b.e23_alpha16b.BLT1850.ne30_t232.033 #356
Comments
|
The simulation ran 15 years. Justin plotted the ATM timeseries and ADF diags. The RESTOM is about -0.4 W/m2 which a bit surprising as it is 4.3 W/m2 lower than the equivalent F case. (usually there is only 2 W/m2 between the F case and the B case). The lab sea seems to be building up. There might be something buggy in the setting and I am looking into this. I am stopping the simulation for now. |
|
ADF diags vs Obs and vs f.cam6_3_119.FLTHIST_ne30.r328_gamma0.33_soae.001 |
|
ADF reports that mean RESTOM is about -0.4 W/m2, while the time series plot is around -1.0 W/m2, and always < -0.5 W/m2. |
|
I agree that the plots are not consistent. @justin-richling: could you check the time series plot. Please make sure not to include any running average in RESTOM. We would like to see what is happening at the very beginning of the run. |
|
Thanks for highlighting this discrepancy @klindsay28, after some exploration this is what I think is going on: I ran the way the ADF gets RESTOM for the tables and plotted it by year for a time series and this is what I get:
notice the huge spike at the end of the last climo year, which should be 12, but the ADF is calculating year 13, just one month of January. [cftime.DatetimeNoLeap(2, 2, 1, 0, 0, 0, 0, has_year_zero=True), Then when I average on a yearly step:
The mean of all the yearly means is: array(-0.42395615) However, my time series calcs do not have this spike and average (non-rolling) is -1.0073648435570808. If I ignore the ADF spiked value and average that I get -1.01189266 |
|
Here is a non-rolling average of RESTOM time series
|
|
Update on the discrepancies between tables and timeseries: |
|
@PeterHjortLauritzen @cecilehannay I did some analysis to characterize what is different about the B-F transition in b-case-33 compared with b-case-32. We want to know why RESTOM changes by 4+ W/m2 in 33, but ~2 W/m2 in 32. Such a large sensitivity is unexpected, I'm told. Here's a spreadsheet looking at global mean values (not from the tables, but from the map plots): https://docs.google.com/spreadsheets/d/1QUj_rLSttMfrCuuxJzrPNnxpMlB7J986kW1OSMYUhQ8/edit?usp=sharing A robust response of both is the ~2 W/m2 reduction in FSNT when coupled to the ocean due to thickening clouds. There is also a robust warming of the stratosphere, discussed more at the end. In 32, FLNT goes down (warms) to offset the decrease in FSNT. This primarily occurs through thickening liquid (TGCLDLWP) and ice clouds (TGCLDIWP), reducing outgoing LW, but a substantial drying (TMQ) offsets this by a slight increase in outgoing clearsky longwave (FLNTC). The reduction in TMQ is consistent with the cooler troposphere. In 33, FLNT goes up (cools) which amplifies the reduction in FSNT. This occurs due to a large (2.44 W/m2) increase in clearsky longwave. Unlike 32, the atmosphere remains moist (TMQ) and the troposphere does not cool much at all. Without a cooler troposphere, the stratospheric warming likely increases clearsky outgoing longwave. See here the Temperature response in 32 Temperature response in 33
I don't know why the Temperature in the stratosphere increases when coupled, but it could possibly be due to the reduction in CO2 in 1850 compared to ~2000 values used in the F-case. This could be tested with a F1850 run, but perhaps someone already knows the cause. In 33, there is 3X more ice (TGCLDIWP) than 32. In 33, SWCF is excessive over the southern ocean, even know the global mean SWCF is about right. Provided there is nothing buggy with our b-case-33 run, I propose a run identical to 33, but with wsub_scale=0.5, which will reduce SWCF over the mid-latitude regions, while also decreasing the amount of cloud ice. This will increase RESTOM by 1-2 W/m2. |
|
For the stratospheric warming, I think it could make sense that it's the difference in CO2 concentrations. Plus maybe ozone differences in the SH. It's probably not necessary to test with an F1850 run as it could just be assessed by looking at how much stratospheric temperatures change over this time period in CMIP6 runs. Those are probably the dominant drives of stratospheric temperature change, at least in the global average. |
|
That makes sense, supported by that bullseye near the southpole perhaps reflecting an increase in ozone in 1850. Now I'd like to figure out why the troposphere doesn't cool in 33 like it does in 32. I am seeing a ~1K warmer SST/TS in 33: Run 32 comparison:
Run 33 comparison
These differences seem pretty darn large locally, such as the W.Pacific. |
|
@adamrher are there plots of the tape recorder, just to make sure water vapor is OK? |
|
|
|
@PeterHjortLauritzen is 33 the same as 356? If so, it would make sense that 33 warms more, since there is more water vapor in the stratosphere. |
|
Here's a comparison of water vapor between the 1850-1859 periods and the 1990-2009 periods of the CESM2 large ensemble. Indeed, water vapor increases quite a lot between these two time periods in these simulations that I don't think have a representation of methane oxidation.
|
Description:
Coupled simulated with cesm2_3_alpha16b with setting requested in #349
Detailed description:
The setting is as follow:
ATM
Bugfix and tuning (we are matching setting from #353):
SourceMods for HB mods
LND
user_nl_clm
SourceModsfromOCN
diag_tablein:add to
input.nmladd to
user_nl_momICE
Case directory:
Locally (if still available):
/glade/p/cesmdata/cseg/runs/cesm2_0/b.e23_alpha16b.BLT1850.ne30_t232.033
On github:
https://github.com/NCAR/amwg_dev/tree/b.e23_alpha16b.BLT1850.ne30_t232.033
Sandbox:
Locally (if still available):
/glade/work/hannay/cesm_tags/cesm2_3_alpha16b
On github:
Diagnostics:
AMWG diags (if available)
https://webext.cgd.ucar.edu/BLT1850/b.e23_alpha16b.BLT1850.ne30_t232.033/
Contacts:
@PeterHjortLauritzen, @adamrher, @JulioTBacmeister, @cecilehannay
@olyson, @wwieder and @slevis-lmwg
@gustavo-marques and @klindsay28
@jedwards4b
@dabail10
The text was updated successfully, but these errors were encountered: