Basil Copeland and Anthony Watts
Image from NASA GSFC
Many WUWT readers will remember that last year we presented evidence of what we thought was a “solar imprint” in globally averaged temperature trends. Not surprisingly, given the strong interest and passion in the subject of climate change and global warming, our results were greeted with both praise and scorn. Some problems were pointed out in our original assessment, and other possible interpretations of the data were suggested. Some WUWT readers have wondered whether we would ever follow up on this.
We have been quietly working on this, and having learned much since our initial effort, are as persuaded as ever that the basic premise of our original presentation remains valid. We have tried out some new techniques, and have posted some preliminary trials on WUWT in the past few months, here, and here.
However, questions remain. Since a lot of bright and capable people read WUWT, rather than wait until we thought we had all the answers, we have decided to present an update and let readers weigh in on where we are at with all of this. We have, in fact, drafted a paper that we might at some point submit for peer review, when we are more comfortable with some of the more speculative aspects of the matter. What follows is taken from that draft, with some modification for presentation here.
For those that prefer to read this in printed form, a PDF of this essay is available for download here
Introduction
Evidence of decadal and bidecadal variations in climate are common in nature. Classic examples of the latter include the 20 year oscillation in January temperature in the Eastern United States and Canada reported by Mock and Hibler [1], and the bidecadal rhythm of drought in the Western High Plains, Mitchell, Stockton, and Meko [2], and Cook, Meko, and Stockton [3]. Other examples include a bidecadal (and pentadecadal) oscillation in the Aleutian Low, Minobe [4]; rainfall and the levels of Lake Victoria, East Africa, Stager et al. [5]; and evidence from tree rings along the Russian Arctic, Raspopov, Dergachev, Kolstrom [6], and the Chilean coast, Rigozo et al. [7].
Evidence of decadal or bidecadal oscillations in temperature data, however, especially upon a global scale, has proven to be more elusive and controversial. Folland [8] found a spectral peak at 23 years in a 335 year record of central England temperatures, and Newell et al. [9] found a 21.8 year peak in marine air temperature. Brunetti, Mageuri, Nanni [10] have reported evidence of a bidecadal signal in Central European mean alpine temperatures. But the first to report bidecadal oscillations – of 21 and 16 years – in globally averaged temperature were Ghil and Vautard [11]. Their results were challenged by Eisner and Tsonis [12], but were later taken up and extended by Keeling and Whorf [13, 14].
No less unsettled is the issue of attribution. Currie [15], examining U.S. temperature records, reported spectral peaks of 10.4 and 18.8 years, attributing the first to the solar cycle, and the latter to the lunar nodal cycle. In the debate over the bidecadal drought cycle of the Western High Plains, Mitchell, Stockton, and Meko [2] concluded that the bidecadal signal was a solar phenomenon, not a lunar one. Bell [16, 17] and Stockton, Mitchell, Meko [18] attributed the bidecadal drought cycle to a combined solar and lunar influence, as did Cook, Meko, and Stockton [3]. Keeling and Whorf [13], working with globally averaged temperature data, reported strong spectral peaks at 9.3, 15.2, and 21.7 years. Eschewing a simpler combination of solar and lunar influences, they proposed a complex mechanism of lunar tidal influences to explain the evidence [14].
The past decade has seen only sporadic interest in the question of whether decadal and bidecadal variations in climate have a solar or lunar attribution, or some combination of the two. Cerveny and Shaffer [19] and Treloar [20] report evidence of tidal influences on the southern oscillation and sea surface temperatures; Yndestad [21, 22] and McKinnell and Crawford [23] attribute climate oscillations in the Arctic and North Pacific to the 18.6 year lunar nodal cycle. But interest in discerning an anthropogenic influence on climate has largely eclipsed the study of natural climate variability, at least on a global scale. There continue to be numerous reports of decadal or bidecadal oscillations in a variety of climate metrics on local and regional scales, variously attributed to solar and or lunar periods [3-7, 10, 19-27], but little has been done to advance the state of knowledge of lunar or solar periodic cycles on globally averaged temperature trends since the final decade of the 20th Century.
Besides the shift in interest to discerning an anthropogenic influence on global climate, the lack of agreement on any kind of basic physical mechanism for a solar role in climate oscillations, combined with the apparent lack of consistency in the relation between solar cycles and terrestrial temperature trends perhaps has made this an uninviting area of research. The difficulty of attributing temperature change to solar influence has been thoroughly surveyed by Hoyt and Schatten [28]. In particular, there are numerous reports of sign reversals in the relationship between temperature and solar activity in the early 20th century, particularly after 1920 [28, pp 115-117]. More recently, Georgieva, Kirov, and Bianchi [29] surveyed comprehensively the evidence for sign reversal in the relationship between solar and terrestrial temperatures, and suggested that these sign reversals are related to a long term secular solar cycle with solar hemispheric asymmetry driving the sign reversals. Specifically, they argue that there is a double Gleissberg cycle in which during one half of the cycle the Southern solar hemisphere is more active, while during the other half of the cycle the Northern solar hemisphere is more active. They argue that this solar hemispheric asymmetry is correlated with long term terrestrial climate variations in atmospheric circulation patterns, with zonal circulation patterns dominating in the 19th and early 20th century, and meridional circulation patterns dominating thereafter (see also [30] and [31]).
In our research, we pick up where Keeling and Whorf [13, 14] leave off, insofar as documenting decadal and bidecadal oscillations in globally averaged temperature trends is concerned, but revert to the explanation proposed by Bell [16] and others [3, 18], that these are likely the result of a combined lunisolar influence, and not simply the result of lunar nodal and tidal influences. We show that decadal and bidecadal oscillations in globally averaged temperature show patterns of alternating weak and strong warming rates, and that these underwent a phase change around 1920. Prior to that time, the lunar influence dominates, while after that time the solar influence dominates. While these show signs of being correlated with the broad secular variation in atmospheric circulation patterns over time, the persistent influence of the lunar nodal cycle, even when the solar cycle dominates the warming rate cycles, implicates oceanic influences on secular trends in terrestrial climate. Moreover, while analyzing the behavior of the secular solar cycle over the limited time frame for which we have reasonably reliable instrumental data for measuring globally averaged temperature should proceed with caution, if the patterns documented here persist, we may be on the cusp of a downward trend in the secular solar cycle in which solar activity will be lower than what has been experienced during the last four double sunspot cycles. These findings could influence our expectations for the future regarding climate change and the issue of anthropogenic versus natural variability in attributing climate change.
