TerraPulse Theoretical Framework

"Nature is the ultimate optimization engine, seeking to minimize the action." — B. Isenbek

The Premise

There are feedback loops between every natural system. The Earth is not a collection of independent domains — it is a single coupled dynamical system, perpetually out of equilibrium, always seeking balance. Every adjustment in one domain triggers responses in others. The structure of these couplings is the puzzle we are trying to piece together.

The Engines

Three fundamental engines drive the Earth system:

Engine	Domain	Energy Source	Timescale
Solar	Electromagnetic radiation, solar wind, UV, X-ray	Fusion (4.6 billion years)	Minutes to centuries
Geothermal	Mantle convection, plate tectonics, volcanism	Radioactive decay + primordial heat	Years to billions of years
Gravitational	Tidal forcing (Moon + Sun), orbital mechanics	Gravitational potential energy	Hours to millennia

And possibly others:

Engine	Domain	Notes
Cosmic	Galactic cosmic rays, supernovae, gamma-ray bursts	Svensmark hypothesis: cosmic rays seed clouds
Biological	Photosynthesis, decomposition, ocean carbon pump	Life itself is a geological force (Gaia hypothesis)
Anthropogenic	CO2 emissions, land use, urbanization	The newest engine, now dominant in some domains

Each engine injects energy into the system. The system distributes that energy through coupled pathways — atmosphere, ocean, crust, magnetosphere, biosphere — always dissipating, always adjusting, never at rest.

The Network

The Cross-Domain Granger Causality Network is our window into the coupling structure. What it reveals:

Temperature → Air Quality (2-day lag): atmospheric stability couples weather to chemistry
Earthquakes → Tides (6-day lag): crustal deformation couples solid Earth to ocean
Tides → Waves (2-day lag): ocean dynamics are internally coupled
Temperature → Tides (1-day lag): thermal expansion couples atmosphere to ocean

This network has a structure. It has a pulse. It has cascading effects — periods of quiet, a drip, then an adjustment triggering responses in other domains. The Granger network captures the linear, lagged, pairwise projections of this structure. But the underlying system is:

Nonlinear (transfer entropy, not just Granger)
Multi-variate (joint effects, not just pairwise)
Multi-scale (hourly tidal forcing, daily weather, decadal climate, century solar cycles)
Non-stationary (the coupling structure itself evolves)

The Manifold

The true state of the Earth system at any moment is a point in a high-dimensional phase space. Each axis is a variable: temperature, pressure, Kp index, earthquake rate, CO2 concentration, tidal stress, solar wind speed, wave height...

The system traces a trajectory through this space. The trajectory is constrained to a manifold — the set of states that are physically realizable given the conservation laws (energy, mass, angular momentum) and the coupling structure.

Our data is a projection of this manifold. Every time series we measure is a shadow of the higher-dimensional trajectory. The Granger network reveals which shadows are correlated. Transfer entropy reveals which shadows share nonlinear information. But the manifold itself — the underlying shape of the generator — is what we're after.

The Backward Pass

In machine learning, a neural network learns by:

Forward pass: compute predicted output from inputs
Loss function: measure how wrong the prediction is
Backward pass: propagate the error back through the network, adjusting weights to reduce loss

Nature does something analogous:

Perturbation: a solar flare, an earthquake, a volcanic eruption
Response: the system adjusts — magnetosphere deflects, crust deforms, atmosphere absorbs
Relaxation: energy dissipates through coupled pathways until a new quasi-equilibrium is reached

The "loss function" Nature minimizes is the action (in the physics sense: the integral of the Lagrangian over time). Every physical process follows the path of least action. The "backward pass" is the relaxation cascade — energy flowing from the perturbation through the coupling network until it is dissipated.

Our job is to build the correct model to capture and adjust the variables during this backward pass — to infer the coupling structure, the time constants, the nonlinearities — and get a better view of the underlying shape of the generator that sustains life on Earth.

The Investigation

We follow the evidence to see where it leads:

Solar wind speed correlates with earthquake rate at r=0.09 (real but tiny)
Earthquakes cluster at New Moon (Schuster test, p<10⁻⁸, after declustering)
Temperature predicts air quality at 2-day lag (Bonferroni-surviving)
70,000 tornadoes are migrating eastward at 0.26°/decade
The Keeling Curve rises at 431.89 ppm and accelerating
19 near-Earth objects passed inside the Moon's orbit in March 2026

Each finding is a constraint on the manifold. Each null result (solar cycle doesn't drive tornadoes, Bz doesn't predict earthquakes daily) eliminates a dimension. The manifold becomes better defined with every honest analysis.

The Path Forward

Phase	Method	What We Learn
Granger Causality	Linear, pairwise, lagged	Which domains are linearly coupled
Transfer Entropy	Nonlinear, pairwise	Which couplings are nonlinear
Vector Autoregression	Linear, multivariate	Joint dynamics of the full network
Phase Space Reconstruction	Takens embedding	Manifold geometry from time series
Convergent Cross Mapping	Nonlinear, causal	True causality (not just prediction)
Neural ODE / Physics-Informed ML	Learned dynamics	Data-driven differential equations

Each method peels back a layer. Granger is the first pass — we're here now. Transfer entropy (#52) is the next step. The endgame is a learned dynamical model of Earth's coupled systems — a digital twin of the planet's vital signs.

The Principle

A system that is always unbalanced, seeking balance. The wave function collapses one honest analysis at a time.

TerraPulse Lab — M. Isenbek, E. Isenbek, B. Isenbek March 2026