Replies: 31 comments 6 replies
-
Sunspot models as a climate science strawman. |
Beta Was this translation helpful? Give feedback.
-
https://www.realclimate.org/index.php/archives/2024/01/unforced-variations-feb-2024/#comment-818894 GM said:
That's phrased incorrectly. Models can do just fine with non-linear phenomena, but more problematic is the process of solving the models to generate unique results. The fact is that through history, humans have developed powerful mathematical techniques to solve linear systems, Once in a linear regime, just about any system can be solved, very often in closed form or using matrix math. Consider that all semiconductor amplification operates in a non-linear fashion, yet electronic circuit designers make certain that feedback and operating conditions are such that it operates in a linear regime. The problem comes about when we are trying to understand nature and its behaviors, which may be non-linear. There are theoretically infinitely many more non-linear formulations possible than linear, as this is just a result of combinatorics of interactions, adding squared, cubed, etc infinitely, with many of these not chaotic. Humans are helpless in being able to solve even a fraction of these as straightforwardly as in a linear system. Not only is it challenging to arrive at a correct solution, but the solution may not be unique, as many distinct non-linear model formulations may give the same answer. Yet, humans are the stubborn types and keep trying to frame models in linear terms so they can be easily solved, much like the drunk who is looking for his missing car keys under the street=lamp because that's where the light is. Alas, machine learning is stubborn in a different way. Neural networks are powerful non-linear solvers that don't try to frame problems in linear terms. Instead, they will keep trying different combinations in conjunction with cross-validation methods that can zoom in on the most unique parsimonious solutions. So that once a non-linear model is identified and validated as being deemed correct, there should be less problems with it being inherently non-linear (see above regarding semiconductors). In summary, I would gladly deal with a non-linear model that I knew was correctly mapped to some natural behavior as then I could put effort into solving that, knowing that I wasn't chasing down some blind alley maze of non-linear combinatorics. For tricky aspects of climate such as natural climate variability and the non-linear geophysical fluid dynamics that drives this, machine learning may be just the approach that finds a plausible and parsimonious working model. Here is a ChatGPT4 response to the above, input as a prompt: BTW, there is a paper under open discussion review that I added comments to, related along these same lines "Moving beyond post-hoc XAI: Lessons learned from dynamical climate modeling" If you have something to add, feel free, as the non-linear phenomena don't reveal their secrets easily and certainly won't get solved on their own. Update: including a section on tidal locking |
Beta Was this translation helpful? Give feedback.
-
Beta Was this translation helpful? Give feedback.
-
Beta Was this translation helpful? Give feedback.
-
https://www.realclimate.org/index.php/archives/2024/03/more-solar-shenanigans/#comment-820094 |
Beta Was this translation helpful? Give feedback.
-
Piotr, do you have a theory as to the sudden rise in oceanic temperatures? I’m not a climate researcher, my background is in engineering and later applied mathematics so I claim no authoritative opinion. However, obviously exceedingly large amounts of thermal energy cannot appear by magic. Also it seems counterintuitive to me that these phenomena are a result of changes in atmospheric phenomena such as aerosols or other gaseous causes due to thermal inertia of the oceans. This makes me conjecture that this energy is already present somewhere in the physical system. Interested to get your thoughts on this sudden temperature rise. |
Beta Was this translation helpful? Give feedback.
-
Note where there are strong internal tidal hotspots, near the Fiji-Samoa-Tonga triangle off OZ-NZ Sensitivity of Internal-Tide Generation to Stratification and Its Implication for Deep Overturning Circulations https://english.news.cn/20220911/3dd8be58415c4a5c8e4cb0bf0f6c371a/c.html
Research paper in Acta Oceanologica Sinica.: https://www.researchgate.net/profile/Xiaofeng-Li-16/publication/362566910_Cover_Story_Oceanic_internal_waves_generated_by_the_Tongan_volcano_eruption/links/62f3a0c0c6f6732999beb5ef/Cover-Story-Oceanic-internal-waves-generated-by-the-Tongan-volcano-eruption.pdf This next one is strange, a PhD thesis. They mention the moon in the intro and never bring it up again. Are they not interested in determining a transfer function? Always seems like they dance around and just derive some dispersive wave equations. Doesn't accomplish anything. https://www.diva-portal.org/smash/get/diva2:1842347/FULLTEXT01.pdf |
Beta Was this translation helpful? Give feedback.
-
Beta Was this translation helpful? Give feedback.
-
April 1 |
Beta Was this translation helpful? Give feedback.
