init

clean

format

path fix

spinup

clean

pip

revs

rev

climacommon up

.buildkite/hierarchies/pipeline.yml

@@ -1,7 +1,7 @@
 agents:
   queue: new-central
   slurm_time: 24:00:00
-  modules: climacommon/2024_03_18
+  modules: climacommon/2024_05_27
 
 env:
   JULIA_LOAD_PATH: "${JULIA_LOAD_PATH}:${BUILDKITE_BUILD_CHECKOUT_PATH}/.buildkite"
@@ -68,6 +68,16 @@ steps:
           slurm_gpus: 1
           modules: common
 
+      - label: "Clima: GPU ClimaCoupler Cloudless Aquaplanet"
+        command:
+          - "julia --color=yes --project=experiments/ClimaEarth/ experiments/ClimaEarth/run_cloudless_aquaplanet.jl"
+        artifact_paths:  "cloudless_aquaplanet/cloudless_aquaplanet/clima_atmos/*"
+        agents:
+          queue: clima
+          slurm_mem: 20GB
+          slurm_gpus: 1
+          modules: common
+
       - label: "Clima: GPU ClimaCoupler Cloudy Aquaplanet"
         command:
           - "julia --color=yes --project=experiments/ClimaEarth/ experiments/ClimaEarth/run_cloudy_aquaplanet.jl"

.buildkite/longruns/pipeline.yml

@@ -1,7 +1,7 @@
 agents:
   queue: new-central
   slurm_time: 24:00:00
-  modules: climacommon/2024_04_30
+  modules: climacommon/2024_05_27
 
 env:
   JULIA_LOAD_PATH: "${JULIA_LOAD_PATH}:${BUILDKITE_BUILD_CHECKOUT_PATH}/.buildkite"

.buildkite/pipeline.yml

@@ -1,7 +1,7 @@
 agents:
   queue: new-central
   slurm_time: 4:00:00
-  modules: climacommon/2024_04_30
+  modules: climacommon/2024_05_27
 
 env:
   JULIA_LOAD_PATH: "${JULIA_LOAD_PATH}:${BUILDKITE_BUILD_CHECKOUT_PATH}/.buildkite"
@@ -397,6 +397,16 @@ steps:
         agents:
           slurm_mem: 20GB
 
+      - label: ":construction: Cloudless Aquaplanet"
+        key: "cloudless_aquaplanet"
+        command:
+          - sed 's/t_end = "1000days"/t_end = "1days"/' experiments/ClimaEarth/run_cloudless_aquaplanet.jl > experiments/ClimaEarth/run_cloudless_aquaplanet_short.jl
+          - "julia --color=yes --project=experiments/ClimaEarth/ experiments/ClimaEarth/run_cloudless_aquaplanet_short.jl"
+        artifact_paths:  "cloudless_aquaplanet/cloudless_aquaplanet/clima_atmos/*"
+
+        agents:
+          slurm_mem: 20GB
+
       - label: ":construction: Cloudy Aquaplanet"
         key: "cloudy_aquaplanet"
         command:
@@ -417,7 +427,14 @@ steps:
         agents:
           slurm_mem: 20GB
 
-
+      - wait
+      - label: ":construction: Hierarchy plots"
+        key: "hierarchy_plots"
+        command:
+          - "julia --color=yes --project=experiments/ClimaEarth/ experiments/ClimaEarth/hierarchy/climate_plots.jl"
+        artifact_paths:  "paper_figs/*"
+        agents:
+          slurm_mem: 20GB
 
   - group: "GPU integration tests"
     steps:

experiments/ClimaEarth/hierarchy/README.md

@@ -0,0 +1,16 @@
+# Hierarchy Experiments for Global Climate Models
+
+This directory contains a series of experiments that demonstrate the use of hierarchical modeling using ClimaEarth, with a focus on the atmospheric component. The hierarchy spans from a simple dry atmosphere to a more complex moist atmosphere with clouds and an Earth-like surface. The experiments are designed to be run in sequence, with each experiment building on the previous one. Each experiment is a self-contained run script that contains all necessary configurations.
+
+## Experiments
+- Dry Held-Suarez
+- Moist Held-Suarez
+- Cloudless Aquaplanet
+- Cloudy Aquaplanet
+- Cloudy Slabplanet
+
+## Postprocessing
+We provide a simple postprocessing script `climate_plots.jl` that can be used to visualize the results of each experiment, with some helper functions in `plot_helper.jl`.
+
+## Associated publication
+- (in preparation)

