Hello,

I'm currently running a version of CESM 1.2.2 using the fully coupled B_2000_STRATMAM7_CN componet set configureation. My research group has also added several new aerosol and cloud interaction treatments to the model but we have noticed that our global mean temperature bias compared to NCDC observation is around -2.9 deg C. This is bias is greater than what is typically accepted for global climate models and I was wondering if there are any parameters or namelist options that can be tuned or adjusted to bring the model into better agreement with the observations?

Thanks,

Tim Glotfelty

What grid resolution are you using? Have you compared against an unmodified version of the model to insure that you did not introduce the bias?

The grid rsolution I use is f09_g16 so 0.9x1.25 degree for the land and atmosphere and gx1v6 for the ice and ocean components. I did a run using the default out of the box B_2000_STRATMAM3_CN componet set. I couldn't use the B_2000_STRATMAM7_CN component set because there was some namelist error in the default version. The default temperature bias is still a little high -2.3 deg C which indicates my additinal aerosol and cloud changes add about -0.6 deg C to the mean bias.

Tim Glotfelty

I've moved this since I think that the people who frequent the CAM forums are more likely to have an answer.

You may need to retune your modified version of the model.

From user:

1) Firstly I was wondering what radiation variables from CAM5 are used to compute the radiative balance. Is is something like FSNTOA+FSNIRTOA = RESTOM or is it calculated using other variables?

2) Is the method any different because I will be tuning a current year time period (2001-2010) instead of pre-industrial (1850)?

3) I also did some checking on the LWCF and SWCF from my simulations compared against CERES data. It appears that on global average the SWCF from my simulations is around 5.0 W m-2 greater (more negative) than CERES and the LWCF is about 2.0 W m-2 smaller. Based on that is there any parameters to tune that make the SWCF less negative and increase the LWCF?

4) Lastly, is there an acceptable range of values with which to adjust the parameters that you suggested to be physically realistic?

1) Firstly I was wondering what radiation variables from CAM5 are used to compute the radiative balance. Is is something like FSNTOA+FSNIRTOA = RESTOM or is it calculated using other variables?In CAM5: RESTOM = FSNT - FLNT

2) Is the method any different because I will be tuning a current year time period (2001-2010) instead of pre-industrial (1850)?The difference would be that you are not in radiative equilibrium for current time

1850 => RESTOM = 0 W/m2

2000 => RESTOM will be positive (in CERES_EBAF it is close to 0.8 W/m2)

3) I also did some checking on the LWCF and SWCF from my simulations compared against CERES data. It appears that on global average the SWCF from my simulations is around 5.0 W m-2 greater (more negative) than CERES and the LWCF is about 2.0 W m-2 smaller. Based on that is there any parameters to tune that make the SWCF less negative and increase the LWCF?increasing rhminl will allow you to decrease SWCF (less negative) without affecting too much LWCF, but 5 W/m2 is a lot so it depends where your rhminl is now

Decreasing Dcs might help too

FOR SWCF: are you comparing to CERES or CERES-EBAF ?

4) Lastly, is there an acceptable range of values with which to adjust the parameters that you suggested to be physically realistic?rhminl is not an observable. This is the threshold from low-level cloud and it is used in the model as a tuning parameter. Personally, I have pushed it up to 0.93.

Dcs = [90-500] is probably a reasonable range (these values are coming from Andrew Gettelman). Dcs is the effective diameter at which we autoconvert to precipitation for ice crystals. So it could be refined based on observations.

Thanks for the information it was quite helpful. I compared my SWCF against CERES-EBAF. By adjusting rhminl I was able to get an RESTOM of 0.7 W/m2 so it is comparable to CERES-EBAF. This adjustment also decresed the bias in SWCF to around -1.0 W/m2 and brought the temperature bias back down to around -2.0 degrees C annually.

Thanks for the assistance,

Tim Glotfelty

Hi Tim,

Thanks for your feedback. It is nice to hear that it works for you.

Cecile

Hi,

Regarding the way you mentioned that RESTOM is calculated.

I am running the B_1850 compset with CESM 1.2.2

I am trying to calculate RESTOM by doing FSNT minus FLNT, then I do a global spatial average and I get around -29 W/m2 (in average for 30 years of simulation).

I have the feeling that I am doing something wrong but not sure what.

You mentioned that FSNT- FLNT is for CAM5, does this may have to do with the fact that B_1850 uses CAM4 and then I need to do the calculation differently?

Thanks

*******

UPDATE:

*******

I just realized I had to do a weighted spatial average and I got -0.02 W/m2.