-
JD says:
I don't think it's overkill when one considers that most newsworthy advanced physics concerns intensely argued topics such as fundamental particles and string theory, etc. By comparison, condensed matter physics is rather tame, notwithstanding the difficulties with fluid dynamics. Hence, Terry Tao's interest in trying to solve Navier-Stokes, which I give him props for. Sabine and other popular physics commentators such as John Carlos Baez have stated that condensed matter physics is a great example of unqualified success in matching theory to experimental verification, so it's often a safer bet to explore. I started getting interested in climate topics like ENSO, when now over 10 years ago I wrote a blog post comparing the general shape of a SOI time-series with a Bloch wave from solid state physics. A Bloch wave is essentially a wave that interacts with the periodic potential of a crystal lattice, as you would find in a semiconductor. The waveforms are often complex , which is appropriate for the erratic nature of ENSO. From there, I've tried to drum up interest in pursuing this association, noting that a few other physics groups, such as Marston at Brown U are also working this angle. I think the most interesting connection is to that of a Brillouin zone in solid state lattice theory. That model is of a spatial nature, and leads to standing waves that have wave numbers that fit into modes of a bounding volume. That's well understood and not exactly applicable, but there is also the concept of a temporal Brillouin zone whereby a time modulation of the matter can induce a temporal periodicity that other forces can interact with. A stimulated wave in such a state will fold over with the modulated periodicity so as to appear as a reduced frequency using what's called modulo arithmetic. When I was originally looking at this, I also applied it to the atmospheric QBO, which has a more distinct periodicity of about 2.3 to 2.4 years. The only tidal cycle that matches the wavenumber=0 mode of QBO is the 27.2122 day lunar nodal cycle interacting with the modulating annual solar periodicity on the ecliptic plane. Using the modulo arithmetic of the Brillouin zone, this comes out to a reduced frequency of (365.242/27.2122) mod 1 = 0.422, which is 2.37 years, matching that of QBO. This also works for the oceanic indices such as ENSO but with a different tidal forcing. Again, I started on this topic over 10 years ago and have presented and published on it, yet have seen it get little traction or acceptance. That's essentially why I want to see physicists such as Sabine , Terry Tao, and Brad Marston do more. They know more physics than others practicing climate science, based on their education, experiences, and mathematical insight, and so are not afraid to apply foreign concepts to a related field. Who knows, they may also say that my analysis is misguided. The point is that you can't make progress in a scientific discipline unless you explore the boundaries. |
Beta Was this translation helpful? Give feedback.
-
Piotr said: “I don’t think gravitational attraction by the Moon has changed, and in one direction, all that much over last several decades …” Common misconception here. Because of the incommensurate nature of the major long-period orbital cycles, the gravitational forcing with respect to the Earth is continuously changing and the repeat pattern is measured in millennia, see Keeling & Whorf, 2000 — yes that Keeling. Consider the lunar and solar declination and perigee cycles (4 possibilities right there), the approximate least common multiple repeat period is about 1000 years for just 2 of these. Take the 18.6 year lunar nodal declination cycle and 8.85 year lunar perigee cycles and that only repeats every 1097.4 years. But then that does not align on an annual cycle which is required for ENSO, so that 5 x 1097.4 = 5487 years is repeat for a hypothetical ENSO cycle that is triggered by a seasonal impulse modulated by a lunar tidal forcing. In practical terms, this means that if ENSO behavior is aligned to a nonlinear response of the fluid dynamics to the precise tidal forcing (ala Laplace’s Tidal Equations used in GCMs) then the 150 years of measurements is not even close to being enough data to discern a pattern from inspecting the data alone. That’s why I think that machine learning neural network experiments that are trying to extract patterns from only the data are potentially misguided. The evaluation of NNs occurs from fitting nonlinear interactions within the data to the data itself, but if the important external forcing component is not included then the real pattern won’t be discovered. Now here’s where the breakthrough is: consider that ocean cycles are pinned to specific Earth geographic configurations, which enforce certain lunar and solar orbital cycles (think in terms of frequency of eclipses in a certain region VS occurring anywhere in the world). However, an atmospheric cycle such as QBO has longitudinal invariance which means that some of the incommensurate periods are removed from consideration. In fact, because of this higher topological symmetry and thus reduced DOF involved in the QBO behavior , the lunar+solar cycles are more easily revealed in a short time-series. And that’s exactly the case as I have published previously; as a succinct overview, here is a mathematical explanation that I recently wrote up: https://geoenergymath.com/2024/03/25/proof-for-allowed-modes-of-an-ideal-qbo/ That’s why I haven’t given up on ENSO, as I realize that the lunar +solar patterns have to be more complex. Perhaps the more trivial nature of QBO gave me too much hope of finding a related pattern in ENSO, yet I haven’t thrown in the towel yet and blamed it on chaos or randomness. It’s getting easier to fit the data and the cross-validation results are improving so there is light at the end of the tunnel. Yet, I still remain mystified by the climate research effort being expended assuming internal variability and the big guns at Google and NVIDIA that are doing machine learning w/o including the external factors. And in the context of this RC post, it’s possible that the 2023 heat spike may be the result of a confluence of lunar + solar factors with a nonlinear response that emerges from a comprehensive analysis. |
Beta Was this translation helpful? Give feedback.
