster & Rahmstorf 2011 analysis is correct.”
Plainly that won’t happen. But from a selfish viewpoint one good thing that came from this very confused post and discussion was that it inspired me to download the F&R data and start to replicate their analysis, finding out how robust that actually is. For example, with respect to the TSI correction Leif suggested above.
Leif Svalgaard says:
January 3, 2012 at 6:01 pm
None that you reference deal with the effect of EUV on the climate.
EUV has a direct effect on the heat and size of the upper atmosphere. The stratosphere is known to couple both ways with the thermosphere.
Again: what chemical reactions do you think EUV are responsible for, and how do they change the climate?
As above and also the formation of O molecules used in the production of ozone at lower levels.
EUV is one part of the spectrum that I am interested as it shows the greatest variance during the solar cycle (100%) and has not been measured properly until recently. The other parts of the spectrum FUV and MUV (30%,1%) interact with the lower parts of the atmosphere in the ozone process and also show much larger variances than TSI as a whole which you conveniently forget to mention. There is growing evidence that these processes are linked to the QBO and planetary waves that have impact on the polar vortexes. These are real world observations over the past 3-4 years.
Geoff Sharp says:
January 3, 2012 at 8:24 pm
EUV has a direct effect on the heat and size of the upper atmosphere. The stratosphere is known to couple both ways with the thermosphere.
It is not known that that coupling influences the climate. Show some studies that say that directly.
As above and also the formation of O molecules used in the production of ozone at lower levels.
The O atoms do not migrate downwards against the density gradient. The density changes by a factor of 1000 for each 50 km.
EUV is one part of the spectrum that I am interested as it shows the greatest variance during the solar cycle (100%) and has not been measured properly until recently.
That you are interested does not prove that EUV changes the climate.
which you conveniently forget to mention.
I’m specifically objecting to your unsupported claim that the climate changes are due to EUV, which they are not.
Leif Svalgaard says:
January 3, 2012 at 8:53 pm
Lets cut to the chase Leif, we have done this before and groundhog day is boring.
Are you prepared to stand by your claim that the chemical reactions that occur across the UV spectrum in no way have any bearing on climate?
I’ve made a couple of comment on this thread on some aspects of the discussion, but in coming back to it there seems to have emerged a bit of a theme suggesting the author should withdraw and apologise for his statements about F&R in respect of their treatment of the linear trend component.
I must say I think Bob Tisdale is basically right to draw attention to the issue he does.
BT’s criticism is that F&R set out to “remove the influence of these factors [ENSO, volcanoes, sun] on the temperature data sets, not only to isolate the longer-term changes, but also to identify whether different data sets show meaningful differences in their response to these factors.” This has them concluding their analysis “enables us to remove an estimate of their influence, thereby isolating the global warming signal.” Elsewhere they claim “It is noteworthy that the noise reduction from removing the influence of exogenous factors enables warming to be established using shorter time spans than with raw data.” As they end: “This is the true global warming signal.”
BT’s argument is that if you are going to partial out the effect of these factors then you should look at how these factors relate to the temperature series independent of the residue. If you do you find an illogical model. The sun ain’t heating things. I’ve already commented earlier that this was worth some discussion by F&R IMHO. I’d like to be boring and ram home the point.
The technocrats who know all about the method have simply chorused “it’s obvious look at the residues, the model has a linear trend in it, we know how to fix that” etc etc.
But this is an empirical science. Data mining is good for marketing consultants and financial wide boys. We (including hopefully F&R) are actually interested in what is really happening out there (although the number of approving comments about the willingness to use F&R for forecasting may suggest I’m just old fashioned, and the real money lies in picking energy futures using this stuff).
The issue is about causality. F&R don’t seem to be claiming that they are just producing a better way to forecast future temperatures in the short-term (andif this is their intent they go about doing it in a strange way). Their aim is to remove the obviously non-anthropogenic short-term “noise” so as to isolate the other factors “(most likely these are exclusively anthropogenic)”.
So to repeat, their purpose is about attributing causality.
Now data mining no matter how sophisticated is different from isolating causality. Those queuing up to say that we got all the techniques right in analysing this data should pause and ask (as I think Bob Tisdale did) what was the model of causality we were hypothesising and how should it be tested it?
Had this been done the mindless use of computer muscle to process multiple models to come up with best-fit would have been unnecessary, the use of linear terms wouldn’t have occurred, and the change in sign on the sun term wouldn’t have been an issue because the causal model wouldn’t have been ill specified. And I think BT is right to say the change in sign is a symptom of not all being well with the real world.
Science comes before stats.
Geoff Sharp says:
January 3, 2012 at 10:07 pm
Are you prepared to stand by your claim that the chemical reactions that occur across the UV spectrum in no way have any bearing on climate?
I’m objecting specifically and only to your unfounded claim that EUV causes chemical reactions that changes the climate.
Geoff Sharp says:
January 3, 2012 at 10:07 pm
Are you prepared to stand by your claim that the chemical reactions that occur across the UV spectrum in no way have any bearing on climate?
Are you prepared to stand by your claim that the chemical reactions that occur due to EUV is the major driver of climate?
Leif Svalgaard says:
January 3, 2012 at 10:30 pm
I’m objecting specifically and only to your unfounded claim that EUV causes chemical reactions that changes the climate.
The gist of my argument that you are running away from is that solar output can have a chemical as well as radiative component in regard to climate forcing. EUV is but one example used by Baldwin who is considered an expert in the QBO mechanics.
http://www.nwra.com/resumes/baldwin/pubs/Baldwin_et_al_2001_QBO.pdf
So I will ask the question another way. Can varying solar UV output have an effect on stratospheric ozone levels?
Thanks Dr S.. Corrected. Data file is auto-loaded into calculating ‘routine’, for some reason 45 got truncated to 4, it never failed before, back to normal.
http://www.vukcevic.talktalk.net/LFC6.htm
If you assume that each living creature on Earth consumed or produced greenhouse gases as part of normal survival, then the artificiality of a break point atrributed to 1970 and the start of the carbon pollution cycle becomes a little obvious, because the count of living things on Earth does not show a break point or mechanism for one at 1970 in any literature I have read. If you are going to toss GHG into multiple regressions, then take them back to times before GHG were measured, you are fighting alligators. You know that you cannot use a linear estimator with time for GHG because that would extrapolate back from today to zero GHG in the atmosphere at a time when we were reasonably sure that there was plant growth at a derived 180 ppm or more CO2 to allow life. For these and related reasons, I am suspicious of a GHG model that shows a sudden rise in trend about 1970. At the very least, it urges caution in the use of regressions with GHG to explain temperature. It is far from clear that the subtraction of other effects leaves a residual attributable – with acceptable accuracy – to CO2. A factor of similar magnitude might plausibly arise from a switch in the way Tmean was calculated each day, after the change from thermometers to multiple daily readings about 1990.
