Noah land surface model

Noah is the most commonly used land-surface model in WRF. It carries four soil layers to a depth of two metres, a single bulk vegetation layer, and a simple snow representation, and returns the sensible heat, latent heat, and upward longwave fluxes the atmosphere needs.

Its job, as with any LSM, is deciding how available energy splits between heating the air and evaporating water. Everything else it does serves that.

Noah-MP is the successor, and its distinguishing feature is offering multiple options for each process — several stomatal resistance formulations, several runoff schemes — rather than one fixed choice. That is more flexible and more work: the option combinations interact, and defaults are not obviously better than alternatives for any particular application.

The thing that matters more than the scheme

Soil moisture initialisation. Noah’s soil moisture has a memory of weeks to months, so a simulation starting with the wrong soil moisture drifts for weeks while the land adjusts — and that drift looks exactly like a model bias.

The trap is subtler than it first appears. Soil moisture is usually initialised from a reanalysis, but a volumetric moisture value is not physically transferable between land-surface schemes. Field capacity and wilting point depend on how each scheme parameterises soil hydraulics, so the same number means different things in different models. Copying 0.25 m³/m³ from a reanalysis into Noah does not reproduce the reanalysis’s wetness state — it reproduces its number.

The fix is unglamorous and effective: spin the land surface up, ideally for months, so soil moisture reaches equilibrium with Noah’s own physics and the atmospheric forcing. In downscaling work this is among the highest-return steps available, and skipping it has cost me more time than doing it would have.

Its structural limits

Bulk vegetation means one canopy per grid cell, so a cell containing forest and field gets a compromise that behaves like neither. Urban surfaces are not represented at all without an added canopy model — the reason urban work needs explicit surface representation or a scheme like PALM-4U.

Groundwater is absent in Noah proper, which matters wherever a shallow water table sustains evaporation through dry periods.

See also: land-surface models generally, soil moisture, surface-layer schemes above it, and evapotranspiration.