-
Piotr claims: "– at a longer end – the same Pukite lecturing me: “ Common misconception here” and then coming up with a cycle …. “measured in millennia“. and then trying to gain credibility by association (with Keeling, “yes that Keeling” [(c) P. Pukite] You said "I don’t think gravitational attraction by the Moon has changed". Again, what do you mean by this? From the text by Pugh and Woodworth (2014), "Sea-Level Science: Understanding Tides, Surges, Tsunamis and Mean Sea-Level Changes," "Levels defined by analysis of long periods of sealevel variations are used to define a reference level known as a tidal datum. These datums are often used for map or chart making, or for referring subsequent sea-level measurements. For geodetic surveys the Mean Sea Level (MSL) is frequently adopted, being the average value of levels observed each hour over a period of at least a year, and preferably over about 19 years, to average over the cycles of 18.61 years in the tidal amplitudes and phases (see Chapter 3), and to average out weather. " This value is often chosen because two perigean cycles of 8.85 years will fit into this 18.6 year nodal precessional interval, But of course this is not an exact commensurate match, which is why the tidal analysis has to be updated regularly. See also country-specific guidelines such as the NOAA Tech Report NOS CO-OPS 068 https://tidesandcurrents.noaa.gov/publications/NOAA_Technical_Report_NOS_COOPS_68.pdf and Figure 1 in particular. And from a recent article, it's worse than this : "Analyses of coastal tide gauge measurements spanning the past ~30–100 years have shown that ocean tides undergo unexpected changes on interannual, decadal, and secular time scales. The changes themselves are rather subtle—typically ~1–3 cm in amplitude over a century—but large enough to rule out a connection to variations in the astronomical forcing." -- Opel, L., Schindelegger, M. & Ray, R.D. A likely role for stratification in long-term changes of the global ocean tides. Commun Earth Environ 5, 261 (2024). https://doi.org/10.1038/s43247-024-01432-5 IMO it is straightforward to model ocean indices such as ENSO and AMO using a common tidal forcing. The periods of 18.6 years or 18 years (Saros) and ~4.5 years (about 1/2 of the perigee cycle because of the antipodal tide) always appear in the forcing modulation patterns. Yet, one has to understand that these ocean cycle instrumental records extend for over 100 years, and the challenge is to keep track and calibrate these so-called long-period tidal factors. Yet, I think the reason that it will be hard to get buy-in to the significance of this approach is because ocean cycle events occur on a interannual basis, which is the calibration scale, a far cry from the daily tidal patterns that most don't even blink an eye over. |
Beta Was this translation helpful? Give feedback.
-
Susan said: It’s a fact of life now. Even for the use of supercomputers for solving climate/MET models it’s a fact. It’s also sad that there’s a Green(!!!) 500 list of energy efficient supercomputers –https://en.wikipedia.org/wiki/Green500 . The most efficient ones on the list around 1 Exaflops are still comparable to the power consumption of a Boeing 737 flying at altitude, and that’s considering the extrinsic factor of cooling the systems with water while running non-stop for maximum utilization. Hopefully they use the waste heat for the buildings. Just think about the number of electrons traveling through all those chips with nothing accomplished but colliding with lattice defects, phonons, etc to dissipate that kind of energy through heat! That’s why I have adopted as my recent crusade to advertise working smart and not hard in solving climate models. There’s absolutely no excuse for researchers to not cross-check model results that can potentially capture ocean cycles by computations running on a laptop. https://geoenergymath.com/2024/09/23/amo-and-the-mt-tide/ OK, so what if this model doesn’t pan out? How much energy is wasted? Is it the equivalent of fleets of 737’s running around the clock while not making any progress in explaining what caused the 2023 temperature or being unable to predict an El Nino beyond a year? Face it. If nothing else, these natural climate patterns will be found out by a machine learning experiment run by Google or NVIDIA. And like what happened with the most recent physics and chemistry Nobel prizes, it won’t go to a traditional physicist or chemist but to someone applying AI. Just think if someone the equivalent of Susan’s father https://en.wikipedia.org/wiki/Philip_W._Anderson will no longer receive Nobel physics awards at the expense of machines? Let’s get the recognition before that plays out ;) |
Beta Was this translation helpful? Give feedback.
-
https://www.realclimate.org/index.php/archives/2025/03/we-need-noaa-now-more-than-ever We’re ignoring the fact that other countries are still going to support their scientists in doing climate research. So, let’s continue to highlight efforts around the world, and do what we can given the circumstances. One thing that I continue to harp on is to push for an advanced Earth science discussion forum, and especially one that is not hamstrung like RC in posting capabilities (which lacks images, equations, charts, etc ). They are out there, but they actually require PARTICIPATION to make a difference — who’d a thunk that? For your consideration, The Azimuth Project, which was started over 10 years ago, but was then deleted by its owner. I recovered the remnants to it with links to archives here, https://github.com/azimuth-project and a GitHub discussion page, which is very capable of supporting any kind of post featuring technical details. People that don’t want to see the USA start falling behind in Earth science are all welcome to join.. BTW, The Azimuth Project forum has no moderation for posting as long as you have a GitHub account, but if Mr. Know It All or Rob Bradley shows up, they are toast. |
Beta Was this translation helpful? Give feedback.