Geoff Sharp says:
January 3, 2012 at 11:29 pm
The gist of my argument that you are running away from is that solar output can have a chemical as well as radiative component in regard to climate forcing. EUV is but one example used by Baldwin who is considered an expert in the QBO mechanics.
That is a false argument, as I specifically object to your unfounded claim that EUV is the major driver of climate. Baldwin does not mention EUV at all, as the QBO does not extend above the mesosphere. And, BTW, the chemistry is controlled more by the QBO than the other way around.
So I will ask the question another way. Can varying solar UV output have an effect on stratospheric ozone levels?
That is not the issue, of course UV has an effect on stratospheric ozone [which in turn has a tiny effect on surface temperatures, less than 0.05 K]. The issue is: does EUV have an influence? and the answer is no.
M.A.Vukcevic says:
January 3, 2012 at 11:30 pm
Thanks Dr S.. Corrected. Data file is auto-loaded into calculating ‘routine’, for some reason 45 got truncated to 4, it never failed before, back to normal.
There is an important issue: ‘due diligence’. That is, to look twice when things are changing too much or otherwise look ‘strange’.
HAS writes,
“I’ve made a couple of comment on this thread on some aspects of the discussion, but in coming back to it there seems to have emerged a bit of a theme suggesting the author should withdraw and apologise for his statements about F&R in respect of their treatment of the linear trend component.”
He won’t, but he should.Tisdale’s effort to “take on” F&R starts out with and builds upon his own statistical confusion. This is not a “technocrat” issue but a really basic, first-year student mistake that undermines the whole post.
“But this is an empirical science. Data mining is good for marketing consultants and financial wide boys.”
That’s inverting what happened. F&R are not data mining, Tisdale is — and poorly, because he is new to statistics and does not understand the tools. The F&R paper is grounded both in statistical competence (statistician F’s specialty) and in physical understanding (oceanographer R’s specialty).
F&R is the kind of paper that will invite replication, improvement, or challenge from other scientists who also understand the tools and the substance. We’ll all learn more in that process.
KR: You wrote in your January 3, 2012 at 9:11 am comment, “Regarding ENSO – Tisdale seems to be claiming (here and in previous posts) that the ENSO variation is releasing more heat to the surface than absorbing it over the last 30 years…”
I replied in my January 3, 2012 at 9:38 am comment, “Reread the post above. You’ve misunderstood the discussion.”
KR, you then wrote and quoted my post in your January 3, 2012 at 11:08 am comment: “Then please, explain what you meant in the post above by:
“Sometimes La Niña events “overcharge” the tropical Pacific, inasmuch as they create more tropical Pacific ocean heat than was discharged during the El Niño that came before it. That was the case during the 1973/74/75/76 and 1995/96 La Niña events.”
And you continued: “You appear to be claiming a long term imbalance in ENSO heat exchanges with the deep ocean. “
That brings us up to date, KR. Now for my current reply:
Deep ocean? For the life of me, I can’t see how you can read that the portion of what you quoted and turn around and state that I “appear to be claiming a long term imbalance in ENSO heat exchanges with the deep ocean.” I provided a graph, Figure 9, to illustrate the “overcharging” of tropical Pacific Ocean Heat Content that took place during the 1973/74/75/76 and 1995/96 La Niña events.
http://bobtisdale.files.wordpress.com/2012/01/figure-9.png
The graph is clearly marked that it represents the tropical Pacific (24S-24N, 120E-90W) Ocean Heat Content for the depths of 0-700meters. So the “deep ocean”, which typically indicates depths greater than 1000 meters, is, first, not discussed in the sentence you quoted and, second, not included in the dataset being presented. Also, the majority of the ENSO-related temperature variations in the tropical and equatorial Pacific take place in the upper 300 meters. This is illustrated on a number of ENSO-related web pages, such as:
http://www.cpc.ncep.noaa.gov/products/analysis_monitoring/enso_update/wkxzteq.shtml
Again, the “deep ocean” does not normally come into play during a discussion of ENSO.
Earlier in the post, during discussion of how La Niña events recharge the tropical Pacific, I did use a form of the word “deep” when I wrote, “And the reason that water warms so quickly as it is carried west is because the stronger-than-normal trade winds reduce cloud cover, and this allows more downward shortwave radiation (visible sunlight) to warm the ocean to depth.” But sunlight only penetrates the ocean, in decreasing strength with depth, to about 200 meters:
http://oceanservice.noaa.gov/facts/light_travel.html
So again, the “deep ocean” is not being discussed.
So if you’re not aware of these basics, or if you misunderstand them, your challenges to what I’ve presented in the post are unfounded. If you were aware of these basics, then the portion of your January 3, 2012 at 11:08 am comment that I included above is an attempt on your part at misdirection or a fabrication.
@Bob: OK, I think I’ve got what you’re saying. I think statistical and causal arguments are getting blended.
For example, Equation 2 is F&R’s statistical model for GISS, while Equation 1 is a model for something different. (Equation 1, when mistaken as a model for GISS, is a very poor model in every way.) In Figure 3, you graph the residuals from Equation 1. In Figure 5, you graph the residuals from Equation 2, with the trend added back in. You note that the residuals of Equation 2 would be flat otherwise, which it should be noted is in general the goal of statistical modeling for time series: to have residuals that are essentially white noise (no trend, normal distribution with zero mean, etc).
So confusion arises between Equation 1’s purpose and Equation 2’s purpose, and also between Figure 3 and Figure 5, both of which show “adjusted” values which are adjusted in different ways. I’m think I’m de-confused now, so moving on…
Your main argument with Equation 2 is that it includes “a trend”, which is what F&R call their variable tau. In actuality, tau is the time since 1990. It’s not a trend, per se, it allows for a trend, depending on the value of its coefficient in the regression. In my mind, time is not a dependent variable, and tau’s coefficient could turn out to be (statistically not significantly different from) zero, which would indicate no linear trend, so it’s not “inserting a trend”. This issue may be beyond my skill level, though.
The real issue is not that a linear trend is possible in Equation 2. Rather, the issues is are: 1) is a straight line really a reasonable approximation for GISS after controlling for the three exogenous variables, 2) if a straight line *is* a reasonable approximation, what other components does this line represent and how do they each contribute, 3) are the three exogenous variables actually independent variables, and 4) the three exogenous variables are simply lagged, but: a) are the calculated lags correct, and b) do the variables only have an effect only at the single lag or might their be cumulative effects?
Issue 1 is where F&R makes their fundamental error. They say, “We focus on the period since 1979, since satellite microwave data are available and the warming trend since that time is at least approximately linear.” But the problem is that while warming may be approximately linear over a fairly short time period that does not mean that it *is* linear, and even if it is linear that doesn’t mean that any particular component that contributes to the overall trend is itself linear.
Similarly, they assume (Issue 2) that all warming other than the three variables is AGW. No need for an Equation 1 to attempt to argue Issues 1 or 2. You make a strong argument (Issue 3) that ENSO is in fact not an independent variable. And along the way you also argue Issues 4a and 4b.