In our original presentation, we utilized Hodrick-Prescott smoothing to reveal decadal and bidecadal temperature oscillations in globally averaged temperature trends. While originally developed in the field of economics to separate business cycles from long term secular trends in economic growth, the technique is applicable to the time series analysis of temperature data in reverse, by filtering out short term climate oscillations, isolating longer term variations in temperature.
For the mathematically inclined, here is what the HP filter equation looks like, courtesy of the Mathworks
The Hodrick-Prescott filter separates a time series yt into a trend component Tt and a cyclical component Ct such that yt = Tt + Ct. It is equivalent to a cubic spline smoother, with the smoothed portion in Tt.
The objective function for the filter has the form
where m is the number of samples and λ is the smoothing parameter. The programming problem is to minimize the objective over all T1, …, Tm. The first sum minimizes the difference between the time series and its trend component (which is its cyclical component). The second sum minimizes the second-order difference of the trend component (which is analogous to minimization of the second derivative of the trend component).
For those with an electrical engineering background, you could think of it much like a bandpass filter, which also has uses in meteorology:
Outside of electronics and signal processing, one example of the use of band-pass filters is in the atmospheric sciences. It is common to band-pass filter recent meteorological data with a period range of, for example, 3 to 10 days, so that only cyclones remain as fluctuations in the data fields.
(Note: For those that wish to try out the HP filter on data themselves, a freeware Excel plugin exists for it which you can download here)
When applied to globally averaged temperature, the HP filter works to extract the longer term trend from variations in temperature that are of short term duration. It is somewhat like a filter that filters out “noise,” but in this case the short term cyclical variations in the data are not noise, but are themselves oscillations of a shorter term that may have a basis in physical processes.
This approach reveals alternating cycles of weak and strong warming rates with decadal and bidecadal frequency. We confirm the validity of the technique using a continuous wavelet transform. Then, using MTM spectrum analysis, we analyze further the frequency of these oscillations in global temperature data. Using sinusoidal model analysis we show that the frequencies obtained using HP smoothing are equivalent to what are obtained using MTM spectrum analysis. In other words, the HP smoothing technique is simply another way of extracting the same spectral density information obtained using more conventional spectrum analysis, while leaving it in the time domain. This allows us to compare the secular pattern of temperature cycles with solar and lunar maxima, yielding results that are not obvious from spectral analysis alone.
Using the Hodrick-Prescott Filter to Reveal Oscillations in Globally Averaged Temperature
We use the open source econometric regression software gretl (GNU Regression, Econometrics, and Time Series) [34] to derive an HP filtered time series for the HadCRUT3 Monthly Global Temperature Anomaly, 1850:01 through 2008:11 [35].
Figure 1 is representative output in gretl for a series filtered with HP smoothing (λ of 129,000). In the top panel is the original series (in gray), with the resulting smoothed trend (in red). In the bottom panel is the cyclical component. In econometric analysis, attention usually focuses on the cyclical component. Our focus, though, is on the trend component in the upper panel, and in particular the first differences of the trend component. The first differences of a trend indicate rate of change.
By taking the first differences of the smoothed trend in Figure 1, we obtain the series (in blue) shown in Figure 2, plotted against the background of the original data (gray), and the smoothed trend (red).
What does this reveal? At first glance, we see an alternating pattern of decadal and bidecadal oscillations in the rate of warming, with a curious exception circa 1920-1930. We will return to this later. Concentrating for now on the general pattern, these oscillations in the rate of warming are representations, in the time domain, of spectral frequencies in the temperature data, with high frequency oscillations filtered out by the HP smoothing.
As evidence of this, Figure 3 presents the result of two Morelet continuous wavelet transforms, the first (in the upper panel) of the unfiltered HadCRUT3 monthly time series, and the second (in the lower panel) of results obtained with HP smoothing.
The wavelet transforms below a frequency of ~7 years (26.4 ≈ 84 months) are visually identical; the HP filter is simply acting as a low pass filter, filtering out oscillations with frequencies above ~7 years, while preserving the decadal and bidecadal oscillations of interest here. In the next section, we investigate these oscillations in further detail, supplementing our results from HP filtering with MTM spectrum analysis, and a sinusoidal model fit.
Frequency Analysis
Figure 4 is an MTM spectrum analysis of the unfiltered HadCRUT3 monthly global temperature analysis.
A feature of MTM spectrum analysis is that it distinguishes signals that are described as “harmonic” from those that are merely “quasi-oscillatory.” In MTM spectrum analysis a harmonic is a more clearly repeatable signal that passes a stronger statistical test of its repeatability. Quasi-oscillatory signals are statistically significant, in the sense of rising above the background noise level, but are not as consistently repeating as the harmonic signals.
The distinction between harmonic and quasi-oscillatory signals is well illustrated in Figure 4 by the two cycles that interest us the most – a “quasi-oscillatory” cycle with a peak at 8.98 years, and a “harmonic” signal centered at 21.33 years. Also shown are a harmonic, and a quasi-oscillatory cycle, of shorter frequencies, possibly ENSO related. The harmonic at 21.33 years in Figure 4 encompasses a range from 18.96 to 24.38 years, and the quasi-oscillatory signal that peaks at 8.93 years has sidebands above the 99% significance level that range from 8.53 to 10.04 years. These signals are consistent with spectra identified by Keeling and Whorf [13,14].
Figure 5 is an MTM spectrum analysis of the HP smoothed first differences.
The basic shape of the spectrum is unchanged, but it is now well above the background noise level because of the HP filtering. Attention is drawn in Figure 5 to four oscillatory modes or cycles because they correspond to the four strongest cycles derived from using the PAST (PAleontological STatistics) software [36] to fit a sinusoidal model to the HP smoothed first differences.
Shown in Figure 6, the sinusoidal fit results in periods of 20.68, 9.22, 15.07 and 54.56 years, in that order of significance. These periodicities fall within the ranges of the spectra obtained using MTM spectrum analysis, and yield a sinusoidal model with an R2 of 0.60.
Discussion
The first differences of the HP smoothed temperature series, shown in Figure 2 and Figure 6, show a pattern of alternating decadal and bidecadal oscillations in globally averaged temperature. From the sinusoidal model fit, these cycles have average frequencies of 20.68 and 9.22 years, results that are consistent with the MTM spectrum analysis, and with spectra in the results published by Keeling and Whorf [13, 14]. But to what can we attribute these persistent periodicities?
