MYNN-2.5 as the next PBL scheme to test against YSU
We ship YSU. MYNN-2.5 is the most credible alternative for daytime convective forecasting. Here is the comparison we plan to run, and what makes it worth running rather than picking from the published evidence alone.
MYNN-2.5 is the most often-cited alternative to YSU for daytime convective conditions in the mid-latitudes. It is a local TKE-closure scheme: it carries a prognostic turbulent kinetic energy field and mixes properties based on the local TKE rather than a bulk transport term. The literature pattern is that MYNN-2.5 produces shallower, more structured boundary layers than YSU, with later convective onset (because TKE has to build up before transport begins) and slightly more faithful representation of small-eddy structure.
YSU is what we ship today, with the reasoning in the companion post. This post is about the comparison against MYNN-2.5 that we plan to run, and what makes it worth running rather than picking from the published evidence alone.
What that would mean for soaring forecasts, if true. A few specific differences worth measuring. Boundary layer height is consistently shallower under MYNN, which would lower `hglider_agl_m` on most days. Convective onset is later, which would push `thermal_trigger_temp_c` later in the morning. Cloudbase response is more sensitive to the structure of the moisture profile, which could go either way for `cloudbase_agl_ft` accuracy. Each of those is a measurable thing if you have observations to compare against.
Why we have not run the comparison. Two practical reasons. First, the validation pipeline does not exist yet (see the validation post). Running a head-to-head without anything to compare against would just produce two sets of model output we could not score. Second, the MYNN-2.5 surface layer scheme pairs with a different surface flux formulation than the one we use today, so a clean comparison is not just a one-line namelist swap - it requires getting the coupled surface-PBL setup right and re-validating that nothing else has shifted.
What the test would look like when it runs. A multi-week period of UK convective days with both configurations running on the same cycles, identical in every other respect (same domain, same microphysics, same land surface, same initial conditions). Sounding comparisons against selected UK upper-air sites for boundary layer height. Pilot-facing comparison through `day_rating` correlated with observed flying activity. A clear runtime cost measurement on our specific hardware.
What would push us to switch. A clean improvement in either boundary layer height bias or convective onset timing on the validation set, with a runtime cost we can afford. The literature is mixed enough that the comparison genuinely could go either way - that is exactly why running it is worthwhile rather than picking from the published evidence alone.
PBL scheme work sits behind the validation pipeline on the roadmap. The shipping scheme is YSU, the reasoning lives in the companion post, and the MYNN-2.5 comparison gets written up here once it has been run.