All of these issues are non-statistical. They are causal. F&R’s statistics seem reasonable to me, given their assumptions, and arguing with tau or arguing sign flips, etc, is questionable and weak. Their weakness is their assumptions and their physics/chemistry, not their math.
@HAS: “BT’s argument is that if you are going to partial out the effect of these factors then you should look at how these factors relate to the temperature series independent of the residue. If you do you find an illogical model. The sun ain’t heating things. I’ve already commented earlier that this was worth some discussion by F&R IMHO.”
You say “illogical model”, my question would be “illogical model of what?” It’s not a model for GISS temperature, but for what? It may in fact not be a model of anything real (even though its component variables are real), in which case you really can’t insist that its individual coefficients make physical sense. Or am I just confused here?
Gneiss’ comment shows us rational skepticism at its best.
It’s a good thing that so many WUWT posters have replicated the F&R calculations for themselves. Moreover, it’s a good thing that the majority of WUWT posters (Bob Tisdale being the main exception) approve of the overall methodology of the F&R article.
Not mentioned in this discussion (so far) is that the method of the F&R paper is a strong prediction method. Namely, that if we continue to apply the F&R analysis methods in the next decade (or two decades, or five decades), and if atmospheric CO2 levels continue to rise (which is a safe bet), and supposing that it is CO2 that is driving the global warming trend (as most scientists believe), then resulting F&R-processed climate data should show us with greater-and-greater clarity the physical reality of that climate warming trend.
Rational skepticism has to approve the F&R style of predictive, testable, independently-verifiable science. And if it be possibly the case that AGW is real, such that the F&R methods show us that reality with ever-increasing clarity, then rational skepticism has to accept that possibility too.
Wayne2 says:
January 4, 2012 at 7:29 am
Your main argument with Equation 2 is that it includes “a trend”, which is what F&R call their variable tau. In actuality, tau is the time since 1990. It’s not a trend, per se, it allows for a trend, depending on the value of its coefficient in the regression. In my mind, time is not a dependent variable, and tau’s coefficient could turn out to be (statistically not significantly different from) zero, which would indicate no linear trend, so it’s not “inserting a trend”.
The coefficient is for tau not zero, it is almost exactly 1. Which basically says that temperature is driven by time, or time is driven by temperature. We know this, temperatures have been generally increasing since the little ice age. F&R have proven that GW is well correlated with time, as CO2 is nowhere in the equation.
The Big Bang suggests it is in fact temperature that is driving time.
Intrigued, they checked out long range observations of silicon-32 and radium-226 decay, both of which showed a slight but definite variation over time. Intriguingly, the decay seemed to vary with the seasons, with the rate a little faster in the winter and a little slower in the summer.
http://io9.com/5619954/the-sun-is-changing-the-rate-of-radioactive-decay-and-breaking-the-rules-of-chemistry
ferd berple says:
January 4, 2012 at 8:34 am
the decay seemed to vary with the seasons, with the rate a little faster in the winter and a little slower in the summer.
So radium atoms know about winter in the Northern Hemisphere or do they change their decay rate when they cross the equator…
So I took the moderator’s advice and looked up the other contributions by “A physicist”. After wasting 15 minutes of my life that I will never get back, I am forced to sadly conclude that he is nothing more than a concern troll. He has one pet issue which is to comment, generally sarcastically, on the concept of “rational skepticism” on this site. Substantive contributions by our friend are a rarity. He is also conveniently blinkered and appears to have never questioned any aspect of “mainstream” AGW theory. This thread is a good example where he is quick to proclaim support for the “overall methodology of the F&R article” but somehow completely ignores BT’s primary argument that ENSO is incorrectly treated.
It is a testimony to the tolerance of the administrators here that “A physicist” is still allowed to post. He would have been banned from RealClimate or Open Mind in the proverbial heartbeat.
Wayne2, regarding your well-worded and well-structured comment at January 4, 2012 at 7:29 am:
Thanks for the clarification and summation from your point of view.
Regards
@fred: No, the coefficient for tau is not nearly 1.
Wayne2 writes,
“@fred: No, the coefficient for tau is not nearly 1.”
I wondered that too. The whole point of F&R is that the coefficient on time (i.e., that linear trend) is about 0.017 (or 0.17 degrees C/decade), which is in the same ballpark as all other estimates of the recent trend. Yet Tisdale’s equation 2 above gives a coefficient on time of 1.024, as Berple has repeatedly mentioned:
“GISS = -91.43 + 1.024Trend + 0.0761MEI(4m lag) + 0.06694TSI.PMOD(1m lag)- 2.334AOD (7m lag)”
If 1.024 really were the coefficient on time, it would imply a 1.024 degrees C/year rate of warming, which nobody anywhere has claimed. What’s going on with Tisdale’s equation? It can’t be what it seems.
Actually, the WUWT moderators have created a special queue for my posts, and a pretty fair fraction of them are not allowed to see the light of day.
Particularly likely to be censored (in my experience) are posts that link to survey articles — meaning, articles that broadly survey and summarize the existing scientific evidence — and that invite WUWT readers to read-and-reflect upon those articles.
No other scientific and/or skeptical forum (that is known to me) practices a brand of selective censorship that seeks to discourage citizens from reading the scientific literature (although many political forums do this, obviously).
And it is troubling that some WUWT readers may not even realize that this censorship is routinely imposed upon them.
WUWT, indeed?
REPLY: Dr. Sidles, you have 188 approved comments here, so your claims of “censorship” are baseless. Since you have had a long history of thread bombing, off topic rants, and a tendency to denigrate not only other commenters, but also the moderators and the host on a regular basis, you indeed have been assigned “troll bin status”. That means every one of your posts goes into the moderation que to receive an extra level of attention for consideration of approval. A good number of your posts don’t meet the criteria for publication here (though your ego will no doubt say they are all valid) so they simply are culled for not meeting policy. You are not singled out, commenters on both sides of the issue find their comments snipped for failure to adhere to policy, as is my right.
As I say in my policy page, this is my home on the Internet, and generally I expect people to act like they are talking to me in my living room. You have on occasion made comments that if you were in my living room, I’d kick your butt out the door. OTOH you have on occasion made some rational arguments, so you live in limbo.
This situation is of your own making, and it has do to with your attitude, condescension, and off topic thread bombing. You can improve the situation by improving your attitude and learn to respect others. If you don’t wish to change, then consider this an invitation to leave, permanently – Anthony
P.S. Whining about this won’t become another off-topic discussion
@A Physicist: I’m not sure your summary is quite right. I think many of us are arguing the F&R is statistically reasonable, but that doesn’t mean that their model isn’t flawed by lumping into a single linear trend a bunch of unidentifiable stuff it labels as “AGW”. The causal/physical aspects of the F&R model are more problematic than BT’s statistical objections.
