From Dr. Judith Curry’s Climate Etc.
Posted on September 1, 2019 by curryja
by Javier
By knowing or estimating where in the solar cycle we are we can get an estimate of the chances of a particular outcome even years ahead.
El Niño Southern Oscillation (ENSO) is the main source of interannual tropical climate variability with an important effect on global temperature and precipitation. Paleoclimatic evidence supports a relationship between ENSO and solar forcing. Moy et al. (2002) attribute the long-term increasing trend in ENSO frequency to orbitally induced changes in insolation (figure 1). The ENSO proxy record described by Moy et al. (2002) displays a millennial-scale oscillation that in the middle Holocene shifts its variance from a 1000-1500-yr period to a 2000-2500-yr period (Moy et al. 2002, their figure 1c). Both frequencies correspond to known solar periodicities, the Eddy and Bray solar cycles. As it has been shown previously (see “Centennial to millennial solar cycles“) the 1000-yr Eddy solar cycle became weaker at the Mid-Holocene Transition regaining strength in the last 2000 years. This 14C-deduced solar behavior corresponds to the ENSO behavior described by Moy et al. (2002).
Figure 1. a) Inverted global average temperature anomaly reconstruction (black line, right scale) from the 73 proxies used by Marcott et al. 2013. The temperature scale has been rescaled to produce a difference of 1.2 °C between the Holocene Climatic Optimum (HCO) and the Little Ice Age, supported on a consilience of glaciological, biological and marine sedimentary evidence that supports a 1-1.5 °C difference. b) Inverted obliquity (purple line, left scale). c) ENSO frequency (black, left scale) measure as the number of strong El Niño events in a 100-yr sliding window, from Moy et al. 2002. ENSO activity was very low during the HCO and has been increasing as the planet cooled during the Neoglacial, following changes in insolation caused by orbital changes in precession and obliquity.
In 2000 Theodore Landscheidt published an article in the proceedings from a meeting presenting his hypothesis of a solar forcing of El Niño and La Niña. He was not the first to defend such hypothesis, as 10 years earlier Roger Anderson (1990) had published some evidence for a solar cycle modulation of ENSO as a possible source of climatic change. Landscheidt’s (2000) article contains two observations and two predictions. The first observation is that most extreme ENSO events correlate with the ascending or descending phase of the solar cycle. He predicted the following El Niño based on the sun’s orbital angular momentum for 2002.9 (± 0.4). It was a 2-year ahead accurate prediction, as the next El Niño started in 2002.67. The second observation was the alternating preponderance of El Niño and La Niña following the 22-year Hale magnetic solar cycle. The 1954-76 Hale cycle showed Niña preponderance, and was followed by the 1976-96 that presented Niño dominance. While this is based only on two complete Hale cycles for which there is instrumental ENSO data it is interesting to read Landscheidt other prediction:
“If the pattern holds a preponderance of La Niña is to be expected during the Hale cycle that began in 1996.”
The Hale cycle-ENSO association is unclear to me due to insufficient data but it is undeniable that both of Landscheidt predictions were correct. Anderson’s and Landscheidt’s articles were completely ignored by the scientific community and they are rarely cited even by authors studying the same subject.
In 2008 van Loon & Meehl showed that the Pacific Ocean displayed a response to peak solar activity years similar to La Niña event years in the Southern Oscillation, but with a different stratospheric response. Haam & Tung (2012), however, failed to find an association between solar peak and La Niña years and warned that two autocorrelated time series might present a spurious correlation by chance. As I will show the problem is in the assumption that ENSO must display a linear response to solar activity with ENSO extremes at maximal and minimal solar activity. This assumption turns out to be false and the analysis of Haam & Tung (2012) using peak-solar years is misleading.
ENSO is usually described as a 2-7-year oscillation, while the Schwabe solar cycle is an 11 ± 2-year oscillation, so no linear relationship is obvious. White & Liu (2008) defend that most El Niño and La Niña episodes from 1900–2005 are grouped into non-commuting pairs that repeat every ~ 11 years, aligned with rising and falling transition phases of the solar cycle as Landscheidt (2000) described (they don’t cite him). These alignments arise from non-linear phase locking between an 11-year solar forced first harmonic and the 3rd and 5th 3.6 and 2.2-year harmonics in ENSO. These solar-forced 3rd and 5th harmonics explain ~ 52% of inter-annual variance in the Nino-3 temperature index. White & Liu (2008) propose “a new paradigm for ENSO, with El Niño and La Niña driven by the solar-forced quasi-decadal oscillation via non-linear processes in the tropical Pacific delayed action/recharge oscillator.”
