# This script plot metric vs parameter # + (optional) best simulations or choosen higlighted simulations # + (optional) ensembles highlighted in colors # # by default plot ensembles from iwavemin to iwavemax (you must specify at the begining of the script) # you can also specify ensemble files and their corresponding parameters files and highlighted simulations # with options # # Ensemble files must be csv files : # - containing an header with metric_name=head_metric+"_"+metric (head_metric can be WAVEX) # - the head of the first line must be "SIM" # - separated by "," # # Parameters files must be asc files : # - containing an header with parameters name # - the head of the first line must be "t_IDs" # - separated by " " # Adapted from CCCma_scripts and scatter_plot.py (almost same arguments) # # Stuff to be done : # - maybe add an option to plot on log scales (when parameter are explores in log scale ?) # - pdfjam by metrics ? -> installer pdfjam sur ricard # - gérer les colors bars (gestion à la main pour l'instant) # - gérer les logscale pour les paramètres explorés en log # # Maelle Coulon--Decorzens - Frédéric Hourdin - Najda Villefranque - Mars 2024 import pandas as pd import re import matplotlib.pyplot as plt import os import argparse import sys import matplotlib.colors as mcolors import math #------------------------------------------------------------------------------ # Some previous variables the user might change #------------------------------------------------------------------------------ iwavemin=1 iwavemax=1 path_fig="fig/param_vs_param/" #path where figures are saved (from here) #param_user="param" #------------------------------------------------------------------------------ # Script optional arguments #------------------------------------------------------------------------------ parser=argparse.ArgumentParser() parser.add_argument("-o",help="Observation csv file",metavar='OBS.csv') parser.add_argument("-e",help='List of ensemble csv files "ENS1.csv,ENS2.csv,..."',metavar="ENS1.csv,ENS2.csv,...") parser.add_argument("-p",help='List of param csv files corresponding to ensemble "PARAM1.asc,PARAM2.asc,..."',metavar="PARAM1.asc,PARAM2.asc,...") parser.add_argument("-param",help='Name of initial parameter file',metavar="param_NAME or param", default="param") parser.add_argument("-l",help='List of ensemble csv files you want to be highlighted in colors ENS1.csv,ENS10.csv,...' ,metavar="ENS1.csv,ENS10.csv,...") parser.add_argument("-b",help='List of names of best simulations sim1,sim2,...',metavar="sim1,sim2,...") parser.add_argument("-n",help='Number of best simulations to be automatically displayed',metavar="N",default='10') parser.add_argument("-fig",help='path where figure are saved' ,metavar="dirfig/dir1/dir2/",default='"fig/param_vs_param/"') parser.add_argument("-x",help="xlim of the plots", metavar="xlim", default=None) parser.add_argument("-y",help="ylim of the plots", metavar="ylim", default=None) #parser.add_argument("-c",help='matplotlib colormap : eg, BrBG PuOr RdYlGn Spectral_r bwr coolwarm_r earth_r ocean_r twilight_shifted Accent',metavar="cmap",default='Paired') args=parser.parse_args() xlim = [float(u) for u in args.x.split(",")] if args.y is not None else None ylim = [float(u) for u in args.y.split(",")] if args.y is not None else None ## Quelques trucs pour mes plots ## darkblue="#1155cc" lightblue="#6fa8dcff" red="#cc0000ff" darkgreen="#314e23" green="#5b8d43cc" ## User parameter files ## #if args.param: # param_user = args.param # df = pd.read_csv(param_user, sep=r"\s+", header=None, engine="python") # #file_param_user = df.set_index(0).T.reset_index(drop=True) # file_param_user = df.set_index(0).T.reset_index(drop=True) # file_param_user=df.drop(0) #if