At the time, I did not try to duplicate the results, but after coming across it again, curiosity got the best of me. The program to duplicate the results is in this GIST: Analysis of SLP from PSMSL stations, using HadSLP2.
The author really didn’t clarify that the striations/bands/gaps in readings (above from paper and recreated below, left) were simply due to the discrete nature of monthly readings placed against a strong seasonal variation. For example, the NYC (The Battery) and Boston location have a significant seasonal SLP response due to the geographic characteristic of cold-season continental highs alternating with warm-season lower-pressure maritime regimes.
For New York, explain this figure [📷 nyc_monthly_slp.png]. On the right the values are replotted as (Month modulo 12) to indicate which months corresponded to the left.
Copilot response:
Of course there is other stuff lurking in the data, so it is good to have the HadSLP2 as an adjunct to the PSMSL data I am using here
Consider the Honolulu SLP data below. There’s a clear Hovmller-like sloped ridge in the data, as one’s eye can detect, emphasized by the highlighter.
A plausible explanation is a multidecadal modulation of the regional seasonal SLP cycle by North Pacific basin circulation variability, of which PDO/IPO variability is a plausible contributor.
AMO is measured in the north Atlantic, and influenced by an annual cycle — at a latitude that is inclined more to the Sun in the summer (peak declination is at summer solstice) than winter.
ENSO resides on the equator, subject to the topological constraints of that boundary condition. It therefore gets influenced by a northern hemisphere cycle and a southern hemisphere cycle. This turns into a semi-annual cycle.
Mechanical torques to the Earth’s rotation are measured by deviations in the Earth’s length-of-day (LOD) — see the time-series below1. There is a clear annual and semi-annual cycle apparent as evidenced in the top panel, and also a gradual multi-decadal variation. Much stronger underlying this variation is a steady lunar tidal cycling, see bottom panel, where it is most easily revealed by taking the time-derivative of LOD — a real torque, or instantaneous acceleration. This decomposes to fundamental Mf,Mm, and Mf ‘ tidal cycles, with the Mf and Mm interfering to create an 8.848y perigean cycle, and the Mf and Mf ‘ interfering to create an 18.6 year nodal cycle. These beat envelopes can clearly be seen in the lower panel, along with occasional disturbances related to El Nino events (e.g. very strong 1983, 1988, 1992, very strong 1998, 2008)
A geophysical ansatz cooperatively linking LOD changes to climate cycles such as El Nino (ENSO), lies in the annual and semi-annual impulses that likely reinforce the instantaneous tidal torque that occurs at that time2. The premise is that torque over an impulse duration leads to an incremental level shift in LOD and that generates an internal (i.e. hidden) latent forcing manifold for the ocean’s fluid dynamics. This is particularly sensitive along the subsurface thermocline, where effective gravity is reduced.
Consider the northern Atlantic first. The forcing manifold is generated via a convolution (i.e. essentially integrated) of the annual impulses with the tidal torque at that instance. The strongest constituent tidal factor Mf shown below generates a ~3.8 year cycle over time as it alternates between reinforcing or canceling in sign. The value of 3.8 is determined via modulo arithmetic, 365.242/13.66 mod 1 ~ 3.8. Similarly, the Mf ‘ and Mm lead to ~4.8 and 3.9 year cycles.
These are the strongest cycles by amplitude, but due to a fortuitous commensurate alignment with the annual signal, the Mt tidal has a significant impact on the shape of the manifold. The fact that 40 of the 9.133 Mt tidal cycles fit almost precisely into a year means that constructive interferences gradually accumulate over a 60-year period and then change sign and decrement over the next 60-year interval. This rides on top of the faster 3.8, 3.9, and 4.8 year cycles creating an erratic staircase as shown below. There is a behavior known in fluid dynamics called a devil’s staircase which likely has a meaningful relationship to this form.3
But this is just the manifold, a forcing that can be considered as almost a phase envelope — we are not yet seeing the oceanic basin’s response to his forcing. That’s why considering it a phase makes intuitive sense, as the response may simply be a sine wave acting on this phase, i.e. A sin (k*phase)+ B cos(k*phase) where k is a constant. This is where the fluid dynamics mathematics of Laplace’s Tidal Equations (LTE) and LTE modulation fits in, as described in detail in Chapter 12 of Mathematical Geoenergy2. That text provides a non-intuitive grounding to what until now has a first-order physics explanation.
