Artificial Intelligence Oceanography

Preface
Contents
Acronyms
Artificial Intelligence Foundation of Smart Ocean
1 The Development of Artificial Intelligence
1.1 The First-Generation AI
1.2 The Second-Generation AI
1.3 The Third-Generation AI
2 The Architecture of Deep Neural Networks
2.1 Deep Feedforward Neural Network
2.2 Deep Convolutional Neural Network
2.3 Deep Recurrent Neural Network
2.4 Deep Generative Adversarial Network
3 Perceptual Understanding Based on Neural Network
3.1 Recognition Based on Neural Network
3.2 Segmentation Based on Neural Network
3.3 Prediction Based on Neural Network
References
Forecasting Tropical Instability Waves Based on Artificial Intelligence
1 Sea Surface Temperature and Tropical Instability Waves
2 Data and Model of SST Forecasting
2.1 Satellite Remote Sensing SST Data
2.2 Architecture and Training of the DL Model
3 SST Forecast of TIW Motion Using the DL Model During the Testing Period (2010/01–2019/03)
4 Interannual Variation in TIW Westward Propagation
5 Zonally Westward Propagation of TIWs
6 Accuracy During the Testing Period (2010/01–2019/03)
7 Conclusions
References
Sea Surface Height Anomaly Prediction Based on Artificial Intelligence
1 Significance of Sea Surface Height Anomaly Prediction
2 Review of SSHA Predicting Methods
3 Multi-Layer Fusion Recurrent Neural Network for SSHA Field Prediction
3.1 A Classical Spatiotemporal Forecasting Architecture: ConvLSTM
3.2 Architecture of MLFrnn
3.3 Multi-layer Fusion Cell
4 Experimental Results and Discussion
4.1 Study Area and Dataset
4.2 Implementation Detail
4.3 Experiment Results and Discussion
4.4 Ablation Study
5 Conclusion
References
Satellite Data-Driven Internal Solitary Wave Forecast Based on Machine Learning Techniques
1 Introduction
2 Satellite Observations of ISWs
3 Machine-Learning-Based ISW Forecast Model
3.1 Model Establishment
3.2 Model Training
3.3 Model Validation
4 Influence Factors on the ISW Forecast Model
5 Conclusions and Future Works
References
AI-Based Subsurface Thermohaline Structure Retrieval from Remote Sensing Observations
1 Introduction
2 Study Area and Data
3 Retrieving Subsurface Thermohaline Based on Ensemble Learning
3.1 EXtreme Gradient Boosting (XGBoost)
3.2 Random Forests (RFs)
4 Predicting Subsurface Thermohaline Based on Deep Learning
4.1 Bi-Long Short-Term Memory (Bi-LSTM)
4.2 Convolutional Neural Network (CNN)
5 Conclusions
References
Ocean Heat Content Retrieval from Remote Sensing Data Based on Machine Learning
1 Introduction
2 Data
3 Method
3.1 Neural Network
3.2 Design of Experiments
4 Results and Analysis
4.1 Optimization of Feature Combinations
4.2 Deep, or Shallow—That Is the Question
4.3 Data Reconstruction
5 Summary and Conclusions
References
Detecting Tropical Cyclogenesis Using Broad Learning System from Satellite Passive Microwave Observations
1 Introduction
2 Data Description
3 Broad Learning System for Tropical Cyclogenesis Detection
3.1 Broad Learning Model
3.2 Incremental Learning of BLS
4 Results
4.1 Basic BLS Results
4.2 Incremental Learning Results
4.3 Case Study: Hurricane Wilma (2005)
5 Conclusion
References
Tropical Cyclone Monitoring Based on Geostationary Satellite Imagery
1 Introduction
2 Data and Methodology
2.1 Data
2.2 Data Pre—Processing
2.3 Methodology
3 TC Center Location
4 TC Intensity Estimation
5 Summary
References
Reconstruction of pCO2 Data in the Southern Ocean Based on Feedforward Neural Network
1 Introduction
1.1 Observations of pCO2 in Southern Ocean
1.2 Comparison of Reconstruction pCO2 Data
2 Data and Methods
2.1 Data