experiments/ClimaEarth/hierarchy/climate_plots.jl

@@ -0,0 +1,89 @@
+# paper_figs
+using NCDatasets
+using Statistics
+using Plots
+
+import Interpolations: LinearInterpolation
+import DelimitedFiles: writedlm, readdlm
+
+include("plot_helper.jl")
+
+for job_id in ["dry_held_suarez", "moist_held_suarez"]
+    if isinteractive()
+        DATA_DIR = "experiments/ClimaEarth/$job_id/$job_id/clima_atmos/output_active/"
+    else
+        build = ENV["BUILDKITE_BUILD_NUMBER"]
+        DATA_DIR = "/central/scratch/esm/slurm-buildkite/climacoupler-ci/$build/climacoupler-ci/$job_id/$job_id/clima_atmos/output_active/"
+    end
+
+    reduction = "6h_inst"
+    PLOT_DIR = "paper_figs"
+
+    mkpath(PLOT_DIR)
+
+    # SUPPLEMENTAL: animation of surface temperature
+    ta_sfc, lat, lon, z, time = get_nc_data_all("ta", reduction, DATA_DIR)
+    anim = Plots.@animate for i in 1:size(ta_sfc, 1)
+        Plots.contourf(
+            lon,
+            lat,
+            ta_sfc[i, :, :, 1]',
+            xlabel = "Longitude",
+            ylabel = "Latitude",
+            title = "$var",
+            color = :viridis,
+            clims = (260, 315),
+        )
+    end
+    Plots.mp4(anim, joinpath(PLOT_DIR, "anim_ta_sfc.mp4"), fps = 10)
+
+    # Figure 2: climatology
+    # this plots the time-mean (upper-level and surface) slices and zonal means of
+    # the mass streamfunction, zonal wind, meridional wind, temperature, max. Eady growth rate, and vertical velocity
+    vars = ["mass_strf", "va", "ua", "ta", "egr", "wa"]
+    for var in vars
+        plot_climate(var, DATA_DIR, PLOT_DIR, job_id, reduction = reduction, interpolate_to_pressure = true)
+    end
+
+    # Figure 4: storm track diagnostics
+    # this extracts the eddy heat flux and plots its climatology
+    lev_st = 6
+    ta_zm, ta_sfc, lat, lon, z = mean_climate_data("ta", reduction, DATA_DIR, lev_i = lev_st)
+    va_zm, va_sfc, lat, lon, z = mean_climate_data("va", reduction, DATA_DIR, lev_i = lev_st)
+    vT_zm, vT_sfc, lat, lon, z = mean_climate_data("vt", reduction, DATA_DIR, lev_i = lev_st)
+    heat_flux_zm = vT_zm .- va_zm .* ta_zm
+
+    pa_zm, ~, ~, ~, ~ = mean_climate_data("pfull", reduction, DATA_DIR)
+    pa_zm = pa_zm ./ 100 # convert to hPa
+    pa_grid = [950, 800, 700, 600, 500, 400, 300, 200, 50]
+
+    heat_flux_int_zm = interpolate_to_pressure_coord_2d(heat_flux_zm, pa_zm, pa_grid)
+    Plots.contourf(
+        lat,
+        -pa_grid,
+        heat_flux_int_zm',
+        xlabel = "Latitude (deg N)",
+        ylabel = "Pressure (hPa)",
+        title = "Heat flux",
+        color = :viridis,
+        ylims = (-pa_grid[1], -pa_grid[end]),
+        yticks = (-pa_grid, pa_grid),
+    )
+    png(joinpath(PLOT_DIR, "$(job_id)_heat_flux.png"))
+
+    # Figure 5: storm track diagnostics reduced to timeseries
+    # this plots the eddy heat flux and max. Eady growth rate in a sectorial selection
+    lev_i, lat_s_i, lat_n_i, lon_w_i, lon_e_i = lev_st, 60, 75, 1, 30
+    println(
+        "Sectorial selevtion for timeseries: \n level: $(z[lev_i]), lat: $(lat[lat_s_i]) to $(lat[lat_n_i]), lon: $(lon[lon_w_i]) to $(lon[lon_e_i])",
+    )
+
+    egr_all, lat, lon, z, time = get_nc_data_all("egr", reduction, DATA_DIR)
+    egr_t = point_timeseries_data(egr_all, [lon_w_i, lon_e_i], [lat_s_i, lat_n_i], lev_i)
+
+    vT_all, lat, lon, z, time = get_nc_data_all("vt", reduction, DATA_DIR)
+    va_all, lat, lon, z, time = get_nc_data_all("va", reduction, DATA_DIR)
+    ta_all, lat, lon, z, time = get_nc_data_all("ta", reduction, DATA_DIR)
+    heat_flux_all = vT_all .- va_all .* ta_all
+    heat_flux_t = point_timeseries_data(heat_flux_all, [lon_w_i, lon_e_i], [lat_s_i, lat_n_i], lev_i)
+end

