Climate Oscillations and GISS Temperature Anomaly Trends

In this post, I examine the combined impacts of Pacific Decadal Oscillation (PDO), Atlantic Multidecadal Oscillation (AMO) and El Nino – Southern Oscillation (ENSO)   on the long-term GISS Land and Ocean Temperature Anomaly (LOTA) trend.

Introduction

Professor Don Easterbrook of Western Washington University  has stated …

“The PDO cool mode has replaced the warm mode in the Pacific Ocean, virtually assuring us of about 30 years of global cooling, perhaps much deeper than the global cooling from about 1945 to 1977.”  source

Easterbrook’s PDO theory is repeated  here and here. Clearly he believes that the shift in PDO phase from warm to cool will have a significant impact on global temperatures for the next 30 years.

In this post I take a closer look at PDO, AMO and ENSO indexes to see how they are related to the GISS anomaly trends.

PDO, AMO, ENSO Climate Oscillation Combinations

I have retrieved the GISS, PDO , AMO and  ENSO data and constructed a consolidated file (available here) of monthly values since 1900.  I have assigned a PAE phase code to classify each month since 1900 based on the combined positive/negative status of each oscillation, using the codes in this table:

Since there are several ENSO phase codes, I used the Nino34 index as the ENSO index.

GISS LOTA Boxplots by PAE Phase

Figure 1 shows the  GISS LOTA boxplots for each of the 8 combined PDO-AMO_ENSO phases. (click to enlarge)

pae_boxplot_by_pae_code

These boxplots show the distribution of monthly GISS anomalies by PAE phase without considering the year of the anomaly. Three important characteristics of the monthly GISS LOTA series:

  1. The median temperatures of the PAE  phases vary, with  PAE 1 (- – -) considerably less than PAE 8 (+++).
  2. There is considerable  overlap among the PAE phases. For example, there are PAE 1 months with GISS anomalies greater than Phase 8 months.
  3. Knowing the PAE phase does not provide sufficient information to assess the monthly temperature anomaly by itself.

GISS LOTA Trends by PAE Phase

Let’s take a look at the role of both the year and PAE phase on the GISS anomalies. By separating the monthly GISS data by PAE phase and plotting on the same overall trend chart we can see how the PAE phases compare over time.

The Figure 2 trend charts show monthly GISS anomalies by PAE phase over the 1900-2010 time period.

giss_trend_by_PAE_phase

  1. GISS anomalies have increased during all 8 phases, from PAE 1 (- – -) to PAE 8 (+++).
  2. The trend rates vary by PAE phase, from a low of 0.00565 / year for Phase 1 (- – -) to a high of 0.00737 for Phase 7 (++-).
  3. Knowing the PAE phase and the year provides much more information than just the  PAE phase. Look at the  PAE 1 trend chart.  PAE 1 anomalies in 1900 – 1920 were much lower than PAE 1 anomalies in 1960-1980.  The same is true for the 7 other PAE phases.

GISS Anomaly Regressions Using Year, PDO, AMO, and ENSO Phases

I have developed a series of regressions to see how powerful year and the PDO, AMO and ENSO phases are in predicting actual monthly GISS LOTA anomalies in for 1900-2010. The 6 regressions are shown in Figure 3.

Figure 3 provides several additional insights into the role of year and oscillation phase on the GISS trend:

GISS_regressions_yr_pdo_ano_enso

  1. PDO and AMO have essentially no explanatory power in predicting GISS anomalies. This is to be expected since both are detrended series. See Atmoz here for a more detailed explanation.
  2. The year regression follows the overall GISS trend, with fluctuations about the trend line.
  3. Adding the PDO-AMO-ENSO phases to the year enhances the regression’s reproduction of the month to month variability.

Conclusions

Based on my analysis of the 1900-2010 monthly GISS LOTA, PDO,AMO and ENSO data, I see no indication that the PDO shift to the cool phase will  assure “… us of about 30 years of global cooling” as predicted by Prof. Easterbrook.  While the rate of temperature increase is slightly affected by the specific PAE phase, the long term temperature anomaly trend has been upward under all PAE phases in the 1900-2010 period.