-
This comment won't post on RC https://www.realclimate.org/index.php/archives/2025/03/wmo-update-on-2023-4-anomalies The superposition of spikes in 3 climate indices -- ENSO in Pacific, AMO in Atlantic, and IOD in Indian -- culminated in the massive global temperature spike. It seems to have now ended as AMO was the last to recede. Now, as far as sophisticated modeling of these climate indices, I could arguably say that my published approach accounts for all of the indices concurrently. The basic fluid dynamics math is outlined in Mathematical Geoenergy (Wiley/AGU, 2019) and I've presented the forcing calibration technique at the 2018 AGU (https://www.authorea.com/users/527249/articles/604670-ephemeris-calibration-of-laplace-s-tidal-equation-model-for-enso). This has more recently been greatly simplified by using a model of the Earth's variation in Length-of-Day (LOD) to represent the angular momentum forcing. The other simplifying ansatz is an application of a biennial impulse as described at the 2017 AGU (https://agu.confex.com/agu/fm17/meetingapp.cgi/Paper/221914). Nothing in regards to the basics I've outlined has changed since, and the subsequent cross-validation effort put into the modeling has only further substantiated the ideas. To put it succinctly, the common-mode effect of the forcing creates a pattern of climate index spikes that will occasionally synchronize, and when all of these also happen to be large, a massive spike impacts the entire Earth. p.s. Attended 5 conferences in 6 years to drum up interest in the approach: |
Beta Was this translation helpful? Give feedback.
-
https://www.realclimate.org/index.php/archives/2025/05/unforced-variations-may-2025 My comments on a recent symbolic regression machine learning seminar The EGU is going on this week but the big breakthroughs in fluid dynamics may be happening at the Royal Society meeting on “Symbolic Regression in the Physical Sciences”, This one by Nathan Kutz of U of Washington. The video starts at a model of Pacific SST 2nd day of meetings
In machine learning circles, that’s referred to as cross-validation and is the workhorse of training. One can arguably assert that advances in cross-validation techniques is what make neural network and symbolic regression tools practical for real-world use. And as I have said before, climate science prediction must get on board with cross-validation. It’s misguided to create long-range predictive models and then have to wait years or decades to observe whether the observations match the model predictions. Behaviors such as ENSO are perfectly amenable to cross-validation with the already existing historical data. That’s why the machine learning community is so excited about mining historical climate data, as they may understand better the value of cross-validation than the ordinary climate scientist, who have been taught to rely on the conventional forecast/observe cycle for validating models. Sure, that’s fine for behaviors such as tidal SLH forecasting, where one just has to wait a few days to months to calibrate a model, but becomes hopeless for a long-range ENSO model that requires correct forecasting of the next 5 El Nino peaks over the coming decades. That’s the career span of a climate researcher. That brings me to the point of why climate scientists gave up so quickly on a tidal forcing basis for ENSO modeling. It’s painfully obvious that the main spectral peaks of an ENSO index such NINO4 align precisely with the lunar tidal sub-bands around the annual frequencies (and the annual harmonics). Is it again that the average climate scientist is not up-to-date with signal processing skills, just as they are not skilled with cross-validation approaches? Sorry to beat the drum here, but I published all this in late 2018 and yet no one seems to be paying attention. Lots of insight in these ML presentations. Nathan Kutz suggested that just because you measured it, it doesn’t make it the right variable. Pick the right coordinate system and the answer may be in some intermediate representation that’s a transform away from matching the results. When Prof. Kutz essentially pointed out via Zebra’s observation “look, we have these three equations about Pacific SST, but now we have to assign physical identities to the symbols and construct a causal narrative”, his other objective is to raise awareness of just looking at the data from another perspective, i.e. this transformed representation that doesn’t on the surface have anything to do with the SST that everyone is looking at. That’s part of the power of ML — to go down different paths. One of Kutz’s collaborators, Steve Brunton has many educational youtube videos on various applied math and applied physics topics such as fluid dynamics. Remember that fluid dynamics is one of the original “big data” sciences, and many advances in ML came out of fluids. These scientists are serious about connecting the AI results to physics and math, which is why I am following them closely. And in terms of what many consider to be “spurious correlations”, there are simply too many signal processing artifacts that point to tidal forcing as guiding the natural variability of the climate indices, both in the ocean and atmosphere. The issue here is that AFAIK, none of the ML training sets are including tidal info, and so is likely finding any patterns within the data itself. That’s the hard way of doing things, and one is left without a causal connection (if one exists).
