Planetary Boundary Layer Scheme

A PBL scheme represents vertical turbulent transport of heat, moisture, and momentum through the boundary layer. It takes the fluxes computed by the surface-layer scheme and distributes them upward, determining boundary-layer depth, near-surface temperature and humidity, and how much moisture reaches the levels where cloud forms.

Because none of the turbulence is resolved at typical resolutions, the scheme is responsible for all of it. That makes it one of the highest-leverage choices in a configuration, and the one whose effects are hardest to attribute — a PBL error surfaces as a cloud error, a convection error, or a near-surface bias, rarely as something recognisably about turbulence.

Local versus non-local

The main distinction between schemes is how they treat convective boundary layers.

Local schemes compute mixing between adjacent levels from the local gradient, usually through a turbulent kinetic energy budget. This is physically clean and struggles with convection, where large thermals transport heat from the surface to the inversion in a single coherent motion. A local scheme has no representation of that: air moving past intermediate levels without exchanging with them cannot be expressed as a gradient-driven flux. MYNN is the common local option.

Non-local schemes add an explicit term for that transport, acknowledging that convective mixing depends on the boundary layer as a whole rather than on local gradients. YSU is the common choice. They handle convective conditions better and are less well founded in stable conditions, where the non-local assumption does not apply.

Neither is uniformly better. Local schemes tend to produce shallower, moister boundary layers; non-local ones deeper and drier. Where a model sits near a regime boundary — as it does over the Galapagos, between stratocumulus and deep convection — that difference can decide which regime the simulation lands in.

The grey zone

All of this assumes the turbulence is entirely sub-grid. At convection-permitting resolutions of 1-4 km, the largest boundary-layer eddies approach the grid spacing and are partially resolved. The scheme then mixes turbulence the dynamics is already mixing, and the total is wrong.

This is unresolved. Scale-aware PBL schemes exist and are less mature than their convection-scheme equivalents. It is a real reason to be careful about claims from runs in this range: the convection may be permitted, but the turbulence scheme is outside its own valid range at the same time.

See also: the WRF PBL options in detail, surface-layer schemes beneath, LES for resolving rather than parameterising, and the physics map.