Analyzing the Relationship Between Ocean Temperature Anomalies and Coral Bleaching Events

This repository explores the impact of ocean temperature anomalies on coral bleaching events. Leveraging a comprehensive dataset of global coral reef observations, this project examines environmental and geographical factors contributing to coral stress and develops predictive models to assess bleaching severity.

---

Key Highlights

1. Dataset Overview

• Source: Global observations of coral reefs, covering temperature, environmental stressors, and bleaching percentages.
• Key Features:
  • Environmental: Sea Surface Temperature (SST), Temperature Stress Anomalies (TSA), Degree Heating Weeks (DHW), Turbidity.
  • Geographical: Latitude, Longitude, Depth, Distance to Shore, Exposure.
  • Temporal: Date (Year, Month, Day extracted for modeling).
  • Target Variable: Percentage of coral bleaching.

---

2. Project Goals

• Analysis: Identify key environmental drivers of coral bleaching.
• Prediction: Build a robust machine learning model to forecast bleaching severity.
• Conservation Impact: Aid in understanding and mitigating the risks posed by climate change to coral ecosystems.

---

3. Insights from Data

• Temperature Stress: Coral bleaching severity increases significantly when SST exceeds 27°C, indicating a critical threshold.
• Geographical Vulnerability: Coral reefs in tropical regions, especially the Caribbean and Indo-Pacific, are highly prone to bleaching.
• Sustained Stress: Prolonged exposure to heat (measured by DHW) is strongly associated with severe bleaching.

---

Methods and Models

1. Data Preprocessing

• Removed irrelevant columns and addressed missing values.
• Encoded categorical variables (e.g., Ocean Name, Exposure) numerically.
• Normalized numerical features using PowerTransformer and RobustScaler.
• Engineered features like:
  • Interaction terms (e.g., SST × TSA).
  • Rolling averages for time-series effects.
  • Clusters for geographical and environmental patterns using K-Means.

---

2. Exploratory Data Analysis

• Visualized the relationship between SST, anomalies, and bleaching.
• Analyzed geographical distributions of bleaching events and temperature anomalies.
• Explored temporal trends in coral bleaching across decades.

---

3. Modeling Approach

• Evaluated multiple regression models:
  • Ridge, Elastic Net, Random Forest, Support Vector Regressor (SVR), Decision Tree, CatBoost, and AdaBoost.
• Best Model: Random Forest Regressor, tuned via Random Search.
  • R² Score (Post-Tuning): 0.7765
  • MSE: 0.0525
  • MAE: 0.1574
  • Hyperparameters:

    {
        'n_estimators': 500,
        'max_depth': 30,
        'min_samples_split': 10,
        'min_samples_leaf': 1,
        'max_features': 'sqrt',
        'bootstrap': False
    }

---

Results and Visualizations

• Key Predictors: SST, anomalies (SSTA, TSA), DHW, and geographical factors (Depth, Distance to Shore).
• Geographical Hotspots: Coral bleaching is most severe in the Caribbean, Southeast Asia, and Red Sea regions.
• Temporal Trends: Coral bleaching events peaked in the late 1990s and 2005, corresponding to global thermal stress events.

Model Performance Metrics:

Metric	Initial Model	Tuned Model
MSE	0.0685	0.0525
RMSE	0.2618	0.2291
MAE	0.1978	0.1574
R² Score	0.7048	0.7765

---

Topic Modeling Insights

Topic modeling was performed using published news articles about the damaging consequences of climante change on coral reefs due to bleaching events. Seven topics rlated to coral bleaching were identified in published articles over the last 5 years.

Who benefits

The revealed topics are beneficial to readers, including

1. members of public looking to contribute to recovery efforts to preserve this natural resource (coral reefs)
2. policy makers looking to understand factors involved in implementing reforms

How it aids recovery efforts

Identifies ongoing global efforts to offset impact of bleaching events on coral reefs, and possible obstacles that are being encountered.