# requirements: # pip install psutil # conda install -c plotly plotly-orca import pandas as pd import plotly.graph_objects as go import numpy as np import textwrap # Read in CSV files and set the 'original_name' column as the index for both DataFrames df_fr = pd.read_csv('../table/CropDeepTrans_FigS1_Sankey_FR.csv').set_index('original_name') df_nl = pd.read_csv('../table/CropDeepTrans_FigS1_Sankey_NL.csv').set_index('original_name') # Set the number of codes to display in the chart, or 'All' if all codes should be displayed. otherwise use integer like Num_code_to_display=40 # Num_code_to_display = 'All' Num_code_to_display = 40 # Add a 'new_name' column to each DataFrame and assign it the same values as the index df_nl['new_name'] = df_nl.index df_fr['new_name'] = df_fr.index # If Num_code_to_display is an integer, limit the number of codes displayed and group the remaining codes together as 'Anderen' for BRP and 'Autre' for RPG if isinstance(Num_code_to_display, int) : df_nl =df_nl.reset_index() df_nl.new_name.loc[df_nl.sort_values('area',ascending=False).iloc[Num_code_to_display:,:].index] = 'Anderen' df_nl =df_nl.set_index('new_name') df_fr =df_fr.reset_index() df_fr.new_name.loc[df_fr.sort_values('area',ascending=False).iloc[Num_code_to_display:,:].index] = 'Autre' df_fr =df_fr.set_index('new_name') List_HCAT2 = list(np.unique(\ df_fr.sort_values('area',ascending=False).iloc[:Num_code_to_display,:].HCAT2_code.unique().tolist() + \ df_nl.sort_values('area',ascending=False).iloc[:Num_code_to_display,:].HCAT2_code.unique().tolist() )) rpg = df_fr.sort_values('area',ascending=False).iloc[:Num_code_to_display,:].index.unique().tolist() brp = df_nl.sort_values('area',ascending=False).iloc[:Num_code_to_display,:].index.unique().tolist() else:# If display All codes df_nl =df_nl.set_index('new_name') df_fr =df_fr.set_index('new_name') List_HCAT2 = list(np.unique(\ df_fr.HCAT2_code.unique().tolist() + \ df_nl.HCAT2_code.unique().tolist() )) rpg = df_fr.index.unique().tolist() brp = df_nl.index.unique().tolist() # Create a list of labels for the nodes in the Sankey diagram label = rpg+brp+List_HCAT2+['Missing RPG code','Missing BRP code','Autre','Anderen'] # Iterate over each HCAT2_code in List_HCAT2 and create links for each code in both the French and Dutch DataFrames source=[] target=[] value=[] # Iterate through each HCAT2 code in the list for i,h in enumerate(List_HCAT2): # Filter the dataframe to only include rows with the current HCAT2 code and group by new_name while summing the area df_temp = df_fr.loc[df_fr.HCAT2_code == h].reset_index('new_name').groupby('new_name').sum() # If there are any rows remaining in the filtered dataframe, iterate through each row if df_temp.index.shape[0]>0: # Iterate through each row in the filtered dataframe for c in df_fr.loc[df_fr.HCAT2_code == h].index: # If the current row's index is in the RPG list, append its label index to the source list if c in rpg: source.append(label.index(c)) # Otherwise, append the label index for "Autre" to the source list else: source.append(label.index('Autre')) # Append the label index for the current HCAT2 code to the target list target.append(label.index(h)) # Append the area value for the current row to the value list value.append(df_temp.loc[c].area) # If there are no rows remaining in the filtered dataframe, append a small value to the value list and the label index for "Missing RPG code" to the source list else: target.append(label.index(h)) source.append(label.index('Missing RPG code')) value.append(0.0001) # Iterate through each HCAT2 code in the list for i,h in enumerate(List_HCAT2): # Filter the dataframe to only include rows with the current HCAT2 code and group by new_name while summing the area df_temp = df_nl.loc[df_nl.HCAT2_code == h].reset_index('new_name').groupby('new_name').sum() # If there are any rows remaining in the filtered dataframe, iterate through each row if df_temp.index.shape[0]>0: # Iterate through each row in the filtered dataframe for c in df_temp.index: # Append the label index for the current new_name to the target list target.append(label.index(c)) # Append the label index for the current HCAT2 code to the source list source.append(label.index(h)) # Append the area value for the current row to the value list value.append(df_temp.loc[c].area) # If there are no rows remaining in the filtered dataframe, append a very small value to the list and the label index for "Missing BRP code" to the target list else: source.append(label.index(h)) target.append(label.index('Missing BRP code')) value.append(0.0001) # Check if Num_code_to_display is an integer, ie not "All" if isinstance(Num_code_to_display, int): # If so, update the label for 'Autre' and 'Anderen' label[label.index('Autre')] = 'Other RPG codes (>' + str(Num_code_to_display) + 'ᵗʰ main ones)' label[label.index('Anderen')] = 'Other BRP codes (>' + str(Num_code_to_display) + 'ᵗʰ main ones)' # Shorten label text to a maximum of 40 characters with ellipses label = [textwrap.shorten(x, width=40, placeholder='...') for x in label] # Create Sankey diagram with specified node and link attributes fig = go.Figure(data=[go.Sankey( node = dict( pad = 15, thickness = 20, line = dict(color = 'black', width = 0.5), label =label, color = 'blue', hovertemplate='%{value}%' ), link = dict( source = source, # indices correspond to labels, eg A1, A2, A1, B1, ... target = target, value=value ))]) # Update layout with title, font size, and dimensions fig.update_layout(title_text='FR / NL crops harmonization ('+str(Num_code_to_display)+' crops selected)',\ font_size=10,width=900, height=len(label)*7) # Display Sankey diagram - may not work everywhere. better to save html first # fig.show() # Write Sankey diagram as HTML file fig.write_html('../data/sankey_'+str(Num_code_to_display)+'_crops.html') # Write Sankey diagram as PNG file with double resolution fig.write_image('../data/sankey_'+str(Num_code_to_display)+'_crops.png', scale=2)