Despite the evidence for a solar forcing of ENSO the accepted paradigm from model studies is that ENSO is self-excited or driven by internal variability random noise.
More recently two solar physicists, Leamon & McIntosh (2017), reported on the coincidence of the termination of the solar magnetic activity bands at the solar equator every ~ 11 years since the 1960s with a shift from El Niño to La Niña conditions in the Pacific. Their report prompted me to examine the issue, observing a pattern repetition since 1956 (figure 2). The solar minimum is preceded by Niña conditions, followed by Niño conditions, and afterwards Niña conditions accompany the rapid increase in solar activity.
Figure 2. Top: Six-month smoothed monthly sunspot number from SILSO. Bottom: Oceanic El Niño Index from NOAA. Red and blue boxes mark the El Niño and La Niña periods in the repeating pattern. This figure was published in July 2018 in an article at WUWT. Since then the Niño prediction has been confirmed.
If we assign 50% probability for seasonal positive or negative ONI (Oceanic Niño Index) values, the probability that the solar minimum will be preceded by Niña conditions, and followed by Niño conditions for six consecutive solar minima by chance is of only 0.024% (1 in 4000). The probability of the entire pattern (Niña-Niño-Niña) repeating six times at a specific time is even lower, indicating that the association between solar activity and ENSO is not due to chance. Solar control of ENSO has led to the prediction of El Niño conditions in 2018-19 by me, and to La Niña conditions in 2020-21 by Leamon & McIntosh (2017). The 2018-19 Niño prediction has been correct.
To perform a no-assumptions analysis of solar activity-ENSO correlation it is necessary to correct for the irregularities in the solar cycle, that can last from less than 9 years to more than 13. Since the sunspot dataset is very noisy I have chosen the 13-month smoothed monthly total sunspot number from SILSO.
The smoothed monthly number results from an averaging of monthly mean values over 13 months, from 6 months before to 6 months after a base month. All months are weighted equal except for the extreme ones, which are weighted by 1/2. This smoothing has been used since the early 20th century to define the times of maximum and minimum for each cycle.
Since solar minima have different levels of activity and different length, the starting and ending months for each solar cycle are defined for the purpose of this study not from the solar minimum, but from the first month that presents >30 smoothed monthly sunspots. This point in the cycle, at the beginning of the rapid ascending phase is more unambiguously defined to a single month that the solar minimum allowing for more confidence in a proper alignment of the solar cycles.
Defined in this way the last six complete solar cycles (SC18-23) have durations between 121 and 159 months. To correct for this variable length each solar cycle is divided into 22 bins that for a regular 11-year solar cycle would contain 6 months, but depending on the cycle length they can have from 5 to 8 months. After the procedure the variable solar cycle length has been normalized into a solar cycle unit (Figure 3).
Figure 3. Thin lines, solar cycles 18-24 with their respective durations normalized in terms of a full cycle and divided in 22 bins. The average monthly smoothed sunspot number for each bin is represented as a point. Thick line, average solar activity for the normalized solar cycles. Grey area standard deviation.
The analysis is restricted to the period 1950-2018, when ONI data and the smoothed monthly sunspot number were available.
ONI values are also grouped into bins corresponding to the solar activity bins. For each 1/22 fraction of the solar cycle we have six 5-8 month sunspot bins from the six solar cycles considered, and the corresponding six 5-8 month ONI bins for the same dates. The values in each bin are averaged and the mean and standard deviation for the six bins corresponding to the same solar cycle fraction obtained. The ONI dataset has a near-normal distribution with a mean very close to zero (figure 4).
Figure 4. Distribution of the 816 twelve-month periods in the 1950-2018 ONI database according to their average ONI value. The distribution follows a near-normal distribution with a mean of 0.025.
