Noah, not Noah-MP: the land surface scheme we ship today
We run the original Noah land surface model in the Convek pipeline, not the more sophisticated Noah-MP. Here is why, what it costs us, and what would push the upgrade onto the roadmap.
The land surface model controls how much of the sun's energy goes into heating the air versus evaporating water. That partition (the Bowen ratio in the literature) drives everything downstream: surface temperature, boundary layer growth, trigger temperature, cloudbase. Get the land surface wrong and every field in the soaring forecast is wrong consistently, in ways that are hard to diagnose because no individual number looks obviously broken.
The shipping namelist runs `sf_surface_physics = 2`, which is the original Noah land surface model. Noah is a four-layer scheme with a single canopy layer, no dynamic vegetation, and the soil hydrology and thermodynamics that have been the WRF default for years. It is the simpler, older scheme. The newer Noah-MP (`sf_surface_physics = 4`) adds multiple canopy layers, prognostic vegetation, a more sophisticated snow model, and a modular structure that lets you choose between several physics options for each sub-process.
Why ship Noah rather than Noah-MP. Three reasons. First, Noah pairs cleanly with the GFS-derived initial conditions we use - GFS itself runs Noah, so the model state we inherit is already in Noah's expected format and the spin-up adjustment is small. Switching to Noah-MP introduces a soil-state translation step that is non-trivial to get right. Second, Noah is faster per timestep, which matters when you are running four cycles a day on a single Hetzner box. Third, the well-trodden path: Noah is what most operational WRF setups for daytime convective forecasting still use, so the failure modes are well-documented and fixes are easy to look up.
What Noah gets right for soaring. Surface temperature evolution through a normal day. The basic latitude-and-season pattern of how much of the incoming radiation goes into sensible heat. The diurnal cycle of soil temperature and moisture under most conditions. For the average summer convective day in the UK, Noah is enough.
Where Noah is weaker than Noah-MP. Snow-covered conditions (less of an issue for our forecast season, but real). Vegetation transitions during the growing season (Noah uses fixed land use, so leaf-out and senescence are not represented). Mixed forest canopies, where the single-canopy assumption smears out the multi-layer structure that Noah-MP can model. And, more subtly, soil moisture response to recent precipitation - Noah-MP's hydrology is more responsive to the wet-then-dry cycles that drive the worst land-surface forecast errors.
What would push us to upgrade. The honest trigger is a validation comparison showing Noah-driven surface heat flux biased systematically against observations on a meaningful fraction of days. Without that, switching to a more expensive scheme is speculative work. With that, the upgrade has a clear motivation and a measurable target.
There is a related upstream issue worth flagging. Soil moisture inherited from GFS can lag reality after fast-moving wet fronts, regardless of which land surface scheme you run downstream. That is an initial-condition problem rather than a Noah-vs-Noah-MP problem. Pulling soil moisture from a more responsive source (ERA5 reanalysis is the obvious candidate) would address it independently of the land surface upgrade. Both are on the roadmap.
For now, Noah is the shipping choice. A defensible default with known weaknesses, and a clear upgrade path when the validation work justifies the cost. The model page lists it alongside the rest of the physics stack.