Skip to content

Mean Seasonal Cycle for a single pixel

julia
using CairoMakie
CairoMakie.activate!()
using Dates
using Statistics

We define the data span. For simplicity, three non-leap years were selected.

julia
t =  Date("2021-01-01"):Day(1):Date("2023-12-31")
NpY = 3
3

and create some seasonal dummy data

julia
x = repeat(range(0, , length=365), NpY)
var = @. sin(x) + 0.1 * randn()
julia
fig, ax, obj = lines(t, var; color = :purple, linewidth=1.25,
    axis=(; xlabel="Time", ylabel="Variable"),
    figure = (; size = (600,400))
    )
ax.xticklabelrotation = π / 4
ax.xticklabelalign = (:right, :center)
fig

Define the cube

julia
julia> using YAXArrays, DimensionalData

julia> axes = (Dim{:Time}(t),)
(Time Date("2021-01-01"):Dates.Day(1):Date("2023-12-31"))
julia
julia> c = YAXArray(axes, var)
╭──────────────────────────────────╮
1095-element YAXArray{Float64,1}
├──────────────────────────────────┴───────────────────────────────────── dims ┐
Time Sampled{Date} Date("2021-01-01"):Dates.Day(1):Date("2023-12-31") ForwardOrdered Regular Points
├──────────────────────────────────────────────────────────────────── metadata ┤
  Dict{String, Any}()
├─────────────────────────────────────────────────────────────────── file size ┤
  file size: 8.55 KB
└──────────────────────────────────────────────────────────────────────────────┘

Let's calculate the mean seasonal cycle of our dummy variable 'var'

julia
function mean_seasonal_cycle(c; ndays = 365)
    ## filterig by month-day
    monthday = map(x->Dates.format(x, "u-d"), collect(c.Time))
    datesid = unique(monthday)
    ## number of years
    NpY = Int(size(monthday,1)/ndays)
    idx = Int.(zeros(ndays, NpY))
    ## get the day-month indices for data subsetting
    for i in 1:ndays
        idx[i,:] = Int.(findall(x-> x == datesid[i], monthday))
    end
    ## compute the mean seasonal cycle
    mscarray = map(x->var[x], idx)
    msc = mapslices(mean, mscarray, dims=2)
    return msc
end

msc = mean_seasonal_cycle(c);
365×1 Matrix{Float64}:
 -0.03130398287109743
 -0.06059678887906461
  0.08125105691657612
  0.019357118933250848
 -0.03069412445788597
  0.09901993559674853
  0.2642946301720222
  0.0724785553501333
  0.21244822988355164
  0.17115548211003082

 -0.21072435162838157
 -0.15326750862815794
 -0.12667116885178392
 -0.03650284859673286
 -0.06920309673015511
  0.006977886140959999
 -0.10555520526290235
  0.022285515523567546
 -0.005783221950539453

TODO: Apply the new groupby funtion from DD

Plot results: mean seasonal cycle

julia
fig, ax, obj = lines(1:365, var[1:365]; label="2021", color=:black,
    linewidth=2.0, linestyle=:dot,
    axis = (;  xlabel="Day of Year", ylabel="Variable"),
    figure=(; size = (600,400))
    )
lines!(1:365, var[366:730], label="2022", color=:brown,
    linewidth=1.5, linestyle=:dash
    )
lines!(1:365, msc[:,1]; label="MSC", color=:dodgerblue, linewidth=2.5)
axislegend()
ax.xticklabelrotation = π / 4
ax.xticklabelalign = (:right, :center)
fig
current_figure()