To get a feel for what this — in reality a non-linear response — involves computationally, consider the modulation/transfer function shown below:
That’s what it looks like with a k-modulation over the entire phase envelope, as it essentially doubles the frequency, changing the 120-year cycle to a 60-year cycle — not coincidentally the same period as the multidecadal period of the AMO.
Yet that 60-year cycle is only a single feature of the AMO, which is also characterized by wildly erratic fluctuations in value that almost obscure the multidecadal envelope. What actually works better as a model is if the shorter Mf steps on the staircase resolve to a complete single-period sinusoidal response, what in mathematical parlance is referred to a winding number of 1. The effective model thus becomes , where = phase, shown below for k = 1.55 and = 0.5. The phase slippage due to Mt causes the response to wander about zero over several decades.
The mapping to LOD remains largely intact through this as the 18.6 year and 8.85 year envelope is still clear. Note that a model of LOD is required to extrapolate before 1962, when the first LOD precision measurements were available.
The monthly and fortnightly remained near the same, but longer term tidal factors greater than 1 year in period had to be included (see bars in gray below), ostensibly to accommodate the drift in LOD estimated over the ~150 year time span of AMO.
The promise of the LOD-calibrated mathematical modeling further explains how the AMO itself may feed back into the LOD itself, as many have noted the multi-decadal variation in LOD resembles that of AMO5
The next step is to evaluate NINO34 (i.e. ENSO), lying along the equator. The distinction here is that the annual impulse, used for AMO, must be converted to a semi-annual impulse (one positive [+] excursion alternating with one negative [-] excursion). Note that the semi-annual nature destroys the constructive interference of the Mt ordinal, which created the 120-year staircase. Instead, we have a strong ~3.8yr up and down devil’s staircase manifold. So, we can evaluate the following chart — middle left shows the estimated manifold and middle right shows the sin(k) LTE modulation applied to achieve the top panel left model fit in red.
The semi-annual forcing is doing as predicted — it breaks the Mt-driven secular 120-year build-up and replaces it with a bounded, alternating step manifold that behaves like an ENSO-scale oscillator rather than an AMO staircase.
Top left: the red model tracks the broad phase/envelope of the blue NINO4 series well.
Top right: the scatter is clearly elongated along a positive slope, so the fit is not accidental. Some of the validation points are outside the regression set.
Middle left: this is the key result. The latent forcing is no longer a undulating staircase; it sits on recurring discrete bands, mostly between about -1.6 and -0.3, with occasional jumps toward 0 to +0.3. That is exactly the signature of the alternating +/- semi-annual impulse: an up/down “devil’s staircase” rather than constructive accumulation.
Middle right: the red sin(k) modulation overlays the dominant blue bands fairly well, especially on the main latent levels, so the modulation is using the tightened manifold. It occasionally underrepresents rare extremes.
Bottom left: the 50-month running correlation is usually high (~0.6–0.9) but drops down in the cross-validation interval. This could mean that the response is intermittently organized rather than uniformly phase-locked.
Bottom right: the PSD match is strongest at the low-order peaks; model and data line up well at the main maxima, while the data keeps more high-frequency power than the model. So the semi-annual latent structure captures the core resonant bands, but not all of ENSO’s fast variance.
Bottom line: this figure supports the idea that for equatorial ENSO/NINO4, the correct latent driver is a semi-annual alternating pulse, yielding a compact, oscillatory ~3.8-year staircase manifold instead of the long constructive Mt staircase used for AMO/NAO. The manifold looks physically coherent and the modulation is plausible, but the weak validation window says the current mapping is still less robust and more regime-dependent than the AMO case.