The numerical treatment of ONI values is not expected to produce values significantly different from zero if solar activity has no significant effect on ENSO. Also ONI values are expected to deviate randomly from zero at each solar cycle fraction without presenting an 11-year pattern if solar activity has no effect on ENSO. By contrast what we find is clear departures from zero, whose statistical significance will be analyzed later, and an 11-year pattern. The 22 ONI averaged values organize in two periods of more probable El Niño and two periods of more probable La Niña (figure 5). This is likely a reflection of the non-commuting El Niño and La Niña pairs that repeat every ~ 11 years found by White & Liu (2008).
Figure 5. a) Average (black line) and standard deviation (grey area) solar activity in monthly smoothed sunspot number (left scale) for the solar cycles between 1950-2018 divided in 22 fractions of a solar cycle. b) Average (dark red and blue areas) and standard deviation (pink and light blue areas) ONI values (right scale) for the same periods. The plot has been divided in five phases (dashed vertical lines) labeled in roman numbers (see text).
Attending to ONI zero-line crossing and variability I have divided the solar cycle into five ENSO phases. Phase I starts when the peak in solar activity is reached (c. 2.5 years into the cycle as defined on average), and lasts around two and a half years during which El Niño conditions are more probable, following the peak in solar activity. Phase II, of another two and a half years length, coincides with declining solar activity. This is a highly variable period when strong La Niña conditions might take place, but during most cycles it has presented strong El Niño conditions. This might be related to the string of very active solar cycles that between 1935 and 2004 have constituted the Modern Solar Maximum, and might represent a delayed response to above average insolation. This phase corresponds to Landscheidt’s observation of strong events during the declining phase of the solar cycle that would correspond preferentially to Niña or Niño depending on the Hale cycle. Phase III, of around three years, coincides with the final decline in solar activity towards the solar minimum, and usually presents La Niña conditions. Particularly frequent is a La Niña right before the solar minimum (figure 2). Phases I to III correlate in general terms with solar activity and might represent a response to solar irradiance. The last two phases display anti-correlation to solar activity. Phase IV, a short period of about 1.5 years, starts around the time of minimal solar activity, but it results in El Niño conditions that can be considerably strong at times, like in 1998. Afterwards, phase V coincides with the period of rapidly rising solar activity, that very reproducibly presents La Niña conditions. Phase IV and V do not appear to follow changes in TSI, so the suggestion by Leamon & McIntosh (2017) that they could depend on other solar parameters or galactic cosmic rays appears reasonable. Certain solar wind properties change trend at the solar minimum, like its electric field strength, or the Alfvén Mach number that reflects plasma wave speed, that peaks at the minimum. Solar wind effects on the magnetosphere have been known for long, and solar-wind-induced changes in the global electric circuit affect weather parameters at the troposphere (Lam & Tinsley 2016).
Now we can see why Haam & Tung (2012) could not find a correlation between solar peak years and La Niña, as this condition takes place preferentially during the 1.5-2 years prior to the solar peak.
To analyze how statistically significant is the solar effect on ENSO I chose Phase V that presents the biggest ONI departure from zero and even its standard deviation range does not include the zero value. A specific procedure was followed for the statistical test. The smoothed monthly sunspot number was set to zero at the solar minimum and 100 at the solar maximum for each cycle. The months when the sunspot number increased from 35% to 80% of the maximum for each cycle were selected as they define the rapid ascent in solar activity that on average lasts one year. The 76 ONI values corresponding to those months from six solar cycles (SC19-24) have an average ONI value of -0.65. Consider this: six periods of ~ 1 year selected on a solar activity criterion display an average full fledged Niña condition. What are the chances of that? To find out I made a dataset with all the 12 consecutive months averages in the ONI database (816 instances, figure 4) and then randomly picked six of them and averaged them. I did that 100,000 times in a Monte Carlo analysis and only in 0.7% of the tests I obtained an equal or lower ONI average. The La Niña -0.65 ONI value at 35 to 80% solar activity has a 99.3% probability of not being due to chance. ENSO is under solar control.
