Arguments
SkS Analogy 22 - Energy SeaSaw
Posted on 22 April 2021 by Evan, jg
Tag Line
Energy exchange vs. Energy change.
Elevator Statement
Energy change represents a change of the total global energy. Energy exchange represents an exchange of energy between two parts of the Earth’s systems, without necessarily increasing the total energy.
A seesaw exchanges the potential energy of Person A on one side of the seesaw to Person B on the other side of the seesaw. As Person A falls, they lose potential energy, causing Person B to rise, gaining the potential energy lost by Person A. In this process there is no change in the total energy of Person A+B: energy is merely exchanged between two people (there is a small amount of energy lost due to friction in the center pivot, requiring each person to push off the ground a small amount).
Climate Science
The oceans and the atmosphere represent two objects that also exchange energy, engaging in their own SeaSaw, where in some years the oceans lose energy to the atmosphere, and in other years the atmosphere loses energy back to the oceans. Just as two people using a seesaw merely exchange potential energy back and forth, without changing their total potential energy, so too, during energy SeaSaw cycles the oceans and atmosphere exchange energy. Although the long-term trend is that global warming causes both the oceans and the atmosphere to gain energy, because the energy exchanged between the oceans and the atmosphere can be up to 10 times as much as the energy gained by the atmosphere during a given year, it is usually difficult to estimate the amount of global warming from the temperature change of the atmosphere from one year to the next. To estimate the magnitude of global warming typically requires looking at atmospheric temperature trends from one decade to the next, or more commonly, by using a 10-year moving average. Try out the SkS Temperature Trend Calculator to see how the averaging period affects the temperature trend line.
Examples of SeaSaws are the El-Nino/La-Nina cycles (also referred to as El-Nino Southern Oscillation, ENSO), and the Pacific Decadal Oscillation (PDO). El-Nino cycles typically occur near Christmas. For this reason, South-American fisherman referred to this weather phenomenon as El-Nino de Navidad, which in Spanish means the Christ child. In Spanish, El-Nino therefore means “the little boy”, whereas La Nina means “the little girl”.
During an El-Nino cycle, the oceans transfer sufficient energy to the atmosphere to cause globally-averaged atmospheric temperature to rise by as much as 0.2°C. During La-Nina cycles, the atmosphere transfers sufficient energy to the oceans to cause as much as 0.2°C cooling. The net effect of El-Nino/La-Nina cycles is not global warming, just energy exchange between the oceans and the atmosphere.
A good way to think of Earth’s energy system is to think of using a seesaw on an elevator. While the elevator is on the ground floor, not moving, a seesaw will simply cause two people to go up and down in rhythm, but there is no net upward motion. An upward motion only occurs when the elevator starts moving upwards, regardless of whether the seesaw is being used.
Temperatures going up and down from year-to-year mean very little: it may mean the elevator is moving up, but it may also represent the action of a SeaSaw. Temperatures going up consistently from decade-to-decade are more concerning, because it likely means the elevator is going up. Decadal temperatures have been increasing for the last five decades. And not just going up, but rising rapidly in geological terms, at a rate of about 0.2°C/decade (0.36°F/decade). SeaSaw-style fluctuations in temperature cause local changes to weather that can seriously impact local communities for a year or two. Elevator-style increases in temperature cause global effects that last for a millennium or two.
For additional information, see Questions 9, 10, and 11 of “Climate Change: Evidence and Causes.”
Spoiler Alert
You may hear reference to natural 1500-year temperature cycles. Such natural temperature cycles are topics of past and continuing research, and represent changes to global SeaSaw cycles. These 1500-year cycles usually represent local warming and cooling: where there is a 1500-year cycle of warming in one part of Earth, there will usually be a 1500-year cycle of cooling elsewhere. These well-known 1500-year cycles do not represent global warming, but rather a very complicated energy redistribution on a global scale. Although some will refer to these 1500-year temperature cycles as global warming/cooling, they actually represent global energy redistribution: if the northern hemisphere warms, the southern hemisphere cools in response. See Richard Alley (2007), Annual Review of Earth Planetary Science, 2007. 35:241-272.
