Why we ship WSM3 microphysics (and why we have not moved off it yet)
WSM3 is the cheapest microphysics scheme in WRF. It is also what we ship. Here is why that was the right starting point and what would push us to change it.
Microphysics is the WRF scheme that decides how water vapour becomes cloud and what happens to the cloud after. Cloud droplets, rain, snow, ice. How they form, how they fall, how they evaporate. For soaring you mostly care about one output: at what altitude does the model decide air has saturated and a cumulus has formed. That altitude, post-processed, becomes `cloudbase_agl_ft`.
The shipping namelist runs `mp_physics = 3`, which is WSM3 - a single-moment three-class scheme tracking water vapour, cloud water or ice, and rain or snow. It is the simplest serious microphysics in WRF, the cheapest to compute, and an obvious starting point for a new operational pipeline that needs to prove the rest of the stack works before tuning any one component.
Why WSM3 first, rather than going straight to something fancier like Thompson or Morrison. Two reasons. First, the bottleneck for soaring forecasts is rarely microphysics quality - it is PBL scheme behaviour, surface heating, and how cleanly the cloudbase post-processing reads the moisture profile. Spending compute on a six-class double-moment scheme before any of those upstream parts have been validated is a classic premature optimisation. Second, every minute of WRF wall-clock matters when you are running four cycles a day on a single Hetzner box. The cheap scheme buys headroom for everything else.
What WSM3 gets you. A working cloudbase signal that is in the right ballpark on most days, derived in our pipeline from surface temperature, mixing ratio, and pressure rather than from the model's cloud field directly. A working precipitation field. A model that does not blow up on convective days at 4 km. None of that is glamorous, but it is what a v1 needs.
What WSM3 does not get you. No double-moment treatment of cloud droplets, which means no airmass-dependent variation in cloud droplet number concentration. In clean maritime air (most UK days), this is known to bias modelled cloud properties relative to richer schemes. No graupel class, so deep convection with hail signatures is poorly represented - acceptable for soaring, less so for severe weather. Less faithful behaviour around mixed-phase cloud at moderate altitudes, which can matter for high-base wave days.
What would push us to upgrade. Two specific things. First, validation against pilot-reported cloudbase showing a systematic bias that tracks airmass cleanliness - that is the textbook WSM3 failure mode and would be the signal to move to Thompson. Second, a future product use case where mixed-phase or icing detail matters, which would push us further again. Neither is the bottleneck today, but both are on the roadmap to revisit when the validation pipeline gives us numbers to chase.
Microphysics is a place where we explicitly chose the cheap option to save the expensive optimisations for the parts of the pipeline that are doing more for the forecast. WSM3 is shipping, the upgrade path to Thompson is known, and the trigger for taking it is a validation number we will measure rather than guess.