Their summary says, “Perhaps most important, it enables us to remove an estimate of their influence, thereby isolating the global warming signal. The resultant adjusted data show clearly, both visually and when subjected to statistical analysis, that the rate of global warming due to other factors (most likely these are exclusively anthropogenic) has been remarkably steady during the 32 years from 1979 through 2010.”
First, they don’t establish that there are no other factors that are stronger than those they considered. Second, they don’t prove that a linear trend is physically appropriate, nor that it is superior to alternatives (i.e. non-linear trends). Third, they assume but don’t prove that the warming trend is “exclusively” man-made (I guess this is similar to my first objection). In short, they have a nice model that passes various diagnostics, but there’s no proof that their model is actually explanatory or realistic. My several postings have been to tell Bob that he needs to focus his assault on their assumptions rather than leading with a probably-mistaken attack on their statistical methods. (Though I imagine he would not view it this way.)
@Gneiss: It appears to me that Bob’s units for time is 60 years. Tau doesn’t appear in the .cvs files, but rather in the R files, and I imagine Bob just made tau range from 0 to 1 over the length of the .csv file. That’d make it about 60 times what F&R (and mine): 0.01710741 * 60 = 1.026445.
Please let me do so by thanking you for your sustained hard work (and the sustained hard work of the moderators too) in making WUWT a forum where skeptic and non-skeptic engage one another — vigorously of course! — upon a basis of reason and mutual respect.
It is very important that such forums exist, and so the above appreciation is extended with complete sincerity and gratitude, in which (IMHO) pretty much everyone who posts here is joined.
Wayne2, thanks for the explanation, odd though it is. Tisdale can perhaps explain why he chose to rescale time instead of keeping years for the units, as F&R and everyone else does.
A couple of thoughts on your comments:
“First, they don’t establish that there are no other factors that are stronger than those they considered.”
You know that’s not possible, right? What is possible is to suggest other factors that you think might be stronger, test whether they are, then interpret and write up the results for discussion, replication, discussion, and close scrutiny just as F&R did.
“Second, they don’t prove that a linear trend is physically appropriate, nor that it is superior to alternatives (i.e. non-linear trends).”
Linearity is always a simplifying assumption. So are all kinds of nonlinear curves, though they become increasingly un-simple until eventually they fit perfectly, being just as complicated as the data. With more parameters a curve will fit better, so the question is not whether linearity is superior, but whether the curve (given its greater complexity) is superior. But yes, they did test simple nonlinear (quadratic) alternatives, and found them no better. Quoting F&R:
Only one of the data sets, the UAH series, showed a statistically significant quadratic term (p-value 0.03). It indicates acceleration of the warming trend at a rate of 0.006 C/decade/yr. However, we regard this acceleration with skepticism because it shows in no other data set, not even the other satellite record.
“Third, they assume but don’t prove that the warming trend is “exclusively” man-made (I guess this is similar to my first objection).”
You keep saying “prove” but that word doesn’t appear in their article. What F&R do is demonstrate empirically that three natural factors well known to influence global temperature cannot account for the observed warming trend, using any of the 5 indexes.
Think they left something out? Suggest your own factors, do the math, and show the world what you get. People surely are working on that right now.
This thread has probably done its dash but a couple more comments to help some of the recalcitrant ones who think data mining is a substitute for science.
Gneiss @ January 4, 2012 at 6:52 am
“This [Tisdale’s effort] is not a ‘technocrat’ issue but a really basic, first-year student mistake that undermines the whole post. “
I went looking for some resources that might help you understand some of the subtleties at play here and found “The role of causal reasoning in understanding Simpson’s paradox, Lord’s paradox, and the suppression effect: covariate selection in the analysis of observational studies” Onyebuchi A Arah, Emerg Themes Epidemiol. 2008; 5: 5. You’ll have to read the paper being commented on – Tu Y-K, Gunnell DJ, Gilthorpe MS. “Simpson’s paradox, Lord’s paradox, and suppression effects are the same phenomenon – the reversal paradox” Emerg Themes Epidemiol. 2008;5:2
“F&R are not data mining”
If you go back and look at the F&R code quoted by SRJ @ January 3, 2012 at 1:59 am, just how many times did they go around that loop? How much data did they hold out? What are the odds of a good result regardless?
“The F&R paper is grounded both in statistical competence (statistician F’s specialty) and in physical understanding (oceanographer R’s specialty).”
Statisticians and oceanographers can make good data miners I’m sure.
“F&R is the kind of paper that will invite replication, improvement, or challenge from other scientists who also understand the tools and the substance.”
It would be a pity if too much effort goes into replicating simple linear models of global climate temperature such as this and the earlier Lean and Rind contributions. We need (and science can deliver) better hypotheses to test than GISS is a linear function of time, MEI, TSI and AOD (with an adjustment for a residue annual cycle in the data). As I said useful for the chartists, perhaps good for the politics of climate change, but hardly something that will carry forward the science.
I would finally note the appropriate outcomes from a fishing expedition like this are hypotheses for testing, not conclusions.
Wayne2 @ January 4, 2012 at 7:29 am reinforces my point about the separation between causal and statistical models, but I’d further make the point (in line with the discussion by Arah mentioned above) that the statistical model is subservient to the causal model in these circumstances. Therefore it is generous of Wayne2 to conclude “F&R’s statistics seem reasonable to me, given their assumptions, and arguing with tau or arguing sign flips, etc, is questionable and weak”. As I’ve noted the flipped sign and the inclusion of time tells you we have a potentially poorly specified casual model, and hence any subsequent statistical analysis is hand waving.
I hope that last comment also clarifies the question asked by Wayne2 @ January 4, 2012 at 7:36 am. I think the issue may be just semantic – it is on the face of it an “illogical model of GISS”. While in this case we can easily reject it as a physical model I can imagine universes where it made perfect physical sense.
Finally in respect of Wayne2 @ January 4, 2012 at 11:19 am “they don’t prove that a linear trend is physically appropriate, nor that it is superior to alternatives” the fact that Ln(YEAR) is virtually as good in GISS is perhaps An Inconvenient Truth.
A physicist @ January 4, 2012 at 7:47 am while indulging in similar back scratching of Gneiss extols the virtues of rational scepticism, and adds: “Rational skepticism has to approve the F&R style of predictive, testable, independently-verifiable science.”
Model building not grounded in the body of empirical science hardly seems to pass muster as rational scepticism – but that aside at least as A physicist says, the forecasts are testable. But they are limited in their testability to their performance as a forecast tool. If the forecast is wrong we have no physical underpinnings to refer back to help tell us what went wrong.
I should perhaps note that I am not arguing against stochastic model building per se (in fact I think GCMs are at the point of diminishing returns, and stochastic modelling is required to move them forward). I am saying however that doing data mining on a limited dataset and a very limited model with no reference to the body of scientific knowledge we have of the system in question is a doubtful use of bandwidth.