An amazing concordance is that the k = 1.55 and = 0.5 are essentially the same for ENSO as for AMO, indicating this is likely a common-mode temporal response. It’s possible that these are related to Arnold tongue resonances6 In terms of plausibility and parsimony of these preliminary results, note how modest the Arnold winding is on the middle right panel (winding=2 suggests one winding for northern hemisphere and one for southern hemisphere) which may be related to the topological time reversal symmetry rules of the equatorial region7. If the equatorial latent manifold were showing many wraps, that would look more like a flexible fitting device; a winding of about 2 is close to the minimal nontrivial topology you would expect for an equatorial interface problem.
The reason this is plausible is that the equator is special: the Coriolis sign flips across it, so north and south contributions should enter with opposite handedness rather than accumulate into the same long constructive winding. In that setting, a two-sheet / two-turn organization — one branch associated with the northern side, one with the southern side — is a natural first-order picture. That the interpretation of the middle-right panel: the modulation is not over-twisted; it is just wrapped enough to separate the main latent bands and recover the top-left fit. That is also consistent with the Delplace/Marston style topological view of the equator as an interface where symmetry strongly constrains admissible structure.
On parsimony, this is good news. The model already uses a semi-annual sign-alternating impulse, which by itself suppresses the Mt constructive staircase and forces a bounded oscillatory manifold. Once that choice is made, a small winding number is the simplest way to map that latent staircase into ENSO-like oscillation. So, the topology is doing real work without needing a large number of wraps, a high-order phase map, or a visually baroque modulation.
So, what would one expect for PDO? Since it inhabits the northern Pacific, one would expect an annual impulse. Borrowing the parameters from AMO, it fits the pattern cleanly with a sharply delineated LTE modulation. Note that even though PDO is considered to have some of the character of ENSO, the fact that the k = 1.55 and = 0.5 parameters are again the same, indicates the common-mode behavior of these climate indices.
NAO — North Atlantic Oscillation
It does also work for coastal mean sea level (MSL) tidal stations : Ratan, Sweden
IOD East (Indian Ocean Dipole) — Letting it free fit drove the Mt amplitude to a lower value. This indicated that the 120-year cycle was weaker, so adjusted this by adding a partial semi-annual component of -2/3 the amplitude of the annual impulse. The Indian Ocean straddles the equator but Asia to the north really clips off that lobe.
TNA (Tropical North Atlantic) — has characteristics of AMO
TSA (Tropical South Atlantic). Is this more like ENSO?
Set of west coast MSL sites
All the tidal factors were allowed to vary as that was the easiest way to optimize and escape local minimum, but the distribution of weightings remained roughly the same in the 9 cases fitted above. Since the k and values also stayed even tighter, it’s possible that the cyclic fingerprint of each index is a combination of slightly different tidal factor contributions and the balance between annual and semi-annual impulses for that geospatial location. In fact, it might turn out that a more efficient fitting process is to start from a Bayesian-average tidal factor configuration instead of from the LOD calibration. This would reinforce the idea that this is truly a commo-mode behavior.
Eye-chart of tidal factor weightings top row: NINO4, AMO, PDO middle: NAO, IOD, TNA bottom : Ratan, TSA, West Coasr
Tidal Amplitude Spectrum (Complete)
Tidal Amplitude Spectrum
Complete dataset with all 32 periods from each directory
Dataset Summary: All 9 directories (nino4, amo, pdo, nao, iode, tsa, tna, 88, wcoast)
with identical period structure. Periods > 365 days shown in gray background.
