Abrupt changes in GHCN station-level temperature records contradict the anthropogenic global warming (AGW) claims.

Guest post by Jens Raunsø Jensen

Preamble

Inspired by a statement by Dr. Kevin Trenberth in the e-mails referred to as Climategate 2.0 (#3946 discussed here), it is hoped that climate scientists will have “an open enough mind to even consider” that the global warming of the 20th century could have occurred mainly as abrupt changes in mean temperature linked with natural events. Observational data supports that claim, at variance with the AGW “consensus view”.

Summary

Abrupt or step changes in temperature regime has been the subject of many discussions on this and other blogs and in the peer reviewed literature. The issue is not only statistical. More importantly, any presence of major step changes in mean temperature regime may contradict the claims of the AGW theory and models, i.e. the claims of increasing and accelerating temperature and of human emissions of GHGs being the major cause for the relatively high temperatures in the second half of the 20th century.

In this post, 232 complete and unadjusted GHCN station records are analysed for step changes in the period 1960-2010, and it is argued that:

• Abrupt changes in temperature linked with natural climate events may be widely responsible for the “global warming” during the second half of the 20th century.
• 50% of sample stations have not experienced increased mean temperature (”warming”) for more than 18 years.
• 70% of Europe stations have not experienced warming for more than 20 years.
• The relative role of natural processes in global warming is very likely underestimated by IPCC.
• The global average temperature curve is ”apples and oranges” and is widely misinterpreted using linear trend and smoothing techniques as indicating a pattern of widespread uniformly increasing temperature.

Objective and methodology.

The post is in continuation to my earlier post on the subject (http://wattsupwiththat.com/2011/08/11/global-warming-%e2%80%93-step-changes-driven-by-enso/ ), now including a near-global station level analysis. The post is based on a ppt presentation including additional details given at a researcher’s workshop at University of Copenhagen, 15th November 2011 (http://www.danishwaterforum.dk/activities/Researchers_Day_Climate_Change_Impact_2011.html ).

The objective with this analysis has been (i) to examine the land-based temperature records at station and higher levels for the presence of step changes during the period 1960-2010, and (ii) to assess the implications for our assessment of global warming during that period. Please note that the objective has not been to dismiss a (likely) presence of an anthropogenic warming signal, or to establish a climate model, or to make projections for the future. The issue is step changes in observational data during 1960-2010.

I have used the documented Regime Shift Detection tool of Rodionov (2004, 2006; www.beringclimate.noaa.gov/ ). The results are considered to be statistically robust (ref. the ppt presentation for details on parameter settings and a verification of the assumptions of constant variance and a likely negligible influence of autocorrelation).

The station level data is from GHCN (“after combine”, http://data.giss.nasa.gov/gistemp/station_data/ ) and include ALL stations with a complete record in the period 1960-2010 in broadly defined sampling regions (ref. Fig. 1).

A total of 232 stations were identified, with 54% located in Europe and Russia. The sampling criteria result in wide differences between the “regions” in terms of station number, density and distribution. Also, the “regions” are more or less homogeneous climatologically. However, this is not of material importance for the following discussion and conclusions.

Fig. 1. Distribution of sample stations according to sampling criteria.

Results

Significant step changes are widely found in the T-records and representative examples for 3 “regions” are shown in Fig. 2a-c. The temperature increase in the steps is typically of a size which is comparable to the often quoted global warming during the 20th century.

Fig. 2a. Alaska T-anomaly (n=9). Step, 1977; T-change = 1.5 oC; significance 0.000001

Fig. 2b. Fichtelberg, Europe. Step, 1988; T-change = 1.0 oC; significance 0.00009

Fig. 2c. Malacca, South-East Asia. Steps: 1978, 1990 and 1998; T-change = 0.4+0.3+0.4 = 1.1 oC; significance, 0.0004, 0.0007 and 0.003.

Warming during 1960-2010 was clearly a non-linear process at station level, with the step pattern differing among the “regions”. The global average T-anomaly curve, constructed by averaging across station-level T-anomaly curves, is therefore highly deceptive in propagating a message of near-linearly increasing temperatures, contrary to the actual processes at station level. Thus, the global T-anomaly curve is inherently “apples and oranges” and can not be used to identify a meaningful global AGW trend if the step changes are neglected. Then, the apparent AGW trend will in reality mainly capture the aggregated effect of the sudden step changes (as e.g. in Foster and Rahmstorf, 2011).

The steps are concentrated in few short periods. Disregarding 39 steps after 2005 (considered highly uncertain and “in progress”; 2/3 ups and 1/3 downs), it is found that:

• The steps occur predominantly (58%) in three 3-year periods: 1977/79, 1987/89 and 1997/99 (Fig. 3).
• 72% of all stations, and more than 50% of stations in each “region” (except Arctic), have one or more steps during these periods (e.g. 89%, 56% and 93% of Europe, Russia and South-East Asia stations, respectively; Fig. 4).
• 78% of Europe stations have a step change in 1987/89, during which the major part of the entire warming of the 2nd half of the 20th century apparently took place.
• 2 or 3 steps are common in South-East Asia (especially 1987/89 and 1997/99), but one step only is common in records from Alaska (1977/79), Europe (1987/89) and Russia (1987/89).