args.param: # param_user = args.param # file_param_user = pd.read_csv(param_user, delim_whitespace=True, header=None) # file_param_user.columns = file_param_user.iloc[0] # file_param_user = file_param_user[1:].reset_index(drop=True).T if(args.param) : param_user=args.param file_param_user = pd.read_csv(param_user, delim_whitespace=True, header=None).T file_param_user.columns = file_param_user.iloc[0] file_param_user = file_param_user[1:].reset_index(drop=True) print(file_param_user) ### Highlighted ensembles ##i# #wave number or ens file metric name you want to be in color if args.l : enstoprint=args.l.split(',') print('Highlighted ensemble from arguments : ',enstoprint) elif(iwavemax > 3 ) : enstoprint=[1,40,iwavemax] #wave number or ens file metric name you to be in color print('Highlighted ensemble from WAVES '+str(iwavemin)+' to '+str(iwavemax)) else : enstoprint=[1,iwavemax] print('Highlighted ensemble from WAVES '+str(iwavemin)+' to '+str(iwavemax)) ### Obs file ### if args.o : obs_file=args.o print('Obs file from arguments : ',obs_file) else : #obs_file="WAVE"+str(iwavemin)+"/metrics_REF_"+str(iwavemin)+".csv" obs_file="metrics_REF_"+str(iwavemin)+".csv" print('Obs file from WAVE'+str(iwavemin)+' : '+obs_file) obs=pd.read_csv(obs_file,sep=",") obs_mean=obs.loc[0] obs_var=obs.loc[1] ### Metrics and Parameters ensembles ### if ( ((args.e is not None) & (args.p is None)) | ((args.p is not None) & (args.e is None) )) : print("if you give specific ensemble in argument you must give their parameters files with -p option") print("Stop the program here") sys.exit() if args.e : metric_files=args.e.split(",") print('Ensemble metrics files from arguments : ',metric_files) nwave=len(metric_files) file_metrics_ref=metric_files[0] param_files=args.p.split(",") file_params_ref=param_files[0] print('Ensemble parameter files from arguments : ',param_files) else : metric_files=[] param_files=[] for iwave in range(iwavemin,iwavemax+1) : metric_files.append("WAVE"+str(iwave)+"/metrics_WAVE"+str(iwave)+"_"+str(iwave)+".csv") param_files.append("WAVE"+str(iwave)+"/Par1D_Wave"+str(iwave)+".asc") file_metrics_ref="WAVE"+str(iwavemin)+"/metrics_WAVE"+str(iwavemin)+"_"+str(iwavemin)+".csv" file_params_ref="WAVE"+str(iwavemin)+"/Par1D_Wave"+str(iwavemin)+".asc" print('Ensemble metrics from WAVES : ',metric_files) print('Ensemble parameter files from WAVES : ',param_files) metrics_ref=pd.read_csv(file_metrics_ref,sep=",") params_ref=pd.read_csv(file_params_ref,sep=' ') ### Bests simulations ### list_bests=[] nbests_auto=int(args.n) if args.b : list_bests=args.b.split(",") elif(args.e is None) : #controled by waves if(nbests_auto > 0 ) : #file_score="score"+str(iwavemin)+"to"+str(iwavemax)+".csv" file_score="score41to42.csv" #"+str(iwavemin)+"to"+str(iwavemax)+".csv" if(os.path.exists(file_score) == False) : os.system(" head -1 score"+str(iwavemin)+".csv > "+file_score) os.system(" for i in $( seq "+str(iwavemin)+" "+str(iwavemax)+") ; do tail -n +2 score$i.csv >> "+file_score+" ; done") score=pd.read_csv(file_score, sep=",") sorted_score=score.sort_values(by="MAX", ascending=True) selected_simu=sorted_score[0:nbests_auto]['SIM'] for simu in selected_simu : list_bests.append(simu) else : print("if you specify ensembles you must specify with option -b the list of simulations you want to highlight, or it will be none of them") list_bests=[] ### Dico bests -> file metrics et bests -> file params ### dico_bests={} dico_bests["metrics"]={} dico_bests["params"]={} imet=0 for fm in metric_files : isimu=0 for simu in list_bests : res=os.system("res=`grep "+str(simu)+" "+fm+"`") if( res == 0 ) : dico_bests["metrics"][simu]=fm dico_bests["params"][simu]=param_files[imet] isimu=isimu+1 imet=imet+1 print("list_bests = ", list_bests) ### Plot colors ### #colors_ens=['#d0d0d0', '#3bafba', '#d56a6a', '#cdc25b'] listcoul=['red', 'green','fuchsia','blue','lime','darkviolet','cyan','darkorange','slateblue','brown','gold'] #colors_ens=['lightgrey','mistyrose','lemonchiffon','palegreen','paleturquoise'] colors_ens=['lightgrey','#5fb6be', '#d56a6a','#cdc25b','#72BB72','#3bafba'] # darkblue colors_ens=[lightblue,darkblue,'darkblue',darkblue,lightblue, '#d56a6a','#cdc25b','#72BB72','#3bafba'] #colors_ens=[red,lightblue,darkblue,'darkblue',darkgreen,green,lightblue, '#d56a6a','#cdc25b','#72BB72','#3bafba'] markeredgecolor='k' marker_ens=["o","o","D","*","*","*"] marker_ens=["o","D","*","*","*"] #marker_ens=["o","o","o","o","*","*"] fact=1.5 ms_ens=[fact*10,fact*10+2,fact*10,fact*36] ms_ens=[fact*10+2,fact*10,fact*40,fact*40] #ms_ens=[fact*10,fact*10,fact*10,fact*10] label_ens=["Membres de $\Lambda_{10}^1$", "Membres de $\Lambda_{10}^2$", "Membres de $\Lambda_{10}^3$"] if(args.fig) : path_fig=args.fig+"/" if(os.path.exists(path_fig) == False) : os.system("mkdir -p "+path_fig) #------------------------------------------------------------------------------ # Loop on metrics and parameters #------------------------------------------------------------------------------ for imetric,metric_ in enumerate(metrics_ref.columns.values[1:]): print(imetric,metric_) metric_split=metric_.split('_') metric="" for x in metric_split[1:] : metric=metric+'_'+x if(len(metric_split) > 1) : metric=metric[1:] metric_head=metric_split[0] else : metric=metric_split[0] metric_head="" print("metric = ", metric) #normalisaiton à la main !!! if("precip" in metric) : min_norm=1 max_norm=8 min_norm=0 max_norm=10 #min_norm=1 #max_norm=7 elif("conv" in metric) : min_norm=-160 max_norm=-30 min_norm=-140 max_norm=-30 #min_norm=-140 #max_norm=-80 else : min_norm=-70 max_norm=-15 min_norm=-50 max_norm=-15 #min_norm=-60 #max_norm=-40 print(min_norm,max_norm) norm = mcolors.Normalize(vmin=min_norm, vmax=max_norm) cmap = plt.get_cmap("magma") print(params_ref.columns.values[1:]) for iparam,param in enumerate(params_ref.columns.values[1:]) : for iparam2,param2 in enumerate(params_ref.columns.values[1:]) : #print(param, param2) fig,ax=plt.subplots(figsize=(5,5)) #plt.subplots_adjust(left=0.1, right=0.7, top=0.6, bottom=0.1) #plt.subplots_adjust() ax.set_box_aspect(1) print("param = ", param) if(file_param_user[param].iloc[3]=="log") : plt.xscale("log") if(file_param_user[param2].iloc[3]=="log") : plt.yscale("log") xmin=file_param_user[param].iloc[0] xmax=file_param_user[param].iloc[1] xprior=file_param_user[param].iloc[2] ymin=file_param_user[param2].iloc[0] ymax=file_param_user[param2].iloc[1] yprior=file_param_user[param2].iloc[2] #------------------------------------------------------------------- # Plotting ensembles #------------------------------------------------------------------- iens=0 ii=1 for file_ens in metric_files : metrics=pd.read_csv(file_ens,sep=",",index_col="SIM") file_param=param_files[iens] params=pd.read_csv(file_param,sep=" ",index_col="t_IDs") if "WAVE" in metric_head : iwave=metrics.keys()[imetric].split('_')[0][4:] metric_head="WAVE"+str(iwave) label='Wave'+str(iwave) else : label=metric+" "+str(iens) iwave=-1 if(metric_head=="") : name_metric=metric else : name_metric=metric_head+'_'+metric #name_metric=metric if ( (("WAVE" in metric_head) & (int(iwave) in enstoprint)) | (file_ens in enstoprint) ) : zorder=ii color=colors_ens[ii] ii=ii+1 else : zorder=-100 color=colors_ens[0] #plt.scatter(params[param],metrics[name_metric],label=label,s=20,marker='x', color=color, zorder=zorder) #plt.scatter(params[param],params[param2], c=metrics[name_metric], label=label,s=20, marker='x', cmap=cmap, norm=norm, zorder=zorder) plt.scatter(params[param],params[param2], marker=marker_ens[iens],color=colors_ens[iens], s=ms_ens[iens], label=label_ens[iens], zorder=20-iens) #, cmap=cmap, norm=norm, zorder=zorder) iens=iens+1 