2.2 Nonlinear Neural Network Model for the pCO2 Reconstruction in the Southern Ocean
2.3 Calculation of Carbon Dioxide Flux in the Southern Ocean
2.4 Evaluation
3 Results and Discussion
3.1 Seasonal Variation in Southern Ocean Sea Surface pCO2
3.2 Annual Variation in Southern Ocean Sea Surface pCO2
3.3 Variability in Sea—AirCO2 Flux
4 Conclusion
References
Detection and Analysis of Mesoscale Eddies Based on Deep Learning
1 Introduction
2 DL–based Eddy Detection Model Based on SSHA Data
2.1 Data
2.2 Method
2.3 Experiment and Performance
3 DL–based Eddy Detection Model Based on SSHA and SST Data
3.1 Data
3.2 Method
3.3 Experiment and Performance
4 Characterization Analysis of Mesoscale Eddies in the Global Ocean
4.1 Spatiotemporal Distributions of Eddies in the Global Ocean
4.2 Long-term Variations in Derived Eddy Parameters
5 Conclusions
References
Deep Convolutional Neural Networks-Based Coastal Inundation Mapping from SAR Imagery: with One Application Case for Bangladesh, a UN-defined Least Developed Country
1 Introduction
2 Dataset
2.1 Data Description
2.2 Data Preparation
3 Model
4 Performance Evaluations and Discussions
4.1 Performance Evaluations
4.2 Spatial and Temporal Characteristics
4.3 Discussions of Performance
5 Application Case in Bangladesh
6 Conclusions
References
Sea Ice Detection from SAR Images Based on Deep Fully Convolutional Networks
1 Introduction
2 Data
2.1 Study Area
2.2 SAR Images
2.3 NSIDC Sea Ice Products
2.4 Data Preprocessing
3 Method
3.1 Overall Structure of DAU—Net
3.2 Encoder
3.3 Attention
3.4 Decoder
3.5 Output
4 Experiments
4.1 Experiments Setting
4.2 Evaluation Metrics
4.3 Comparison Experiments Against Other Models Performances
4.4 Effectiveness of IA
4.5 Effectiveness of Dual—Polarization Information
4.6 Performances of Different ResNet-Based Encoders
5 Discussions
6 Conclusions
References
Detection and Analysis of Marine Green Algae Based on Artificial Intelligence
1 Introduction
2 Data and Methodology
2.1 Satellite Images and Labels
2.2 UNet-Based Algae Detection Network (AlgaeNet)
2.3 Model's Performance
3 Results and Discussion
4 Conclusions
References
Automatic Waterline Extraction of Large-Scale Tidal Flats from SAR Images Based on Deep Convolutional Neural Networks
1 Introduction
2 Study Area and Data
3 Methodology
3.1 U-Net
3.2 Data Preparation
3.3 Training
4 Results
4.1 Model Performance
4.2 Automatic Topographic Mapping of Tidal Flats
5 Discussions
6 Conclusions
References
Extracting Ship's Size from SAR Images by Deep Learning
1 Introduction
2 Traditional Methods
2.1 Typical Procedure of Traditional Methods
2.2 Representative Traditional Methods
2.3 Issue to be Further Addressed
3 Deep Learning Method
3.1 Ship Detection Based on DL
3.2 SSENet: A Deep Learning Model to Extract Ship Size from SAR Images
3.3 Experiments on SSENet
4 Discussions
4.1 ML versus DL
4.2 Errors's Sources
5 Conclusions
References
Benthic Organism Detection, Quantification and Seamount Biology Detection Based on Deep Learning
1 Overview
1.1 Backgrounds
1.2 Related Works
1.3 Research Content and Innovation
2 The Target Detection Techniques
2.1 Introduction on Target Detection
2.2 The Single-Stage Target Detection
2.3 The Two-Stage Target Detection
2.4 Summary
3 DQBO Based on Faster R-CNN with FPN
3.1 Introduction on DQBO
3.2 The Faster R-CNN with FPN Framework for DQBO
3.3 Experimental Results and Discussions
3.4 Summary
4 DMS Based on SSD
4.1 Introduction on DMS
4.2 Seamount Macrobenthos Dataset
4.3 The SSD Framework for DMS
4.4 Experimental Results and Discussions
4.5 Summary
5 Conclusions and Future Works
References