Of course ENSO is not exclusively under solar control as it is a very complex phenomenon, and thus we shouldn’t expect that the patterns are always reproduced. However it is clear from paleoclimatic data (Moy et al., 2002), solar physics (Leamon & McIntosh 2017), Modeling and reanalysis (van Loon & Meehl 2008), frequency analysis (White & Liu 2008), and the present analysis, that solar activity has a clear strong effect on ENSO, probably being its main forcing. The reported 2-7-year ENSO periodicity appears to be an 11-year periodicity with several occurrences. The present (mid-2019) position in the solar cycle is at the transition between phases III-IV, close to the solar minimum. With some uncertainty due to the irregularity of the 11-yr solar cycle, a La Niña can be projected for phase V, by mid-2020 (Leamon & McIntosh 2017). The failed El Niño projection from February 2017 by ENSO models (figure 6) took place at the transition between phases II and III in figure 5, a time when the solar cycle favors La Niña conditions that finally developed a few months later. This is an instance when ENSO prediction from solar activity would have been superior to models.
Figure 6. ECMWF ENSO forecast for February 2017 indicating Niño conditions for late summer at a time solar activity favored Niña conditions due to the transition from phase II to III. Finally Niña conditions developed.
The solar effect on ENSO could be responsible for the detected global temperature variation of 0.1-0.2 °C between solar cycle maximum and minimum (Tung and Camp, 2008), attributed to tropical evaporative feedback (Zhou and Tung, 2013). ENSO is the leading mode of interannual variability in the tropical climate system, with a global impact on surface temperature and precipitation. Its frequency has been implicated in interdecadal shifts in the tropical Pacific climate (Kumar and Hu 2013). The latest shift in 2000 has been related to the subsequent reduced rate of warming observed during the pause, a period characterized by a higher frequency of La Niña, until the 2015 El Niño put an apparent end to it. Given the clear association between solar activity and ENSO, an interesting question is if long-term changes in solar activity could be responsible for long-term changes in ENSO frequency. To compare them, 10.7 cm solar flux data, and ONI data were smoothed with a Gaussian filter equivalent to an 11-year moving average (figure 7).
Figure 7. Gaussian smoothed 1950-2018 Oceanic Niño Index (black line delimiting red and blue areas, right scale), and Gaussian smoothed 1950-2018 10.7 cm solar flux, a proxy for solar activity (thick dashed line, left scale). A 4-year lagged 10.7 cm solar flux (thick continuous line) shows that periods of high solar activity tend to coincide with periods of predominant Niño conditions, and periods of low solar activity tend to coincide with periods of predominant Niña conditions.
Long-term changes in ENSO frequency are compatible with long-term changes in solar activity. Peaks and troughs in ENSO frequency follow with a ~ 4-year lag peaks and troughs in long-term solar activity. Since ENSO activity is clearly directed by solar activity (figure 2), it is likely that the long-term correlation between both has a physical basis. If the effect of long-term changes in solar activity has to account for this lagged long-term effect on ENSO, its effect on global temperature must be much higher that the effect detected over a single solar cycle. By altering ENSO frequencies, solar activity might alter the decadal rates of warming, leading to periods of increased warming and periods of reduced warming (pauses). The 2015 El Niño has put an apparent end to the La Niña-predominant period since 2000, and to the period of reduced warming. However, if the relation between solar activity and ENSO frequency is maintained, we can project that both should continue until long-term solar activity increases again in the future.
How can we use solar activity to improve our ENSO predictions?
Figure 5 shows a probabilistic plot of ENSO in terms of solar activity. By knowing or estimating where in the solar cycle we are we can get an estimate of the chances of a particular outcome even years ahead. This can then be compared to models output when they become available. If they disagree like it happened in February 2017 we should reduce our confidence on the model prediction, and increase it when there is agreement. Landscheidt method also deserves closer attention to examine how well it has performed since 2000. Leamon & McIntosh (2017) have predicted a Niña for 2020, and over a year ago I predicted the 2018-19 Niño based on solar activity at an article at WUWT.
It is clear that even in a crude form solar activity is useful for ENSO prediction and no doubt the method can be improved enormously as Landscheidt suggested: “these problems can only be solved by a joint interdisciplinary effort of open-minded scientists.”
Data sources and bibliography [ link]
Javier,
Excellent post, and analysis. To me this analysis provides the critical link between solar activity and climate change.