I’m also surprised at the lack of willingness for many to put their brains in gear on this issue. I say it is the fault of all these damn computers. If you’d had to invert all those matrices by hand we wouldn’t have this over fitting nonsense, throwing everything at everything.
We’d hopefully have people thinking about what is actually going on out there.
A physicist says: “It’s a good thing that so many WUWT posters have replicated the F&R calculations for themselves. Moreover, it’s a good thing that the majority of WUWT posters (Bob Tisdale being the main exception) approve of the overall methodology of the F&R article.”
I have not counted the number of bloggers here who have expressed approval of F&R’s methods, nor do I intend to. But I will note that those who have approved, like you, have apparently failed to understand the significance of the second portion of my post. That part plainly illustrates and discusses how the impacts of ENSO on Global Temperatures cannot be removed through regression analyses using an ENSO index as an independent variable. Some on this thread have argued quite well for the benefits of the F&R model with great understanding of its content, but the basic fact is, ENSO cannot be accounted for as F&R has attempted to do, and that voids all of their arguments.
Have a nice day.
HAS writes,
“How much data did they hold out? What are the odds of a good result regardless?”
The fact that I and others have so easily replicated their basic conclusion, using varied and generally simpler approaches, confirm what F&R said: their results are robust. The iterative methods for chosing lags made marginal improvements but are not needed for the main conclusions. Some alternative but substantively less definitive measures of solar, volcanic and ENSO effects were tested and again reached essentially the same conclusions. None of this is what “data mining” means.
“fact that Ln(YEAR) is virtually as good in GISS is perhaps An Inconvenient Truth.”
It’s obvious and nothing of the kind. Since a nonlinear trend offers no improvement, Occam votes for linear.
Wayne2 says: “It appears to me that Bob’s units for time is 60 years.”
And Gneiss says: “Wayne2, thanks for the explanation, odd though it is. Tisdale can perhaps explain why he chose to rescale time instead of keeping years for the units, as F&R and everyone else does.”
My dependent and independent variables include only 384 months of data. When I included a linear trend as an independent variable, I used the monthly values of the trend that EXCEL determined for the dependent variable.
Bob Tisdale writes,
“My dependent and independent variables include only 384 months of data. When I included a linear trend as an independent variable, I used the monthly values of the trend that EXCEL determined for the dependent variable.”
OK, maybe now I get it. Do you mean that you regressed GISS on MEI, AOD, TSI and — as your fourth predictor — on on the predicted values from a simple regression of GISS on year?
@ fred berple says:
By including Trend(GISS) as an independent variable they have eliminated GISS. [etc.]
Thanks! I’ve just read Bob’s most excellent analysis and was almost surprised myself by F&R’s need for a “Trend” to allegedly explain or reveal a trend, but recovered quickly enough – given that it is, after all, the “methods” of genuine Climate “Science” that we are witnessing before us.
JPeden writes,
“@ fred berple says:
By including Trend(GISS) as an independent variable they have eliminated GISS. [etc.]
Thanks! I’ve just read Bob’s most excellent analysis and was almost surprised myself by F&R’s need for a “Trend” to allegedly explain or reveal a trend, but recovered quickly enough – given that it is, after all, the “methods” of genuine Climate “Science” that we are witnessing before us.”
No, F&R didn’t do it that way, Tisdale did. He thought that “including a trend” meant literally that you used predicted values calculated from that trend as one of your predictor variables. So on the left hand side of Tisdale’s equation he has GISS, and on the right hand side as one of the predictors of GISS he has GISS predicted from time.
Hence the strange near-1 coefficient that confused berple several times above, and the strange way Tisdale talked about “trend” that confused me and others.
Gneiss says:
January 4, 2012 at 3:55 pm
No, F&R didn’t do it that way, Tisdale did.
No , Bob Tisdale is saying that including the mentioned but not discussed trend is either one of the ways, or maybe even the only way F&R can get the results they want, once they’ve also decided to make the solar forcing sign negative. For example, Bob says and quotes:
Foster and Rahmstorf (2011) have added a fourth variable: linear trend. The last sentence of the third paragraph under the heading of “Introduction” reads:
“The influence of exogenous factors will be approximated by multiple regression of temperature against ENSO, volcanic influence, total solar irradiance (TSI) and a linear time trend to approximate the global warming that has occurred during the 32 years subject to analysis”
Etc., etc..
Wake me up when Climate Science gets anything of a relevant nature to the alleged hypotheses involving “CO2 = CAGW” right. Especially empirical predictions.
Leif Svalgaard says:
January 4, 2012 at 9:16 am
So radium atoms know about winter in the Northern Hemisphere or do they change their decay rate when they cross the equator…
According to GR, time passes slowest during the NH winter.
“JPeden says:
January 4, 2012 at 5:08 pm
a linear time trend to approximate the global warming that has occurred during the 32 years subject to analysis”
Which has the effect of removing the trend from GISS, so that the other time dependent factors such as PDO will appear to have no effect.
solve
GISS (t) = b + delta(GISS(t)) / delta(t) + PDO(t)
The solution is
-b = PDO(t)
Therefore PDO(t) is constant, therefore PDO has no effect on temperature.
correction:
Therefore PDO(t) is constant, therefore PDO has no effect on temperature CHANGE over time.
Leif Svalgaard says:
January 4, 2012 at 4:21 am
That is a false argument, as I specifically object to your unfounded claim that EUV is the major driver of climate. Baldwin does not mention EUV at all, as the QBO does not extend above the mesosphere. And, BTW, the chemistry is controlled more by the QBO than the other way around.
The claim is not unfounded and the main point is the chemical solar connection to climate that is not being addressed in this whole exercise. Any study that tries to isolate climate drivers without looking at the solar chemical connection is unfounded. But glad to see you finally admit this connection even if you wrongly think there is only a small effect on temperature.
Baldwin talks about UV processes that occur in the mesosphere that contribute to ozone. What you are missing is that this process happens above 120 km where an O2 molecule is split by strong UV radiation. Only UV in the band lower than 242 nm is capable of this process that makes available single oxygen molecules for later production of ozone. This firmly puts EUV in the important class. EUV is not the major driver but it and others forms of UV make up one aspect of the natural climate drivers.
Gneiss says: “OK, maybe now I get it. Do you mean that you regressed GISS on MEI, AOD, TSI and — as your fourth predictor — on on the predicted values from a simple regression of GISS on year?”
In addition to MEI, AOD, and TSI as independent variables, I used the values of the linear trend, which EXCEL calculated with its LINEST function from the monthly GISS data, as the fourth independent variable. Yes, this apparently was confusing for some. But there were others early on who understood what I had done and verified the results with other regression software. In retrospect, I should not have included the equations. I should have included a table that listed the coefficients instead.
