Mixed-layer LCL: the cloudbase formulation upgrade we have not shipped
Today's cloudbase comes from a surface-parcel LCL. The standard upgrade is a mixed-layer LCL averaged over the diagnosed PBL depth. Here is what that would change and what is in the way.
The known weakness of the surface-parcel LCL we ship today (covered in the companion post) is that it samples the boundary layer at its coldest, moistest point. Real cumulus cloudbase is set by parcels that have mixed up through the PBL, lost some moisture to entrainment of drier air aloft, and warmed slightly. So a pure surface-parcel formulation tends to undersell cloudbase on actively-mixing days. The bias direction is helpful for safety (errs toward keeping pilots clear of cloud) but it is still a bias.
The standard upgrade is a mixed-layer LCL: instead of lifting the surface parcel, you average temperature and moisture over the diagnosed PBL depth and lift that averaged parcel. The averaged parcel is warmer and drier than the surface parcel, which lifts the LCL closer to where actual cumulus forms. This is the formulation most operational soaring products use.
What is in the way of just shipping it. A few things, none of them fundamental but each non-trivial. First, the mixed-layer formulation depends on a clean PBL height field from the model. WRF's PBLH diagnostic is reasonable but has known noise, and a 50 m wobble in the diagnosed top can move the averaged moisture by enough to shift cloudbase by 200 ft. A robust implementation needs a smoothing or sanity step. Second, on days with residual moisture layers above the mixed layer (overnight cloud that has not fully dispersed), naive averaging can pick up moisture from the wrong layer and put cloudbase too low. Third, we do not yet have a validation comparison to confirm that the mixed-layer version is actually more accurate than the surface-parcel one in our specific configuration.
What we would want to validate before shipping the upgrade. Side-by-side runs on a representative set of days, both formulations producing cloudbase fields, both compared against pilot-reported observed cloudbase. Particular attention to the edge cases (very shallow PBL, residual moisture, very dry days where the surface parcel and mixed-layer parcel are similar) where the two formulations should agree. Anywhere they disagree by more than a couple of hundred feet is a place to look hard at the implementation.
What would push us to ship. Clean validation signal that the mixed-layer LCL improves cloudbase RMSE in our setup, with the edge cases handled cleanly enough not to introduce new failure modes. Or pilot feedback that the current surface-parcel low bias is hurting forecast credibility on specific airmass types.
The mixed-layer LCL upgrade sits behind the validation work in the queue. Once the validation pipeline can score the upgrade cleanly against the live surface-parcel baseline, this is the next post-processing change to land. Update will follow here when it does.