Intermediate-complexity climate models often contain feedback loops with long time delays associated with the transport of energy across the ocean. In the case of El Niño, delays arise from the time it takes for Kelvin and Rossby waves to cross the Pacific Ocean. Traditional studies relied on long numerical integrations, which do not reveal the full structure of possible regimes. This work places the ENSO model in the context of modern delay dynamical systems theory, where bifurcation analysis becomes a standard tool.
The authors use the delayed oscillator model proposed by Tziperman and colleagues, which reduces all the physics to a single scalar equation with two delays and parametric annual forcing. Using the DDE-Biftool package, they numerically continue equilibrium positions and periodic orbits as solutions of boundary value problems, tracking their stability via the spectrum of the characteristic function and Floquet multipliers. Resonance tongues are computed as surfaces in the extended space, which bypasses the slow convergence near the excitation threshold.
A complete bifurcation diagram is constructed in the parameter plane (dk, k0), where dk is the amplitude of the annual variation of ocean–atmosphere coupling and k0 is its mean value. All resonance tongues k:ℓ with ℓ ≤ 11 are found, including the main 1:3 and 1:4 tongues, as well as torus bifurcation and period-doubling curves. It is shown that the trivial solution is stable in a large region below the torus curve, while above it stable resonant periodic orbits emerge. The parameter point used in the original Tziperman study lies slightly to the left of the 1:3 tongue, which explains the non-periodic dynamics observed there, close to a three-year cycle.
This work demonstrates that numerical bifurcation analysis of delay differential equations is now quite feasible and provides information inaccessible by simple simulation. Unlike neighboring studies that either limit themselves to simulations or consider simpler models without delays, here the resonance surfaces are systematically traced for the first time in the full parameter plane. This allows a rigorous determination of the predictability limits of ENSO and the sensitivity of the model to poorly known parameters.
The methodology can be extended to more realistic ENSO models with multiple delays and spatial structure. Also promising is the study of the overlap of resonance tongues at large forcing amplitudes, which may lead to chaotic dynamics. Another avenue is the inclusion of stochastic perturbations to compare with noise-driven linear models.
The results have direct implications for climatology, as they provide a rigorous tool for sensitivity analysis of ENSO models. They are also useful in other fields where delay differential equations arise, such as laser physics and population biology.
The next step could be the bifurcation analysis of more detailed models, such as the Zebiak–Cane model, taking into account spatial variability. It is also important to study the effect of stochastic noise on resonance tongues and compare with observational data.
In climatology, an open question remains whether El Niño is deterministic chaos or a linear system excited by noise. This work provides a bifurcation basis for distinguishing these hypotheses, showing which parameter regions correspond to stable periodic regimes and which to quasi-periodic or chaotic ones.
Kelvin waves that carry warm anomalies across the Pacific Ocean travel at about 2–3 meters per second, crossing 10,000 kilometers in roughly two months.
In Kim Stanley Robinson's novel 'Forty Signs of Rain', climate models with feedbacks and delays play a key role in predicting abrupt weather changes, echoing the idea of ENSO's sensitivity to parameters.