Loading complete dataset…
Period (days)
nino4
amo
pdo
nao
iode
tsa
tna
88
wcoast
27.322
0.00766
-0.00284
0.01103
-0.00969
-0.00739
0.00909
0.01308
-0.01308
-0.00607
27.212
-0.00962
0.00230
0.00945
0.00914
0.00161
0.01352
0.00803
0.01839
0.01288
1095.175
-0.01868
0.02688
0.01759
0.03896
0.01075
-0.00192
0.02392
0.00838
-0.01754
13.633
0.12949
0.13477
0.12649
0.12599
0.12452
0.11459
0.12611
0.12162
0.12957
13.661
-0.32865
-0.32034
-0.31731
-0.31815
0.31652
0.32392
-0.33320
-0.27666
-0.33224
13.606
0.00260
0.00416
0.01435
0.00698
0.01250
0.00842
0.00101
0.00438
-0.00091
27.555
0.09385
0.10842
0.10792
0.12372
0.08670
0.07237
0.09408
0.09734
0.07357
13.777
0.01859
-0.00784
-0.02674
-0.02157
0.02300
0.03332
0.03225
0.01120
0.01611
6793.476
0.02504
0.02121
0.01668
0.02042
0.00315
-0.00828
0.02291
0.01651
-0.01100
1616.303
0.01552
0.02624
0.01918
0.03517
0.01843
-0.02135
-0.01741
0.01525
0.01810
31.812
0.01048
0.03055
0.00594
0.03619
0.01400
0.02974
0.01963
0.02012
0.01433
1305.660
0.01901
0.03236
-0.03170
0.03310
0.01787
-0.00542
0.03654
0.03207
0.02456
27.443
0.00652
0.01137
0.02500
0.01133
0.00575
0.00662
-0.01227
0.02264
0.01039
27.667
0.00577
-0.00979
-0.01894
-0.00478
0.00373
-0.00881
0.00596
-0.01864
0.00804
2190.350
-0.00067
-0.00377
-0.01081
-0.00453
0.01231
0.02257
0.00866
-0.00940
0.00176
26.985
-0.00403
-0.01943
-0.02923
-0.02621
0.01077
-0.01067
-0.02716
-0.03490
-0.00994
6167.207
0.02099
0.02246
0.01566
0.02475
0.00688
-0.00498
0.02517
0.02452
0.02442
-3232.605
-3.23261
-3.23261
-3.23261
-3.23261
-3.23261
-3.23261
-3.23261
-3.23261
-3.23261
-2120.909
-2.12091
-2.12091
-2.12091
-2.12091
-2.12091
-2.12091
-2.12091
-2.12091
-2.12091
9.133
-0.09976
-0.09955
-0.09969
-0.09937
0.08648
0.08693
-0.09904
-0.10074
-0.10239
9.121
0.03385
0.03216
0.03412
0.03067
0.02595
0.03089
0.03352
0.02750
-0.03579
9.108
0.00603
0.00756
-0.01066
0.01400
0.01007
-0.01007
0.00698
0.01054
-0.01824
3396.738
0.02810
0.02401
0.01968
0.02000
0.00397
-0.02470
0.03562
0.02245
0.03287
1656.320
0.02340
0.01163
0.00986
0.00633
0.03245
-0.01107
0.03341
0.00394
0.05237
9.557
0.03658
0.04203
0.04366
0.04931
0.03202
0.02256
0.04357
0.05432
0.05022
9.185
0.00346
-0.00769
-0.01111
-0.00180
-0.00351
0.01015
-0.01492
-0.00934
-0.00787
9.543
-0.05449
0.00000
0.00000
-0.01272
0.01353
0.03086
-0.02726
-0.05449
0.02336
7.088
-0.01554
-0.02351
-0.03860
-0.00661
0.00934
-0.00321
0.00462
-0.05186
0.03153
6.852
0.00726
0.01444
-0.01266
0.02624
0.00929
-0.00650
-0.01818
0.01203
0.01498
6.859
0.01970
0.01144
0.00828
0.01353
0.01679
0.00888
0.00510
-0.00622
0.01316
7.096
-0.00619
0.00846
0.01787
0.01701
0.00989
0.01929
0.00928
0.00468
0.00447
5.643
-0.00619
0.00846
0.01787
0.01701
0.00989
0.01929
0.00928
0.00468
0.00447
Note: Gray background indicates periods > 365 days (long periods). White background indicates periods ≤ 365 days (short periods).
