Addressing high surface fluxes and near surface wind speeds in CESM3

[adamrher]

CESM3 has excessive near surface wind speeds and surface fluxes. The lowest model level wind biases in the latest baseline (#198) relative to ERA5 are shown here:

There are high biases over the equatorial Pacific ocean in JJA and the W.Pac region in DJF. The UA surface flux algorithm (#224) seems to reduce this excessive easterlies in W.Pac.

The latent heat fluxes are high in our model when compared to ERA5, Large&Yeager and in-situ booey measurements. Below shows the ERA5 bias in #198, and it also shows the reduction in LHFLXs and improved RMSE using the UA scheme.

@swrneale @JulioTBacmeister @islasimpson @fobryan3 @jfasullo @dabail10 @dlawrenncar @gustavo-marques @iangrooms @klindsay28 @duvivier

[jfasullo]

I find the same - I also see RH as being too low in the lowest model level vs ERA5. Attached are plots from CMAT. Recall that comparisons against obs flow in the left column and against CESM2 in the right with biases on the bottom. I think these plots support the idea the LH is too high and that it is driven in part by wind speeds that are too high and RH values that are too low (over ocean).

[adamrher]

Thanks John. The RH (@993 hPa) plots show that we are too dry near the surface, but I'm having a hard time distinguishing the temperature effect from the humidity effect. I'd also imagine the difference of comparing a PI to a PD dataset creates some inconsistency as well. Does CMAT also plot T and Q @993 hPa?

In addition, there is this study that indicates the RH is biased high in ERA5, when compared with radiosondes (in the tropics).

https://www.tandfonline.com/doi/full/10.1080/16000870.2021.1929752

[jfasullo]

Yes, the RH constraint is perhaps more meaningful as Q992 differences are so sensitive to T. Note that CMAT does add in an 'increment' on what it thinks the PI and PD difference should be based on the CESM1LE. And no CMAT doesn't show Q992 (due to the T sensitivity). CMAT also uses an observational uncertainty of +/- 5% and the hatching exceeds that. Also note the changes from CESM2, which are of the same sign/pattern. The stippling is where the differences exceed 5% range based on internal variability.