And I just noticed that I failed to explain another portion of the post, and it may have been confusing to those who are looking at equations and graphs. I performed the regression analyses with the “raw” monthly data; then I made the adjustments to the monthly data. (I had prepared a graph using monthly data, similar to F&R’s Figure 4, with 1979-2010 as base years. But I felt my version was an unintelligible spaghetti graph with little value, so I didn’t include it.) I then converted the adjusted data to annual data; and last, changed the base years to 1979-2010. The bottom line: It’s difficult to see any difference between my Figure 7 and F&R’s Figure 5, which I included as my Figure 8.
Gneiss says: “The fact that I and others have so easily replicated their basic conclusion, using varied and generally simpler approaches, confirm what F&R said: their results are robust.”
I was also able to replicate F&R’s basic results using my primitive methods. My Figure 7 is pretty close to being a duplicate of F&R’s Figure 5, (which I included as my Figure 8). But I disagree that F&R’s results are robust for the simple reason that they assumed including an ENSO index in their regression analyses would eliminate the effects of ENSO on global surface temperatures. The second part of this post clearly illustrated that it cannot be done. The results of F&R are erroneous for that reason.
Geoff Sharp says:
January 4, 2012 at 7:00 pm
The claim is not unfounded and the main point is the chemical solar connection to climate that is not being addressed in this whole exercise. Any study that tries to isolate climate drivers without looking at the solar chemical connection is unfounded. But glad to see you finally admit this connection even if you wrongly think there is only a small effect on temperature.
You are completely off the wall here. The issue is EUV, not UV. But this is your usual straw man style.
Baldwin talks about UV processes that occur in the mesosphere that contribute to ozone. What you are missing is that this process happens above 120 km where an O2 molecule is split by strong UV radiation.
And any O3 molecule would be destroyed by the same process. And you falsely claimed that Baldwin talked about EUV. He did not. Either you did not even read his paper or you are deliberately lying and hoping nobody would notice. Which is it?
Only UV in the band lower than 242 nm is capable of this process that makes available single oxygen molecules for later production of ozone.
The O molecules above 150 km do not participate in creation of ozone. There is no ozone to speak of up there in the Thermosphere. Do you believe the O atoms move downwards? From 150 km down to the stratosphere the density increases a million-fold. And EUV is below 120 nm, BTW.
This firmly puts EUV in the important class.
This is something you make up. EUV does not create ozone, rather EUV destroys ozone.
EUV is not the major driver
You are backing off. You did call it a ‘viable’ driver of climate. That implies a significant effect. But it is good to see that you don’t think that any more.
but it and others forms of UV make up one aspect of the natural climate drivers
EUV does not. You have no evidence of that. Just what you make up. Produce some links, if you can. UV produces ~0.05 K effect at the surface, so is also not important.
You are wiggling mightily to get out of the pickle you got yourself into with this silly statement:
Geoff Sharp says:
January 2, 2012 at 9:01 pm
Using TSI as a solar proxy is one mistake made in this process and one that the warmist brigade are happy to use. Solar influence on climate (isolating PDO) is more likely a result of the large fluctuations in solar EUV
and this one:
Geoff Sharp says:
January 2, 2012 at 10:28 pm
Strange how a scientist after being shown by many that solar effects can be chemical and radiative that he continues with the same line. This can only mean an agenda is involved.
and this one:
Geoff Sharp says:
January 3, 2012 at 2:06 pm
I and others have shown you the chemical reaction along with the change in atmospheric heights are the components of EUV that science papers and bodies are now strongly suggesting as a viable climate driver.
You can check my article from June 2010 which has many links to articles and papers related to EUV and climate.
Your article does not contain a single reference or link to any such papers.
It is time to take your ball and go home, rather than try to play with the big boys.
Leif Svalgaard says:
January 4, 2012 at 8:54 pm
You really get your knickers in a knot when proven wrong dont you. You are hardly a big boy when it comes to climate science.
Tamino’s the ignorant clown who’s REPEATEDLY STATED HE THINKS MIKE MANN’S STATISTICS are REAL.
That’s all you need to know about him, right there.
Plus he actually has said he thinks trees are treemometers, too.
He’s a total greenway belt wackjob.
“Bob Tisdale says:
January 4, 2012 at 7:38 pm
But I disagree that F&R’s results are robust for the simple reason that they assumed including an ENSO index in their regression analyses would eliminate the effects of ENSO on global surface temperatures.”
If F&R’s regression has skill, then it should be able to hindcast the temperature outside the range of study and accurately predict the temperatures from 1945-1975, and from 1915-1945.
The law of large numbers says that random fluctuations should cancel out over time. So if their regression is accurately describing the real world, then their temperature forecasts should become more accurate the further back they hindcast.
For example, looking at the famous equation (2)
GISS = 123.6 + 0.06769MEI(4m lag) – 0.09025TSI.PMOD(1m lag)- 3.837AOD (7m lag)
Equation 2 tells us that if we hindcast the regression back to the birth of Christ, the temperature will be -123.6C +/- a few months of lag. Can’t get much more accurate than that.
After all, what is the point of the regression if it is simply to tell you about the data your already have? The reason you do a regression is to understand the data and thereby make a more accurate prediction about something that has not yet been observed. This allows you to take advantage of the increased accuracy to improve efficiency, increase yields, reduce waste, save lives, reduce costs, improve profits, etc., etc.
So, the while you can use various tests within the regression to tell you about obvious faults in the model, the one and only test that in the end that has any meaning is predictive skill. The ability of the model to accurately and reliably predict outside the boundaries of the model’s data. The most reliable way to do this is to test the model against a range of data that it has not yet seen.
For example, you can teach a child to add by showing them examples, 1+1, 2+2, etc. However, if you want to test if they understand addition, you can’t give them the teaching material as a test, because you may only be testing their memory. In the end, the only reliable test is to give them questions they have not yet seen and see if they get the correct answer. Model testing is not a whole lot different.
Geoff Sharp says:
January 4, 2012 at 9:40 pm
You really get your knickers in a knot when proven wrong dont you. You are hardly a big boy when it comes to climate science.
You have failed to substantiate your unfounded claim that EUV drives climate: Solar influence on climate (isolating PDO) is more likely a result of the large fluctuations in solar EUV. That is all. You can’t even respond meaningfully to the points I bring up.
BTW, I’m giving an invited talk at the 2nd Nagoya Workshop on the Relationship between Solar Activity and Climate Change (Jan 16-17, 2012). Other invited speakers include: Judith Lean, Henrik Svensmark, Adam Scaife. We are all big boys and gals.
Leif Svalgaard says:
January 4, 2012 at 8:54 pm
And any O3 molecule would be destroyed by the same process. And you falsely claimed that Baldwin talked about EUV. He did not. Either you did not even read his paper or you are deliberately lying and hoping nobody would notice. Which is it?
Most authors speak of UV in general of which EUV is one part of the spectrum. It stands to reason as I have outlined that any process of ozone creation in the mesosphere will involve UV radiation at values lower than 242 nm. EUV is a big part of that spectrum (50%). I think you are coming on a bit strong and creating your own strawman, I stated solar influence on climate is more likely a result of EUV fluctuations.