A bidecadal frequency of 20.68 years is too short to be attributed solely to the double sunspot cycle, and too long to be attributed solely to the 18.6 year lunar nodal cycle. There is indeed evidence of a spectral peak at ~15 years, which Keeling and Whorf combined with their evidence of a 21.7 year cycle to argue for attributing the oscillations entirely to the 18.6 year lunar nodal cycle.
But our evidence indicates that the ~15 year spectrum is much weaker, is not harmonic, and probably derives from the anomalous behavior of the spectra circa 1920-1930, something Keeling and Whorf could not appreciate with evidence only from the frequency domain. Especially in light of the evidence presented below, and because the bidecadal signal is harmonic, and readily discernible in the time domain representation of Figure 2 and Figure 6, we believe that a better attribution is the beat cycle explanation proposed by Bell [16], i.e. a cycle representing the combined influence of the 22 year double sunspot cycle and the 18.6 year lunar nodal cycle.
As for the decadal signal of 9.22 years, this is too short to be likely attributable to the 11 year solar cycle, but is very close to half the 18.6 year lunar nodal cycle, and thus may well be attributable to the lunar nodal cycle. Together, the pattern of alternating weak and strong warming cycles shown in Figure 2 and Figure 6 suggest a complex pattern of interaction between the double sunspot cycle and the lunar nodal cycle.
This complex pattern of interaction between the double sunspot cycle and lunar nodal maxima in relation to the alternating pattern of decadal and bidecadal warming rates is demonstrated further in Figure 6 with indicia plotted to indicate solar and lunar maxima. Since circa 1920, the strong warming rate cycles have tended to correlate with solar maxima associated with odd numbered solar cycles, and the weak warming rate cycles with lunar maxima.
Prior to 1920, the strong warming rate cycles tend to correlate with the lunar nodal cycle, with the weak warming rate cycles associated with even numbered solar cycles. The sinusoidal model fit begins to break down prior to 1870. Whether this is a reflection of the poorer quality of data prior to 1880, or indications of an earlier phase shift, we cannot say, though the timing would be roughly correct for the latter. But the anomalous pattern circa 1920, when viewed against the shift from strong warming rate cycles dominated by the lunar nodal cycle, to strong warming rate cycles dominated by the double sunspot cycle, has the appearance of a disturbance associated with what clearly seems to be a phase shift
These results agree with the evidence mustered by Hoyt and Schatten [28] and Georgieva, Kirov, and Bianchi [29] for a phase shift circa 1920 in the relationship between solar activity and terrestrial temperatures. However, we can suggest, here, that the supposed negative correlation between solar activity and terrestrial temperatures prior to 1920 rests on a misconstrued understanding of the data. As can be seen in Figure 6, the relationship between the change in the warming rate and solar activity is still positive, i.e. the warming rate is peaking near the peaks of solar cycles 10, 12, and 14, but at a much reduced level, indicative of the lower level of solar activity during the period. Indeed, for much of solar cycle 12, and all of solar cycle 14, the “warming” rate is negative, but the change in the warming rate is still following the level of solar activity, becoming less negative as solar activity increases, and more negative as solar activity decreases. Still, there is a strong suggestion in Figure 6 of a phase shift circa 1920 in which the relationship between solar activity and terrestrial temperatures changes dramatically before and after the shift. Before the shift, the lunar period dominates, and the solar period is much weaker. After the shift, the solar period dominates, and the lunar period becomes subordinate.
Speculating
To this point, we believe that we are on relatively solid ground in describing what the data show, and the likelihood of a lunisolar influence on global temperatures on decadal and bidecadal timescales. What follows now is more speculative. To what can we attribute the apparent phase shift circa 1920, evident not just in our findings, but as reported by Hoyt and Schatten [28] and Georgieva, Kirov, and Bianchi [29]? While the period of data is too short to do more than speculate, the periods before and after the phase shift appear to be roughly equivalent in length to the Gleissberg cycle.
Since 1920, we’ve had four double sunspot cycles with strong warming rates ending in odd numbered cycles. Prior to 1920, while the results are less certain at the beginning of the data period, there is a reasonable interpretation of the data in which we see four bidecadal periods dominated by the influence of the lunar cycle. These differences may be attributable to the broad swings in atmospheric “circulation epochs” discussed by Georgeiva, et al. [30], characterized either predominantly by zonal circulation, or meridional circulation. With respect to the period of time shown in Figure 6, zonal circulation prevailed prior to 1920, and since then meridional circulation has dominated. These “circulation epochs” may have persistent influence on the relative roles of solar and lunar influence in warming rate cycles.
While the role of variation in solar irradiation over the length of a solar cycle on the broad secular rise in global temperature is disputed, we are presenting here evidence primarily of a more subtle repeated oscillation in the rate of change in temperature, not its absolute level. As shown in Figure 2 and Figure 6, the rate of change oscillates between bounded positive and negative values (with the exception circa 1920 duly noted). Variations in solar irradiance over the course of the solar cycle are a reasonable candidate for the source of this variation in warming rate cycle. As WUWT’s “resident solar physicist”, Leif Svalgaard, has pointed out, variations in TSI over a normal solar cycle can only account for about 0.07°C of total variation over the course of a solar cycle. The range of change in warming rates shown in Figure 2 and Figure 6 are at most only about one-tenth of this, or about ~0.006°C to ~0.008°C. If anything, we should be curious why the variation is so small. We attribute this to the averaging of regional and hemispheric variations in the globally averaged data. On a regional basis, analysis [not presented here] shows much larger variation, but still within the range of 0.07°C that might plausibly be attributed to the variation in TSI over the course of a solar cycle.
So variations in solar irradiance over the course of the solar cycle are a reasonable candidate for the source of this variation in warming rate cycle. At the same time, the lunar nodal cycle may be further modulating this natural cycle in the rate of change in global temperatures. As to the degree of modulation, that may be influenced by atmospheric circulation patterns. With zonal circulation, the solar influence is moderated and the lunar influence dominates the modulation of the warming rate cycles. With meridional circulation, the solar influence is stronger, and the warming rate cycles are dominated by the solar influence.
