(see later posts here)
A simple model of the Southern Oscillation Index (SOI) does not exist. I find it important to understand the origin of the SOI fluctuations, not only because it is an interesting scientific problem but for its potential predictive value -- in particular, I could use a model to extrapolate the SOI factor needed by the CSALT model to make global surface temperature projections. This has implications not only for long-term climate projections but for medium-term seasonal weather predictions, particularly in predicting the next El Nino.
The current thinking is that the index that characterizes the presence of El Nino and La Nina conditions (also known as ENSO) is unpredictable enough to make any prediction beyond a year or two pointless. That makes it a challenging problem, to say the least.
So although the SOI is defined as oscillatory (thus the name), these oscillations are not the typically sinusoidal, perfectly periodic waveforms that we are used to dealing with, but consist of uneven, sporadic pulses that remain virtually impossible to deconstruct. Yet, the problem may not be as intractable as we are lead to believe. The key to understanding the SOI is to decode the characteristics of the waveform itself shown below in Figure 1.
The waveform is periodic alright but this is the periodicity that lies in the strange mathematical world of crystal lattices and warped coordinate systems. Follow the math on the next page and the mystery is revealed.
The first piece of evidence that suggests a promising derivation is shown by the slanted dotted line in Figure 1. Although only faintly visible, an envelope of a higher order period is discernable within the fluctuating waveform. This is very reminiscent of a Bloch wave, which occurs as a solution to a particle residing within potential energy wells modulated by a periodic lattice.
What happens is that the waveform of an electron propagating within a crystal lattice will interfere with the waveform of the lattice itself to define the band-gap structure of the electronic material. Since crystal lattices are my area of academic expertise, I know enough that solving Schrodinger's equation doesn't apply to the vast expanse of the Pacific Ocean. But we do know that boundary conditions and the shape of the ocean itself are enough to set up the equivalent of standing waves and propagating waves of complicated structure within the ocean medium. The motivating factor is that once one has seen how much work goes into calculating the strangely periodic band-structure of Silicon and Gallium Arsenide semiconductors, the SOI looks a bit like child's play in comparison (only half-joking).
The usual math formulation applied to these kinds of problems is referred to as the wave equation. This is a partial differential equation that includes spatial and temporal dimensions.
in two dimensions, this becomes
Solutions to this class of partial differential equation are critically dependent on the boundary conditions of the problem. The other important aspect is to be able to separate the dimensions of variability in the problem domain. This means that if we can separate the temporal from the spatial, and then separate the two spatial dimensions (x and y), we are off and running with a potential solution that we can check against the data.
The observation that I turn into a premise relies on the similarity of the Pacific Ocean basin to a portion of an elliptical coordinate system, as shown in Figure 3 below.
Such an elliptical coordinate system is used to solve problems that do not exhibit rectangular, cylindical, or perfectly spherical symmetry. A decomposition of the wave equation onto an elliptical coordinate system is aided by the application of the foreboding world of Mathieu functions. Keep in mind that since Arfken  has warned us that "Mathieu functions to be among the most difficult special functions used in physics", we proceed with caution and trepidation, but ultimately realize that you often have to do what you have to do. And so this is where we start.
Wave Equation in Elliptical Coordinates
I will get into more detail in a future post but with the aid of the excellent resource of Arfken's "Mathematical methods for physicists"  and the Mathematica web pages on Mathieu Functions, we can skip many of the derivation steps to arrive at a suitable formulation that we can evaluate. As a few starting definitions, we introduce a variable relating space to time via a propagating or traveling wave:
and to represent a forcing strength a linear dispersion relating the wave vector to frequency:
These wave-based identities are applied to transform the partial differential wave equation into an ordinary differential equation (ODE), which is more easy to solve -- if you don't mind working with hideous transcendental functions not available within an Excel spreadsheet.
The following ODE is known as the Mathieu equation. Although relatively innocuous-looking, the nonlinear recursive interaction of the Φ = Phi term (which represents the observable measure) and the cosine factor makes it impossible to solve in terms of other trigonometric identities.
Fortunately, we have an out available and we can find analytical solutions if we are not afraid to apply the special-case functions known by
MathieuS. These are the elliptical trigonometric analogues to the familiar Cosine and Sine function used to form the basis set of periodic solutions in a linear ODE. (Note that for a small enough q, we recover the perfect sinusoidal result).
As a kernel solution to the above Mathieu equation, we can theoretically assume a linear combination of the two basis Mathieu functions, where a is the characteristic value (a form of eigenvalue) and q is the parameter as described in the ODE. The constants c1 and c2 are determined by the initial conditions.
z(t) = c1 MathieuC(a, q, t) + c2 MathieuS(a, q, t)
The next step is to apply this result to a physical phenomenon; in this case that of the measurable SOI time series parameter. Since the SOI is formally defined as the atmospheric pressure difference between measurements made at Tahiti and Darwin, we begin with the historical data set from the Darwin location.
As an initial exploratory step, we create a parametric plot relating the SOI value against its first derivative. A purely sinusoidal waveform will appear as a circle. However, a Mathieu function will appear as an elliptically-convoluted orbit as shown in Figure 4. The Darwin data is shown in the two lower panels.
The agreement is reassuring so the next step is to linearize the data according to the ODE. Rearranging the ODE, we want to see if we can extract the characteristic value, strength parameter, and the basic frequency:
By numerically calculating the second derivative from the Darwin time-series data as a LHS value and then minimizing the error with respect to the RHS value, we can estimate these values.
The R^2 values are not very good but we are trying to find any kind of agreement so that we will live with these values for the moment. The following values come from applying a solver:
- a = 2.83
- q = 2.72
- T = 6.30 years
Note that T is close to the third harmonic of the diurnal Kola cycle or Lunar standstill used in the CSALT model.
So the last step is the most amazing of all. We take the values as estimated from the solver and use these to parameterize the Mathieu functions, and then compare it to an interval of the Darwin data from the years
This looks reasonable, so what happens if we project the Mathieu SOI model to 2080?
The model does not capture the extended El Nino conditions of the 1980-2000 interval and it actually inverts the early 1990's high pressure Darwin anomaly. But notice that it does have a long term repeat interval of 40 years as indicated by the Bloch-wave-like envelope. This 40-year number has long been identified as a significant climate periodicity. The elliptical-premise SOI Model clearly has potential.
The SOI is defined as the atmospheric pressure difference between Darwin, Australia and Tahiti in the south Pacific. Atmospheric pressure has the strong physical property that it must revert to a long-term mean of the earth's sea-level pressure. This means that an excursion beyond the mean for any length of time can not be sustained and that it eventually has to settle back to zero and then oscillate in the other direction. Since the SOI is a significant factor in the CSALT model of global temperature, and that it goes a long way to explaining the pause in current temperatures, any model of SOI is welcome as a tool to making climate projections. Other researchers have modeled the SOI as red noise , while there is a strong contingent that calls the chaos card and asserts that models are impossible due to irreducible uncertainty within the natural environment.
One thing left to do is fit to the actual SOI -- as we apply Mathieu models with the correct aligning phase shifts for each location, and then difference the two time-series. This can be compared to the actual data. More on that later.
In summary, given that we have measurements in SOI that span over 130+ years, there is now a strong possibility that we can crack the code and decipher the fluctuating waveform. Mathieu functions and their brethren have long been used to model the complex solid-state physics behind all modern-day electronics. And so it appears that a similar Mathieu-based SOI Model gives us a clue to the mathematics behind the unpredictable nature of the El Nino. It's another of those predictably unpredictable behaviors that seem to proliferate in nature, but this one may be cracked completely.