Your article does not contain a single reference or link to any such papers.
There is a link that reports on the shrinking thermosphere and EUV. There are links between the stratosphere and thermosphere so it is reasonable to suggest there may me climate implications.
http://wattsupwiththat.com/2010/07/15/earths-thermosphere-collapses-film-at-11/
This is something you make up. EUV does not create ozone, rather EUV destroys ozone.
This not something I make up, it is standard climate physics. EUV both destroys ozone and makes the building blocks for creating ozone.
http://wps.prenhall.com/wps/media/objects/3312/3392285/blb1802.html
Perhaps you can explain how you think ozone is formed.
As I said before it doesnt matter what part of the UV spectrum is involved as both EUV and FUV play their roles. What is important is that solar influence on climate is more than any radiative effect that you always push out when only considering TSI. You are now forced to acknowledge a UV based chemical solar influence on climate which has to be higher than the 0.5K value you suggest. I dont see how you can calculate the effect of changes to atmospheric teleconnections or any cloud albedo effect that may be present. Yes it is new science, but not something you can dismiss with a throw away line.
Leif Svalgaard says:
January 4, 2012 at 11:37 pm
BTW, I’m giving an invited talk at the 2nd Nagoya Workshop on the Relationship between Solar Activity and Climate Change (Jan 16-17, 2012). Other invited speakers include: Judith Lean, Henrik Svensmark, Adam Scaife. We are all big boys and gals.
I like to see the result at that workshop if you state EUV is not involved in the formation of ozone. Give it a try and see what happens.
Bob,
@HAS seems to have put it better recently than I could have. But taking it slightly further, the argument has been about paradigms. F&R have done a calculation, or model, using some indices they reckon are used by sceptics to explain global warming. They may have used established statistical methods, but do their calculations prove anything? If so what do they prove? I minored in stats at undergraduate level, and know that stats can be mis-used for any purpose.
Problem is that climate “science” seems to be to grab a headline, and don’t let the details get in the way of a good headline. So they’ve done a calculation, and made it appear that the sun, ENSO and volcanoes have not affected the “global warming trend”. This works beautifully with the paradigm they live in.
Your multiple posts further exploring the ENSO effects, (and I was interested when crosspatch, I think, asked if they are the root cause, or an effect of something else, I’m curious on that point as well) demonstrate quite clearly that ENSO is much bigger and the effects more global than the “team” like to think. As I understand it, the MEI is more an index of the ENSO status at a point in time but less an index of the ENSO effect on the global temperature. So using the MEI and taking it off the GISS trend (not even going there) is just nonsense!
Your description of the ENSO effect in the post, was a cogent, complete and easily understood description of it, IMO. A nice summation of your many posts on the topic, as it related to this. I’ve read them all.
There’s a lot of pain, I understand in taking on the climate “science” establishment, and this little foray shows how far they’ll go to knock you down before you even get to an attempt at peer review. I am impressed though by the comments in the latter part of this thread showing some genuine curiosity about what is going on in the world’s climate. Sooner or later this knowledge will become important, and it will be encouraged.
Thanks again Bob.
Geoff Sharp says:
January 4, 2012 at 11:48 pm
Most authors speak of UV in general of which EUV is one part of the spectrum. It stands to reason as I have outlined that any process of ozone creation in the mesosphere will involve UV radiation at values lower than 242 nm. EUV is a big part of that spectrum (50%).
None of them mention EUV. EUV doesn’t reach the mesosphere. Your 50% is meaningless. What is important is not the extent in wavelength, but the energy in the band. For EUV that totals to about 0.001-0.002 W/m2. For UV lower than 242 nm, the integrated energy is 50-100 times larger, so EUV is not ‘a big part’.
I stated solar influence on climate is more likely a result of EUV fluctuations
Meaning that it is less likely that some other thing is causing solar influence.
There is a link that reports on the shrinking thermosphere and EUV. There are links between the stratosphere and thermosphere so it is reasonable to suggest there may be climate implications.
None of the links make that suggestion, you make it up.
Perhaps you can explain how you think ozone is formed.
This is basic climate physics. Sydney Chapman explained that a long time ago:
http://www.columbia.edu/itc/chemistry/chem-c2407/hw/ozone_kinetics.pdf
As I said before it doesnt matter what part of the UV spectrum is involved as both EUV and FUV play their roles.
No, EUV does not play any role in ozone formation.
What is important is that solar influence on climate is more than any radiative effect that you always push out when only considering TSI.
TSI raises the temperature 250K, chemical reactions raises the temperature 0.05K, so yes, both have effect.
You are now forced to acknowledge a UV based chemical solar influence on climate which has to be higher than the 0.5K value you suggest.
Not 0.5K, but 0.05K, see e.g. http://lasp.colorado.edu/sorce/news/2011ScienceMeeting/docs/presentations/6b_Cahalan_Sedona_9-15-2011.pdf
I dont see how you can calculate the effect…
That you can’t see something is not important.
Geoff Sharp says:
January 4, 2012 at 11:54 pm
I like to see the result at that workshop if you state EUV is not involved in the formation of ozone. Give it a try and see what happens.
No need to. Nobody with even a minimum of knowledge in this field would know that.
Try to find a link that shows how much ozone is created by EUV, then show us.
Geoff Sharp says:
January 4, 2012 at 11:54 pm
I like to see the result at that workshop if you state EUV is not involved in the formation of ozone. Give it a try and see what happens.
No need to. Anybody with even a minimum of knowledge in this field would know that.
Geoff Sharp says:
January 4, 2012 at 11:54 pm
I like to see the result at that workshop if you state EUV is not involved in the formation of ozone. Give it a try and see what happens.
No need to. Anybody with even a minimum of knowledge in this field would know that.
Geoff Sharp says:
January 4, 2012 at 11:54 pm
I like to see the result at that workshop if you state EUV is not involved in the formation of ozone. Give it a try and see what happens.
No need to. Anybody with even a minimum of knowledge in this field would know that.
Here you can see how the ozone concentration falls to insignificance as we approach 140 km altitude. EUV is absorbed above ~150 km. There is no ozone produced there. http://www.leif.org/research/Oxygen-Ozone-Thermosphere.png
JPeden writes,
“F&R can get the results they want, once they’ve also decided to make the solar forcing sign negative.”
Unfortunately this still gets it backwards. That “negative solar forcing sign” was not a decision by F&R, it does not come from them at all. It’s entirely from Tisdale’s mistakes, arising from his unfamiliarity with statistical methods and consequent misreading of what F&R actually did.
I don’t know whether Tisdale understands yet why that unexpected negative sign appears in his own misspecified model. His insinuation that that it’s evidence of deception by F&R deserves an apology.
“That means, it appears Foster and Rahmstorf (2011) needed to include the trend of the GISTEMP data in the regression analysis only to assure the sign of the solar influence they sought.”
