# ======================================================================
## Fig. 2: pie chart & beta distribution
blue_c = '#0072B2'
brown_c = '#E69F00'
green_c = '#009E73'
order_gen = ['detectable', 'non-detectable']
color_dic = {'Gc': blue_c, 'GPP': brown_c, 'iWUE': green_c, 'Gs': blue_c}
beta_values = {
'Gc': df_SNR['Gc_beta'].values,
'GPP': df_SNR['GPP_beta'].values,
'iWUE': df_SNR['iWUE_beta'].values
}
data = [beta_values['GPP'], beta_values['Gc'], beta_values['iWUE']]
data = pd.DataFrame(beta_values)
melted_data = data.melt(var_name='Variable', value_name='Beta Value')
fontsize = 6.5
plt.rcParams["legend.frameon"] = False
plt.rcParams['savefig.dpi'] = 800
plt.rcParams['font.size'] = fontsize
plt.rcParams['font.family'] = 'Arial'
def pie_chart(data, fig, ax, colors, startangle, order_gen, a, b, fontsize=fontsize):
wedges, texts, _ = ax.pie(data, autopct=lambda x: '{:.0f}'.format(x * df_SNR.shape[0] / 100),
wedgeprops={"alpha": 1, "edgecolor": "none", 'linewidth': 0.5, 'antialiased': True},
textprops={'size': fontsize, 'weight': 'bold'}, colors=colors, startangle=startangle)
bbox_props = dict(boxstyle="square,pad=0.3", fc="w", ec="none")
kw = dict(arrowprops=dict(arrowstyle="-", lw=0.5),
bbox=bbox_props, zorder=0, va="center", fontsize=fontsize)
for i, p in enumerate(wedges):
ang = (p.theta2 - p.theta1) / 2. + p.theta1
y = np.sin(np.deg2rad(ang))
x = np.cos(np.deg2rad(ang))
horizontalalignment = {-1: "right", 1: "left"}[int(np.sign(x))]
connectionstyle = f"angle,angleA=0,angleB={ang}"
kw["arrowprops"].update({"connectionstyle": connectionstyle})
ax.annotate(order_gen[i], xy=(x, y), xytext=(a * np.sign(x), b * y),
horizontalalignment=horizontalalignment, **kw)
width_old, height_old = 8.2, 7.5
width_new = 3.8
height_new = height_old / width_old * width_new
fig = plt.figure(figsize=(width_new, height_new), dpi=200)
shape = (2, 2)
ax_1 = plt.subplot2grid(shape, loc=(0, 0), colspan=2, rowspan=1)
ax_2 = plt.subplot2grid(shape, loc=(1, 0), colspan=1, rowspan=1)
ax_3 = plt.subplot2grid(shape, loc=(1, 1), colspan=1, rowspan=1)
blue_pie, red_pie = '#4292C5', '#FFC20A'
df_group = df_SNR.groupby('Gc_cluster').count().iloc[:, 0]
data = df_group.loc[order_gen].values
pie_chart(data, fig, ax_2, [blue_pie, red_pie], 90, order_gen, 0.89, 1.2)
df_group = df_SNR.groupby('GPP_cluster').count().iloc[:, 0]
data = df_group.loc[order_gen].values
pie_chart(data, fig, ax_3, [blue_pie, red_pie], 15, order_gen, 0.89, 1.2)
targets = ['Gc', 'GPP']
for i, ax in enumerate([ax_2, ax_3]):
ax.set_ylabel('')
ax.set_title(f'The $\\mathbf{{CO_2}}$ effect on {targets[i]}',
fontweight='bold', fontsize=6.5)
violin_order = ['iWUE', 'Gc', 'GPP']
violin_parts = sns.violinplot(x='Beta Value', y='Variable',
data=melted_data, scale='area',
palette=color_dic, cut=0,
order=violin_order, ax=ax_1, inner=None, width=1, zorder=1)
for i, violin in enumerate(violin_parts.collections):
violin.set_edgecolor(None)
violin.set_alpha(0.25)
size = 3.5
alpha = 0.35
jitter_strength = {'iWUE': 0.08, 'Gc': 0.08, 'GPP': 0.16}
seeds = [42, 42, 46]
for i_v, variable in enumerate(violin_order):
x_data = melted_data[melted_data['Variable'] == variable]['Beta Value']
np.random.seed(seeds[i_v])
y_data = np.random.normal(loc=violin_order.index(variable), scale=jitter_strength[variable], size=len(x_data))
ax_1.scatter(x_data, y_data, color=color_dic[variable], s=size ** 2, alpha=alpha, zorder=10, edgecolors='none')
quantiles = melted_data.groupby('Variable')['Beta Value'].quantile([0, 0.5, 1])
iqr_values = quantiles.unstack()
for i, var in enumerate(violin_order):
q1 = iqr_values.loc[var, 0]
median = iqr_values.loc[var, 0.5]
q3 = iqr_values.loc[var, 1]
violin_color = violin_parts.collections[i].get_facecolor().flatten()
linecolor = (0.2, 0.2, 0.2, 1)
edgewidth = 1
ax_1.errorbar(x=median, y=i, xerr=[[median - q1], [q3 - median]], fmt='o', color=violin_color,
ecolor=linecolor,
elinewidth=edgewidth,
capthick=2,
capsize=1.7,
markersize=7,
alpha=1,
markeredgewidth=edgewidth,
markeredgecolor=linecolor,
zorder=10)
x_dic = {'iWUE': -0.05, 'Gc': 0.25, 'GPP': 0.01}
h_dic = {'iWUE': 0.3, 'Gc': 1.3, 'GPP': 2.6}
for var in ['iWUE', 'Gc', 'GPP']:
median = df_SNR[f'{var}_beta'].median()
ax_1.text(median + x_dic[var], h_dic[var], r'$\tilde \beta_{%s}=%.2f$' % (var, median),
color=color_dic[var],
verticalalignment='center',
horizontalalignment='center',
fontsize=fontsize)
ax_1.set_ylabel('')
ax_1.set_xlabel(r'$\mathbf{\beta \:}$ (dimensionless)', fontsize=fontsize)
ax_1.set_ylim([2.8, -0.4])
ax_1.set_xlim([-5.5, 5.5])
ax_1.tick_params(axis='x', labelsize=fontsize + 0.5)
for label in ax_1.get_yticklabels():
label.set_fontweight('bold')
label.set_fontsize(fontsize + 0.5)
for ax in [ax_1, ax_2, ax_3]:
ax.tick_params(width=0.3, length=2)
for spine in ax.spines.values():
spine.set_linewidth(0.3)
ax_1.annotate('a', xy=(4, 83), xycoords='axes points', fontweight='bold', fontsize=fontsize + 0.5)
ax_2.annotate('b', xy=(4, 100), xycoords='axes points', fontweight='bold', fontsize=fontsize + 0.5)
ax_3.annotate('c', xy=(4, 100), xycoords='axes points', fontweight='bold', fontsize=fontsize + 0.5)
plt.tight_layout()
# ===================== 自动保存图片 =====================
save_path = os.path.join(save_fig_dir, 'Fig2.png')
plt.savefig(save_path, dpi=800, bbox_inches='tight')
print(f"✅ 图片已保存到：\n{save_path}")
# ================================================================
plt.show()