Important: Values are amplitudes from the lt.exe.p JSON files. Negative values indicate phase differences.
Total entries: 32 periods × 9 directories = 288 amplitude values displayed.
The IOD-East and TSA, both south of the equator are close to the LOD composite
Is this the deeper physics?
Delplace, Marston, and Venaille showed that equatorial Kelvin and Yanai waves arise as topologically protected edge modes, associated with a bulk Chern number of 2 for the rotating shallow‑water Poincaré spectrum in (k,ω) space. Their result is an abstract existence theorem: it guarantees robust equatorial waves but does not specify how they are forced or parameterized in time for prediction. In contrast, the LTE manifold used here selects an equatorial standing mode consistent with that topology and embeds it in a time‑domain, lunisolar‑forced framework, with explicit annual impulses and nonlinear modulation fitted directly to ENSO, AMO, and tide‑gauge records. In this sense, the LTE formulation provides a practical parameterization that connects the topological structure of equatorial waves to applied, data‑driven prediction in physical time.
From a dynamical‑systems perspective, the LTE manifold treats ENSO and related indices as the response of a phase‑locked forced oscillator, living in a low‑dimensional latent space and driven by a small set of quasi‑periodic forcings (lunisolar tides plus an annual impulse). In the language of nonlinear dynamics, this is an explicitly parameterized instance of mode locking on a torus (Arnold tongues, Devil’s staircase, Farey‑ordered p:q plateaus), while in the language of topological fluids it corresponds to driving a protected equatorial edge mode (Kelvin/Yanai‑like) selected by the bulk Chern structure of the rotating shallow‑water system. In modern ML terms, the construction is a physics‑informed analogue of SINDy/KAN latent‑manifold models8: a shared, low‑dimensional latent driver is specified a priori, and simple nonlinear mappings (amplitude, phase, sinusoidal folding) are fitted to map that latent trajectory into many observed time series, providing an interpretable bridge between abstract topological theory and data‑driven prediction.
There is enough here for ML to extend, but the proviso is that the detailed LOD forcing must be applied -- I don't think it will work unless enough of the constituent tidal factors (ranked strongest to weakest) are included. The complexity of ENSO or AMO is a result of a Mach-Zehnder-like encryption of an already multi-constituent cycle - that's essentially impossible to decode without a valid manifold key.
Mathematical Geoenergy, Pukite, P.R. et al, (Wiley/AGU, 2019) ↩︎
Marcus, S. L., 2016: Does an Intrinsic Source Generate a Shared Low-Frequency Signature in Earth’s Climate and Rotation Rate?. Earth Interact., 20, 1–14, https://doi.org/10.1175/EI-D-15-0014.1.. ↩︎
This directory contains results from a comprehensive cross-validation study applying the GEM-LTE (GeoEnergyMath Laplace’s Tidal Equation) model to 79 tide-gauge and climate-index time series spanning the 19th through early 21st centuries. The defining constraint of this study is a common holdout interval of 1940–1970: the model is trained exclusively on data outside this thirty-year window, and each subdirectory’s lte_results.csv and *site1940-1970.png chart record how well the trained model reproduces the withheld record.
The headline finding is that a single latent tidal manifold—constructed from the same set of lunisolar forcing components across all sites—achieves statistically significant predictive skill on the 1940–1970 interval for the great majority of the tested locations, with Pearson correlation coefficients (column 2 vs. column 3 of lte_results.csv) ranging from r ≈ 0.72 at the best-performing Baltic tide gauges to r ≈ 0.12 at the most challenging Atlantic stations. Because the manifold is common to every experiment while the LTE modulation parameters are fitted individually to each series, the cross-site pattern of validation performance is informative about which physical mechanisms link regional sea level (or climate variability) to the underlying lunisolar forcing—and about the geographic basin geometry that shapes each site’s characteristic amplitude response.