From your Figure 5, Phases 1 and 2 appear to be a straight forward response of ENSO to solar activity, with a lag that is due to the ‘thermal mass’ of the oceans. Likewise for Phases 3 and 5. Phase 4 is the most intriguing, as it appears to be counter-intuitive. However, your explanation that it might be caused by solar wind effects at the change-over between solar cycles seems reasonable.
The other element of the link between solar activity and climate change is the link between ENSO and Global Atmospheric Temperatures. Many climate scientists seem not to be able to find this link when they try to correlate ENSO and temperature.
This, I believe, is because these scientists make a fundamental error in their analysis. Most scientists appear to try to correlate temperature with ENSO using concurrent data. Doing so implicitly sets the ‘thermal mass’ of the atmospheric system and transport times, for the transfer of energy from ENSO throughout the atmosphere, to zero., which is a physical impossibility.
To properly correlate temperature with a forcing such as ENSO, when there are significant time lags caused by thermal mass and the time required for heat transport, it is necessary to use a filter on the ENSO data, such as an exponentially weighted moving average. The EWMA is defined as follows:
EWMA(i) = EWMA(i -1) x (1- λ) + λ*Data(i)
As Allan MacRae recently described (wattsupwiththat.com/2019/06/15/co2-global-warming-climate-and-energy-2) most short term variations in global temperatures can be correlated to ENSO, with a lag of about 4 months. This correlation is readily shown when correlating temperature with ENSO, using an EWMA filter on monthly ENSO data with a value of λ of 1/8.
However, Willis Eschenbach explains (wattsupwiththat.com/2019/06/14/a-second-look-at-radiation-versus-temperature), that this short term effect is related to the transport of energy from ENSO to the global climate system by the process of convection. A second, and more important, process for transferring energy from ENSO to the global atmosphere is the process of advection, the process ‘periodically pump billions of cubic meters of the warmest Pacific equatorial water towards the poles’. The ‘advection’ process is much slower than the ‘convection’ process, and correlates with gradual, long term, changes to global temperatures, what Bob Tisdale and others have described as the ‘permanent effect’ of ENSO. To capture this correlation one must use an EWMA filter on the monthy ENSO data with a value of λ of approximately 1/160.
Ideally the processes of the transfer of thermal energy from ENSO to the global atmosphere by ‘convection’ (short term) and by ‘advection’ (long term) should be separated and treated as separate elements in multiple regression analyses with global atmospheric temperatures. When this is done it is readily seen that ENSO is the forcing function behind the vast majority of global temperature changes over the last half century.
To resume, according this paper by Javier, variations in solar activity drive ENSO, and ENSO can readily be shown to be the mechanism to transfer these variations in solar activity to global atmospheric temperatures, i.e. : Solar activity drives climate change through ENSO.
Yes ENSO defined as 3.4 Region SST shows a clear correlation to global surface average temperature changes, and in fact can be used to predict GSAT a few months in advance as Allan MacRae among others has shown.
I have little doubt that solar variability has a disproportionate effect on climate and one of the ways it does that is through ENSO, but it is not the only way. Atmospheric effects are very important too. The Polar Vortex is under solar activity control as I showed in another article:
And Stephen Wilde has been showing for a long time the importance of solar mechanisms mediated by ozone:
http://joannenova.com.au/2015/01/is-the-sun-driving-ozone-and-changing-the-climate/
Javier,
I agree that atmospheric effects are important, in addition to ENSO, for the transmission of climate forcings to Global Atmospheric Temperatures, but atmospheric effects tend to be short term and transitory.
The ‘advection ‘ mechanism described by Willis Eschenbach, i.e. the transmission of ENSO throughout the ocean system by ocean currents, appears however to be the principle mechanism for longer term climate change. As I said above, this has been qualitatively described by Tisdale and others. However to quantitatively describe the effect, it is necessary to account for the important lags in the system. This can be done by using a filter such as an EWMA, with a λ value of the order of 1/160 for monthly data. When the ENSO data is thus filtered, the correlations with Global Atmospheric Temperatures is highly significant and is shown to account for the vast majority of the variance in global temperatures associated with long term climate change.
isn’t this, literally, ancient history?
reminds me of when people write “what about … stratospheric/sea/etc. trends!!?”
well, i live on land in 2019. why should i attend to trends that are irrelevant to me (or, more important to me, my grandchildren’s survival)?
chris, your grandchildren’s quality of life will depend entirely on the direction solar cycles take from here. If they live near the ocean they’ll be affected as those areas are now regularly.