This brings me to another question.

The reason I was using B_1850 is because is cheaper than B_1850_CN and the CN is not so relevant for my research purposes. But I have seen that B_1850 is not a scientifically validated compset as opposed to B_1850_CN.

So I was trying to check RESTOM as one of the metrics to see how good this compset is.

My questions are:

1) What else (other metrics) should I check to make sure this COMPSET is as good as a 'scientifically validated' one?

2) In the scientifically validated compstets webpage, B_1850_CN appears with model version 1.0.x.

Does that means that it is only validated for that version of the model? So if I run it on 1.2.2 is not scientifically validated anyway?

Cheers,

Bryam

Scientific validation of CESM consists of a multi-decadal model run of the given component set at the target resolution, followed by scientific review of the model output diagnostics. All scientifically supported component sets are also accompanied by diagnostic and model output data. If a comspet is not scientifically validated, it means we didn't go through thsi process. But it doesn't mean that you would get something incorrect. It just means that we didn't test this configuration.

So you would have to do a run (long enough enough - the length depends which component you are looking at), then analyze the simulation to make sure it looks fine. You could compare with another run. For teh atmosphere, you would loke to cloud forcing, radiative fluxes, precipitation, TS, for instance.

Hello,

I'm having the same issue (before your update), but I'm using CESM version 2.0.

I'm using the BHIST compset, running from 1850 to 2005, and the RESTOM time series is showing considerably different values from the ones available at http://www.cesm.ucar.edu/models/cesm2/scientifically-validated-cesm2.html.

Well, the resolution I'm using is not scientifc validated (f19_g17), but could it be the reason for those considerably differences? Or perhaps I'm missing something in the calculations?

Here are the steps:

1) read FSNT and FLNT from every 1872 CAM output files(from 1850-01-01 to 2005-01-01) and put each one into a 144x96x1872 variable.

2) calculate the means: first over longitude, then latitude for each variable.

3) RESTOM=FSNT(from step 2)-FLNT(from step 2)

I plotted the time series and you can find it attached (b.e20.bhist.f19_g17.gsm.dsc.001.png) as well as NCAR's times series (b.e20.BHIST.f09_g17.20thC.297_05_surf.gif) for comparison.

I don't know what I'm doing wrong ...

Could you, please, give more details about:

"I just realized I had to do a weighted spatial average and I got -0.02 W/m2." ?

Cheers,

Rafael

Hi Cecile,

thanks for you answer.

I calculated it again, and the result for RESTOM (BHIST run with f19_g17 resolution and unsupported) is 0.0095. I think it is too small.

What I did (Matlab code):

1) Read FLNT and FSNT from 1872 monthly results;

>> whos FSNT FLNT gw

Name Size Bytes Class Attributes

FLNT 144x96x1872 103514112 single

FSNT 144x96x1872 103514112 single

gw 96x1 768 double

2) Calculate restom:

restom = FSNT-FLNT;

3) Preparing for the weighted average:

gw_2D=repmat(gw,[1,144])'; %create a variable with the gw first column repeated 144 times

gw_3D = repmat(gw_2D,1,1,size(restom,3)); %create a variable with the previous 2D variable repeated 1872 times

restom_lat_weighted = restom.*gw_3D; %multiplies every latitude and time by gw

4) Averaging:

a=mean(restom_lat,1); %mean over the longitudes

b=mean(a,2); %mean over latitudes

c=mean(b,3); %mean over time

RESTOM=c/(sum(gw))

The result is: RESTOM=0.0095

What am I doing wrong? I checked the code and tried different ways of calculation, but always had the same result.

Here: http://www.cesm.ucar.edu/working_groups/Atmosphere/metrics.html it says that restom for 1850 coupled should be close to zero, but 0.0095 is too close...

The order of magnitude for RESTOM is:

RESTOM very close to zero in 1850

RESTOM ~ 0.8 W/m2 at the end of the 20th century (from CERES-EBAF)

These numbers are just guidance and it will depend on how you tuned your model. You adjust RESTOM by tuning the model.

For instance, in recent runs we have RESTOM = 0.4 W/m2 for the period 1991-2005

http://webext.cgd.ucar.edu/B20TH/b.e20.BHIST.f09_g17.20thC.297_01/atm/b....

I am not sure thsi answers your question. I am not sure what you mean by "0.0095 is too close..."

Also beware RESTOM has large year-to-year variations. This is becasue this is the difference of 2 large numbers very close to each other. You can plot the timeseries of RESTOM to have a sense of the variability.