Ocean and land albedo

[AnnaDSMS]

in pre-definded experiment "Albedo" (notebook uploaded to Pre-Defined_Experiments folder), I set ocean and land albedo in this way:

#Create a ManualAlbedo object
manual = ManualAlbedo(spectral_grid)
#Create the composite albedo object
albedo = Albedo(manual, manual)
#Set land and ocean albedo patterns using the composite object
set!(albedo.land, (λ, φ) -> 0.3 + 0.2 * abs(φ)/90)
set!(albedo.ocean, (λ, φ) -> 0.06 + 0.04 * abs(φ)/90)

But when I look at simulation.diagnostic_variables.physics.ocean.albedo and simulation.diagnostic_variables.physics.land.albedo they are the same. Why?

[AnnaDSMS]

I managed to set albedo for ocean and land that I would like:

#Create a ManualAlbedo object
manual = ManualAlbedo(spectral_grid)
#Create the composite albedo object
albedo = Albedo(manual, manual)
#1. Manual albedo for the ocean
set!(albedo.ocean, (λ, φ) -> 0.06 + 0.04 * abs(φ)/90)
#2. Use existing climatology for land
land_clim_albedo = AlbedoClimatology(spectral_grid)
#3. Combine them
albedo = Albedo(manual, land_clim_albedo)

[milankl]

(Note: put code into ```julia ... ``` blocks for mono spaced font and syntax highlighting)

[milankl]

But when I look at simulation.diagnostic_variables.physics.ocean.albedo and simulation.diagnostic_variables.physics.land.albedo they are the same. Why?

TL;DR: If you haven't run!(simulation) yet then the diagnostic variables haven't been written into yet.

The logic is as follows: model.albedo (or albedo here as you haven't passed it on to the model constructor yet) is considered a pre-defined albedo. So it contains fields that describe albedo globally but are constant in time (or contain a Nx12 albedo field, one for each month). So think of this as read-only constants that are only once initialized at initialize!(model).

On the other hand simulation.diagnostic_variables is just pre-allocated memory before the run!(simulation) so it could be in any state. We do initialize a lot with zeros but it technically doesn't matter because the diagnostic variables are more like scratch spaces to store intermediate calculations (many of which are physically very relevant though!)

Then inside the simulation.diagnostic_variables there are 3 albedo fields

• .physics.land.albedo this is on every time step the albedo for any land grid cell, depending on model.albedo this can be constant or change on every time step. The ocean grid cells can be whatever.
• .physics.ocean.albedo same but for the ocean.
• .physics.albedo is on every time step the land-mask weighted albedo field from the two above. This is the field that's actually used for output.

[milankl]

Note that maybe less confusing you can use keyword arguments when collecting ocean & land albedo into Albedo

spectral_grid = SpectralGrid()
ocean_albedo = ManualAlbedo(spectral_grid)     # constant in time, 0 if not used with set!
land_albedo = AlbedoClimatology(spectral_grid) # loads a climatology (no seasons) from file
albedo = Albedo(ocean=ocean_albedo, land=land_albedo)   # just a container for the independent definitions for ocean/land

[milankl]

Note that this what you wrote

#Create a ManualAlbedo object
manual = ManualAlbedo(spectral_grid)
#Create the composite albedo object
albedo = Albedo(manual, manual)
#1. Manual albedo for the ocean
set!(albedo.ocean, (λ, φ) -> 0.06 + 0.04 * abs(φ)/90)
#2. Use existing climatology for land
land_clim_albedo = AlbedoClimatology(spectral_grid)
#3. Combine them
albedo = Albedo(manual, land_clim_albedo)

is somewhat confusing because you create albedo based on manual twice, note that therefore albedo.ocean and albedo.land point to the same object so if you change the ocean albedo then you also change the land albedo in the same way!!

Then you are defining a new land albedo land_clim_albedo and overwrite albedo from before. Cleaner would be to do

# 1. Create a ManualAlbedo object for the ocean and set! it
ocean_albedo = ManualAlbedo(spectral_grid)
set!(ocean_albedo, (λ, φ) -> 0.06 + 0.04 * abs(φ)/90)

# 2. Use existing climatology for land
land_albedo = AlbedoClimatology(spectral_grid)

# 3. Combine the two 
albedo = Albedo(ocean=ocean_albedo, land=land_albedo)

[milankl]

To show this after these lines of code the ocean albedo is set as desired because set! always applies immediately but otherwise creating a model component like AlbedoClimatology doesn't initialize any fields. That happens at initialize! so land albedo will contain only zeros at this stage

but after initialize! the data will have been read from file

[milankl]

But when I look at simulation.diagnostic_variables.physics.ocean.albedo and simulation.diagnostic_variables.physics.land.albedo they are the same. Why?

Looking back at your original question I now see the crucial misunderstanding

manual = ManualAlbedo(spectral_grid)   # this is one mutable albedo object
albedo = Albedo(manual, manual)    # this uses the same (!) albedo for both ocean and land

# now this changes albedo for both ocean and land as they point to the same `manual` albedo
set!(albedo.land, (λ, φ) -> 0.3 + 0.2 * abs(φ)/90)

# and this changes again both albedos but to something different!
set!(albedo.ocean, (λ, φ) -> 0.06 + 0.04 * abs(φ)/90)

You can think of all our model components as mutable like an array or list in python. In Albedo(manual, manual) no copy of manual is created. It's a reference to the same object!