HAS made a less conspiratorial but also wrong guess about Tisdale’s negative coefficient. Leif unsurprisingly saw the reason for it right away — I’m not sure whether anyone else noticed when he did.
In quoting the above non-skeptic argument, I have redacted the tendentious rhetoric of the original SkepticalScience post, on the grounds that such rhetoric is (IMHO) pointless and regrettable on skeptical and non-skeptical sites alike.
The question asked (when stripped of pointless rhetoric) is a sensible one, though.
Leif Svalgaard says:
January 3, 2012 at 9:23 am
The null-hypothesis is that there is no effect from such an insignificant source.
The null-hypothesis only establishes probability – it does not rule out the possibility. To claim that there is no possibility even though unproven is unscientific.
A physicist says:
January 5, 2012 at 6:01 am
if El Niños cause abrupt temperature step changes upward, why wouldn’t La Niñas cause equivalent abrupt temperature step changes downward?
Why does the stock market never have step wise corrections upwards, but step-wise corrections downward are relatively common?
Gneiss says:
January 5, 2012 at 5:50 am
unsurprisingly saw the reason for it right away — I’m not sure whether anyone else noticed when he did.
The reason for including the trend on the right and left was to set all the other terms to near zero, and thus “prove” that the ocean currents have no time dependent effect on temperature trends. A mathematical parlor trick.
ferd berple says:
January 5, 2012 at 6:50 am
The null-hypothesis only establishes probability – it does not rule out the possibility. To claim that there is no possibility even though unproven is unscientific.
To help with your science education, consider this: In evaluating if something is even possible, one looks at the energy and forces available. If they are too small, the possibility is usually discarded. An example: we receive photons from Jupiter’s moons. Because their intensity is so low we discard the possibility that they cause hurricanes on Earth. We also discard the possibility that they cause hurricanes on our Moon [for the additional reason that the Moon does not have an atmosphere and an ocean]. You see, it falls to them who claim an effect to show there is an effect.
fred berple writes,
Leif Svalgaard says:
January 3, 2012 at 9:23 am
The null-hypothesis is that there is no effect from such an insignificant source.
The null-hypothesis only establishes probability – it does not rule out the possibility. To claim that there is no possibility even though unproven is unscientific.”
A null hypothesis does not establish probability. Testing a null hypothesis yields a probability, but the null hypothesis itself simply makes some claim about facts (as in Leif’s good example). If the test yields a low enough probability, we might decide not to believe that claim.
Open any stats book to see other examples of null hypotheses and how they are used.
@Gneiss: Yes, I over-reached by asking them to prove non-existence. My bad.
Would you say that:
“But yes, they did test simple nonlinear (quadratic) alternatives, and found them no better. Quoting F&R:
Only one of the data sets, the UAH series, showed a statistically significant quadratic term (p-value 0.03). It indicates acceleration of the warming trend at a rate of 0.006 C/decade/yr. However, we regard this acceleration with skepticism because it shows in no other data set, not even the other satellite record.”
indicates that warming is not accelerating? A huge number of sites say that global warming is accelerating, but their statement above seems to indicate that an accelerating (at least polynomially) warming rate is not statistically justified (over the last 30 years).
That, in itself, would be interesting.
Based on your challenge, I’m going to, as someone else suggested, add CO2 levels to their model to see how that changes the catch-all linear trend. Sound reasonable?
I still believe that Bob has shown that their ENSO value is not what they think it is, so we’re perhaps playing with nonsense, but I’m on vacation, so why not give it a shot?
In some sense, I wonder, though, if your challenge is rhetorical. The big problem with climate discussions is that those who know statistics don’t tend to know climate science, and those who know climate science don’t tend to do statistics very well. I certainly don’t know much about oceans and circulation, so really can’t do much towards your challenge in terms of original research.
fred berple writes,
“The reason for including the trend on the right and left was to set all the other terms to near zero, and thus “prove” that the ocean currents have no time dependent effect on temperature trends. A mathematical parlor trick.”
Nope, you still don’t get it. Just for a start, F&R did not “include a trend on the right and the left,” that’s a mistake introduced by Tisdale and others on this thread. Nor did they set other terms to zero, that’s a mistake you made building on Tisdale’s error.
In a number of posts above you commented on Tisdale’s 1.024 coefficient, and how illogical his equation looked. You were right to fret about the 1.024, although not for the reason you thought (it’s an impossible value if Tisdale had understood what “including a linear trend” actually means). You also were right to think Tisdale’s equation looked illogical, although you repeatedly blamed that on F&R instead of realizing it was Tisdale’s mistake.
That fact that you and many others still repeat Tisdale’s original accusation, despite how that’s unraveled in this thread, is why I think he owes an apology — not to F&R, who so far as I know haven’t noticed they’ve been “taken on” here — but to his readers at WUWT.
Leif Svalgaard says:
January 3, 2012 at 9:23 am
there is no effect from such an insignificant source.
Strip naked and lie on the beach unprotected for 12 hours on a sunny summer day after a winter spent inside.
ferd berple says:
January 5, 2012 at 7:53 am
Strip naked and lie on the beach unprotected for 12 hours on a sunny summer day after a winter spent inside.
You call that science? The energy in the UV you receive on the beach is 1000 times larger than the EUV flux.
Wayne2 writes,
“Would you say that … indicates that warming is not accelerating? A huge number of sites say that global warming is accelerating, but their statement above seems to indicate that an accelerating (at least polynomially) warming rate is not statistically justified (over the last 30 years).”
Yes, I would say that, and F&R do as well.
“Based on your challenge, I’m going to, as someone else suggested, add CO2 levels to their model to see how that changes the catch-all linear trend. Sound reasonable?”
It sounded reasonable enough to me that I tried it myself, so I know how that works out. Have a go!
“I still believe that Bob has shown that their ENSO value is not what they think it is, so we’re perhaps playing with nonsense, but I’m on vacation, so why not give it a shot?”
I claim no expertise on ENSO, I’d defer to oceanographers on that. There must be a peer-reviewed literature on the strengths and weaknesses of MEI.
“In some sense, I wonder, though, if your challenge is rhetorical.”
Not at all. My challenge to you is pretty much the way I approach things myself, if they’re interesting and fall within an area I know.
“The big problem with climate discussions is that those who know statistics don’t tend to know climate science, and those who know climate science don’t tend to do statistics very well.”
That’s exactly why Foster (statistician) teamed up with Rahmstorf (oceanographer) to write F&R. That’s why Tisdale (neither) makes such a hash when he tries to use statistics here. I totally agree that good statistics and good science both are needed.
“I certainly don’t know much about oceans and circulation, so really can’t do much towards your challenge in terms of original research.”
Fair enough, but your comments suggested that maybe F&R left out some variables. Of course they did, that critique is always true. But then the followup question has to ask for some content. What variables? If we test them, what results?