Tracking Coastal Ecosystem Shifts in Cable Bay, BC (2018–2024)

Objective

This project aimed to monitor and quantify land cover changes in the Cable Bay coastal ecosystem using Sentinel-2 satellite imagery over a six-year period (2018-2024). The analysis focused on identifying ecological transitions and understanding their implications for coastal management and conservation efforts.

Methods

• Processed annual cloud-free Sentinel-2 Level-2A composites using Google Earth Engine
• Applied Random Forest classification algorithm with 100 trees for robust results
• Created 255 training polygons across multiple land cover classes for accurate classification
• Conducted transition matrix analysis to quantify detailed change patterns
• Performed hotspot analysis (Getis-Ord Gi*) to identify areas of significant change
• Applied Theil-Sen slope estimator for trend analysis of vegetation indices
• Utilized Python, R, ArcGIS Pro, and QGIS for processing and analysis

Technical Highlights

# Example of the Random Forest implementation in Python
from sklearn.ensemble import RandomForestClassifier
from sklearn.metrics import accuracy_score, confusion_matrix
import numpy as np
import matplotlib.pyplot as plt
import seaborn as sns

# Initialize the Random Forest model
rf_model = RandomForestClassifier(
    n_estimators=100, 
    random_state=42,
    max_depth=15,
    min_samples_split=10,
    bootstrap=True,
    oob_score=True
)

# Train the model
rf_model.fit(X_train, y_train)

# Make predictions
y_pred = rf_model.predict(X_test)

# Evaluate accuracy
accuracy = accuracy_score(y_test, y_pred)
conf_matrix = confusion_matrix(y_test, y_pred)

print(f"Model Accuracy: {accuracy * 100:.2f}%")
print(f"Out-of-bag score: {rf_model.oob_score_:.2f}")

# Feature importance visualization
feature_importance = rf_model.feature_importances_
sorted_idx = np.argsort(feature_importance)
plt.figure(figsize=(10, 6))
plt.barh(range(len(sorted_idx)), feature_importance[sorted_idx], align='center')
plt.yticks(range(len(sorted_idx)), [features[i] for i in sorted_idx])
plt.title('Random Forest Feature Importance')
plt.tight_layout()
plt.savefig('feature_importance.png', dpi=300)
                        

Outcomes

• Achieved exceptional classification accuracy between 97.0% and 98.6% across all years
• Revealed significant broadleaf forest loss (1,012 pixels) over the study period
• Documented urban growth (131 pixels) impacting coastal ecosystems
• Identified critical areas of change using hotspot analysis
• Quantified annual transition rates between major land cover classes
• Results support UN Sustainable Development Goal 15 (Life on Land)
• Findings contribute to British Columbia's Biodiversity Strategy

Key Findings

The analysis revealed significant ecological transitions in the Cable Bay area over the six-year period. The most notable changes included:

• Broadleaf forest conversion to urban development along the coastline
• Expansion of impervious surfaces impacting watershed dynamics
• Shifts in wetland vegetation composition due to changing hydrological patterns
• Incremental coastal erosion affecting nearshore habitats
• Correlation between temperature anomalies and vegetation stress patterns

Significance

This project demonstrates the power of multi-temporal satellite imagery and machine learning techniques for monitoring environmental change. The high classification accuracy validates the methodology and provides a framework for similar analyses in other coastal regions. The findings have been shared with local conservation authorities to inform habitat protection strategies and sustainable development planning.