Physics-Guided Generative Adversarial Networks for Sea Subsurface Temperature Prediction

Abstract

Sea subsurface temperature, an essential component of aquatic wildlife, underwater dynamics and heat transfer with the sea surface, is affected by global warming in climate change. Existing research is commonly based on either physics-based numerical models or data based models. Physical modeling and machine learning are traditionally considered as two unrelated fields for the sea subsurface temperature prediction task, with very different scientific paradigms (physics-driven and datadriven). However, we believe both methods are complementary to each other. Physical modeling methods can offer the potential for extrapolation beyond observational conditions, while datadriven methods are flexible in adapting to data and are capable of detecting unexpected patterns. The combination of both approaches is very attractive and offers potential performance improvement. In this paper, we propose a novel framework based on generative adversarial network (GAN) combined with numerical model to predict sea subsurface temperature. First, a GAN-based model is used to learn the simplified physics between the surface temperature and the target subsurface temperature in numerical model. Then, observation data are used to calibrate the GAN-based model parameters to obtain better prediction. We evaluate the proposed framework by predicting daily sea subsurface temperature in the South China sea. Extensive experiments demonstrate the effectiveness of the proposed framework compared to existing state-of-the-art methods.

A review of artificial intelligence in marine science

“…The inputs and outputs of the PBML model are first determined by the main variables in the physicsbased wave model, and then machine learning algorithms are used to train and perform multi-step ahead forecasts. A framework based on a combination of GAN and physical numerical models for predicting SST is also used to train the neural network model using the physical-based numerical model and then calibrate the model parameters using observed data (Meng et al, 2021b).…”

Fusion Classification of HSI and MSI Using a Spatial-Spectral Vision Transformer for Wetland Biodiversity Estimation

“…Therefore, joining the complementary merits of multisource data further improves the accuracy of land-cover classification [19]. In the past decade, extensive classification techniques have been successfully applied to multisource data [20][21][22]. Some of the machine learning methods rely on support vector machine (SVM), extreme learning machine (ELM) and random forest (RF) [23][24][25].…”

Deriving Sea Subsurface Temperature Fields from Satellite Remote Sensing Data Using a Generative Adversarial Network Model

“…Lu et al (Lu et al, 2019) reported a combined pre-clustering process and a neural network approach to determine STAs by using ocean surface temperature, surface height, surface wind observations and gridded monthly Argo data. Meng et al (Meng et al, 2021) proposed a generative adversarial network (GAN)based framework combined with a numerical model to predict sea subsurface temperature. In addition, popular deep learning algorithms, such as convolutional neural networks (CNNs), long short-term memory (LSTM) neural networks, extreme gradient boosting (XGBoost) and light gradient boosting machine (LightGBM), have also been adopted to invert ocean interior thermal structures from multisource satellite observations (Han et al, 2019;Su et al, 2019;Su, Wang, et al, 2021;.…”

“…The maximum number of training epochs is set to 200, and the model with the best performance on the validation dataset is saved as the final model.Second, the trained model at each depth level was fine-tuned with satellite remote sensing data and CORA 2.0 data. Meng et al(Meng et al, 2021) used daily Argo data to fine-tune a trained model which can improve the model performance. However, our study areas are the marginal sea of China with only a few Argo buoys at the edge of the ECS.…”