As long as solar activity maintains a high enough average, El Ninos predominate, which drives up integrated MEI, which is linearly related to sea surface temperature, sea surface temperatures will rise, as will sea level, precipitation extremes, and associated extreme weather events. If solar activity falls deeply quite fast and for the next many decades, we are in for trouble as the ocean will cool, the rains will diminish, the lands will dry out, allowing for the hot high UVI days in the SW US summer that also drive record temps, and the ocean emits less CO2 as the equatorial ocean temperature drops below the outgas temperature range.
I believe humanity will really be tested under those conditions, whereas a progressive warming trend will imply more of the same kind of weather we’ve had over the last 30 years, with more outgassed CO2, all net beneficial for greening the earth and feeding us, as it has already. I think ancient cultures have repeatedly risen then fallen on the feast or famine conditions brought on by climate extremes driven by long-duration solar changes.
The question is open whether humanity will successfully ride out a long cooling wave without population(s) collapsing as always appears to happen. Many grand ancient cultures grew and thrived during long-duration higher solar activity periods, only to face an impossible task of feeding such a large population into a subsequent deep cooler drier downturn, and cultures crashed.
Things won’t change so much if NASA is wrong about low solar activity in SC25. Stay tuned on that.
Arguing against human emissions is futile and wasteful, counter-productive to real societal needs. Emissions can’t cause the wind to blow let alone change the climate.
Integrate all the information presented here to know our future depends on the sun.
I believe your grandchildren’s children will be taught the sun controls the climate, top to bottom.
When data are viewed through the prism of as many unorthodox, ad hoc assumptions as are invoked here, the likelihood of getting a misleading conclusion becomes high.
ENSO is very far from being quasi-periodic and shows very little power density near the ~11 yr Schwabe cycle of sunspots. Its power spectrum displays no distinctive harmonics and bispectrum analysis shows no significant phase-locking. Despite the unwarranted restriction to post-1956 data, there simply is no real consistency in what happens in the ENSO record near the nadir of that cycle. Positive and negative anomalies in the ENSO record occur quite irregularly and strong El Nino episodes are by no means restricted to any phase of the Schwabe cycle. Nor is there any demonstrated success in rigorously predicting such episodes.
All in all, the entire line of reasoning here smacks of mere visual impressions being mistaken for solid scientific evidence–an all too common fault in the “soft” sciences when dealing with hard problems.
I can’t disagree all that much except to say when you base ENSO activity predictions as I do on changes and levels of TSI, things work out, as opposed to the more common probabilistic mechanism-free method.
It is very hard to express all this concisely and definitively as a proof, so I do like the use of odds that do help for that reason, but doing that won’t tell when to expect the next top-of-cycle El Ninos or why one might falter, unless solar indices are used.
ENSO is very far from being quasi-periodic and shows very little power density near the ~11 yr Schwabe cycle of sunspots.
1sky1, what kind of crap are you spouting. What you say is the equivalent of saying that a violin string does not resonate at the frequency with which the violinist plays. No! Its frequency depends on the length, diameter and tension of the string. Only the amplitude is dependent on the violinist.
Similarly, the frequency of a ENSO is governed by the internal dynamics of the ocean flows. The amplitude of the system response however is a result forcing. In this case, Javier has shown very convincingly that the solar cycle has a strong effect on the amplitude of ENSO.
sky isn’t off base so much. The fact is the sunspot record is not synchronized at 11 years, it’s an average, and furthermore even if the average for each cycle could somehow be the same, the development through each cycle to attain the same average is always going to be different for every cycle as sunspot emergence is random.
The great correlations and statistical properties between SN, F10.7, and TSI have to be weighed against the knowledge that each cycle’s development is unique and as of this date unpredictable, other than the general cycle size per Leif S, and the fact that for a given sunspot number at any given time, the TSI up to that same number can develop as differently as there are combinations of sunspot number leading to that same average just due to the basic aperiodic nature of sunspot activity.
