Loading and Visualizing Output

A simulation that saves diagnostics writes NetCDF (.nc) files to its output directory. The recommended tool for reading them is ClimaAnalysis.jl, which loads variables, reduces them (averages, slices, windows), and plots them through Makie.

ClimaAnalysis and a Makie backend are separate packages; add them to the environment you work in (pkg> add ClimaAnalysis CairoMakie).

Loading the output

Point SimDir at the output directory. With the default directory style, the latest run is behind the output_active link:

using ClimaAnalysis

simdir = SimDir("output/my_script_run/output_active")

SimDir catalogs every variable the simulation wrote; get retrieves one as an OutputVar by its short name:

ta = get(simdir, "ta")   # air temperature

Reducing dimensions

OutputVars reduce with functions such as average_time, average_lon, average_lat, and slice:

# Time average over the simulation, then zonal mean
ta_avg = average_lon(average_time(ta))

# Vertical profile of the zonal mean at the equator
ta_profile = slice(ta_avg, lat = 0.0)

Each reduction returns a new OutputVar with the reduced dimension removed and the attributes updated, so results stay self-describing.

Plotting

The ClimaAnalysis.Visualize module plots OutputVars through a Makie backend such as CairoMakie or GLMakie:

import ClimaAnalysis.Visualize as viz
import CairoMakie

fig = CairoMakie.Figure()
viz.plot!(fig, ta_avg)   # latitude-height heatmap of the zonal-mean temperature
CairoMakie.save("zonal_mean_temperature.png", fig)

viz.plot! picks a heatmap or a line plot from the number of remaining dimensions; viz.heatmap2D!, viz.sliced_heatmap!, and viz.line_plot1D! give explicit control. The ClimaAnalysis.jl documentation covers the full feature set, including windowing, units, and comparison against observations.

Where the files come from

ClimaDiagnostics.jl computes the diagnostics during the simulation and writes the NetCDF files; ClimaAnalysis reads and interprets them. To add diagnostic outputs or change how variables are interpolated to the output grid, see Computing and Saving Diagnostics.