Mean buoyancy budget in 2Day suite simulations

[sandreza]

Plotting the mean buoyancy in time for some of the 2Day suite simulations seems to not satisfy the mean buoyancy budget. In particular

using JLD2, Statistics, Plots
filename = pwd() * "/three_layer_constant_fluxes_hr48_Qu0.0e+00_Qb1.2e-07_f1.0e-04_Nh256_Nz128_free_convection_averaged_statistics.jld2"
file = jldopen(filename)

αg = 0.00196133

keys(file["timeseries"]["u"])
u = []
v = []
b = []
e = []
t = []
for key in keys(file["timeseries"]["u"])
    push!(u,file["timeseries"]["u"][key]*1.0)
    push!(v,file["timeseries"]["v"][key]*1.0)
    push!(b,file["timeseries"]["b"][key]*1.0)
    push!(e,file["timeseries"]["e"][key]*1.0)
    push!(t,file["timeseries"]["t"][key]*1.0)
end

plot(t, mean.(b))

yields a plot that isn't a straight line

[sandreza]

[sandreza]

Actually, this may be due to the fact that what I grabbed are the time-averaged statistics as opposed to the instantaneous. This would explain the factor of two discrepancies in the slope as compared to the expected budget, as well as initial transient

[sandreza]

That was indeed the case. Using statistics.nc (instead of averaged_statistcs.nc)

using NCDatasets, Statistics, Plots
files = [
    pwd() * "/three_layer_constant_fluxes_cubic_hr48_Qu0.0e+00_Qb1.0e-08_f1.0e-04_Nh256_Nz128_free_convection_2days_Qb1e-8.nc",
    pwd() * "/checkthis.nc"
]
file = files[2]
ds = Dataset(file,"r")
field = ds["T"]
t = ds["time"][:]
meanT = zeros(length(t))
key = keys(field)
for i in eachindex(t) 
    meanT[i] = mean(field[:,i])  
end

plot(t, meanT)

αg = 0.00196133
for  i in 2:length(t)
    println(αg * (meanT[i-1] - meanT[i]) / (t[i] - t[i-1]))
end

yielded

[glwagner]

Maybe we should use time_averaged_statistics.nc and instantaneous_statistics.nc as filenames?

[ali-ramadhan]

Yeah I'd be on board with these clearer names.

[sandreza]

Same here!

[ali-ramadhan]

These names are now in use on the master branch (switch was made in PR #56).