Sea surface temperature (SST) prediction has attracted increasing attention, due to its crucial role in understanding the Earth’s climate and ocean system. Existing SST prediction methods are typically based on either physics-based numerical methods or data-driven methods. Physics-based numerical methods rely on marine physics equations and have stable and explicable outputs, while data-driven methods are flexible in adapting to data and are capable of detecting unexpected patterns. We believe that these two types of method are complementary to each other, and their combination can potentially achieve better performances. In this paper, a space-time partial differential equation (PDE) is employed to form a novel physics-based deep learning framework, named the space-time PDE-guided neural network (STPDE-Net), to predict daily SST. Comprehensive experiments for SST prediction were conducted, and the results proved that our method could outperform the traditional finite-difference forecast method and several state-of-the-art deep learning and physics-guided deep learning methods.

Advances in data assimilation (DA) methods and the increasing amount of observations have continuously improved the accuracy of initial fields in numerical weather prediction during the last decades. Meanwhile, in order to effectively utilize the rapidly increasing data, Earth scientists must further improve DA methods. Recent studies have introduced machine learning (ML) methods to assist the DA process. In this paper, we explore the potential of a four‐dimensional variational (4DVar) constrained neural network (NN) method for accurate DA. Our NN is trained to approximate the solution of the variational problem, thereby avoiding the need for expensive online optimization when generating the initial fields. In the context that the full‐field system truths are unavailable, our approach embeds the system's kinetic features described by a series of analysis fields into the NN through a 4DVar‐form loss function. Numerical experiments on the Lorenz96 physical model demonstrate that our method can generate better initial fields than most traditional DA methods at a low computational cost, and is robust when assimilating observations with higher error outside of the distributions where it is trained. Furthermore, our NN‐based DA model is effective against Lorenz96 physical models with larger variable numbers. Our approach exemplifies how ML methods can be leveraged to improve both the efficiency and accuracy of DA techniques.

The evaporation duct is a special atmospheric stratification that significantly influences the propagation path of electromagnetic waves at sea, and hence, it is crucial for the stability of the radio communication systems. Affected by physical parameters that are not universal, traditional evaporation duct theoretical models often have limited accuracy and poor generalization ability, e.g., the remote sensing method is limited by the inversion algorithm. The accuracy, generalization ability and scientific interpretability of the existing pure data-driven evaporation duct height prediction models still need to be improved. To address these issues, in this paper, we use the voyage observation data and propose the physically constrained LightGBM evaporation duct height prediction model (LGB-PHY). The proposed model integrates the Babin–Young–Carton (BYC) physical model into a custom loss function. Compared with the eXtreme Gradient Boosting (XGB) model, the LGB-PHY based on a 5-day voyage data set of the South China Sea provides significant improvement where the RMSE index is reduced by 68%, while the SCC index is improved by 6.5%. We further carried out a cross-comparison experiment of regional generalization and show that in the sea area with high latitude and strong adaptability of the BYC model, the LGB-PHY model has a stronger regional generalization performance than that of the XGB model.

Accurate retrieval of ocean surface wind speed (OSWS) has a vital impact on maritime transportation planning and extreme weather forecast. Current models leveraging deep-learning (DL) techniques have demonstrated considerable potential for satellite remote-sensing wind retrieval. However, these models tend to focus on synchronizing the retrieved wind speeds with the label, neglecting the inherent absolute error (AE) embedded within the label and thus resulting in retrieval errors. To mitigate the disruptive impact of AE on retrieval accuracy, we introduce a novel network called TCNet, which retrieves observations of cyclone global navigation satellite system (CYGNSS) as OSWS. The network constructs an AE module (AEM), guided by triple collocation (TC) method for improved accuracy in real-time wind retrieval by calculating the AE as loss value. These calculations guide the network training process, thereby enhancing retrieval accuracy. Meanwhile, the wind speed dataset imbalance and inherent averaging characteristics of networks frequently result in wind speed uncertaintines in extremes. Notably, this occurs as a gross underestimation of high-speed winds. Therefore, TCNet incorporates an adaptive penalty module (APM) to solve this problem. By assigning higher penalty factors to high-speed winds, the sensitivity of network to its retrieval is improved. Experimentally, the APM in TCNet exhibited a remarkable reduction of AE in high-speed wind retrieval and mitigates the understating of high-speed scenarios while maintaining an overall error that is not significantly increased. Importantly, TCNet demonstrated notable resistance to noise and portrayed excellent generalizability, providing fresh insights into weather forecasting, climate research, and other marine applications.

Data‐driven weather prediction (DDWP) has made significant advancements in recent years. However, weather prediction using DDWPs still requires an accurate initial field as the input. To fulfill this requirement, the four‐dimensional variational (4DVar) approach can offer initial fields. Recent studies have demonstrated the potential of deep learning (DL)‐based methods in accelerating 4DVar. In this study, we propose a novel model called the 4DVar‐informed generative adversarial network (4DVarGAN), which combines prior knowledge from 4DVar with the conditional generative network (CGAN). We employ a CGAN to non‐iteratively solve the 4DVar cost function and utilize a cycle‐consistent adversarial learning framework for data augmentation. Additionally, we incorporate a 4DVar‐based adaptive adjustment to the output of the proposed model's analysis increment‐generating component, which promotes reasonable stabilization. Experimental results using 500 hPa geopotential fields from the WeatherBench data set demonstrate that our approach achieves a 73‐fold acceleration compared to the 4DVar implemented by the DDWP model. Furthermore, our model exhibits the lowest initial and forecast errors, outperforming state‐of‐the‐art DL‐based data assimilation (DA) methods. Moreover, our method demonstrates effective performance when starting from background fields of varying qualities, consistently achieving stable results. These findings highlight the potential of CGANs in enhancing the reliability of data‐driven DA by incorporating the prior knowledge of the 4DVar method.

In the field of numerical weather prediction, fine-grained precipitation fields play a crucial role in forecasting and analyzing the spatial distribution and intensity of the precipitation. Historically, it is customary to employ the interpolation technique to downscale the low-resolution initial field output by assimilation systems, aligning with the requirements of a high-resolution forecasting model. Currently, data-driven deep learning methods offer novel solutions to address this challenge. In this letter, we propose a spatial downscaling algorithm for precipitation data generated from the North American Land Data Assimilation System (NLDAS), called Gaussian-Perceptual-based Generative Adversarial Network (GP-GAN). Specifically, the GP-GAN introduces a Siamese Gaussian-Perceptual Module (SGPM) which maps the data reconstructed from the generator and ground-truth to Gaussian latent space to learn the distribution of precipitation. Moreover, the Adaptive Weighted Loss Function (AWLF) is proposed to strengthen the emphasis and understanding of extreme precipitation events. Experimental results on the RainNet dataset comprising hourly precipitation over the USA demonstrate that GP-GAN provides better performance than other generative adversarial networks and diffusion models in improving spatial resolution.