order_of_magnitude = 10 ** math.floor(math.log10(abs(xmin))) ax.set_xlim(xlim) #xmin-0.1*order_of_magnitude, xmax+0.1*order_of_magnitude) print(xmin,xmax) ax.set_ylim(ylim) #ax.set_ylim(0.5, ymax) lw=2 #plt.axhline(y=0.81,color='k', ls='--', lw=lw) #opt-3T 0.836 plt.axvline(x=0.81,color='k', ls='--', lw=lw) #opt-3T 0.836 #plt.axhline(y=0.84,color='k', ls='--') #opt-3T 0.836 #plt.axhline(y=0.91,color='k', ls='-.') #opt-2T #plt.axhline(y=1,color='k', ls=':') #opt-T #plt.axhline(y=1.11, color='k', lw=2) #opt #plt.axhline(y=1.25,color='k', ls=':') #opt+T #plt.axhline(y=1.41,color='k', ls='-.') #opt+2T #plt.axhline(y=1.61,color='k', ls='--') #opt+3T plt.axvline(x=1.69,color='k', ls='--', lw=lw) #opt+3T ## end loop on ensembles (or waves) #plt.colorbar(label=metric) #------------------------------------------------------------------- # Plotting targets and errors #------------------------------------------------------------------- #obs_head=obs.keys()[imetric+1].split('_')[0] #name_obs_metric=obs_head+"_"+metric #plt.axhline(obs_mean[name_obs_metric],xmin=0,xmax=1,color='k') #plt.axhline(obs_mean[name_obs_metric]+(obs_var[name_obs_metric])**0.5,xmin=0,xmax=1,color='k', ls=':') #plt.axhline(obs_mean[name_obs_metric]-(obs_var[name_obs_metric])**0.5,xmin=0,xmax=1, color='k', ls=':') ##------------------------------------------------------------------- ## Plotting bests simulations ##------------------------------------------------------------------- isimu=0 for simu in list_bests : file_metrics=dico_bests["metrics"][simu] file_params=dico_bests["params"][simu] metrics=pd.read_csv(file_metrics,sep=",", index_col='SIM') params=pd.read_csv(file_params,sep=" ", index_col="t_IDs") if "SCM" in simu : iwave=simu.split('-')[1] metric_head="WAVE"+str(iwave) name_metric=metric_head+'_'+metric #name_metric=metric met_value=metrics.loc[simu][name_metric] param_value=params.loc[simu][param] param2_value=params.loc[simu][param2] color=listcoul[isimu] #plt.scatter(param_value,met_value,label=simu,s=20, marker='o', edgecolor="k", zorder=100,color=color) #plt.scatter(param_value,param2_value, c=met_value, label=label,s=20, marker='o', cmap=cmap, norm=norm, zorder=100, edgecolor='k') plt.scatter(param_value,param2_value ,s=20, marker='o', color="tab:red", zorder=100, edgecolor='k') isimu=isimu+1 ##end loop on bests simu #------------------------------------------------------------------- # Closing grpahics #------------------------------------------------------------------- fontsize=12 plt.xlabel(param,fontsize=fontsize) plt.ylabel(param2,fontsize=fontsize) #plt.colorbar(label=metric) #plt.legend(ncol=2,loc=(1.05,0.),fontsize=6) ax.tick_params(labelsize=fontsize) plt.legend(loc='upper center', bbox_to_anchor=(0.5,-0.15), fontsize=fontsize, frameon=False, ncol=2) #plt.legend(loc=(1.05,0.8),fontsize=fontsize) plt.savefig(path_fig+metric+"_"+param+"_"+param2+"wleg.png",bbox_inches='tight') ax.get_legend().remove() plt.savefig(path_fig+metric+"_"+param+"_"+param2+"noleg.png", bbox_inches='tight') plt.close() ## end loop on parameters # on regroupe les figures par métrique (idem Plots_Metric.pdf mais avec toutes les vagues) #os.system("pdfjam --fitpaper true "+path_fig+metric+"_*.pdf --outfile "+path_fig+metrics+"_vs_param.pdf") #print(path_fig+metrics+"_vs_param.pdf") ##end loop on metrics # on regroupe les figures par paramètres #for iparam,param in enumerate(params_ref.columns.values[1:]) : # os.system("pdfjam --nup 3x3 --landscape "+path_fig+"*"+param+"*.pdf --outfile "+path_fig++param+"_vs_metrics.pdf") # print(path_fig+param+"_vs_metrics.pdf")