The ocean's past is not a simple archive. It is a set of unresolved equations, and the Last Glacial Maximum offers one of the clearest tests we have for whether our climate models can actually solve them. The new comparison of MIROC-ES2L and MITgcm-REcoM2 against proxy reconstructions does more than map disagreement; it exposes a fundamental truth about how we understand carbon cycling. Productivity in the glacial North Pacific was not uniformly higher or lower, and the models that disagree are not merely imprecise. They are telling us that the system's response depends on a coupled balance among iron supply, macronutrient resupply, and physical mixing. That is the real story, and it is one we should read carefully before trusting any single simulation of past or future oceans.
The key finding here is that dust-driven iron fertilization alone cannot explain the proxy record. MIROC-ES2L simulates widespread productivity gains and higher subarctic nitrate, while MITgcm-REcoM2 produces a more muted subarctic response and a subtropical increase that aligns better with the spatial contrast seen in reconstructions. The models diverge sharply in mixed-layer depth, background nitrate, and dissolved iron, which means their outcomes are shaped by initial conditions and circulation dynamics, not just atmospheric deposition. The high-resolution ROMS-CoSiNE simulation sharpens this point: enhanced iron availability can shift the subarctic Pacific from iron limitation toward nitrogen or phosphorus limitation, with productivity rising where physical processes replenish macronutrients, such as the western subarctic Pacific and the Kuroshio Extension, and falling in the northeastern Pacific where intensified consumption depletes those nutrients. This reconciles the apparent paradox of greater glacial iron supply yet lower productivity at some sites. The system is not driven by one lever; it is driven by the interaction of several, and that interaction is precisely what coarse models often miss.
This matters beyond paleoceanography. If we cannot reproduce the glacial North Pacific's productivity response without realistic nutrient fields, circulation, and mixing, then our confidence in future projections of ocean carbon uptake is weaker than we would like. The same mechanisms that govern this past end-member state, iron limitation, macronutrient resupply, and vertical mixing, are active today in the Southern Ocean and the equatorial Pacific, where models struggle with similar issues. The study's implication is direct: any simulation that relies on dust forcing alone, without resolving the physical oceanography that redistributes nutrients, is building on sand. We should treat model agreement with proxies as a necessary but insufficient condition, and model disagreement as a diagnostic tool rather than a failure.
The practical takeaway is precise. For researchers, this means that initial nutrient fields and mixing parameterizations are not tuning details; they are first-order controls on simulated carbon cycle outcomes. For those interpreting paleo-proxies, it means that a single site's productivity signal cannot be read as a direct measure of iron supply. The glacial North Pacific was not a simple iron-fertilized system. It was a place where iron and macronutrients competed for control, and the winner varied by region. The open question now is whether Earth system models can adopt the high-resolution physical dynamics that ROMS-CoSiNE reveals without sacrificing computational efficiency. That is the next hurdle, and it is one we should watch closely as the field moves toward more integrated ocean intelligence.
