DARWIN WRF experiment suite

A set of WRF sensitivity experiments for the DARWIN Galapagos work. The organising decision was to treat physics options as variables to be tested rather than settings to be chosen once, and the reasoning behind that is worth stating because it changes what the suite is for.

The goal is not to find the best configuration. Tuning until output matches observations produces a configuration that reproduces the observations and tells you nothing about why. It also invites compensating errors: a convection scheme that rains too much paired with a boundary-layer scheme that dries too fast can give a defensible seasonal total through two errors cancelling, and the result will fail the moment the regime changes.

The goal is to find out which conclusions survive the choice. If a feature of the modelled climate appears under every reasonable physics combination, it is a property of the system. If it appears only under one, it is a property of that configuration, and saying so is more honest than presenting it as a result.

Why the Galapagos makes this acute

The archipelago sits close to a regime boundary. Cold SST and a strong inversion keep the atmosphere in a shallow, stratiform state through the garúa season; warmer conditions tip it into deep convection. A model near that boundary can fall on either side depending on how convection is triggered and how strongly the boundary layer mixes.

So the sensitivities that would be second-order elsewhere are first-order here. A convection scheme that fires too readily does not merely rain too early — it can put the model in the wrong seasonal regime entirely.

What was varied

• Baseline: the tropical WRF setup.
• Convective triggering: Kain-Fritsch against KF-CuP, which ties the trigger to boundary-layer turbulence rather than resolved vertical velocity — directly relevant given how much rides on when convection initiates.
• Boundary-layer mixing: MYNN, where the question is low-cloud structure and near-surface stability, both of which determine whether the stratocumulus deck survives in the model as it does in reality.

What it supports

The suite is the reason the refined analysis can be read as more than one model’s opinion. Where the analysis makes a claim, the suite is what establishes whether that claim is configuration-dependent.

Run identifiers, production bookkeeping, and intermediate testing stay in private notes; the physics reasoning is the part that generalises.

See also: dynamical downscaling, cumulus convection schemes, and the model physics map.