[Python] log-log time series plot with event markers

I want to create a log-log plot showing the growth of publications over time (papers/year), similar to scientific figures like Fig. 1a in "A Century of Physics" (Nature Physics 2015).

Requirements:

1. Y-axis (number of papers) on a log scale (10²–10⁶) with ticks at each power of 10.
2. X-axis (year) with decade ticks (e.g., 1960, 1970, 1980...).
3. Multiple vertical dashed lines indicating important trend changes.
4. Serif font, light dashed grid, no top/right spines.
5. Line colors and styles similar to journal figures (e.g., matplotlib, Seaborn).
6. A panel label (e.g., "Web of Science") at the top left.

How can I best style my matplotlib plot to match this?

The following is what I did (in Python):

#!/usr/bin/env python3

"""
replicate_figure1a_misinformation.py

Count the yearly number of WoS publications (total) versus
the number of misinformation publications (from combined_journals.xlsx)
from 1960 to 2024, and plot both curves (log‐scale) to reproduce
“Figure 1 | The Evolution of Physics” analogous to Sinatra et al.
"""

import os
import gzip
import xml.etree.ElementTree as ET
from lxml import etree

import pandas as pd
import matplotlib.pyplot as plt

# 1) Configuration
# ──────────────────────────────────────────────────────────
# Path to your combined misinformation Excel file
MISINFO_XLSX = 'combined_journals.xlsx'

# Root directory containing files like 1960.xml.gz … 2024.xml.gz
WOS_DIR = '/Users/DataSets/Documents/DataSets/WoS/merged_yearly'

# Years to include in the plot
START_YEAR = 1960
END_YEAR = 2024
YEARS = list(range(START_YEAR, END_YEAR + 1))


# 2) Load misinformation paper counts by year
# ──────────────────────────────────────────────────────────
misinfo_df = pd.read_excel(MISINFO_XLSX, dtype={'UT (Unique WOS ID)': str,
'Publication Year': int,
'WoS Categories': str})

# ensure only years in our window
misinfo_df = misinfo_df[
(misinfo_df['Publication Year'] >= START_YEAR) &
(misinfo_df['Publication Year'] <= END_YEAR)
]

# count unique UT per year
mis_counts = (
misinfo_df
.drop_duplicates(subset='UT (Unique WOS ID)')
.groupby('Publication Year')
.size()
.reindex(YEARS, fill_value=0)
.to_dict()
)

print(f"Loaded misinformation counts for {len(mis_counts)} years")


# 3) Parse WoS XML to get total counts per year
# ──────────────────────────────────────────────────────────
total_counts = {yr: 0 for yr in YEARS}

for yr in YEARS:
gz_path = os.path.join(WOS_DIR, f'{yr}.xml.gz')
if not os.path.exists(gz_path):
continue

with gzip.open(gz_path, 'rb') as f:
# directly enable recovery and huge_tree here
for _, rec in etree.iterparse(
f,
events=('end',),
tag='REC',
recover=True,
huge_tree=True
):
total_counts[yr] += 1
rec.clear()


# 4) Build DataFrame and plot
# ──────────────────────────────────────────────────────────
df_counts = pd.DataFrame({
'Year': YEARS,
'WoS total': [total_counts[yr] for yr in YEARS],
'Misinformation': [mis_counts.get(yr, 0) for yr in YEARS],
})
df_counts.set_index('Year', inplace=True)

plt.figure(figsize=(10, 6))
# plt.plot(df_counts.index, df_counts['WoS total'], label='WoS total papers')
plt.plot(df_counts.index, df_counts['Misinformation'], label='Web of Science')
plt.yscale('log')
plt.xlabel('Year')
plt.ylabel('Number of papers')
plt.title('The Evolution of Misinformation (1960–2024)')
plt.legend()
plt.grid(True, which='both', linestyle='--', alpha=0.3)
plt.tight_layout()

# To save the figure instead of showing:
plt.savefig('figure1a_misinformation.png', dpi=600)
# Or to display:
# plt.show()


Attached is my plot and the plot of the article I am trying to replicate exactly same.



Continue reading...