experiments/ClimaEarth/hierarchy/plot_helper.jl

@@ -0,0 +1,119 @@
+# plot_helper
+
+"""
+    get_nc_data_all(var, red, DATA_DIR)
+
+Reads the netcdf file for the variable `var` and reduction `red` from the directory `DATA_DIR` and returns the variable, lat, lon, z, and time.
+"""
+get_nc_data_all = (var, red, DATA_DIR) -> begin
+    ds = NCDataset("$DATA_DIR/$(var)_$red.nc")
+    var = ds["$var"][:, :, :, :]
+    lat = ds["lat"][:]
+    lon = ds["lon"][:]
+    z = ds["z"][:]
+    time = ds["time"][:]
+    close(ds)
+    return var, lat, lon, z, time
+end
+
+"""
+    mean_climate_data(varname, reduction, DATA_DIR; lev_i = 1, spinup=1)
+
+Postprocesses the climate data for the variable `varname` and `reduction` from the directory `DATA_DIR`. Returns the zonal mean and horizontal surface slice mean of the variable.
+"""
+mean_climate_data =
+    (varname, reduction, DATA_DIR; lev_i = 1, spinup = 1) -> begin
+
+        var, lat, lon, z, time = get_nc_data_all(varname, reduction, DATA_DIR)
+        @assert spinup < size(var, 1)
+
+        var_time_zonal_mean = mean(var[spinup:end, :, :, :], dims = (1, 2))[1, 1, :, :]
+        var_time_mean_sfc = mean(var[spinup:end, :, :, :], dims = (1))[1, :, :, lev_i]
+
+        return var_time_zonal_mean, var_time_mean_sfc, lat, lon, z
+    end
+
+"""
+    point_timeseries_data(variable, lon_i, lat_i, lev_i)
+
+Returns the time series data for the variable `variable` at the indices `lon_i`, `lat_i`, and `lev_i`.
+"""
+point_timeseries_data =
+    (variable, lon_i, lat_i, lev_i) -> begin
+
+        variable_time_mean = mean(variable[:, lon_i[1]:lon_i[2], lat_i[1]:lat_i[2], lev_i], dims = (2, 3))[:, 1, 1]
+
+        return variable_time_mean
+    end
+
+"""
+    plot_climate(var, DATA_DIR, PLOT_DIR, job_id; reduction = "inst", interpolate_to_pressure = false)
+
+Plots the zonal mean and horizontal surface slice mean of the variable `var` from the directory `DATA_DIR` and saves the plots in the directory `PLOT_DIR`.
+"""
+function plot_climate(var, DATA_DIR, PLOT_DIR, job_id; reduction = "inst", interpolate_to_pressure = false)
+    strf_zm, strf_sfc, lat, lon, z = mean_climate_data(var, reduction, DATA_DIR)
+    strf_zm, strf_upper, lat, lon, z = mean_climate_data(var, reduction, DATA_DIR, lev_i = 10)
+
+    # vertical-lat plot of zonal and time mean
+    if interpolate_to_pressure
+        pa_zm, ~, ~, ~, ~ = mean_climate_data("pfull", reduction, DATA_DIR)
+        pa_zm = pa_zm ./ 100 # convert to hPa
+        pa_grid = [950, 800, 700, 600, 500, 400, 300, 200, 50]
+        strf_zm = interpolate_to_pressure_coord_2d(strf_zm, pa_zm, pa_grid)
+        Plots.contourf(
+            lat,
+            -pa_grid,
+            strf_zm',
+            xlabel = "Latitude (deg N)",
+            ylabel = "Pressure (hPa)",
+            title = "$var",
+            color = :viridis,
+            ylims = (-pa_grid[1], -pa_grid[end]),
+            yticks = (-pa_grid, pa_grid),
+        )# , clims=(-1e10, 1e10))
+        png(joinpath(PLOT_DIR, "$(job_id)_$(var)_pa.png"))
+    else
+        Plots.contourf(
+            lat,
+            z,
+            strf_zm',
+            xlabel = "Latitude",
+            ylabel = "Height (km)",
+            title = "$var",
+            color = :viridis,
+            ylims = (0, 3e4),
+            yscale = :log10,
+            yticks = ([1e3, 5e3, 10e3, 20e3, 30e3], ["1", "5", "10", "20", "30"]),
+        )# , clims=(-1e10, 1e10)
+        png(joinpath(PLOT_DIR, "$(job_id)_$var.png"))
+    end
+
+    # horizontal slices
+    Plots.contourf(lon, lat, strf_sfc', xlabel = "Longitude", ylabel = "Latitude", title = "$var", color = :viridis)#, clims=(-1e10, 1e10))
+    png(joinpath(PLOT_DIR, "$(job_id)_$(var)_sfc.png"))
+
+    Plots.contourf(lon, lat, strf_upper', xlabel = "Longitude", ylabel = "Latitude", title = "$var", color = :viridis)#, clims=(-1e10, 1e10))
+    png(joinpath(PLOT_DIR, "$(job_id)_$(var)_10km.png"))
+end
+
+"""
+    interpolate_to_pressure_coord_2d(var_zm, pa, pa_grid)
+
+Interpolates the 2D variable `var_zm` to the pressure grid `pa_grid` using the pressure values `pa`.
+"""
+function interpolate_to_pressure_coord_2d(var_zm, pa, pa_grid)
+    var_on_pa = zeros(size(var_zm, 1), length(pa_grid))
+    for lat_i in collect(1:size(var_zm, 1))
+        # Extract ua and corresponding ta values
+        var_values = var_zm[lat_i, :]
+        pa_values = pa[lat_i, :]
+
+        # Interpolate ua onto ta_grid
+        for (pa_j, pa_val) in enumerate(pa_grid)
+            itp_var = LinearInterpolation(-pa_values, var_values)
+            var_on_pa[lat_i, pa_j] = itp_var(-pa_val)
+        end
+    end
+    return var_on_pa
+end