Deploying Deep Learning for Real-Time Tsunami Monitoring: A CNN-LSTM Model for Sensor-Based Prediction

Early detection of tsunamis is essential for disaster risk reduction and minimizing loss of life. This research introduces an advanced deep learning-based tsunami detection model that integrates Convolutional Neural Networks (CNN) with Long Short-Term Memory (LSTM) networks. The dataset comprises seismic waveform signals, oceanographic sensor data, and satellite imagery, incorporating critical earthquake attributes such as magnitude, depth, epicenter location, and past tsunami occurrences. To enhance feature extraction, Short-Time Fourier Transform (STFT) is applied, converting time-domain seismic signals into frequency-domain spectrograms. The CNN component captures spatial characteristics in seismic spectrograms, while the LSTM component processes sequential oceanic sensor readings to analyze long-term dependencies in water level fluctuations and seismic activity patterns. The model is optimized using binary cross-entropy loss and the Adam optimizer, ensuring efficient learning while minimizing false alarms. Evaluation metrics, including accuracy, precision, recall, F1-score, and AUC-ROC, confirm that the CNN + LSTM framework surpasses standalone deep learning models such as CNN, LSTM, and GRU. Experimental findings highlight the model's capability to provide accurate tsunami predictions with reduced false positives. This approach can be implemented in real-time tsunami early warning systems, IoT-based monitoring platforms, and disaster risk assessment tools to improve preparedness and response strategies.