The fact that sunspot cycle peak timing(s) vary so much is important. SC24 peaked in late 2014, and now the ocean heat generated by that TSI spike is just about fully offset by the solar decline since then. Imagine if the peak was earlier, like just about everyone thought prior to 2012, the ocean would’ve had more time to cool by now. Timing is as important as level and duration, as solar effects are layered and time-dependent.
For those reasons the power density at 11 years is low and the R values of my TSI-ENSO indice cross-correlations are low. The signal is there regardless. I find the ocean warms and cools at nearly mirror-image rates, so the real variability of interest here is the sun’s.
I also agree that the ENSO has an internal dynamic, based on circulation and upwell time. There is a lag between sunspot activity and maximum TSI from that activity. Then there is some lag at the ocean after peak TSI at the top of the cycle from circulation and upwell time. The lags probably have some sun strength-dependent variability too.
But in almost all cases there is a ‘solar cycle onset’ El Nino, an initialization point, and post-cycle maximum El Nino(s) after monthly solar exceeds SN>94, F10.7>120, SORCE TSI>1361.25.
This is why my data-driven system is better at predictions than the probabilistic method, because it’s based on available incoming solar energy, which is tracked daily.
If you had any comprehension of the dependence of the power density upon the nature of the dynamics involved, you would not spout any nonsensical analogies with a violin string. There is not the slightest indication of any resonant response in the featureless, broad-band power spectrum of ENSO.
Obviously or this article would be unnecessary.
That is not correct. See White & Liu (2008) in the linked file.
Consistency is not a requirement for a real significant effect. We are dealing with a complex non-linear phenomenon.
You are welcome to your opinion, but for what you say in your comment I doubt it is based on reading the article rather than just looking at the pictures.
Not only have I read your posted article, but I’ve done, more than a decade ago, all the rigorous analyses of the relationship between ENSO and sunspot signals that you seem to have never performed.
It requires far more than just peaks in the raw periodogram to establish the presence of any significant harmonics in a broad-band process. What White and Liu find through an unconventional partial variance analysis is harmonics not of the ENSO signal , but of a spatial decomposition of Pacific SSTs between 40S and 40N in latitude. Their supposed applicability to ENSO rests entirely upon a visual phasing comparison of subsequent simulations in the time domain.
I correctly assumed there was no need to do them since multiple researchers must have done them many times, and the lack of reports indicates a lack of positive result. I found a more original approach to the problem than beat the beaten path and then give up.
Their figure 4 shows that the sum of the 3.6- and 2.2-yr harmonics thus obtained is very similar to the Niño 3 SST index. That’s one of the main conclusions from a paper cited 66 times according to Google Scholar and that to my knowledge has not been refuted.
The “more original approach to the problem” here is simplistic and lacks grounding in signal analysis fundamentals.
All that White & Liu really managed to demonstrate is a wholly unsurprising, weakly correlated phase relationship between ENSO3 and the harmonics of a much-wider-area Pacific SST index, encompassing the equator. They present no evidence of cross-spectral coherence with solar activity itself. Their attribution of those harmonics to solar activity is based entirely upon finding a ~11yr spectral peak in their quaint “partial variance” analysis. There is no such peak, nor such harmonics evident in rigorously estimated Nino 3.4 power spectra, although raw periodograms may haphazardly show some. That this solar attribution hasn’t been refuted in a field rife with unchallenged speculation and analytic ineptitude provides more a sociological than a scientific indication.
Javier,
The lag in Figure 5 appears quite clear to me. Have you looked at the cross-correlation between the solar and temperature indeixes?
Also, the first period of La Nina is deeper, where the slope of the solar cycle graph is steeper, the weaker second phase at the end of the cycle the slope is lower. And the El Nino’s are (lagged) after there is very little change of slope during the peak of the solar cycle. So does the gradient of the solar cycle curve relate to the intensity? Once a lag is taken out?
What about integration? A cumulative effect?
Just thoughts. They would be things I would look at.
Nice bit of work, very interesting.
Thank you.
Several scientists have looked at that. Particularly K.K. Tung has several articles. For example:
Zhou, J. and Tung, K.K., 2010. Solar cycles in 150 years of global sea surface temperature data. Journal of Climate, 23(12), pp.3234-3248.
https://journals.ametsoc.org/doi/pdf/10.1175/2010JCLI3232.1
It doesn’t appear so. The intensity of ENSO appears to depend mainly on factors other than solar activity.