Fig. 3. Distribution of step changes by year of change.

Fig. 4. Percent of stations with one or more steps in indicated 3 periods.

Similar step changes are identified in national average records (ref. link to presentation above): US contiguous 48 states (GISS): 1986 and 1998; Australia (BOM): 1979 and 2002; and Denmark (DMI): 1988. The steps in the Global T-records are: Crutem3gl: 1977, 1987 and 1998; GISS L/O: 1977, 1987 and 1998; and Hadcrut3: 1977, 1990 and 1997.

The steps are statistically highly significant. But are they supported by a probable physical cause? The answer must be yes for the majority of steps. The steps occur in a temporal and spatial pattern coinciding with well-documented events and regime changes in the ocean-atmosphere system:

• 1976/77: the great pacific shift from a “cold” to a “warm” mode (e.g. Trenberth, 1990; Hartmann and Wendler, 2005).
• 1987/89 and 1997/99: the two clearly most intense El Niños of the period, 1986/88 and 1997/98, with the intensity here defined as event-accumulated nino3.4 anomalies (NOAA’s ONI index); there were two less intense events in 1982 and 1991, the impact of which was probably occluded by the major volcanoes El Chichon and Mt. Pinatubo.
• A regime shift in NH SST in 1988/89 (Yasunaka and Hanawa, 2005).
• A new regime of constant temperature after the 1997/98 El Niño, i.e. the now widely accepted “hiatus” in global warming.
• Documented step changes and regime shifts in marine ecosystems, e.g. the late 1980s in Europe and in the Japan/East Sea.
• The short-term regionally diverse global impact of ENSO events is generally well-known.

The empirical evidence, from this station level analysis and other sources, is unequivocal: the step changes in mean temperature are likely real and associated with natural events. The physical mechanisms remain to be understood, and this is certainly not to claim, that ENSO events are the only elements of the natural cause-effect chain.

It is therefore concluded, that the major part of the temperature change (global warming) in the 2nd half of the 20th century occurred as abrupt changes in mean temperature associated with natural events in the ocean-atmosphere system. Still, a warming/cooling trend – albeit relatively small compared with the step changes – could of course be hidden by the regime change model. But it seems inconceivable, that steadily increasing CO2 levels could be responsible for the major sudden changes observed as e.g. in Alaska in 1977, Europe in 1988 and South-East Asia in 1998. In principle, the natural events and step changes could have been amplified by human caused warming, but this is currently pure speculation.

Implications when accepting the presence of steps

“Increasing temperature and accelerated warming” : this study does not support general statements like that. The bulk of the “global warming” has likely taken place in abrupt steps, and 50% of the stations analysed has not experienced any significant warming for more than 18 years (Fig. 5). In Europe, 70% of the stations have not experienced significant change in mean temperature for more than 20 years.

In South-East Asia, the median value is 13 years as many stations here also experienced a step change in 1997/98 (Fig. 4).

Fig. 5. Years of constant T-mean prior to 2010. Box-Whisker plot, 1st and 3rd quartiles. (note: uncertain up and down step changes during 2006-2010 are disregarded).

Challenging the IPCC consensus view, i.e.: “Most of the observed increase in global average temperature since the mid-20th century is very likely due to the observed increase in anthropogenic greenhouse-gas concentrations”. However, the finding above, that abrupt changes linked with natural processes likely account for most of the increase in temperature during 1960-2010, contradicts the IPCC claim regarding the relative importance of natural and human causes. Thus, when IPCC (AR4) can only reproduce the T-curve by including GHG effects, then logically

• either the IPCC GCM models do not adequately model the natural processes of high significance for the temperature variations (there is still low confidence in the projection of changes in the ENSO variability and frequency of El Niños, ref. the recent SREX-SPM IPCC report),
• or/and the IPCC has overestimated the climate sensitivity to CO2 changes by eg. attributing natural temperature increases to CO2-induced feed-back processes.

  In either case, the relative importance of natural processes for the T-changes has likely been underestimated by IPCC.

Conclusion

This study has established that step changes in land-based temperature records during 1960-2010 are common and very likely real and linked with natural climate events. The step changes are statistically highly significant and with a systematic yet regionally diverse pattern of occurrence coinciding with major climate events and regime shifts. This finding has far reaching consequences for our analysis of climate records and for our assessment of global warming.

Thus, although many different statistical models can be applied to explore the pattern of T-change, the presence of step changes invalidates the widely used statistical techniques of linear trend and smoothing as means of identifying the pattern of temperature variation during 1960-2010.

Furthermore, the step changes account for the main part of the temperature changes during the 2nd half of the 20th century. The logical consequence is that natural processes have been the major cause for the temperature change during this period, leaving a secondary role to other causes such as the anthropogenic greenhouse effect.