Data and scripts for "Noise-shaped hysteresis cycles of the AMOC under increasing CO2 forcing"

This dataset contains supplementary material for the paper Noise-shaped hysteresis cycles of the AMOC under increasing CO2 forcing (M.Cini, G. Zappa, F. Ragone, S. Corti, 2023) submitted to Chaos, APJ publishing.  Here we uploaded most relevant data and scripts concerning this study. Feel free to contact us for other resources.This datasets contains:

1) The diagnostic output of one 6000-years simulation composed by a ramp up 3000-year simulation with concentration increasing from 354 ppm to 954ppm, and a ramp down 3000-simulation with concentration decreasing from 954 ppm to 354ppm2) All scripts that have been used to plot the figures in the paper.

Simulation Architecture and model setup

The simulations have been performed with an intermediate complexity coupled climate model, composed by the Planet Simulator (PlaSim) and the Large Scale Geostrophic Ocean (LSG). All simulations are perfomed with increasing/dereasing co2 concentrations (0.2 ppm/year) starting from 354 ppm to 954ppm and vice versa for the ramp down.   More information about model setup and scope of the simulations can be found in the paper.We performed 3 single member  simulations. Each simulation is composed of a 3000 years ramp up and 3000 years ramp down fo the co2. Here we upload only the diagnostic output of one of the three simulationAMOC_LD_amoc_pos_changeco2_0.2y1_ntraj1_k0_LBlock360_p1_startIDl207-y2490_r2.tar.gz    &    AMOC_LD_amoc_pos_changeco2_-0.2y1_ntraj1_k0_LBlock360_p1_startIDl207-y2490_r2.tar.gz

These are the ramp up and ramp down part of one simulation. Folders contain the diag files where Amoc index can be extracted. They contain also executbale files and namelists of the model.

time teries_Co2

Script for plotting time series. It has been used to plot fig.1, fig.2, fig.S1, fig.S2, fig.S3 of the paper

EWS_Co2

Scripts for the Early-Warning-Signal analysis. It has been used to plot fig.3, fig.S4, fig.S5, fig.S6 of the paperNB: In the second cell, the first plot (fig 3a of the paper) is perfromed as :

run_fit_a_ar1(stemp[i], ws)

while the others (Fig 3b and 3c) are performed as 

run_fit_a_ar1(stemp[i] - funcfit4(time,*popt31), ws)

But the results are the same with both approaches. This means that the use of funcfit4, in our case didn't change the results, since a linear detrending is already applied inside the window.