This is a naive viewpoint. I’ve been involved with aspects of AI for years in a professional setting and so will just point out that once some feature of AI is incorporated into current technology, it’s no longer referred to as AI. One can go through all sorts of algorithms, such as camera autofocus, spam filters, recommendation systems, speech recognition (Siri, Alexa, etc.), predictive text & autocorrect (even tho many curse it), OCR, and fraud detection (for credit cards, banking) that once fell under the category of AI. So that essentially AI is defined as whatever hasn’t been fully commoditized yet. The moment it works reliably, people stop calling it AI. That’s why AI always seems “just around the corner”—because by definition, once it arrives, it’s no longer considered AI. In terms of image and audio AI, so much of what has been called AI has benefited form advanced signal processing techniques such as Fourier analysis, wavelets, convolution/de-convolution, Kalman filtering, PIDs, Markov chains/HMM, diffusion models, etc. I keep on saying this, but climate scientists have still not scratched the surface on applying signal processing to climate data, especially natural climate variation such as El Nino. Consider just Fourier analysis. I recently revisited a Fourier analysis of the ENSO NINO4 time series, that I first looked at around 10 years ago. I still see the same indications of tidal forcing in the Fourier spectra, with EVERY primary lunar tidal showing up as the ONLY symmetric sidebands of the annual carrier, which I had noted before (and subsequently published). The upshot of this is that the erratic cycling of ENSO is clearly a result of tidal forcing interacting with the annual cycle. Now, here’s the deal … its just a matter of time until some machine learning training session applied to climate data discovers the same signal processing artifacts as I have laid out. Other scientists are getting close, such as this recent paper by a NASA team “There is no six-year periodicity in tidal forcing” https://www.nature.com/articles/s41598-025-97361-0.epdf, but they are missing the forest for the trees. Another aspect of the Kutz presentation is the importance of separation of variables — in time and space. This naturally leads to the idea of standing wave modes, which are ubiquitous whenever a separation is achieved via partial differential equations. The observation in ocean indices is that the spatial modes are fixed and never divert from that configuration, while the temporal modes contain the complexity in their seeming erratic nature. My contention (and which is supported by others) is that since the spatial standing wave modes are observed to be invariant (i,.e, fixed) over time, the temporal modes may be erratic but not necessarily chaotic. Then, it’s a matter of deciphering the temporal pattern based on historical observations and one can obtain a deterministic and potentially predictable time-series. Kutz mentions measuring at fixed points in space and watching how that evolves, which is very natural for a standing wave behavior such as ENSO. And again the concept of a latent variable is important here, which I called an intermediate representation in a previous comment, in that what you are measuring is not the important variable for modeling. IMO, that me be the key, as I use a latent variable that gets transformed to the model of the observational result in my own research. Although I never called it a latent variable — for example in electro-magnetics if one were interested in the electric field but only the magnetic field was available to be measured, that would be a latent variable. Yet, no one ever called it a latent variable in E-M textbooks for engineering or physics. That’s the whole idea of cross-disciplinary learning — much can be transferred from other disciplines. Barry Finch said:
This is all qualitative analysis and narrative not really worthwhile to respond to. If you want to make a difference, the only progress will come from quantitative cross-validation of the hundreds of peaks and valleys of ENSO and other climate indices measured over the past 100+ years. I offer up again that if anyone wants to contribute, join the Azimuth Project discussion forum, where anyone with a GitHub login has free reign to post detailed analysis and charts The Trump regime is trashing all federal science funding so even though this is not a replacement, at some point individuals may have to take collaborations into their own hands — which is what a NSF grant essentially is, a collaboration between a selection panel + funding managers with members of a research team. We can do collaborations in climate science on our own. Of course, one can continue to do hand-wavy arguments and claim that you did AI on an 8-bit CPU in 1975, or you can roll up your sleeves and get your hands dirty. |
Beta Was this translation helpful? Give feedback.
-
I never got resolution over a citation oversight from 2 years ago. A Chinese research group reporting on ENSO predictions wrote this:
Strange citation in the excerpt above :
No such description is found in Keppenne and Ghil, but this analysis does appear paraphrased in a preprint I uploaded to the arXiv site https://arxiv.org/abs/1411.0815 "Sloshing Model for ENSO". From my abstract:
This preprint eventually evolved into a chapter of Pukite et al, Mathematical Geoenergy (Wiley/AGU, 2018), with different wording. This may be just an isolated citation error, but it's possible that they actually believe Keppenne and Ghil did this type of analysis I made a PubPeer entry but never got a response: https://pubpeer.com/publications/38757753E4187899A00B220E81777F |
Beta Was this translation helpful? Give feedback.
-
I didn't post this to RC after writing it: The biggest elephant in the room concerning AGW is still the physical properties and characteristics of CO2. What makes it so effective as a GHG -- it's ability to accumulate in the atmosphere over time -- also makes it really immune to any kind of practical mitigation. So, whatever is in the atmosphere right now, it ain't going anywhere for the long term, and all these strategies for CDR are pipe dreams. The amount of effort that it will take to unwind this is comparable to the amount of effort that it took to emit it all into the atmosphere in the first place. IMO, the issues is so intractable that it's debilitating to even think about. For example, I teed it up for ChatGPT here: And then I added a response "You're wrong about mitigation -- as it's easy to remove all that CO2 that has accumulated" -- to which ChatGPT responded Is there a particular approach or innovation you believe is being overlooked?" This is one of those "talk to the hand" impasses. If you can't convince ChatGPT that the futility of CDR schemes is unwarranted, then the facts and reality of the situation are set in stone. Working on ENSO models is much more fun and less intractable than solving CDR, which explains where my head is at right now. It's possible that this multiple-handle (Poor Peru/Ned Kelly/Dharma) troll keeps at it is to remind everyone that this ain't no picnic, as The Minutemen once sang. |
Beta Was this translation helpful? Give feedback.