At this writing, we are in the transition from solar cycle 23 to 24, a transition that has taken longer than expected, and longer than the transitions typical of solar cycles 16 through 23. Indeed, the transition is more typical of the transitions of solar cycles 10 through 15. If the patterns observed in Figure 6 are not happenstance, we may be seeing an end to the strong solar activity of solar cycles 16-23, and a reversion to the weaker levels of activity associated with solar cycles 10-15. If that occurs, then we should see a breakdown in the correlation between warming rate cycles and solar cycles at bidecadal frequencies, and a reversion to a dominant correlation between temperature oscillations and the lunar nodal cycle.
Interestingly, there was a lunar nodal maximum in 2006 not closely associated with the timing of decadal or bidecadal oscillations in globally averaged temperature. This could be an indication of a breakdown in the pattern similar to what we see in the 1920’s, i.e. the noise associated with a phase shift in the weaker warming rate cycles will occur at times of the solar maximum, and the stronger warming rate cycles will occur at times of lunar nodal maximum.
Repeating, there appear to be parallels between our findings and the argument of Georgieva et al. [29] of a relationship between terrestrial climate and solar hemispheric asymmetry on the scale of a double Gleissberg cycle. Solar cycles 16-23, associated as we have seen with increased solar activity, and strong correlations with the strong terrestrial warming rate cycles of bidecadal frequency, represent 8 solar cycles, a period of time associated with a Gleissberg cycle.
While the existence of Gleissberg length cycles is hardly challenged, their exact length and timing is subject to a debate we will not entertain here at any length. Javariah [37] on the basis of the disputed 179 year cycle of Jose [38] believes that a descending phase of a Gleissberg cycle is already underway, and will end with the end of a double Hale cycle comprising solar cycles 22-25.
While it is true that solar activity, as measured by SSN, is already on the decline, we would include the double Hale cycle 20-23 in the recent peak of solar activity, and not necessarily expect to see the bottom of the current decline in solar activity that quickly.
The issue here can perhaps be framed with respect to Figure 7 below:
Assuming we are on the cusp of a downward trend in solar activity that began circa 1990 according to Javariah, and will decline, say, to a level comparable to the trough seen in the early 1900’s, will it be a sharp decline, like that seen at the beginning of the 19th Century, or a more moderate decline like that seen at the beginning of the 20th Century? A naïve extrapolation might be to replicate the more gradual decline seen during the latter half of the 19th Century, suggesting a gradual decline in solar activity through solar cycle 31, i.e. for most of the 21st Century. And based on the prospect of a phase shift in the pattern of decadal and bidecadal warming rate cycles, the bidecadal cycle would come to be dominated by the influence of the lunar nodal cycle, and the influence of the solar cycle would be diminished, leading at least to a reduction in the rate of global warming, if not an era of global cooling.
This is a prospect worthy of more investigation.
Finally, while we readily concede that multidecadal projections are at best little more than gross speculation, in Figure 6 we have carried the sinusoidal model fit out to 2030, and in Figure 8 we use the sinusoidal model of rate changes to project temperature
anomalies through 2030. Assuming a simple projection of the sinusoidal model of rate changes persists through 2030, there would be little or no significant change in global temperature anomalies for the next two decades.
Looking carefully at the sinusoidal model, what we are seeing projected for 2010-2020 are a return to conditions similar to what the model shows for circa 1850-1860, with the period 1853-2020 representing a complete composite cycle of the four combined periods of oscillation. That is, 1853 is the first point at which the sinusoidal model is crossing the x-axis, and at 2020 the model again crossing the x-axis and beginning to repeat a ~167 year cycle. In terms of solar cycle history, that corresponds to a return to conditions similar to solar cycles 10-15, with another phase shift reversing the phase shift of ~1920. If these broad, long term secular swings in solar activity and global atmospheric conditions and temperature anomalies are not random, but reflect solar-terrestrial dynamics that play out over multidecadal and even centennial time-scales, then the early 21st Century may yield a respite from the global warming of the late 20th Century.
Conclusion
There is substantial and statistically significant evidence for decadal and bidecadal oscillations in globally averaged temperature trends. Sinusoidal model analysis of the first differences of the HP smoothed HadCRUT3 time series reveals strong periodicities at 248.2 and 110.7 months, periodicities confirmed as well with MTM spectrum analysis.
Analyzing these periodicities in the time domain with the first differences of the HP smoothed HadCRUT3 time series reveals a pattern of correlation between strong warming rate cycles and the double sunspot cycle for the past four double sunspot cycles. Prior to that, with a phase shift circa 1920, the strong warming rate cycles were dominated by the timing of the lunar nodal cycle.
We suggest that this reversal may be related to a weaker epoch of solar activity prior to 1920, and that we may on the cusp of another phase shift associated with a resumption of such weakened solar activity.
If so, this may result in a reduction in the rate of global warming, and possibly a period of global cooling, further complicating the effort to attribute recent global warming to anthropogenic sources.
References
[1] Mock, SJ, Hibler III, WD. The 20-Year oscillation in Eastern North American temperature records. Nature. 1976; 261: 484–486.
[2] Mitchell, Jr., JM, Stockton, CW, Meko, DM. Evidence of a 22-year rhythm of drought in the western United States related to the Hale solar cycle since the 17th century. In: McCormac, BM, Seliga, TA, editors. Solar-terrestrial influences on weather and climate. Dordrecht, Holland: Kluwer Acad. Publ; 1979. p.125-143.
[3] Cook, ER, Meko, DM, Stockton, CW. A new assessment of possible solar and lunar forcing of the bidecadal drought rhythm in the western U.S. Journal of Climate. 1997; 10:1343-1356.
[4] Minobe, S. Resonance in bidecadal and pentadecadal climate oscillations over the North Pacific: Role in climate regime shifts. Geophysical Research Letters. 1999; 26(7): 855-858.
[5] Stager, JC. Ryves, D. Cumming, BF. Meeker, LD. Beer J. Solar variability and the levels of Lake Victoria, East Africa, during the last millennium. Journal of Paleolimnology. 2005; 33(2):243-251.
[6] Raspopov, OM. Dergachev, VA. Kolstrom, T. .Periodicity of climate conditions and solar variability derived from dendrochronological and other palaeoclimatic data in high latitudes. Palaeogeography, Palaeoclimatology, Palaeoecology. 2004; 209: 127-139.
[7] Rigozo, NR. Nordeman, DJR. Echer, E. Vieira, LEA. Echer, MPS. Prestes, A. Tree-ring width wavelet and spectral analysis of solar variability and climatic effects on a Chilean cypress during the last two and a half millennia. Climate of the Past Discussions. 2005; 1:121-135.