About these ads

16 responses to “Climate Oscillations and GISS Temperature Anomaly Trends

  1. Pingback: UAH Temperature Anomalies Following Predictable Pattern | Climate Charts & Graphs

  2. Pingback: Volcanic Solar Dimming, ENSO and Temperature Anomalies | Climate Charts & Graphs

  3. Hello !

    Sorry for my bad English, I am French and I don’t speak English.

    For use the PDO index data with GISS temperature since 1880, here, http://jisao.washington.edu/pdo/PDO.latest , there are the PDO index data before 1900.

    I have these errors in your R script to consolidate PDO, NINO34, AMO and GISS LOT anomaly data :

    In the script for the AMO index :
    ## dfa$Month is factor, Convert to month number & calc yr_frac
    > amo_mo_num amo_mo_frac yr_mo amo_mo_frac <- as.numeric( (amo_mo_num-0.5)/12 )
    Erreur dans amo_mo_num – 0.5 :
    argument non numérique pour un opérateur binaire

    In the script for the GISS anomaly temperature :
    > peag_df 1900)
    > peag_df peag_df peag_df <- peag_df[order(peag_df$yr_frac), ]
    Erreur dans order(peag_df$yr_frac) : l’argument 1 n’est pas un vecteur

    Thank you very much for your very good work with R and for us teatch it, I am trying to learn R with your blog.
    I put a Loess/30 years in your script (legends in French).

    http://meteo.besse83.free.fr/imfix/RClimate_5_series_month9_trends_latest_loess30.txt

    • ChristianP

      Thanks for letting me know about the AMO index problem. I have fixed it.

      I have also incorporated your suggested loess fit into the September Anomaly Trends. Very good idea, thanks again.

      I appreciate both items, please continue to write. Your R language skills are excellent and your English is very good!

      Kelly

  4. Pingback: Checking Do-It-Yourself Climate Science | Climate Charts & Graphs

  5. Hi Kelly,

    Thanks for you quick response. I really appreciate that. With regards to my project, I am trying to fit a regression between certain signals after performing SVD on daily SWE data and the PDO (which is monthly data) to access variability on SWE.

    Any suggestions will be welcome.

    Thanks again.

  6. Hi Kelly,

    I am working on a project that involves PDO and AMO. Is it possible to obtain daily measurements of the PDO as well as the AMO? If your answer is YES, can you send me a link to that effect or the data if available to you.

    By the way good job to the analyzes, it has really helped me to overcome some challenges I was facing.

    Thanks a lot and look forward to hearing from you soon.

    • James B

      I am not aware of any daily PDO, AMO data series.

      What is your project? Can you share any details?

  7. Thanks Kelly, thats what I thought, just wanted too be sure :)

  8. Steve Hempell

    Kelly,

    While going through your script to give the figures I came across this:

    # Figure 3: GISS REgressions by Year, PDO, AMO, ENSO Phases

    r_yr <- lm(giss~yr_frac , data=df)
    r_pdo <- lm(giss~pdo, data=df)
    r_amo <- lm(giss~pdo, data=df)

    Is the last line correct? Should it not be:

    r_amo <- lm(giss~amo, data=df)?

  9. Steve Hempell

    Hi Kelly,

    Maybe I’m misuderstanding something,but if I download the :
    Consolidated monthly csv data file: GISS, PDO, AMO, ENSO, PAE phases

    I get this:

    1900.29166666667,1900,”Apr”,
    1900.375,1900,”May”,
    1900.625,1900,”Aug”,
    1900.79166666667,1900,”Oct”,

    ?

    • Steve

      OK, I see what you are talking about.

      I’ve fixed the problem so the new data file includes all 12 months.

      The problem was caused by how I merged the individual data.frames.

  10. Steve Hempell

    Hi Kelly,

    While looking at your downloaded consolidated file I noticed that the months June, July and September was missing from every year. Am I correct to assume this is a bug? I fixed it in a BF&I sort of way. It doesn’t change the results much, but I thought you might like to look at it. Also tried the script with Hadcrutv3 data. Some differences – seem to be less outliers.

    • I can’t reproduce your problem.

      You may have downloaded a copy while I was testing some changes to my R script.

      I have it both as a Google file and Processtrends.com file. Both have all 12 months.

  11. In figure 1 there are some points above and under the boxplot, are those part of the data and if so, why are they outside the boxplot and so many?

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s