-
what do you want to know? The frequency spectrum of NINO4 has peaks at 0.43. 0.36, 0.27/yr, which are exactly as predicted for annual aliasing of the main monthly lunar tides. The majority of the other secondary peaks line up as well, in this case when interacting with the lunar perigean cycle. I wrote and published this aliasing analysis in 2018 Can peruse through all the model fits on the GIT GIST site here: One of the brilliant insights IMHO is to do this same analysis at the ocean coastal sites that track sea-level heights (SLH). I have focused on 3 sites that have had near continuous measurements for over 100 years — Ft Denison in Sydney Harbor, Brest on the French coast, and Honolulu. Using monthly averages (which filter out the diurnal and semi-diurnal cycles) and removing the largely predictable annual cycles, one can cross-validate in a similar fashion to NINO4. No one else does this kind of analysis. Maybe it’s been overlooked — jeez, could tidal forces cause tidal changes in these locations, but perhaps in a slightly different way that has been applied to the daily tidal cycles? Nah … that’s too obvious to consider (yes, I know that the inverse barometric effect plays a role here as well, but the root cause may be the same). In any case, the same analysis with nearly identical tidal forcing has been applied to these long term ocean cycles (all monthly and over 100 years in length => The only index that doesn’t match in terms of a common forcing is the atmospheric QBO time series, And that’s OK because it has global group symmetry and so only a few of the lunar forces apply. The intriguing bit about the GIST site is that it’s really not that hard for someone to do a deep learning screen scrape of all the data and try to replicate the modeling results via whatever machine learning algorithm is appropriate. Guaranteed it will pick out the pattern easily. Piotr can try his hand at finding a concordance with sunspots LOL |
Beta Was this translation helpful? Give feedback.
-
The AGU posted this yesterday on BlueSky:
https://bsky.app/profile/agu.org/post/3lr4nrvqk6b2r Please don’t take us for idiots. The two differential equations referred to relate subsurface heat content to temperature at the surface. At best, they describe a typical second-order response of the forcing (i.e. heat content) as it presents as an SST. It still in no way describes the “key dynamics” of subsurface heat redistribution via a thermocline modulation. Some of us know math-based physics and don’t easily get fooled by an assertion backed by “equations” — we’re not that gullible,. Something mechanical is causing the thermocline to vary and it’s certainly not sunspot variations. The fact that the thermocline exists within a reduced effective gravity environment means that any of the expected gravitational or inertial forces are greatly amplified in strength. Sunspot variations are not gravitational or inertial. Yet lunar and seasonal solar variations have plenty of mechanical torque at their disposal. The seasonal modulation is very apparent in the ENSO time-series and associated frequency spectrum, while the lunar modulation is readily apparent if aliasing of lunar cycles with the annual modulation is taken into account. As I said above, the expected spectral locations of aliased lunar periods match those predicted. This really should not be that difficult for a lay-person to understand, in contrast to the AGU playing the “elegant math” card to an unsuspecting public. Everyone at least appreciates how tides are a response to lunar cycles, and so that needs to be at least considered on a more global basis. Consider the over 100 year time series of sea-level-height in Sydney harbor – diurnal cycles account for likely 95% of the variation, while annual comprises perhaps 4%, and the rest (which I assert is lunar nonlinearly modulated by annual) is 1% at most. The key finding is that this 1% residual variation matches exceedingly well the ENSO variation as exemplified by the NINO4 time-series. So why is it that if SLH variation is almost 100% accounted for by lunisolar variation, why can’t the research community even allow for the consideration that the ultimately much more gravitationally-sensitive Pacific ocean thermocline is not impacted by lunisolar forces in a similar fashion? If that’s the case, ENSO may in fact be 100% driven by tidal forces. This is just logical thinking, and not some ego-driven mindset that Piotr believes every curious scientist is afflicted with. Refs:
|
Beta Was this translation helpful? Give feedback.
-
Coriolis is a virtual deflecting force which will lead to complex dynamics when interacting with a varying force, such as the buoyancy driven impact of lunisolar: forces. Take a look at this frequency spectrum of the ENSO NINO4 time-series: This spectrum acts like a fingerprint (or akin to a diffraction pattern of a crystal lattice) where every peak is explained by a specific cyclic tidal interaction interacting with the annual cycle and subharmonics (4/2, 4/3, 4/1). It appears erratic when viewed in the intuitive time frame but reveals regularity when viewed in the reciprocal or 1/time frame. Why don't climate scientists look at the data this way? That's a good question, as Earth scientists and specifically geophysicists have long been innovators in spectral analysis -- just consider Burg's work on Maximum Entropy Spectral Estimation. which was used to great effect for seismic oil exploration. This is actually a toehold into further analysis, as there are other spectral approaches one can use given prior information is available, or at least hinted at. Wish that others were intellectually curious about exploring this path. Submitted for moderation: https://www.realclimate.org/index.php/archives/2025/06/unforced-variations-jun-2025/?unapproved=834484&moderation-hash=71bf0c9e6d3d3b112936e5cff25d2ef9#comment-834484 |
Beta Was this translation helpful? Give feedback.