Nice presentation. There are many possible examples but the sea-saw idea should be familiar to most people.
I just have one minor concern and suggestion.
The following statement at the end of the opening para under the Climate Science sub-heading could be misunderstood to mean that we need to wait until 2030 to see if the warming trend is continuing.
"To estimate the magnitude of global warming typically requires looking at atmospheric temperature trends from one decade to the next."
It may be clarified by ending it "... looking at temperature trends from one decade to the next, of by looking at the change of a 10 year moving average, or longer, as each new data point, typically each month, is obtained."
The SkS Temperature Trend Calculator can be used to see how this works. The default "Moving Average" is 12 months. For any chosen Start and End Date the appearance of the red line for the 12 month moving average can be compared to the appearance of the 120 month (10 year) or longer moving average.
There is no need to wait for the next decade to be completed to see what happened with the temperature trend.
Thanks for the suggestion. I updated the post to include your suggested revisions.
Even a decade may not be enough. There needs to be sufficient time to see a trend appear over fluctuations with scientific significance (to reject the null hypothesis of no warming). That's on the order of 20-30 years given current trends.
Unless you are accounting for cyclical and spike variations like the see-saw so well described above - ENSO, solar cycle, volcanic aerosols, etc. See Foster and Rahmstorf 2011 for an example of that. If you do, significance can be detected in perhaps 10-15 years, although that leaves open quibbling about how those variations were dealt with.
KR thanks for your comments. Because I am not a climate scientist, but I am a chemical engineer, my role is to understand what the experts are saying and then find ways to communicate that message effectively and accurately to non-technical people. One of my goals is to be as consistent as possible with the messaging I hear from the professionals, which you also appear to be.
Whereas I agree technically with what you're saying, what I heard James Hansen say in 1988 is that the warming signal had emerged from the background noise, which was still present in the early 1970's. I hear you saying that James Hansen was able to make his statements because he knew how to remove from the temperature signal the effect of transients, such as ENSO, PDO, volcanoes, ect. I concede that point, but to the casual observer, what they hear is that the 80's were hotter than the 70's. What we commonly hear now is that since the 1970's each successive decade has been hotter than the previous decade. This is the message that I think resonates with people who are really trying to understand what's happening, and not just endlessly argue the points. Considering that globally averaged atmospheric temperatures are increasing about 0.2C/decade, and that the effect of ENSO is to create a transient with a maximum of about 0.2C over a few years, 10 years seems like a suffiiently long time to provide a degree of technical rigour, yet short enough that people can grasp the immediacy of the problem. I can only assume that this is why we hear reports of the trends of decadal, average temperatures
If I try to present all of the nuances, then the presentation also becomes more difficult to follow. Therefore, whereas I concede the point you're making, materially I think it is accurate and consistent with the messaging from climate scientists that the warming signal is clearly seen if we look at the decadal temperature trends.
Evan:
1988 would have been when Hansen was testifying in front of Congress. A pivotal point - at that time a lot of politicians were much more open to the idea that climate was changing. You may be remembering that testimony, or other statements/papers at the time.
Hansen's testimony and predictions are covered here at Skeptical Science:
https://skepticalscience.com/Hansen-1988-prediction.htm
A little earlier, in 1981, was when a key Hansen et al paper was published in Science:
At that time, the paper said "It is shown that the anthropogenic carbon dioxide warming should emerge from the noise level of natural climate variability by the end of the century".
In a paper, authors tend to be a little less assertive about their projections. The late 1980s was also a period of abnormal warmth, so that may have changed the statistics.
Thanks for the clarifications Bob.
So my question to you and KR is this. Is there any reason to direct people to observe a temperature record longer than 10 years, or are we on sufficiently solid ground directing casual readers to consider 10-year atmospheric temperature trends as indicative of what the climate is doing?