[8] Folland, CK. Regional-scale interannual variability of climate – a north-west European perspective. Meteorological Mazazine. 1983; 112: 163-183.
[9] Newell, NE. Newell, RE. Hsiung, H. and Zhongxiang, W. Global marine temperature variation and the solar magnetic cycle. Geophysical Research Letters. 1989; 16(4): 311–314.
[10] Brunetti, M. Maugeri, M. Nanni, T. Study of the solar signal in mean Central Europe temperature series from 1760 to 1998. Il Nuovo Cimento C. 2003; 26(3): 287-295.
[11] Ghil, M. Vautard, R. Interdecadal oscillations and the warming trend in global temperature time series. Nature. 1991; 350: 324-327.
[12] Eisner, JB. Tsonis, AA. Do bidecadal oscillations exist in the global record? Nature. 1991; 353: 551-553
[13] Keeling CD. Whorf TP. Decadal oscillations in global temperature and atmospheric carbon dioxide. In: Natural climate variability on decade-to-century time scales. Climate Research Committee, National Research Council. Washington, DC: The National Academies; 1996; pp. 97-110.
[14] Keeling CD. Whorf TP. Possible forcing of global temperature by the oceanic tides. Proceedings of the National Academy of Sciences. 1997; 94(16):8321-8328.
[15] Currie, RG. Luni-solar 18.6- and solar cycle 10-11-year signals in USA air temperature records. International Journal of Climatology. 1993; 13(1): 31-50.
[16] Bell, PR. The combined solar and tidal influence on climate. In: Sofia, SS, editor. Variations of the Solar Constant. Washington, DC: National Aeronautics and Space Administration, 1981; p. 241–256..
[17] Bell, PR. Predominant periods in the time series of drought area index for the western high plains AD 1700 to 1962. In: Sofia, SS, editor. Variations of the Solar Constant. Washington, DC: National Aeronautics and Space Administration; 1981; p. 257–264.
[18] Stockton, CW. Mitchell Jr., JM. Meko, DM. 1983: A reappraisal of the 22-year drought cycle. In: McCormac, BM., editor. Solar-Terrestrial Influences on Weather and Climate. Boulder, CO: Colorado Associated University Press; 1983, p. 507–515.
[19] Cerveny, RS. Shaffer, JA. The Moon and El Nino. Geophysical Research Letters. 2001; 28(1): 25-28.
[20] Treloar, NC. Luni-solar tidal influences on climate variability. International Journal of Climatology. 22(12): 2002; 1527-1542.
[21] Yndestad, H. The Arctic Ocean as a coupled oscillating system to the forced 18.6 year lunar gravity cycle. In: Tsonis, AA, Elsner, JB, editors. Nonlinear Dynamics in Geosciences. New York, Springer: 2007; p. 281-290.
[22] Yndestad, H. The influence of the lunar nodal cycle on Arctic climate. ICES Journal of Marine Science. 2006; 63(3): 401-420.
[23] McKinnell, SM. Crawford, WR. The 18.6-year lunar nodal cycle and surface temperature variability in the northeast Pacific. Journal of Geophysical Research. 2007; 112, C02002, doi:10.1029/2006JC003671
[24] Kasatkina, EA. Shumilov, OI. Krapiec, M. On periodicities in long term climatic variations near 68° N, 30° E. Advances in Geosciences. 2007; 13: 25-29.
[25] Yimin, Z. Xiuqun, Y. Joint propagating patterns of SST and SLP anomalies in the North Pacific on bidecadal and pentadecadal timescales. Advances in Atmospheric Sciences. 2003; 694-710.
[26] W. H. Berger, WH. von Rad, U. Decadal to millennial cyclicity in varves and turbidites from the Arabian Sea: hypothesis of tidal origin. Global and Planetary Change. 2002; 34: 313-325.
[27] Zhao, J. Han, Y.-B. Li, Z.-A. The Effect of Solar Activity on the Annual Precipitation in the Beijing Area. Chinese Journal of. Astronomy and. Astrophysics. 2004; 4: 189–197.
[28] Hoyt DV, Schatten, KH. The Role of the Sun in Climate Change. New York: Oxford Univ. Press; 1997.
[29] Georgieva, K. Kirov, B. Bianchi, C. Long-term variations in the correlation between solar activity and climate. Memorie della Società Astronomica Italiana (Journal of the Italian Astronomical Society). 2005; 76(4): 965-968.
[30] Georgieva, K. Kirov, B. Tonev, P. Guineva, V. Atanasov, D. Long-term variations in the correlation between NAO and solar activity: The importance of north–south solar activity asymmetry for atmospheric circulation. Advances in Space Research. 2007; 40(7): 1152-1166.
[31] Georgieva, K. Solar dynamics and solar-terrestrial Influences. In: Maravell, N. ed. Space Science: New Research. New York: Nova Science Publishers. 2006; p. 35-84.
[32] Hodrick, R. Prescott, E. Postwar US business cycles: an empirical investigation. Journal of Money, Credit and Banking. 1997. 29(1): 1-16. Reprint of University of Minneapolis discussion paper 451, 1981
[33] Maravall, A. del Río, A. Time aggregation and the Hodrick-Prescott filter. Banco de España Working Papers 0108, Banco de España. 2001. Available: http://www.bde.es/informes/be/docs/dt0108e.pdf[34] Cottrell, A., Lucchetti, R. Gretl User’s Guide. 2007. Available: http://ricardo.ecn.wfu.edu/pub//gretl/manual/en/gretl-guide.pdf
[35] Brohan, P. Kennedy, J. Harris, I. Tett, S. Jones, P. Uncertainty estimates in regional and global observed temperature changes: a new dataset from 1860. Journal of Geophysical Research. 2006; 111: D12106, doi:10.1029/2005JD006548. Data retrieved at: http://hadobs.metoffice.com/hadcrut3/diagnostics/
[36] Hammer, Ø. Harper, D. Ryan, P. PAST: Paleontological Statistics Software Package for Education and Data Analysis. Palaeontologia Electronica. 2001; 4(1): 9pp. http://palaeo-electronica.org/2001_1/past/issue1_01.htm
[37] Javariah, J. Sun’s retrograde motion and violation of even-odd cycle rule in sunspot activity. 2005; 362(4): 1311-1318.
[38] Jose, P. Sun’s motion and sunspots. Astronics Journal. 1965; 70: 193-200.
Discover more from Watts Up With That?
Subscribe to get the latest posts sent to your email.
