Story submitted by Jens Raunsø Jensen
The IPCC dismisses in its AR4 report of 2007 natural climate variability as a major reason for the global temperature increase in the second half of the 20th century. The basic arguments are “greenhouse physics”, increasing and accelerating temperatures in the second half of the 20th century, and the inability of climate models to reproduce the temperature changes if only natural processes are considered.
However, many local, regional and global temperature curves for 1960-2010 may be summarised as consisting of step changes, coinciding with one or more major ENSO-related events (El Niño) and separated by periods of near constant temperature. Thus, the temperature increase (proxy for global warming) in the second half of the 20th century could have taken place in steps driven by major ENSO events. This challenges IPCC’s notion of increasing and accelerating temperatures and IPCC’s modelling argument for accepting the anthropogenic global warming (AGW) hypothesis as the major explanation for the observed temperature changes.
Temperature curves have been analysed with many different tools to establish a perceived underlying pattern for statistical and/or for attribution purposes: smoothing, linear regression, waves and periodicities, break points, shifts etc. They all have their merits and limitations, and there is no general agreement on the pattern except as consisting of a relatively cold period from the mid 1940s to mid 1970s, followed by a warmer period during the 1980s and 1990s.
This post analyses temperature data using a tool for identifying step changes in the mean temperature, focusing on the period 1960-2010. This analysis complements many other similar analyses in the peer reviewed literature and on this and other blogs (see eg. Bob Tisdale here http://wattsupwiththat.com/2011/03/11/tisdale-on-enso-step-changes-in-rss-global-temperature-data/ ). The focus is on the land-based temperature record, the use of data up to 2010, and the application of a statistical tool that does not require a priori assumptions of the time or number of step changes. It is noted, that 1960 was selected as the start year for the analysis in order to cover the main period of interest from a global warming perspective. The step changes presented below remain the same when the entire historical observational records are analysed.
The tool, I have relied upon, is available from NOAA’s homepage and has been documented in the peer-reviewed literature (www.beringclimate.noaa.gov/). Trial runs on different annual temperature datasets suggest, that a robust solution (maximum correlation and low sensitivity to parameter setting) is obtained when using the following settings: a cut-off length parameter in the interval of 8 to 14 years (12 selected), a correction for autocorrelation by the IPN4 method, and an outlier definition of 3 s deviation in order to effectively give equal weight to all observations.
The Fig.1 below shows the result for two of the many cases I have looked at: global (crutem3gl; http://www.cru.uea.ac.uk/cru/data/temperature/ ) and Denmark, DK (t_dk_k, from Danish Meteorological Institute DMI; http://www.dmi.dk/dmi/index/klima/dmi-publikationer/tekniskerapporter.htm ).
The T-anomaly is with reference to 1961-1990 (note: the DK curve has been shifted upwards by 2 oC to avoid overlap). At the bottom in the figure, the warm (red) and cold (blue) state of the pacific decadal oscillation (PDO) is shown together with major volcanoes (squares) and El Niños (triangles). Vertical lines show the PDO shift in 1976 and the start of El Niños in 1986 and 1997.
Notwithstanding the confounding influence of anthropogenic forcings, it is hard not to see this figure as suggesting, that natural processes have had a major influence on the course of the global warming in the second half of the 20th century, contrary to the assessment of the IPCC.
The identified steps are statistically highly significant, and 85% of the variation in the global land temperature during 1960-2010 may be explained by 3 upward steps, separated by periods of near constant temperature and with a lack of warming (insignificant trend) during the most recent 13 years. The step curve for Denmark explains 40% of the variance (as compared to 30% by the Gauss-filtered smoothing model of DMI), with a lack of warming during the most recent 23 years.
The three steps in the global curve occur at 1977, 1987 and 1998. This could be a statistical coincidence as eg. any curve with a true linear trend may be summarised as a step curve. However, the three years have a documented physical significance: 1977, the great pacific shift, with the PDO turning to the warm mode, and 1987 and 1998 being years of major ENSO activity. Thus, in terms of the accumulated nino3.4 anomaly, the El Niños of 1997/98 and 1986/88 were the most extreme on record (NOAA data, 3-month average nino3.4-anomaly). Furthermore, the linear trends of the four periods separating the change points are all non-significantly different from zero, but the power of this test is of course reduced in the periods of shorter length. (It is noted that the hadcrut3 and the GISS land-ocean datasets give essentially the same result, with steps at 1977/1990/1997 and 1977/1987/1997, respectively).
Local and regional temperatures are generally known to be differently affected by ENSO events. Accordingly, many local temperature curves across the globe can similarly be summarised by the step model, with one or more steps at or close to one or more of the steps identified above in the global record. For example, the Denmark curve in Fig. 1 displays one step in 1988; Alaska curves display only one but very significant step in 1977 (GHCN data, 4 stations analysed, not shown); USA have steps in 1986 and 1998 (GISS, contiguous 48 states, not shown); and Australia have steps in 1979 and 2002 (BOM data, not shown).
Finally, sidestepping a bit with some food for thoughts: inspired by the current discussion on the role of natural causes for the changes in the atmospheric CO2 concentration, it may be mentioned, that the annual change in ppmv CO2 at Mauna Loa displays significant upward shifts in 1977 and in 1998, on average increasing the annual concentration increment by 58% and a further 33%, respectively. It seems that there could be a strong influence of ENSO also on the annual increment on the CO2 curve during 1960-2010.
It is demonstrated above that the temperature increase in the second half of the 20th century could have taken place in steps driven by major ENSO events. The significance of the finding does not mainly rest on the statistical significance of the model fit, but on the physical support of the ENSO observations for the step changes, identified without making a priori assumptions on the timing or number of steps.
If this was indeed the case – and it could be, unless proven otherwise – then the following implications arise:
1. Natural processes in the ocean-atmosphere system may have had a major influence on the global temperature change in the second half of the 20th century. If so, then something must be wrong with IPCC’s climate models, as the models according to the AR4 can not at all reproduce the observed temperature curve by considering natural causes only. This could question the climate sensitivity of the models and the models ability to adequately describe the natural processes in oceans and atmosphere (eg. ENSO phenomena). While it is generally accepted, that ENSO events can produce abrupt changes in global temperatures, the IPCC considers such effects to be short lived (albeit based on a poor ability to model ENSO processes), whereas the observational data when summarised as step changes imply a longer term effect on both local and higher-level average temperature curves.
2. The linearity assumption underlying the use of linear regressions for trend analysis of the temperature records is in principle violated by the presence of steps. Thus, the global temperature should not be considered as simply uniformly increasing or accelerating, and claims of average temperature increases and accelerations may be erroneous and misleading. The use of linear regression for analysing temperature (and other climate-related) curves should be reconsidered.
3. Regional and global temperature anomaly curves are “apples and oranges”, as they average over locations differently influenced by natural processes and in different states of the climate system. There is a need to emphasise more on the analysis of local temperature curves.
4. It was recently suggested, that the lack of warming during 1998-2008 was driven largely by natural factors (Kauffmann et al., 2011). Referring to Fig. 1, then what is the explanation for the apparent lack of increase in global temperature during 1977-1986 and 1987-1997? And what is then the conclusion for the overall cause of global warming during 1960-2010?
Finally, I want to make it clear, that I do agree with the presence of an anthropogenic greenhouse effect. But I find reasons in the observational data to doubt, that the IPCC, in its current analysis (AR4, including only data up to 2005), has assessed the relative importance of natural and anthropogenic causes for the temperature changes correctly. The role of natural processes could have been significantly underestimated.