Global Warming FAQ - Alaskan Warming and Pacific Decadal Oscillation (PDO)

Alaska Temperatures and their Relationship with the Pacific Decadal Oscillation (PDO)

Introduction There has been some controversy recently about the temperature trends in Alaska and their relationship with the PDO. Media reports have described alarming levels of Alaskan warming over the past 30 years, but critics have responded that the level of warming is lower than that reported, and that warming over this period is largely a result of a change in the PDO, which resulted in a 'step change' in temperatures around 1976. This hypothesis can be t4ested by comparing Alaskan temperature trends with changes in the PDO. Methods The FANB compilation devised by Dr Sue Bowling at Fairbanks in 1990 serves as a useful measure of temperature across Alaska. FANB is simply the unweighted mean of data from Fairbanks, Anchorage, Nome, and Barrow from 1954 onwards. In his 'Response to the NY Times Article', Professor Wendler explains some of the reasoning behind the choice of these four stations as representative of temperature changes in Alaska overall. The FANB reported by Dr Bowling shows a significant step-change in 1976. For the purposes of this analysis, the FANB was recreated using data from the homogenised dataset maintained by the Goddard Institute for Space Studies (GISS). It is important to note that urban sites in this dataset are 'forced' to follow the same trend as local, rural stations. Therefore, the overall trend in each of the 4 records is as per the local rural sites, although the yearly fluctuations are unique. This was compared with the PDO Index provided by Nathan Mantua at the University of Washington. The annual mean of each series was used to exclude seasonal effects. The PDO index was regressed vs the FANB, and the resultant regression formula used to computed 'predicted' FANB temperatures given the PDO index for each year. By subtracting these predictions from the actual temperatures, a series of residuals are obtained that show the temperature anomalies that are not explained by changes in PDO. Results
Plots of the annual FANB and the PDO index since 1954 are shown in Figure 1. There appears to be a good correspondence between the two, and the step change in temperatures and PDO is evident. The linear trend in temperatures is large and positive, amounting to around 0.37°C/decade.	Figure 1. Alaskan (FANB) temperatures and PDO index since 1954.
The linear regression of FANB vs PDO Index is shown in Figure 2. The relationship is linear, with a correlation (r-squared) of 0.48 (p<0.0001).	Figure 2. Regression of PDO vs FANB
After covarying FANB for PDO, the resultant residuals are plotted in Figure 3. The step change that occurred in 1976 has been eliminated, and the overall trend is reduced by around 60%. The linear trend for this period, excluding the effects of the PDO, is around 0.14°C/decade. Although Sue Bowling only extends the data back to 1954, the GISS data provide records for all four stations stretching back to 1930. Therefore, I applied the above process to these data. Since much of the controversy has focussed on the time period 1971-2001, I did the same for theses data. The full results are shown in Table 1. The GISS global trends for the same time periods are shown for comparison.	Figure 3. Residuals left after covarying FANB for PDO.
Table 1. Comparison of decadal temperature trends (°C/decade) for FANB, FANB corrected for PDO, and GISS global surface temperatures.

	Time Period
	1930-2001	1954-2001	1971-2001
FANB	0.104	0.369	0.447
FANB-PDO	0.085	0.145	0.281
GISS	0.059	0.125	0.177

Modelling FANB Temperatures Based on PDO and GISS Global Temperatures

Table 1 suggests that there is an additional warming trend in the FANB that is not accounted for by changes in the PDO. A reasonable hypothesis is that this additional warming trend is related to the warming trend observed in the global surface temperature record (global warming). Indeed, the GISS global land surface record, like the PDO, is positively correlated with the FANB (r²=0.42, p<0.0001).

To test this hypothesis, multiple linear regression was used to regress both the PDO and the GISS dataset against the FANB. This regression provides a very good fit to the data, and more importantly provides a better fit than either the PDO or the GISS dataset used individually. (r²=0.64, p<0.0001). The following regression formula was obtained:

Predicted FANB = -4.27675 + 0.61686 x PDO + 1.82961 x GISS

Notably, this regression formula suggests that the warming trend in the FANB that is around 1.8 times that of the warming in the global dataset.

A comparison of temperatures predicted based on this formula with the actual FANB dataset is shown in Figure 4. To test the closeness of the fit, the residuals were again calculated (as above). A linear regression of these residuals showed no evidence of any further unexplained trend in the data(r²=0.015, p=0.41).

Figure 4. A comparison of FANB predicted based on a combination of PDO and GISS global temperatures

Conclusions

The FANB compilation suggests that Alaska has experienced a considerable degree of warming over the period of study – around 50-150% greater than the global mean over all time periods.
The Alaskan warming trend shown in the FANB may be expected to have extensive effects on local ecosystems.
Although substantial, this warming is not as great as that reported in the NY Times article. It is conceivable (though it seems unlikely) that a more comprehensive, spatially weighted record (such as that provided by the gridded analysis of the CRU) could be substantially different. E
A substantial part of this warming is a result of changes in the PDO.
However, even after removing the effects of the PDO, there is a residual, unexplained, warming trend.
By including the global dataset in the analysis, this warming trend is explained, suggesting that there is a warming trend in Alaska that is closely associated with global warming.
The rate of warming in Alaska, even after removing the effects of the PDO, remains at least as great (and probably greater) than the global average.
The rate of warming appears to be accelerating. These data are consistent with predictions that Arctic regions will experience a greater degree of warming than the global average as a result of greenhouse gas emissions.

Unique visitors: since 7 July 2001

Last updated 03/07/04. By Tom Rees. Contact the author