I did not have time to go through it throughly yet but I wondered how you ensured that the “bandpass filter” does not introduce ripples into the data. When a applying window function to a high frequency data set it inevitably creates some ripples.
See http://en.wikipedia.org/wiki/Window_function for examples of what windowing can do to data. You end up with side lobes. How did you test to ensure that the oscillations you observe are real and not induced by spectral leakage?
I’ve done a little of this kind of analysis before and have found some of the same repeating cycle timelines.
They seem to be slightly off the timelines one would expect with the solar cycle.
I don’t know if there is a mathematical theorem to describe this but perhaps one should actually expect the spectral analysis peaks to be slightly off the solar cycle timelines.
Instead of an 11 year peak, maybe the strongest peak is at 9 years. Maybe the Hale cycle peak is at 18.6 years rather than 22. If one is trying to detect a signal from 11-year peak to 11-year peak of the cycle, this signal will be less strong than an offset of (3/4 of the peak) to peak.
The problem is the length of the solar cycles are not regular so there will be some drift in the off-set idea timelines. I’ve found the biggest cycle in the temperature series actually occurs at 25 years.
Just to illustrate, I did this chart before on Sunspot Numbers.
http://img269.imageshack.us/img269/3349/ssnspectral.png
I think the main point is that the solar cycle influence on temperatures may not show up exactly at the timelines expected with this kind of analysis.
I have another paper you might want to look at (this comes from before tree-rings took over his research).
http://holocene.meteo.psu.edu/shared/articles/MannLees1996.pdf
Wow! This is serious. I’ve got some things I just have to do outside – weeds, garden, horses – real, sun, wind, weather things. But this is the sort of post that resonates with me so as someone once said, “I’ll be back.”
What a paper! I also need more time to read it. Morlet is spelled without “e” before “let”.
Jeremy, if they actually did Morlet transform, it is based on a Gaussian-modulated plane waves. The Gaussian is the smoothest localized distribution: its Fourier transform is another Gaussian with width determined by a saturated “uncertainty relation”.
I wouldn’t call Gaussians – either in frequency or time coordinates – “ripples”. They’re very different from rectangular windows etc. For purely periodic time series, the Morlet transform always reduces the amplitudes for “wrong” frequencies (by the Gaussian factor): it can never enhance them which seems to solve your worry about the fake ripples.
In this piece, given its controversial nature, I wouldn’t side step oceanic oscillations as the source of global temperature change (both the noise and the trend). I would then speculate what affect the lunisolar source would have on these oscillations. It is too much of a leap for me to contemplate a direct route from temperature to lunisolar and leaves the paper open to harsh, and in my mind, deserved criticism.
Here is a mild one. It is possible that the lunisolar source is going in and out of phase with oceanic oscillations, and you have identified that phenomenon. Nature’s cycles go in and out of phase with each other all the time and present false correlations to those who study these cycles. Have you done that here?
One problem with an kind of analysis looking for periodic effects is that a-periodic effects like volcanoes can serious mess with and impede efforts to identify cycles in the data. If you lag by about 7 or 8 months and multiply by −2.216 or −2.948 the AOD data here:
http://data.giss.nasa.gov/modelforce/strataer/tau_line.txt
Will give you a good volcano signal.
If you analyze the data ~after~ removing volcano effects, solar and others should be more clearly visible.
Tho variations in TSI have been dismissed as insignificant in global temperature variation, figure 7 seems remarkably similar in amplitude (at this scale) with that variation.
In first paragraph “greeting” should be “greeted”.
My first impression also was how do you factor in other known cycles (PDO, etc.). Or, better, how do you account for them. At least you should discuss them.
My initial impression of the HP filter is that it is very simplistic. As lambda-> inf, you get back L_t, the least-squares best regression line, while lambda=0 gives back your data y_t. In general, aren’t you going to get back something akin to alpha*y_t+(1-alpha)*L_t for alpha (in [0,1]) some function of lambda?
There doesn’t seem to be anything in the filter that addresses the cyclic nature of the data. It’s just a smoothing filter. Also, I am concerned about the analysis of the first differences of the trend component. The filter, after all, penalizes those differences. So in effect, you first adjust those differences in some way you think is appropriate (by playing with lambda), then you analyze them. Seems like a potential alorithmic “conflict of interest” if there is such a thing.
Well done fellows!
“…HP filter works to extract the longer term trend from variations in temperature that are of short term duration…somewhat like a filter that filters out “noise,” but in this case the short term cyclical variations in the data are not noise, but are themselves oscillations of a shorter term that may have a basis in physical processes.”
Wish we could tatoo this to the foreheads of statisticians.
“the HP smoothing technique is simply another way of extracting the same spectral density information obtained using more conventional spectrum analysis, while leaving it in the time domain.”
Insightful. Halfway thru and believe a second or third reading will be required.
Figure 7 is the open door to oceanic oscillating cycles as either a criticism if you ignore this essential piece, or an opportunity to talk about these issues in your speculation section, as the graph looks very SSTish familiar.
Good gawdamighty. I meant figure 6, but let me double check again….yes, figure 6. I should just stop posting till I recover from a really good party last night.
“I wouldn’t side step oceanic oscillations as the source of global temperature change (both the noise and the trend).”
Oceans are, I believe, greatly misrepresented in many calculations of both temperature and CO2. While ocean surface (say to 300m) responds quickly to surface changes, the great bulk of the ocean will respond much more slowly. It wouldn’t surprise me to learn that the vast bulk of ocean water is *still* recovering from the LIA in temperature. But the recovery will be in hundredths of a degree per decade. Same with CO2 content. If you have a surface layer that changes 1 degree, you have a certain change in dissolved CO2. But if you have a million times more volume of water than changes only a hundredth of a degree, you have a greater overall change in dissolved CO2. And you can have a situation where the upper layers are cooling, gaining CO2 and the abyssal plains still warming and losing CO2.
Say you have a situation where the surface is warm for several centuries (MWP), then you cool for a few centuries (LIA) … I believe the deep oceans cool slowly in response albeit very slowly. Now things warm up as they have now and the deep responds to that too, but it has only been about a century.
Surface temperatures are now cooling and we see lower SST anomalies but is the surface still warmer than the average surface temperature over the time taken for the deep ocean to respond? If so, then the deep oceans are still warming. If not, then the deep will begin cooling. I speculate that it can take several centuries for the deep ocean to respond completely to a change in surface temperatures and if the surface were to increase 1 degree in a step change and remain there solid as a rock, it would take more than a century for the entire ocean to adjust to that. So if the surface increased, we could see increasing CO2 for several hundred years until that process is complete. This would tend to be supported by the great lag between temperature chance and CO2 response.