-
Read the previous entry in this RC thread. Now notice how Piotr tries to marginalize what I have researched by taking various quotes of mine completely out of context. And then he hammers at it over and over like a skipping record. But .. you know what? It doesn't really matter. Most everything of interest I have reported on here was published before 2019. All I'm doing is ironing out ways to communicate the results so that it eventually sticks, and that's all that Piotr is criticizing - not the foundational substance. He has nothing to say about that.. So yea, sure, I might try out an analogy or a simplification that may help with understanding, but if it doesn't work out, I won't get bent out of shape. I will just try some other analogy to maybe help get the ideas out there.. And eventually it will, because that's the way science always works -- the ground truth always wins out and if the math supports it, that becomes the model that sticks as a consensus view. So when some machine learning result, say from NVIDIA or Google, discovers the same results that I have presented, we all win. What will then happen is that you will prompt a LLM with the findings and then it will match to my results first, and then perhaps mention other results secondarily. As of now, if you prompt an LLM, it will describe my results, but provide a caveat that it''s not a consensus view. Consider that Piotr said: 'And nobody stops you discussing it on your “ENSO blog”. ' So all the effort I have put into scientific social media such as blogs won't go for naught. The training of an LLM will pick all that up and it goes into it's set of pattern matches. Piotr will get steamed by this of course, but welcome to the future of scientific discovery and the process of establishing precedence for citations. |
Beta Was this translation helpful? Give feedback.
-
"Do I understand correctly that variations in SOI records do correlate well with tidal frequencies you have identified by your spectral analysis?" It correlates remarkably well. Better to phrase it as the primary tidal frequencies occur exactly as predicted in a spectral analysis -- all that changes is the amplitude and phase of the peaks, which is always what happens in a practical tidal analysis. I use the GIST site at GitHub as a mini-repo for maintaining results. Recently I went through a quick cross-validation of all the long-term climate indices I could think of, each at least 100 years. There's a huge (blind-spot, mental block, brain freeze, mind control, ???) going on with respect to the impact of tidal factors across the board in climate science. Whatever I am relaying right now is already in the bank, as far as publishing prior findings circa 2018/2019. It's only that the more I look at the data with various signal processing and modeling approaches, the more significant the tidal fingerprints emerge. The blindspot in understanding is likely related to (1) ignorance of long period tidal factors in deference to the daily cycle, (2) non-recognition of the nodal tidal cycle wrt to the ecliptic plane, and (3) modulation of the tidal factors with the annual/biennial cycle leading to aliasing. This is described in detail in Mathematical Geoenergy This is extended to other indices such as the atmospheric QBO and solid-body Chandler wobble The tidal contribution to the Chandler wobble is unmistakable given the fingerprint of the CW frequency of 0.845/y being replicated strongly as a sideband harmonic at 1.845/y. That is impossible to achieve as a natural resonance -- as the conventional wisdom of Chandler wobble theory would have it. In fact, this value can only occur as a forced response. It's a mental block on the entire geophysics community, that's how I can rationalize it. |
Beta Was this translation helpful? Give feedback.
-
![]() www.realclimate.org/index.php/archives/2025/07/unforced-variations-july-2025/#comment-836260 |
Beta Was this translation helpful? Give feedback.
-
ATTP comment: None of this applies to the majority of topics in physics, where controlled experiments are available to verify a new idea, or alternatively to debunk or falsify a wrong new idea. Just look at the room temperature superconductor finding from a few years ago. It was just in a matter of weeks that the claim was debunked — all because others scientists tried to replicate the finding and were unable to do so. https://www.chemistryworld.com/features/superconductivity-the-search-and-the-scandal/4019292.article So that brings up the physics topics that have no controlled experiments — mainly in astrophysics and the Earth sciences, where scale (both in size and time) and lack of a gravity emulation make it impossible to do much in a lab setting. Often one can neither debunk nor validate a novel claim within a scientist’s lifetime, so it sits there in limbo serving as a punching bag to opposing schools of thought. It also explains why Earth scientists are reluctant to take on novel concepts — they know that validation will be slow. On the other hand, condensed matter physicists will gleefully jump on new ideas, eager to scoop other scientists in a highly competitive atmosphere. I’ve been to both kinds of conferences — a Material Research Society or American Vacuum Society conference is full of attendees secretly taking photos of posters and recording talks, while an AGU or EGU is more of a jaded atmosphere where attendees are more interested in beer and meeting up with colleagues than finding the next breakthrough. |
Beta Was this translation helpful? Give feedback.
-
Beta Was this translation helpful? Give feedback.