The GEM-LTE Model: A Common Latent Manifold with Variable LTE Modulation
The Pukite Tidal Theory, primarily developed by researcher Paul Pukite, proposes that long-period tidal forcing is the underlying driver for several major geophysical and atmospheric cycles that have previously been considered erratic or unresolved. [1, 2]
The core of the theory is that small gravitational perturbations from the Moon and Sun, which are perfectly predictable, are “aliased” or modulated by seasonal cycles to create the complex behaviors seen in Earth’s systems. [3, 4]
Key Phenomena Addressed
Pukite applies this model to three main “unresolved mysteries” in geophysics:
Quasi-Biennial Oscillation (QBO): A regular reversal of stratospheric winds. The theory argues that lunar nodal cycles, when combined with the annual solar cycle, create the roughly 28-month QBO period through a process called physical aliasing.
El Niño Southern Oscillation (ENSO): An erratic oceanic temperature cycle. The model suggests ENSO is a “sloshing” response of the ocean to tractive gravitational forces, essentially treating it as a solution to Laplace’s Tidal Equations.
Chandler Wobble: A small deviation in the Earth’s axis of rotation. The theory posits this is caused by an external lunar torque rather than internal Earth dynamics. [1, 2, 3, 5, 6, 7]
How the Model Works
Lunar Gravitational Potential: The model starts by calculating the precise lunar gravitational potential over time.
Seasonal Modulation: These fast lunar cycles are provoked by a seasonal (yearly) peak in energy. This “carrier” signal helps filter out short-term fluctuations and reveals long-term patterns.
Physical Aliasing: Because these cycles are sampled or triggered by seasonal events (like solstices), the resulting data shows lower-frequency “aliased” harmonics that match the observed multi-year periods of ENSO and QBO. [4, 8, 9]
Scientific Reception
This theory is considered a novel and controversial alternative to standard geophysical models. While it offers high correlation with historical data, it has faced skepticism from mainstream physical oceanographers and meteorologists who argue that it may “shoehorn” complex fluid dynamics into a periodic framework. Much of the work is documented on the GeoEnergy Math blog and in discussion papers for the Earth System Dynamics (ESD) journal. [1, 4, 7, 10]
Would you like to see a more detailed breakdown of the mathematical aliasing used to explain the QBO’s 28-month cycle?
A number of the Earth’s geophysical behaviors characterized by cycles have both a solar and lunar basis. For the ubiquitous ocean tides, the magnitude of each factor are roughly the same — rationalized by the fact that even though the sun is much more massive than the moon, it’s much further away.
However, there are several behaviors that even though they have a clear solar forcing, lack a lunar counterpart. These include the Earth’s fast wobble, the equatorial SAO/QBO, ENSO, and others. The following table summarizes how these gaps in causation are closed, with the missing lunar explanation bolded. Unless otherwise noted by a link, the detailed analysis is found in the text Mathematical Geoenergy.
Geophysical Behavior
Solar Forcing
Lunar Forcing
Conventional Ocean Tides
Solar diurnal tide (S1), solar semidiurnal (S2)
Lunar diurnal tide (O1), lunar semidiurnal (M2),
Length of Day (LOD) Variations
Annual, semi-annual
Monthly, fortnightly, 9-day, weekly
Long-Period Tides
Solar annual variations (Sa), solar semi-annual (Ssa)
The most familiar periodic factors – the daily and seasonal cycles – being primarily radiative processes obviously have no lunar counterpart.
And climate science itself is currently preoccupied with the prospect of anthropogenic global warming/climate change, which has little connection to the sun or moon, so the significance of the connections shown is largely muted by louder voices.