What about looking at the cumulative totals of energy in vs energy out over time? It would seem to me that a phenomenon like ENSO intensity, especially the really big Nino would reflect the residual energy stored in the system over time. Looking at the graph, the particularly large Ninos of 98 and 16 follow periods of relatively small Ninos and deeper La Nina’s. Makes me wonder if the energy driving larger warming events is the result of the ocean belching out accumulated energy stored over a period of 10-15 years. So while the cycle effects dictate the when, the energy stored determines the size.
a phenomenon like ENSO intensity, especially the really big Nino would reflect the residual energy stored in the system over time.
I started out thinking this also, that the ’98 ENSO derived from the previous cycle, but it absolutely conforms to the solar TSI impulse at the start of the cycle, after the low solar minimum energy has reduced tropical evaporation thus clearing the skies, re CP OLR:
The ’98 El Nino was a solar impulse response that could not be sustained by the sun and died out, as the ocean didn’t hold the new heat long, kicking it out as fast as it was deposited, ending in a La Nina.
The ’16 El Nino resulted from the year-to-year build-up after the SC24 cycle onset ENSO of ’09/10 of sea surface temperature from year-to-year TSI growth to the top of the cycle, and ended when TSI was no longer sufficient to maintain further warming as defined in my work.
Sorry Javier, I meant cross-correlation of the two curves in your Figure 5 above ie the sunspot curve and the ENSO index.
I did digitise your data and try it, the peak correlation is R=0.52 with a lag of 0.18 of your normalised solar cycle.
If you plot the indices looped round again though, the end point of the ENSO curve gets very high amplitude when the cycle goes to zero. So its like the El Nino ramps on (lagged) after the sunspot rate of change gets small.
That’s an interesting observation. Corresponds to a variable lag of ~ 2 years. Another observation is that variability appears to go up during the declining phase of the solar cycle, with ENSO being more consistent during the low, increasing and high activity phases.
If you are interested your questions were addressed in my several posts above.
My empirically-based data-driven analytical work covers everything necessary to know how the sun controls the climate.
Javier,
Figure 2 only shows a significant correlation between La Nina activity and the solar cycle. I believe that it does not show a convincing long-term correlation between El Nino activity and the solar cycle. Hence,
I am in agreement with you about the likelihood of a La Nina event in late 2020 to early 2021. However, the El Nino event you are claiming for 2018/19 was only a borderline strength event. The Australian BMO listed the 2018/19 event as a marginal El Nino, at best. Most of the other climate groups had to invoke the Modoki definition to get it over the line. I am sticking with my prediction that we will have a moderate to strong El Nino event sometime between July 2019 and January 2020. Obviously, since it now early September 2019, the first half of this period has been ruled out. But I am still standing by my prediction.
As you know, I have proposed that La Nina and El Nino events are produced by different mechanisms. I agree with you that the likelihood of a La Nina event is controlled by the level of solar activity as they both exhibit the same historical trends. However, I believe that El Nino events are initiated by lunar tidal events.
The increase in galactic radiation has a huge impact on pressure changes abover the polar circles. Ionization by GCR causes a local strong temperature rise in the stratosphere.
So Javier thinks he has predicted the El Nino episode that normally peaks just after sunspot minimum, and has more recently peaked a year or more after sunspot minimum, but sunspot minimum hasn’t even happened yet. I’d call that fairly fatal confirmation bias.
Interesting article which is connected to my own ENSO prediction using Artificial Neural Networks. I’m still working on my ENSO predictions improving my calculations by improving my program, in-data and by including additional mathematical methods. I recently got new results where I estimate correlations. I found that the in-data with best correlations to ENSO were the aa-index, followed by tidal influence and QBO. The aa-index gave the best correlation which is consistent with that solar activity has large influence over ENSO. QBO variation is driven largely driven by lunar gravitation where the base QBO frequency is equal to the 18.6 year lunar cycle divided by 8. Right now I’m halfway in my work, which has as its aim to create accurate ENSO prediction and to explain the underlining forces which drives ENSO variations.