Imagine you have a bathtub filling from a faucet (fossil fuel emissions) and draining through a plughole (natural sinks). If you turn off the faucet, how long does it take to drain? The answer depends entirely on how you model the plughole — one drain or five drains at different heights. This paper says one drain, 60 years. The mainstream says multiple drains, some of which barely move, and the answer is centuries to millennia for the last 20-30%. The committed claim: a single-pool exponential model with a 60-year time constant for total CO₂ and an 8-year time constant for the isotopic signature (C × δ¹³C) reproduces the observed atmospheric CO₂ record from 1000 to 2024 and the δ¹³C decline since 1980. The two different time constants arise because isotopic exchange — driven primarily by photosynthesis and respiration on land — recycles ¹³C much faster than it recycles total carbon mass. The model's fit to the observational record is genuinely decent for what it is: a two-parameter exponential relaxation matched to the Mauna Loa CO₂ curve and Antarctic ice-core δ¹³C data. But fit quality on the historical period is a necessary, not sufficient, condition. The critical question is whether one-pool dynamics correctly predict the future — and here the paper collides head-on with the Bern/IPCC multi-pool carbon cycle models, which partition the ocean-atmosphere-biosphere system into reservoirs with time constants ranging from 4 years to thousands of years. Those models predict that roughly 20-30% of a CO₂ pulse persists for millennia because deep-ocean mixing and carbonate chemistry are slow. Van Wijngaarden and Happer's model, by construction, has no such long tail. The architecture is a simple ODE system: dC/dt = S(t) − (C − C₀)/τ, where S is emissions, C₀ is the preindustrial baseline (280 ppm), and τ = 60 y. For the isotopic product, the same structure with τiso = 8 y. This is a zeroth-order box model — no ocean layers, no soil carbon pools, no weathering feedback, no ocean chemistry saturation (the Revelle factor). The authors acknowledge this simplicity but frame the good historical fit as evidence that the complexity isn't needed. The mainstream position is that the complexity is needed precisely for the long-tail prediction that a simple model cannot produce. Integrity is the weak link. The validation is entirely self-consistent: the same historical data used to fit the two parameters is used to evaluate the model. There is no out-of-sample test, no comparison against pulse-response experiments (like bomb-¹⁴C decay, which independently constrains residence times), and no direct comparison against the Bern model predictions on the same data. The authors cite Harde (2017) and Salby, both of whom have been rebutted in the literature, but do not engage with the rebuttals. The paper does not appear to be peer-reviewed in a geosciences journal — it is a physics arXiv preprint. The milestone question is clarifying: if this model is right, the observable consequence is that δ¹³C should recover measurably within a decade of any significant emissions reduction. The Bern model predicts much slower recovery. This is, in principle, testable — but it requires a large, sustained emissions reduction that hasn't occurred. The bomb-¹⁴C record from atmospheric nuclear testing in the 1950s-60s already provides a natural experiment in atmospheric tracer decay; the authors do not discuss it, which is a notable omission since ¹⁴C decay data supports multi-pool models with a long tail. The obvious successor experiment the authors did not run: comparing their model against bomb-radiocarbon (¹⁴C) decay data, which is the gold-standard independent tracer for atmospheric carbon residence time. Honest read: they likely didn't run it because ¹⁴C data does not support a single 60-year decay — it shows a fast component (~4-5 years for atmosphere-ocean surface exchange) and a much slower deep-ocean component, which is exactly the structure the multi-pool models encode. Running this comparison would undermine the paper's central claim.
