Global CO2 Emissions Dashboard

Interactive data visualization dashboard that explores 64 years of global CO2 emissions (1960-2023) across 205 countries, featuring volume analysis, growth rate tracking, and dynamic filtering—deployed live on Streamlit Cloud.

Dashboard Preview

• Live Demo
• GitHub Repository
• EDA Report (Jupyter Notebook)

Objectives

This project aims to:

1. Visualize global emission trends: Build an interactive dashboard to explore CO2 emissions across countries and time periods

2. Identify key patterns: Analyze both volume leaders (who emits the most) and growth leaders (who’s growing fastest)

3. Enable data exploration: Provide intuitive filters and chart options for users to discover insights themselves

Overview

An end-to-end dashboard development—from data wrangling to cloud deployment. With climate change being one of the most pressing global issues, I wanted to created a tool that makes CO2 emissions data accessible and explorable for anyone.

The dashboard answers two fundamental questions:

• Who are the biggest emitters? → China, USA, India dominate
• Who’s growing the fastest? → Southeast Asian nations show explosive growth

Key Features

• Interactive Visualizations: Toggle between Area and Line charts for different perspectives
• Dual Analysis Tabs: Volume Analysis (absolute emissions) + Growth Analysis (% change)
• Dynamic Filtering: Year range slider, country multi-select, Top N selector
• Real-Time Metrics: KPI cards showing total emissions, YoY change, top emitter
• Growth Annotations: Charts display growth percentages directly on data points
• Responsive Design: Works on desktop and mobile browsers

Technical Implementation

Dashboard Architecture

User Input (Sidebar)
  → Data Filtering (Pandas)
  → Visualization (Plotly)
  → Cloud Deployment (Streamlit Cloud)

Data Pipeline

Built a robust data cleaning system to handle World Bank dataset quirks:

# Three-layer filtering to remove aggregate regions
df_clean = df[
    (~df['country_name'].isin(ignore_list)) &      # Layer 1: Name list
    (~df['country_code'].isin(aggregate_codes)) &  # Layer 2: Code list  
    (~df['country_name'].apply(looks_like_aggregate))  # Layer 3: Pattern matching
].copy()

# Result: 12,818 → 11,103 rows (205 actual countries)

Growth Rate Calculation

Implemented growth analysis with statistical safeguards:

# Filter significant emitters to avoid misleading percentages
# (e.g., 10 → 20 kt = 100% growth, but insignificant)
significant = df_end[df_end > 10000].index

# Calculate growth rate
growth_rate = ((df_end - df_start) / df_start * 100)

# Safety: Only include countries with data in BOTH years
valid_countries = df_start.index.intersection(significant)

Visualization Features

Dynamic chart generation with user-controlled options:

# Area vs Line toggle affects both tabs
if chart_type == "Area":
    fig = px.area(df_trend, x='year', y='value', color='country_name')
else:
    fig = px.line(df_trend, x='year', y='value', color='country_name', markers=True)

# Growth annotations on chart endpoints
fig.add_annotation(
    x=end_year, y=value,
    text=f"+{growth_pct:.1f}%",
    showarrow=True
)

Challenges and Solutions

Challenge 1: Aggregate Region Contamination

• Problem: Dataset mixed actual countries with aggregate regions (“World”, “OECD”, “High Income”), causing double-counting.
• Solution: Developed 3-layer filtering system (name list + code list + pattern matching) that removes 13% of rows while preserving all 205 actual countries.

Challenge 2: Area Chart Stacking Order

• Problem: Plotly’s default stacking order made charts confusing—smaller countries appeared on top.
• Solution: Implemented dynamic sorting by total emissions so largest emitters stack at bottom, matching visual intuition.

Challenge 3: Growth Rate Annotation Placement

• Problem: In Area charts, annotations pointed to wrong positions (individual values vs. cumulative stack).
• Solution: Built separate annotation logic for Area (cumulative positions) vs. Line (individual values) charts.

Challenge 4: GitHub Notebook Rendering

• Problem: Plotly visualizations don’t render on GitHub, making EDA notebook useless for repository.
• Solution: Rewrote notebook using Matplotlib/Seaborn for static images that render properly on GitHub.

Technologies Used

• Framework: Streamlit
• Data Processing: Pandas, NumPy
• Visualization: Plotly Express (Dashboard), Matplotlib/Seaborn (EDA Notebook)
• Data Source: World Bank / Our World in Data
• Deployment: Streamlit Cloud
• Version Control: Git, GitHub

Key Learnings

1. Dashboard UX: Small details matter—chart type toggles, proper sorting, and annotation placement significantly improve user experience

2. Data Cleaning: Real-world datasets are messy. The 3-layer filtering approach taught me to think defensively about data quality.

3. Deployment Pipeline: First time deploying a live web dashboard app. Learned about requirements.txt, environment consistency, and Streamlit Cloud workflow.

4. Portfolio Presentation: GitHub doesn’t render interactive charts—always have a static fallback for notebooks.

Performance Metrics

Dashboard Stats

• Countries Covered: 205
• Time Span: 64 years (1960-2023)
• Data Points: 11,103 country-year records
• Load Time: <3 seconds

Key Findings

Future Enhancements

• Per-Capita Analysis: Add population data for fairer country comparisons
• Sector Breakdown: Show emissions by industry, transport, residential
• Predictive Modeling: Forecast future emissions using time series
• Data Export: Allow users to download filtered datasets
• Embedding: Make dashboard embeddable in other websites

Project Structure

global-co2-insight/
├── data/
│   └── co2_emissions_kt_by_country_2023.csv
├── notebooks/
│   └── 01_data_eda.ipynb          # EDA with static visualizations
├── src/
│   └── app.py                      # Streamlit dashboard
├── assets/                         # images for README/portfolio
├── requirements.txt
└── README.md

Try It Out

• Live Dashboard: global-co2-insight-by-ju-ho-kim.streamlit.app
• GitHub Repository: github.com/jurinho17-sv/global-co2-insight
• Data Source: World Bank Data360

Visual Results

Analysis Highlights

• Volume Tab: China’s dramatic rise, surpassing US around 2006
• Growth Tab: Southeast Asian nations leading with >1000% growth rates
• Interactive: Users can filter by year range and select specific countries

Wrapping up

This was my first time creating a dashboard. The data cleaning process was stressful, but seeing the finished dashboard made it all worth it.

See you in another project 😁

Written at Grimes Engineering Center.