-
AS said: Regarding this aspect of natural oscillations, I've been modeling the residual sea-level height oscillations measured in many of the ocean ports across the world with at least 100 years worth of data. The main website keeping track of this is called the Permament Service for Mean Sea Level (https://psmsl.org/data/obtaining/map.html). So far I have tallied 78 sites and am working to cross-validate the models. Summarizing the approach (1) The diurnal cycles in readings are removed by PSMSL resulting in mean monthly values (2) I also filter out the annual cycle by applying a 13 pt boxcar filter and remove any linear trend (some ports such as in the Gulf of Bothnia show a decreasing trend), (3) what remains is an erratic cycle that sometimes reflects the influence of El Nino/La Nina cycles, such as the set in Sydney harbor Ft. Denison in Sydney harbor The region indicated by dashed lines is the cross-validated interval Another data set is from the port city of Brest in France This does not show as much of the ENSO behavior, but the model demonstrates excellent cross-validation over the dashed test interval. There are also some missing readings after the year 1945, which are filled in by the model. If there are any auxiliary readings from this time period, it would provide additional validation. So I have 76 other sea-level sites that show predominately a good cross-validation over a test interval. The novelty of the model is that it applies a lunar period modulation to an annual cycle, resulting in an erratic time-series that though complex, shows predictability (as per the cross-validation results). No other climate researchers are doing anything close to this approach, and now especially in the USA with funding sources drying up due to the Trump MAGA effect, I doubt anybody else will pick it up soon. The final reveal on all this of course is that the approach provides discrimination to the secular trends in sea-level rise, enabling a view without the natural cycles getting in the way. |
Beta Was this translation helpful? Give feedback.
-
“Recent intensification of wind-driven circulation in the Pacific” https://www.nature.com/articles/nclimate2106 Wind is a confounding indicator. Observation of increased winds may be meaningless as it can likely be the result of the formation of an El Nino itself — wind is driven by an atmospheric pressure gradient, and such a gradient is developed during the initiation of a subsurface thermocline ENSO dipole extending spatially along the equatorial Pacific. Incidentally, discussion of confounding indicators is in the recent news. MAGA doctors and scientists are being pushed by Trump and RFK Jr to claim that Tylenol usage is leading to increased autism rates. Yet an obvious confounding indicator to this claimed correlation is that older women (> 35 y) giving birth are more likely to have autistic children, while they are also more likely to take Tylenol as a normal course of aging with accompanying aches&pains. Another confounding possibility is that there is an unknown illness contributing to autism, which is expressed by increased usage of Tylenol. So wind associated with an ENSO spike is akin to Tylenol associated with adverse birth outcomes, both are there more as a result of some other underlying causal factor. Open to further discussion on this, as ENSO has an obvious overlooked causal factor. https://github.com/copilot/share/827d01b4-0144-80f5-b851-1202000f2049 |
Beta Was this translation helpful? Give feedback.
-
Good research on tidal gauge data. I appreciate the work done on filling the gaps in the time-series data, especially for the Brest station, which has a ~12 year gap between the 1940's and 1950's. Using my technique for modeling long-period tides, I can show excellent cross-validation using training up to 1930 and after 1970, and thus filling in everything in between with interpolated values. In the chart above, the gap in missing values is shown as a straight line. I don't show the interpolated missing values, simply because of the plotting algorithm being used, but these can be checked by using data from nearby sites, as your paper describes. Yet, if the already good cross-validation of the other test values - in the 1930's and 1960's validation interval (dashed interval) shown in the chart - - is any indication, the missing values will likely also show a good correlation when filled in. That would provide a completely unbiased validation of the model as the unseen data is completely out-of-band. |
Beta Was this translation helpful? Give feedback.
-
condescendingly smug? ![]() Peer-review? check
Could argue that none of this is actually new. All I'm doing is modeling mean sea-level (MSL) variations with lunisolar tidal cycles. The following sites are all aggregators of MSL data with varying degrees of resolution. PSMSL : https://psmsl.org/data/obtaining/map.html I just found out about the NASA web site above. It has an interesting option to remove subsidence at each station, and it also includes the satellite altimetry plotted along with the tidal gauge measurements. The two (gauge & satellite readings) align very well, as you can note at Stockholm: So why don't they go the extra step and provide the useful model-fitted results, similar to what I've done with the stations that have long enough records? Dunno for certain. Is it that they don't know how to do the harmonic sub-banding that is required to generate a nonlinear tidal response? Possibly. Whatever the reason, once it all gets rediscovered by someone or by ML, it will lead to one of the all-time "DOH! (or V8)" moments in science. Most would think it would be impossible to overlook something this obvious. If that is indeed the case, then the situation would be reversed when or if somebody points out what the models are getting wrong. |
Beta Was this translation helpful? Give feedback.
Uh oh!
There was an error while loading. Please reload this page.
-
https://www.realclimate.org/index.php/archives/2024/01/unforced-variations-jan-2024/
Tidal forces are inexorable and occur day or night. Because of the enormous reduction in the effective gravity at the thermocline interface, the tides can raise or lower the depth at which the thermocline occurs at will.
I find it amazing that people are enthralled by zero-gravity antics paid for by billionaires such as Musk, Branson, and Bezos, yet are not educated on gravitational physics that are impacting the climate variations in day-to-day life -- see El Nino, AMO, PDO, IOD, QBO, MJO, etc.
That's a rhetorical question because fluid dynamics is challenging subject matter and unless the mathematics is correctly formulated and the long-period tidal forces are correctly calibrated, climate scientists will continue to spin their wheels and not gain any traction in making progress in understanding ocean dynamics.
"patrick o 27" whoever that is has tried to engage on this topic in past RC comments but has gotten bogged down in word salad.. There are ways to get beyond this and actually fit the ocean cycles and perform the cross-validation necessary to substantiate the models.
Beta Was this translation helpful? Give feedback.
All reactions