Evan,
I am not a climate science specialist, but as an experienced engineer I have experience in evaluating data to see what is important, and to see through misleading claims (Technical Sales Reps often struggle to answer the questions that good engineers will ask them about technical claims they make to promote their product).
Do not lose sight of the point I make in my comment @1 "... a 10 year moving average, or longer, ..."
When the data is evaluated using a tool like the SkS Temperature Trend Calculator, all of the data should be looked at, starting in 1880 for the surface temperature data sets (called Global and Non-global) and 1979 for the Satellite data sets). Note that manipulations of satellite data do not represent the "Surface Temperature" (They are manipulations of data to represent the averages of temperatures within a thickness of atmosphere, so they are not "surface temperatures").
What is important to avoid is making a statement that could be interpreted to mean that "the Complete Next Decade of data" must be waited for to see if the new "10 year average data point" is warmer or not. The rolling average updated each month gives an indication of the trend, and the longer the rolling average is the less noise there is in the trend line (without needing to remove the variable influences like ENSO).
The National Climate Normal values in Canada are updated every 10 years using the most recent 30 years of data. So from that perspective a 30 year rolling average is better. And this can be seen by comparing the trend line of the 10 year rolling average to the trend line of the 30 year rolling average. However, because the satellite data set only starts in 1978, a 20 year rolling average shows the trend better than the 30 year rolling average, especially 10 years ago when the data set was only about 40 years.
Evan:
"Is the record long enough?" is a characteristic of the data, not so much the method. A rule of thumb for global surface air temperature is that you need to have somewhere around 15-20 years of data to determine a trend - but that is deterimed from the noise characteristics of surface air temperature. It is not generally applicable to other data sets.
In addition to statistical significance, you also need to look at the power of the test. For a small trend, you need more data than for a large trend. With low power, a non-significant result may just be a lack of data, whereas if you have high power, no significant result is more likely to mean no relationship. Dikran Marsupial did a good post on this four years ago:
https://skepticalscience.com/statisticalsignificance.html
In addition to this, we are no longer in a position where the significance of our temperature trend is in question. We are clearly in a warming trend. When we encounter someone claiming "the warming has stopped!" or "No warming since 1998 2006 2010" (or whatever year they pick), asking if the trend is significantly different from zero is the wrong statistical test.
A statistical test involves evaluating the significance of the difference between "observed" and "expected". When testing if the trend is non-zero, the "expected" is zero. When testing if the existing trend has stopped, the "expected" value is the previous trend, not zero. So that is the comparison you need to make when determining something like a t-statistic.
So, to try to directly answer your question, any 10-year period should be looked at in the context of what preceded it. For the "warming stopped in..." argument, the easiest statement to make would be to say "the last 10 years shows the trend continuing".
The Foster and Rahmstorf paper that KR mentions is very good. Foster is the person that writes https://tamino.wordpress.com/, where you will find lots of excellent posts on statistics and climate (He's been on hiatus for while now, but the archives are a treasure-trove of good stuff.)
The approach taken in the Foster and Rahmstorf paper is to quantify known sources of variation and remove them from the data, leaving the long-term trend showing clearly. If you leave that variation in, it is treated as unidentified noise in a regression, which makes the trend harder to see. (It would be wrong to reduce the noise by averaging - that just deceives the regression test. You need to be able to independently say "yes this causes variation and I know how much".)
...and we always can make use of the SkS Escalator image:
OPOF@7 and Bob@8
I am not suggesting that we need to wait until 2030/2031 to assess current trends. Just that it seems that talking about decadal averages is more intuitive for non-technical people than refering to moving averages. That may change now in the age of Covid, because we are hearing more and more about 7-day moving averages, for obvious reasons. If moving averages move into and stay in the modern lexicon, I will consider using them. But decadal averages seems like a concept that is easy to understand, and accurate: for about 50 years each decade has been warmer than the previous, and with CO2 accumulation in the atmosphere continuing its 60-year, upward acceleration, there is no idication this trend is going to stop anytime soon.
I appreciate all of the references that you are suggesting and will consider additional modifications to my messaging.