Now in relation to this article … I believe the surface is more responsive to these changes but there should be a considerable damping impact from the deep. The deep acting as a giant “heat capacitor” connected to the surface through a resistance that represents the relative exchange between the deep and surface layer waters.
Cyclical changes in winds can greatly change that “resistance” by increasing or decreasing upwelling from deep water to the surface. What you might be observing is changes in heat exchange from deep ocean to surface in response to cyclical wind pattern changes.
As already noted, removing the volcanic noise signals would be helpful.
And as noted by Pamela Gray, if you’re looking for some plausible way for lunar cycles to couple with temperature cycles, tidal/temperature interactions might be responsible for cyclic changes in ocean circulation that periodically change the mixing of deep and surface water. This should effect the temperature noise, but not the long-term trend – that is, it might easily mask GHG warming for periods like the last decade 😉
I’ll repeat here my initial comments on an earlier version of this post they sent to me:
In general, there is a bit too much cyclomania for my my taste. And I would EXPECT the 11-yr cycle in TSI to have a clear signal [as some people actually do find] and not a ~22 year variation. In addition, the lunar period remains unexplained. There must be 2000 such papers out there. You paper looks like number 2001.
As you point out the real problem is where the 0.07 K TSI related variation is [although many people claim they see an 11-year variation]. My main point is really that there are just too many studies that show contradictory results and , but all with overwhelming statistical significance.
I’m short of time right now, so I’ll just take a quick stab at responding to Pamela about ocean influences. The ocean currents are merely part of what distributes the energy received from the sun. Atmospheric circulation plays a part too. I don’t see where we are ignoring either. But those are — to use terms from my field of study — endogenous influences, not exogenous influences. If you say that the decadal and bidecadal oscillations in temperatures are “caused” by oscillations in ocean currents, that still begs the question of causes the oscillations in ocean currents. Doesn’t it? Besides vulcanism, the only exogenous forces I can think of are lunar and solar. (OK, somebody is going to interject the anthropogenic as exogenous, but that should be a linear process, not an oscillating one.) Vulcanism is surely important as a climate variable, but it is too random to account for the relatively predictable decadal and bidecadal changes in the rate of temperature change.
For sure, you will see SST’s in what we are analyzing, as they are part of HadCRUT3. But even if we focused only on SST’s, the question would remain: what causes decadal and bidecadal oscillations in the rate of change SST’s?
Basil
Gentlemen, you’re on the right track by examining the ROC of the secular sunspot level. That’s where one finds coherence with terrestrial temperatures. You might be on the wrong track, however, in taking HADCRUT3 anomalies as good, consistent estimates of “global average temperature.” Their constructed time-series has highly peculiar features that show up in cross-spectrum analysis with other indices. I realize that there is no other viable alternative if one seeks the longest such series, but a note of caution might be sounded in your splendid presentation. And the length of the series is what determines the statistical confidence of all spectral analyses at a given resolution. Since you chose high resolution, a note on the expected high variability would help.
My understanding of the “beat” frequency is the difference frequency that shows up as the signal “envelope” when two near-frequency narrow-band signals are superimposed, rather than some intermediate between the two. Also, readers might benefit by an earlier identification of the Hale cycle with the bidecadal double sunspot cycle. I nitpick here only to be helpful in your laudable project.
Oh, and if I am correct in my hunch, what one would see is a “ringing” after a major change such as warming from the LIA or cooling from the MWP. And the longer the period of stability after the change , the more the “ringing” would damp out over time as teh deep sea adjusts to the surface stability. In other words, the longer things are stable, the more they tend to remain stable but when an instability is introduced, I believe things can fluctuate widely until the entire system adjusts to the new condition. So warming in 1998 was not as intense as warming in 1933. Cooling now might not be as much as the early 1970’s. Figure 8 seems to reflect the same response I would expect but for different reasons. That is, unless the solar and lunar influences are impacting tidal and wind patterns that affect winds and upwelling.
Bill Illis (08:01:37) :
I’ve done a little of this kind of analysis before and have found some of the same repeating cycle timelines.
They seem to be slightly off the timelines one would expect with the solar cycle.
I don’t know if there is a mathematical theorem to describe this but perhaps one should actually expect the spectral analysis peaks to be slightly off the solar cycle timelines.
Instead of an 11 year peak, maybe the strongest peak is at 9 years. Maybe the Hale cycle peak is at 18.6 years rather than 22. If one is trying to detect a signal from 11-year peak to 11-year peak of the cycle, this signal will be less strong than an offset of (3/4 of the peak) to peak.
The Hale cycle is at ~22. The lunar nodal cycle is at 18.6. I think the strong peak at ~9 years is most likely a reflection of the lunar nodal cycle. There is a lot of literature on how the lunar nodal cycle, driving tidal forces, could influence SST’s. Keeling and Whorf tried to attribute all of the variation to the lunar nodal cycle. We don’t think that works. The 20-21 year peak, which is also strong, is too long for that. It is not readily accessible, because it was published in a proceedings volume, not in a journal, but the papers by Bell (see #’s 16 and 17 in the reference list) argue for a “beat cycle” from combining the two influences. I think that is where the answer lies, as far as explaining the ~9 and ~20.5 year cycles go, and why we do not see things matching up exactly with an 11 year cycle.
Basil
Since endogenous sources have such strong effects (they create the wide swings and oscillations in noise everyone is so willing to disregard and remove from the data), it seems plausible to me that I would also look for endogenous sources for the trends as well. ESPECIALLY since the noise DETERMINES the statistical trend. Remember, the trend is not data, it is statistical analysis. Therefor the noise cannot be separated from the trend in terms of source, at least not on Earth. The only way one would be able to do that is find some terrestrial planet similar to Earth but without its endogenous features and see what the Sun does to surface land temperatures. If a trend is identified similar to the statistical analysis of the noise here on Earth, I would contemplate such an exogenous factor as the Sun causing a trend that is overlayed by endogenous noise.
WUWT
As much fun as it has been Digging For H2o on Mars, I would have preferred Seeding The planet and moon/s with a number of elemental enviromental (ground and atmopspheric) sensors such as to record an iindependant planetary record during this solar minimum.
