Impacts of clubb diffusion choices

[islasimpson]

Here's a place where we can keep track of the impacts of the clubb diffusion choices.

[islasimpson]

From the project meeting. Turning down diffusion has a dramatic impact on tropical SST variability

[islasimpson]

Here is a look at the vertical profiles of humidity using model level ERA5 interpolated onto CAM's model levels. This is in May 1st initialized predictions with the 194 (diffusion on) and 199 (diffusion off) configuration

First, a look at specific humidity on the lowest model level. Turning off diffusion increases near surface humidity most places, except west Africa.

In general, this makes the bias in the lowest model level worse, except for in the equatorial Pacific during the winter and West Africa.

Here's a model level near 770 hPa. When we turn off diffusion we get too dry there in the sub-tropics.

Here are some vertical profiles at a few locations...

NH sub-tropics in September where turning off diffusion makes the near surface humidity bias worse.

NH equator in December, where in some regions the near surface humidity improves when diffusion is turned off

West Africa during Sept where turning off diffusion leads to an overall improvement.

Overall, turning off diffusion fully leads to more negatives than positives. But if we run these with the 0.75 diffusion, we can see what that looks like.

[adamrher]

This is consistent with my analysis of explicit diffusion in the tropics a few years ago -- turning off the diffusion increases the near surface values of humidity, but decreases it in the 800 hPa ( 2 km) range. Without the added diffusion, moisture doesn't get mixed up vertically as efficiently, and the higher near surface moisture is why the LHFLX's decrease.

[islasimpson]

@cecilehannay has run the following initialized predictions from August 1st:
194 = diffusion fully on
199 = diffusion fully off
199 no tune = diffusion fully off but without the additional tuning
213 = diffusion on with a scaling of 0.75 (also with the old ocean settings).

Here are some plots of the vertical profiles of the RMSE for specific humidity for these simulations.

First, averaged over all grid points in the tropics from 30S to 30N. It's a mixed bag and it depends where you look in the vertical as to where the switching off of the diffusion has a detrimental impact. The RMSE is bigger for 199 at the surface and between 700-800 hPa. But it's smaller between about 850 hPa and 950 hPa. 213 has a consistently bigger RMSE at the surface but it tends to have smaller errors above.

It seems like 199 without the tuning ends up with much smaller RMSE than 199 itself. But I think this should be ignored. The SST goes off the rails in the simulation without tuning as you'll see in the plot below. I've kept 199 no tune on the plots below but it should probably be ignored.

Here is the RMSE for the NH mid-latitudes.

and for the SH mid-latitudes

Turning off the diffusion has more favorable impacts in the mid-latitudes. In the SH mid-latitudes, 199 has lower RMSE almost everywhere. In the NH mid-latitudes 199 has a higher RMSE right near the surface but lower RMSE above about 950 hPa.

Here are the high latitudes:

Again, turning off diffusion seems to have favorable impacts here.

Here are maps that summarize the biases at three different levels. The lowest one, one around 850 hPa, and one around 700 hPa.

At the lowest model level we ens up with too much moisture when diffusion is turned off (as we already saw in the first posting).

Around 850 hPa with diffusion on (194) or at 0.75 (213) we have too much moisture in most places, except for the stratocumulus decks. Turning off diffusion (199) reduces the moisture leading to an overall lower bias in many places

Around 700 hPa turning off diffusion makes it too try in the tropics

Overall, from the perspective of humidity profiles, it's a mixed bag. Turning off diffusion is good for the extra-tropics but it makes it too moist near the surface in the tropics and too dry above. 213 is not that different from 194, but it does have consistently higher RMSE near the surface in the tropics and lower RMSE in the NH mid-latitudes.