Running means represent smoothing, but they are highly auto-correlated and normal significance testing cannot be used.
Decadal averages do provide a clear step-by-step indication that things are warming.
Another convenient tool is looking at the list of warmest years - so many of the top 10/15/20 years are the recent ones.
One last link to an interesting analysis to remove some of the variation: Dana did a post here about results of an analysis by John Nielsen-Gammon, which plotted independent lines for El Nino, neutral, and La Nina years.
https://skepticalscience.com/john-nielsen-gammon-commentson-on-continued-global-warming.html
The key graphic:
Bob@10
"Dana did a post here about results of an analysis by John Nielsen-Gammon, which plotted independent lines for El Nino, neutral, and La Nina years."
We did something similar with our Christmas Dinner analogy where we plotted a trend line through annual temperatures and a trend line through decadal maximum temperatures.
In the Argo age, you could argue that OHC www.ncei.noaa.gov/access/global-ocean-heat-content/ is both a less noisy dataset (and so significance of trends is established over shorter time frames), and a better indicator of climate.
Evan:
I thought your post has an interesting perspective. I don't have anything to add there.
I think one additional reason that ten year averages are used is because of the Sun's cycle. The Sun gets stronger for a while and then gets dimmer. This pattern is clearly seen in the temperature record. Most of the highest temperatures are at the height of the Sun cycle. The average Sun cycle is about 11 years. That means that a 10 year average removes most of the Sun cycle effect.
I think your point that 10 years is meaningful to the person on the street is a good one also.
scaddenp@12 You make a valid point, and perhaps we should plot ocean temperatures together with air temperatures to emphasize their link and that they are driving each other upwards.
michael sweet@13, thanks for your comments. It's good to get affirmation of what is and is not connecting with readers (e.g., concept of decadal averages). The point of these analogies is certainly not to educate the experts (which I am not), but rather to communicate what the experts know to non-technical people.
I appreciate the time all of the experts have given to adding their comments here. Thanks.
Evan and Bob,
We appear to be discussing the same things from different perspectives.
I appreciate that using the term Trend for the impression a person would get from seeing the red-line presentation of the moving-average is not “Technically the Trend”. My perspective is that many people will accept a Trend like a Fashion Trend, something that they can see changing, more readily than the result of a statistical analysis of data.
The SkS Temperature Trend Calculator presents what is being discussed: Trend, Moving Average, and the See-Saw.
Images from the SkS Temperature Trend Calculator would help explain my perspective. But I am not skilled at “image” inserting, so I will try to present it with words.
The difference between the Statistical Trend and a 360-month moving average can be seen when looking at the full set of surface temperature data from 1880 to 2021. The Statistical Trend for the full data set does not show what is happening as well as the moving average. The moving average line shows that the recent temperature Trend is steeper than the Statistical Trend of the full data set.
And the See-Saw effect can be seen by using a shorter moving average like 60 months. So the use of a longer moving average with the full data set provides a visual impression of what is happening that is better than the statistical Trend calculations that are the “Scientifically more precise” way to determine if there is a Trend in the Data. And that presentation does not require the further manipulation of data, and related questions about how it is done, to "attempt to remove variables that create the See-Saw effect".
Whatever the technical merits of what you're saying OPOF (I'm not disagreeing with anything you're saying), it has to be personal for people to feel it enough to move them to action. The problem with going back to 1880 is that we cover a period where other signals were on the same order of magnitude or larger than the warming signal. Although industrial output cranked up after WWII, the warming signal was still muddled by other signals. But ince about 1970 the warming signal has emerged from the noise. It is also convenient that 1970 is within the consciosness of many people alive, so this is not a matter of history but of personal experience. Many people feel a difference in winters now than during their childhood (1970's or earlier). Can you tell how old I am? :-)
For me one of the most striking statistics is this. During the last deglaciation cycle CO2 rose 100 ppm in 10,000 years. CO2 has risen that much in my lifetime!
Perhaps the largest benefit of going back to1880 with the temperature record is to show how strong the variability was compared to the underlying warming signal prior to 1970, and by comparison, how strong the warming signal is now compared to the variability.