It is always instructive to back away periodically and survey a larger view of the landscape upon which we search for reality. Dr. Leonard Weinstein has provided such a view (http://wattsupwiththat.com/2009/05/22/limitations-on-anthropogenic-global-warming/) with cogent descriptions of the scope and the temporal validity of the state of knowledge to date. My apreciation to Dr. Leonard Weinstein
In the spirit of the larger view, I commend the wide readership of WUWT to digest an offering by
Michael Crichton regarding the socio/political landscape on which the world of science must navigate. If you find the master of fiction an ironic choice for commentary on the subject of global warming, I invoke the words of my literature prof, “Irony makes life real”. Mr. Chrichton bluntly describes what we are up against.
Mr. Chrichton can be found here; http://sharpgary.org/ItsAboutTimeToo.htmlhttp://sharpgary.org/
For all you number crunchers out there,
Timo Niroma of Finland constructs an exhaustive analysis of our suns behavior over time. It is incredible what a creative mind can do with a 9 month winter. Bravo Timo Niroma. Right or wrong, Mr. Niroma makes (for me) a compelling case for the astrophysical aspects of the larger solar system that we call Home, and the cumulative effects on our sun. I do not find any conflicts vis a vis Niroda-Svalgarrd, only the degree of precision. Ultimately, I suspect, Dr. Svalgaard will be vindicated as will Timo Niroma.
Timo Niroma can be found here; http://sharpgary.org/ItsAboutTimeToo.html
Enjoy
From eyeballing Figure 2, the 1853 anomaly was about -0.35°C, an from Figure 8, it will be about +0.35°C. +0.7°C/1.67 centuries = +0.42°C/century. In http://wattsupwiththat.com/2009/03/20/dr-syun-akasofu-on-ipccs-forecast-accuracy/ Dr. Aksofu describes a model based on a linear recovery from the Little Ice Age plus a multi-decadal oscillation does a good job describing temperature history. “The linear increase has a rate of ~ +0.5°C/100 years,” a decent match.
Akasofu took the multidecadal oscillation and fit it to the PDO, you took it and fit it to luna and solar effects. Neither group included the CO2 trend in analysis and results, though Akasofu suggests its current effect might be +0.1°C/century simply because the IPCC projects (predicts?) a +0.6°C/century warming.
You have a decent match with Akasofu’s work, and one I think is worth mentioning in your paper and in the references. Next project – look for lunisolar influences that drive the PDO. 🙂
Lots of people have computed power spectra [even fancy ones with wavelets, maximum entropy, etc] of the sunspot series, filtered or unfiltered.
Here is the FFT spectrum based on yearly SSNs with the start of the series varying from 1700 by 1 year [1700, 1701, 702, …] up to some twenty years later to see what the end effects are [wavelets get you around that]:
http://www.leif.org/research/FFT-Power-Spectrum-SSN-1700-2008.png
Using monthly resolution [and no shifting] one gets:
http://www.leif.org/research/FFT-SSN-Monthly-1755-2007.png
(the blue curve is the ‘12.5’-month smoothed monthly SSN which just shows how the smoothing removes the power around 1 year.)
The 22-yr ‘cycle’ and in particular the 18.6-yr cycle are very weak compared to the ~11 year cycles. There are really no solar quantity that has a significant variation on the Hale-cycle scale. The various asymmetries with a 22-yr period are second order effects that would seem to cause second order effects in climate too compared to the first-order effect of the 11-yr cycle. This is my biggest problem with this: ‘where is the 11-yr cycle that is expected: 0.07 degrees? and that countless other papers find.
On Leif’s point, that there are “2000 papers” out there reporting cycles of these kinds, that’s true. But you can count on one hand, by my reckoning, the number of papers that deal with these cycles in the global temperature data. The papers that Leif is talking about are all the papers finding these periodicities in tree rings, varves, and other climate proxies, or if in temperature data, then most likely data that is regional in nature, or limted to SST’s, and not global. Our paper is more like #4 or #5, in a series that starts with Ghil and Vautaurd, the dispute by Tsonis and Eisner, and the further development by Keeling and Whorf. Since the latter is now more than a decade past, and we’re approaching nearly two decades since the original Ghil and Vautaurd, I think our work is relevant just to see how things look with one to two more decades of data.
Beyond that, our putting a time dimension to the cycles, and not just measuring their frequencies, is, we believe, worthy of reporting.
As to Leif’s expectation that the impact of TSI should be seen in an 11 year cycle, and not just a 22 year cycle, we’re suggesting that the lunar tidal influence that Keeling and Whorf saw is modulating or interfering with the “decadal” solar signal. I don’t doubt that the physical processes that Keeling and Whorf, and others, have postulated for a lunar tidal influence on SST’s are partly at work here. I think that is what the ~9 year signal represents. Whereas Leif likes to point out that the Sun is “messy,” so is terrestrial climate. And when you have 9/11 and 18.6/22 year cycles both influencing terrestrial climate, there is bound to be some “messiness” in how they interact, or interfere, with each other.
As to the studies showing contradictory results, I think Leif overlooks the “pattern” in these “contradictory” results. If the contradictions were purely random, Leif’s point would be stronger. But their seem to be “runs” in the results, with periodic sign reversals. Again, this has been thoroughly surveyed by Hoyt and Schatten and Georgieva et al. Rather than dismiss the results as contradictory, if the contradictions are somehow systematic, it is relevant to ask whether or not there is an explanation for the systematic nature of the contradictions (i.e. one sign dominating prior to 1920, and another sign dominating after 1920). Georgieva and her collaborators may not be right in their answers (solar hemispheric asymmetry on a double Gleissburg scale that leads to terrestrial atmospheric circulation “epochs,” but at least they are asking the right questions. It may be coincidence, it may be random, or it may be relevant, but the way in which the weak/strong warming cycles shift between timing dominated by lunar and solar periodicities at around the same time as the solar sign reversal shift of the 1920’s (cf. Figure 6) is what it is.
Anythony Watt and Basil Copeland apply statistical tricks which I don’t understand much of and trust less. Lief Svalgard, as always, and despite being praised by name in the post, makes a valid point when he complains of ‘cyclemania’.
But they may be right. This is evidence is cosistent with basic physics. After the heat of sun the next most powful influence on the earth’s atmosphere is the forces of gravity exerted by the earth and the moon.
Combined and mediated by the next biggest influence, the oceans.
Lunisolar influence, the answer to AGW hoax.