Import CliMAF¶
In [1]:
from climaf.api import *
The example we will illustrate here is:¶
- make a CMIP5 multi-model ensemble of historical runs
- make a CMIP6 IPSL-CM6A-LR ensemble of historical runs with the thirty members
- compute the climatologies and regrid all the members on the same regular grid
- find a member that has an error and exclude it from the ensemble
- compute the ensemble means
- get a reference (observations) for tos
- compute the ensemble mean bias maps
- and plot them
First, build the CMIP5 multi-model ensemble¶
- request of all the models
- check what is available with explore('choices') to ensure we have only one attribut with multiple choices
- and build the ensemble with explore('ensemble')
In [2]:
req_cmip5 = ds(project='CMIP5',
variable='tos',
period='2000',
experiment='historical',
frequency='monthly',
model='*',
version='latest'
)
req_cmip5.explore('choices')
Out[2]:
In [3]:
ens_cmip5 = req_cmip5.explore('ensemble')
summary(ens_cmip5)
Now do the same with the CMIP6 IPSL-CM6A-LR grand ensemble with realization='*'¶
In [4]:
req_cmip6 = ds(project='CMIP6',
experiment='historical',
model='IPSL-CM6A-LR',
period='2000',
variable='tos',
table='Omon',
realization='*'
)
req_cmip6.explore('choices')
Out[4]:
In [5]:
ens_cmip6 = req_cmip6.explore('ensemble')
summary(ens_cmip6)
Compute annual mean climatology of each member on both ensembles¶
In [6]:
anm_ens_cmip5 = clim_average(ens_cmip5,'ANM')
anm_ens_cmip6 = clim_average(ens_cmip6,'ANM')
Regrid over the same regular grid¶
In [7]:
rgrd_anm_ens_cmip5 = regridn(anm_ens_cmip5,
cdogrid='r360x180')
rgrd_anm_ens_cmip6 = regridn(anm_ens_cmip6,
cdogrid='r360x180')
And compute the ensemble mean with ccdo_ens(..., operator='ensavg')¶
In [8]:
ensmean_cmip5 = ccdo_ens(rgrd_anm_ens_cmip5,
operator='ensavg')
ensmean_cmip6 = ccdo_ens(rgrd_anm_ens_cmip6,
operator='ensavg')
Get some observations¶
In [9]:
ref = ds(project='ref_climatos', variable='tos',
product='WOA13-v2').explore('resolve')
summary(ref)
Out[9]:
Regrid the reference as well after computing the climatology¶
In [10]:
rgrd_ref = regridn(clim_average(ref,'ANM'),
cdogrid='r360x180')
Plot the ensemble mean for CMIP6 to check if the result is ok¶
Reminder:
- plot() is the basic CliMAF plot operator
- iplot() is a layer on top of plot used to display a CliMAF plot in a Notebook; use => iplot(plot(dat, ...))
- implot() can be used directly in notebooks to both do the plot and display it; use => implot(dat, ...)
In [11]:
implot(ensmean_cmip6)
Out[11]:
Same for CMIP5¶
In [12]:
implot(ensmean_cmip5)
Out[12]:
Ooops, we have a problem...¶
There is obviously a problem with one (or more) of the members of the ensemble.
Here is a way to display the members one by one
Plot a first series of 12 plots with iplot_members => N = 1¶
In [13]:
iplot_members(rgrd_anm_ens_cmip5, N=1)
Out[13]:
For a better tracking of the errors, specify:¶
- focus = 'ocean' (to focus on the ocean only)
- min, max and delta to give boundaries to levels of values (and track potential problems of units)
In [14]:
iplot_members(rgrd_anm_ens_cmip5, N=1, focus='ocean', min=272, max=302, delta=2)
Out[14]:
This one looks ok.¶
Plot the next series of 12 plots in the ensemble => N = 1¶
In [16]:
iplot_members(rgrd_anm_ens_cmip5, N=2, focus='ocean', min=272, max=302, delta=2)
Out[16]:
In [17]:
# -- Make a copy of the original ensemble
ok_rgrd_anm_ens_cmip5 = rgrd_anm_ens_cmip5.copy()
# -- Remove BNU-ESM
ok_rgrd_anm_ens_cmip5.pop('BNU-ESM')
# -- Add 273.15 to GFDL-CM2p1 (Celsius to Kelvin)
ok_rgrd_anm_ens_cmip5['GFDL-CM2p1'] = fadd(rgrd_anm_ens_cmip5['GFDL-CM2p1'], 273.15)
# -- Might not have seen that, but MIROC5 mask over continents was set to zero; set it to missing value
# with cdo setctomiss,0
ok_rgrd_anm_ens_cmip5['MIROC5'] = ccdo(rgrd_anm_ens_cmip5['MIROC5'], operator='setctomiss,0')
Check that your corrections are effective:¶
In [33]:
iplot_members(ok_rgrd_anm_ens_cmip5, N=4, focus='ocean', min=272, max=302, delta=2)
Out[33]:
And make another clean ensemble mean¶
In [34]:
ok_ensmean_cmip5 = ccdo_ens(ok_rgrd_anm_ens_cmip5, operator='ensavg')
In [35]:
bias_ensmean_cmip6 = minus(ensmean_cmip6, rgrd_ref)
bias_ensmean_cmip5 = minus(ok_ensmean_cmip5, rgrd_ref)
And plot them:¶
- one by one first
- then in a multiplot
In [37]:
implot(bias_ensmean_cmip5)
Out[37]:
In [38]:
implot(bias_ensmean_cmip6)
Out[38]:
In [39]:
# -- Start by making two plots with plot() !! not iplot() !!
cmip5_plot = plot(bias_ensmean_cmip5, focus='ocean', title='CMIP5 ens. mean',
min=-6, max=6, delta=1, contours=1)
cmip6_plot = plot(bias_ensmean_cmip6, focus='ocean', title='IPSL-CM6A-LR ens. mean',
min=-6, max=6, delta=1, contours=1)
# -- Make the multiplot with cpage()
mp = cpage(fig_lines=[[cmip5_plot,cmip6_plot]])
# -- And display the result in the notebook with iplot
iplot(mp)
Out[39]:
Just for fun: add two plots with polar stereographic projections:¶
In [40]:
NH_cmip5_plot = plot(bias_ensmean_cmip5, focus='ocean', title='CMIP5 ens. mean',
min=-6, max=6, delta=1, contours=1,
proj='NH40')
NH_cmip6_plot = plot(bias_ensmean_cmip6, focus='ocean', title='IPSL-CM6A-LR ens. mean',
min=-6, max=6, delta=1, contours=1,
proj='NH40')
# -- Make the multiplot with cpage()
mp = cpage(fig_lines=[ [cmip5_plot,cmip6_plot], [NH_cmip5_plot, NH_cmip6_plot]])
# -- And display the result in the notebook with iplot
iplot(mp)
Out[40]:
Tip: how to get help on cpage?¶
In [41]:
help(cpage)
That's it for this last hour of the CliMAF training session of November 19 - Convergence - CLIMERI!¶
See more on the CliMAF frontpage:¶
Here: https://climaf.readthedocs.io/en/master/#can-climaf-make-my-scientific-life-easier