Evan, I try to learn as I try to help other people understand what is happening. I appreciate your attention to my comments, and your feedback.
I have a few more suggestions for your consideration starting with the last point you made in your comment @16 (potentially only my comments “Regarding 10 years of temperature data being a sufficiently long time to provide a degree of technical rigour” affect what you have presented in your See-Saw item).
I find it helps to expose people to the fuller record of basic data like CO2 levels and Global Average Surface Temperature. That can help them see how unusual or unnatural the recent values are and that CO2 and Temperature are related. That is why I recommend looking at the history of Temperature and CO2 data:
The higher recent rates of warming do indeed over-whelm the impressions of the See-Saw. However, the magnitude of the warming is more important. Even if the decade rate was only 0.10 degrees C, eventually the warming would be clear in spite of the larger swings of the See-Saw.
Regarding how people will perceive a message
There is a diversity of awareness, understanding and perspective. Not everyone will see things the way you intend.
You asked: “Many people feel a difference in winters now than during their childhood (1970's or earlier). Can you tell how old I am? :-)”
What I can tell is how far North you likely live. You are likely part of the small portion of humanity who live north of, or near to, 60 degrees N latitude. The arctic regions have warmed faster than the rest of the global surface. People may legitimately recollect that Northern winters were different decades ago. But global average warming since the 1960s is far less than 1 degree C with non-arctic areas warming less than the average (and there is more warming at night than the daytime. So, people in non-arctic areas may not recall a difference. I was born before 1970 and have lived between 50 an 55 degrees N. In spite of my bias of being aware of the warming and climate change that has occurred, I cannot claim a clear recollection that winters were significantly different when I was younger. So there are likely many people who do not have a legitimate recollection that winters were different decades ago.
Regarding 10 years of temperature data being a sufficiently long time to provide a degree of technical rigour
I do agree you may want to reconsider what you say about the adequacy of a 10 year set of temperature data.
As KR suggests, unless the data has had significant variable influences like ENSO and volcanic impacts scrubbed out of the data, which raises questions about how those impacts are “scrubbed out”, temperature data sets longer than 20 years may be needed to avoid unintended interpretations.
I spent a little time learning about “decades of temperature data using the SkS Temperature Trend Calculator. I looked at the Trend values for sets of 10 years in the GISTEMPv4 and UAHv6.0 TLT data starting in 1979 (everyone can do this to verify the results):
This may explain why the likes of Dr. Roy Spencer focus on their satellite data manipulations and try to claim the superiority of that data over surface temperature data. That run of values from 1997 through 2005 was a long period of being able to claim that the warming had appeared to have ended even though CO2 levels continued to increase (the UAHv6 data set trend for the 19 year period of 1997 to 2015 is negative. In the UAHv5.6 data set the longest negative Trend was for the 11 years 1998 – 2009, and in the RSSv4 TLT data set the longest negative trend is the decade starting with 2003). So shorter sets of data, rather than the fuller story, can be the “Friend” of the likes of Dr. Roy Spencer (and updated manipulations of the data can also be “Friendlier to the likes of Dr. Roy Spencer.).
A final point about presenting decade averages
I do like the presentation of the averages of the 70s, 80s, 90, 2000s, 2010s when a graph cannot be shown. And I agree that such a presentation is not improved by adding earlier decades. But I also consider a “moving average” presentation to be better, but it needs to be Graphed (referring to the SkS Temperature Trend Calculator works). The moving average values can’t be described in words the way the decade averages can be. However, the discrete decade averages are a 120 month "moving average" with the data points being every 10 years (on a graph the decade averages would be points in the middle of each decade). As you can see from the investigation I summarized above, any set of 10 years of data can be a Decade average. And when those averages are done for each new month of data the series of points will look like a line (note that Dr. Roy Spencer presents a 13 month moving average because that makes it easier to present the data points. They go on the middle point of the data set – no need to set the graphic up to present a 12 month average between the middle two months of a 12 item data set).