Accurate prediction of future chlorophyll-a (Chl-a) concentrations is of great importance for effective management and early warning of marine ecological systems. However, previous studies primarily focused on chlorophyll-a inversion and reconstruction, while methods for predicting Chl-a concentrations remain limited. To address this issue, we adopted four deep learning approaches, including Convolutional LSTM Network (ConvLSTM), Convolutional Neural Network-Long Short-Term Memory (CNN-LSTM), Eidetic 3D LSTM (E3D-LSTM), and Self-Attention ConvLSTM (SA-ConvLSTM) models, to predict Chl-a over the Yellow Sea and Bohai Sea (YBS) in China. Furthermore, 14 environmental variables obtained from the remote sensing data of Moderate-resolution Imaging Spectroradiometer (MODIS) and ECMWF Reanalysis v5 (ERA5) were utilized to predict the Chl-a concentrations in the study area. The results showed that all four models performed satisfactorily in predicting Chl-a concentrations in the YBS, with SA-ConvLSTM exhibiting a closer approximation to true values. Furthermore, we analyzed the impact of the Self-Attention Memory Module (SAM) on the prediction results. Compared to the ConvLSTM model, the SA-ConvLSTM model integrated with the SAM module better captured subtle large-scale variations within the study area. The SA-ConvLSTM model exhibited the highest prediction accuracy, and the one-month Pearson correlation coefficient reached 0.887. Our study provides an available approach for anticipating Chl-a concentrations over a large area of sea.

Remote sensing techniques for shoreline extraction are crucial for monitoring changes in erosion rates, surface hydrology, and ecosystem structure. In recent years, Convolutional neural networks (CNNs) have developed as a cutting-edge deep learning technique that has been extensively used in shoreline extraction from remote sensing images, owing to their exceptional feature extraction capabilities. They are progressively replacing traditional methods in this field. However, most CNN models only focus on the features in local receptive fields, and overlook the consideration of global contextual information, which will hamper the model’s ability to perform a precise segmentation of boundaries and small objects, consequently leading to unsatisfactory segmentation results. To solve this problem, we propose a parallel semantic segmentation network (TCU-Net) combining CNN and Transformer, to extract shorelines from multispectral remote sensing images, and improve the extraction accuracy. Firstly, TCU-Net imports the Pyramid Vision Transformer V2 (PVT V2) network and ResNet, which serve as backbones for the Transformer branch and CNN branch, respectively, forming a parallel dual-encoder structure for the extraction of both global and local features. Furthermore, a feature interaction module is designed to achieve information exchange, and complementary advantages of features, between the two branches. Secondly, for the decoder part, we propose a cross-scale multi-source feature fusion module to replace the original UNet decoder block, to aggregate multi-scale semantic features more effectively. In addition, a sea–land segmentation dataset covering the Yellow Sea region (GF Dataset) is constructed through the processing of three scenes from Gaofen-6 remote sensing images. We perform a comprehensive experiment with the GF dataset to compare the proposed method with mainstream semantic segmentation models, and the results demonstrate that TCU-Net outperforms the competing models in all three evaluation indices: the PA (pixel accuracy), F1-score, and MIoU (mean intersection over union), while requiring significantly fewer parameters and computational resources compared to other models. These results indicate that the TCU-Net model proposed in this article can extract the shoreline from remote sensing images more effectively, with a shorter time, and lower computational overhead.

Soil salinization has seriously affected agricultural production and ecological balance in the Yellow River Delta region. Rapid and accurate monitoring of soil salinity has become an urgent need. Traditional machine learning models tend to fall into local optimal values during the learning process, which reduces their accuracy. This paper introduces Circle map to enhance the crayfish optimization algorithm (COA), which is then integrated with the regularized extreme learning machine (RELM) model, aiming to improve the accuracy of soil salinity content (SSC) inversion in the Yellow River Delta region. We employed Landsat5 TM remote sensing images and measured salinity data to develop spectral indices, such as the band index, salinity index, vegetation index, and comprehensive index, selecting the optimal modeling variable group through Pearson correlation analysis and variable projection importance analysis. The back propagation neural network (BPNN), RELM, and improved crayfish optimization algorithm–regularized extreme learning machine (ICOA-RELM) models were constructed using measured data and selected variable groups for SSC inversion. The results indicate that the ICOA-RELM model enhances the $R^2$ value by an average of about 0.1 compared to other models, particularly those using groups of variables filtered by variable projection importance analysis as input variables, which showed the best inversion effect (test set $R^2$ value of 0.75, MAE of 0.198, RMSE of 0.249). The SSC inversion results indicate a higher salinization degree in the coastal regions of the Yellow River Delta and a lower degree in the inland areas, with moderate saline soil and severe saline soil comprising 48.69% of the total area. These results are consistent with the actual sampling results, which verify the practicability of the model. This paper’s methods and findings introduce an innovative and practical tool for monitoring and managing salinized soils in the Yellow River Delta, offering significant theoretical and practical benefits.

In contrast to conventional remote sensing images, hyperspectral remote sensing images are characterized by a greater number of spectral bands and exceptionally high resolution. The richness of both spectral and spatial information facilitates the precise classification of various objects within the images, establishing hyperspectral imaging as indispensable for remote sensing applications. However, the labor-intensive and time-consuming process of labeling hyperspectral images results in limited labeled samples, while challenges like spectral similarity between different objects and spectral variation within the same object further complicate the development of classification algorithms. Therefore, efficiently exploiting the spatial and spectral information in hyperspectral images is crucial for accomplishing the classification task. To address these challenges, this paper presents a multi-scale feature fusion convolutional neural network (MSFF). The network introduces a dual branch spectral and spatial feature extraction module utilizing 3D depthwise separable convolution for joint spectral and spatial feature extraction, further refined by an attention-based-on-central-pixels (ACP) mechanism. Additionally, a spectral–spatial joint attention module (SSJA) is designed to interactively explore latent dependency between spectral and spatial information through the use of multilayer perceptron and global pooling operations. Finally, a feature fusion module (FF) and an adaptive multi-scale feature extraction module (AMSFE) are incorporated to enable adaptive feature fusion and comprehensive mining of feature information. Experimental results demonstrate that the proposed method performs exceptionally well on the IP, PU, and YRE datasets, delivering superior classification results compared to other methods and underscoring the potential and advantages of MSFF in hyperspectral remote sensing classification.

Band selection (BS) identifies key bands from hyperspectral imagery (HSI) for specific downstream tasks, playing a pivotal role in practical applications. eXtreme Gradient Boosting (XGBoost), an interpretable tree-based ensemble learning classifier, explicitly implements the complex nonlinear hyperspectral classification. The interpretable information extracted from the tree structure offers a novel basis for supervised BS. To this end, this article proposes a supervised BS method, named classification task-driven hyperspectral BS via interpretability from XGBoost (XGBS). It leverages prior knowledge to train a classification task-driven XGBoost and interprets the tree structure to extract multivariate interpretable information, encompassing band split gain and two types of band dependencies. Subsequently, a heuristic search framework is employed to evaluate band performance, facilitating the selection of bands with strong classification capability, high interdependency, and low redundancy. Experiments conducted on six real HSI datasets demonstrate the effectiveness and stability of the proposed XGBS.

With the continuous maturity of hyperspectral remote sensing imaging technology, it has been widely adopted by scholars to improve the performance of feature classification. However, due to the challenges in acquiring hyperspectral images and producing training samples, the limited training sample is a common problem that researchers often face. Furthermore, efficient algorithms are necessary to excavate the spatial and spectral information from these images, and then, make full use of this information with limited training samples. To solve this problem, a novel two-branch deep learning network model is proposed for extracting hyperspectral remote sensing features in this paper. In this model, one branch focuses on extracting spectral features using multi-scale convolution and a normalization-based attention module, while the other branch captures spatial features through small-scale dilation convolution and Euclidean Similarity Attention. Subsequently, pooling and layering techniques are employed to further extract abstract features after feature fusion. In the experiments conducted on two public datasets, namely, IP and UP, as well as our own labeled dataset, namely, YRE, the proposed DMAN achieves the best classification results, with overall accuracies of 96.74%, 97.4%, and 98.08%, respectively. Compared to the sub-optimal state-of-the-art methods, the overall accuracies are improved by 1.05, 0.42, and 0.51 percentage points, respectively. The advantage of this network structure is particularly evident in unbalanced sample environments. Additionally, we introduce a new strategy based on the RPNet, which utilizes a small number of principal components for feature classification after dimensionality reduction. The results demonstrate its effectiveness in uncovering compressed feature information, with an overall accuracy improvement of 0.68 percentage points. Consequently, our model helps mitigate the impact of data scarcity on model performance, thereby contributing positively to the advancement of hyperspectral remote sensing technology in practical applications.