Mengxue Zhang

The spatiotemporal coupling between soil moisture and precipitation is a fundamental pillar of the global hydrological cycle. With the escalating risk of severe droughts and pluvial extremes, a critical question arises: whether observed variations in soil moisture and precipitation coupling are the result of anthropogenic Forced Response (FR) or Internal Variability (IV). While recent benchmarks, such as the Forced Component Estimation Statistical Method Intercomparison Project, have advanced the estimation of forced components from observational data, a significant gap remains: how to leverage these diagnostic tools to elucidate the non-stationary and non-linear interactions across the full moisture spectrum.This study introduces a statistical attribution framework that reconciles stationary and non-stationary coupling regimes, allowing for a more robust characterization of shifting climate dynamics. We extend the analysis of direct impacts—where FR and IV drivers linearly alter coupled variables—to the assessment of indirect impacts, where drivers exert non-linear influence on mediating variables, which modulate the dynamic sensitivity and strength of the coupling mechanisms. By decoupling these pathways, we move beyond the simple attribution of trends in moisture states; instead, we identify how anthropogenic forcing and internal variability are fundamentally restructuring the feedback mechanisms of the hydrological cycle.

Unlike other deep learning (DL) models, Transformer has the ability to extract long-range dependency features from hyperspectral image (HSI) data. Masked autoencoder (MAE), which is based on Transformer architecture, employs a “mask-reconstruction" strategy for training, allowing the model to be effective for downstream tasks. However, existing MAE-based methods only apply spectral or spatial masking to HSI and reconstruct them for feature learning, which is too simplistic and insufficient for the model to learn robust features. Additionally, the issue of lacking labeled samples in HSI and the primary objective of MAE to reduce the reliance on labeled samples are often overlooked. To address these issues, we are inspired by diffusion-based representation learning and propose diffusion enhanced masked autoencoder (DEMAE) for HSI classification with few labeled samples. First, an asymmetric encoder-decoder framework is constructed as the backbone by stacking both conditional and standard Transformer blocks. Second, we devise an auxiliary task aimed at simultaneous denoising and reconstruction, facilitating heuristic feature learning from HSI data. Third, the encoder of DEMAE is isolated for training with few labeled samples. Finally, the encoder is used for classification, and a novel signal-to-noise ratio enhanced (SNR-Enhanced) loss function is introduced to regularize the model training process. The performance of DEMAE is evaluated on four benchmark datasets, demonstrating its superiority in classification accuracy and mapping capabilities on unlabeled areas compared to existing state-of-the-art methods with few labeled samples. The source code will be available online: https://github.com/ZhaohuiXue/DEMAE.

Transformer has shown great potential in extracting global features and it can achieve better classification performance with large number of training samples compared with other deep learning (DL) models. However, most of the existing Transformer-based models for hyperspectral image (HSI) classification simply use the multi-head self-attention and channel multilayer perceptron (MLP) modules, which contain many parameters to learn, resulting in poor performance in few-shot learning scenario. To overcome the above issue, a lightweight self-pooling Transformer (SPFormer) is proposed for few-shot HSI classification. First, a one-layer autoencoder based on self-supervised learning is built to reduce the dimensionality of HSI. Second, two parameter-free modules, channel shuffle for multi-head self-pooing with sparse mapping (CSSM-MHSP) and central token mixer (CTM), are proposed for mapping spectral features to higher dimensions and promoting information interaction between pixels, respectively. Third, a light-weight channel embedding is designed to extract deep spectral features. Finally, a fully connected layer is used for classification. The classification performance of the proposed method is evaluated on four benchmark datasets, showing its superiority in classification accuracy, generalization performance and model complexity compared to existing state-of-the-art methods with limited training samples.

Recently, few-shot learning (FSL) has exhibited great potentials in hyperspectral image (HSI) classification due to its promising performance under few training samples. Although existing FSL methods have achieved great success, some limitations can still be witnessed. On the one hand, current methods mainly rely on the single metric to identify, which cannot effectively represent the class distribution with few labeled samples. On the other hand, existing methods usually only use the last deep feature of feature extractor, which may lead to the under-utilization of scarce labeled samples. To overcome the above issues, a novel multistage relation network with dual-metric (DM-MRN) is proposed for few-shot HSI classification. Firstly, a sample recombination strategy is designed to increase the variety of classification tasks in training period. Secondly, an embedding module is employed to extract deep features of the input image patches. Thirdly, we propose two relation modules: image-to-class (I2C) block and image-to-image (I2I) block. I2C block is designed to compute I2C-level relation score between second-order features, and I2I block is conceived to generate I2I-level relation score between first-order features. Finally, DM-MRN is constructed by integrating one embedding module, two I2C blocks, and one I2I block. In addition, an adaptive weighting strategy is designed to fuse the obtained relation scores, and classification can be achieved by assigning each query sample to the class with the highest value of the fused relation score. Extensive experiments carried out on five popular HSI data sets demonstrate that the proposed method outperforms other traditional and advanced models under few training samples in terms of classification accuracy and generalization performance, i.e., the performance improvement in terms of OA is around 0.30%-27.98% under 10 labeled samples per class.

Deep learning (DL) exhibits commendable performance in HSI classification because of its powerful feature expression ability. Siamese neural network further improves the performance of DL models by learning similarities within-class and differences between-class from sample pairs. However, there are still some limitations in siamese neural network. On the one hand, siamese neural network usually needs large number of negative pair samples in training process, leading to computing overhead. On the other hand, current models may lack interpretability because of complex network structure. To overcome the above limitations, we propose a spectral-spatial siamese neural network with Bag-of-Features (S3BoF) for HSI classification. Firstly, we use a siamese neural network with 3D and 2D convolutions to extract the spectral-spatial features. Secondly, we introduce stop-gradient operation and prediction head structure to make the siamese neural network work without negative pair samples, thus reducing computational burden. Thirdly, a Bag-of-Features (BoF) learning module is introduced to enhance the model interpretability and feature representation. Finally, a symmetric loss and a cross entropy loss are respectively used for contrastive learning and classification. Experiments results on four common hyperspectral data sets indicated that S3BoF performs better than the other traditional and state-of-the-art deep learning HSI classification methods in terms of classification accuracy and generalization performance, with improvements in terms of OA around 1.40%-30.01%, 0.27%-8.65%, 0.37%-6.27%, 0.22%-6.64% for Indian Pines, University of Pavia, Salinas, and Yellow River Delta data sets, respectively, under 5% labeled samples per class.

Low-rank and sparse representation (LRSR) has gained popularity in hyperspectral image (HSI) classification. However, existing LRSR models usually treat HSI as a two-dimensional matrix, which may destroy the original 3D intrinsic structure of HSI. Moreover, the dictionary consisting of only training samples lacks completeness and may be suboptimal for representation. To overcome the above issues, we propose an incremental dictionary learning-driven tensor low-rank and sparse representation (TLRSR-IDL) model for HSI classification. First, we represent HSI as a third-order tensor to retain its original 3D intrinsic structure by using the TLRSR model, which also combines both sparsity and low rankness to maintain global and local data structures. Second, we design an optimal reconstruction within regularized neighborhood (ORRN) method to exploit spectral-spatial information by avoiding the interference of heterogeneous samples in the neighborhood. Finally, an incremental dictionary learning (IDL) scheme is designed to iteratively introduce augmented samples into the dictionary, and the final classification map is produced by feeding back the last round of the incremental dictionary into the TLRSR-IDL model. The main innovative contribution lies in that the proposed IDL scheme can leverage supervised and unsupervised information, which greatly enhances traditional LRSR and TLRSR models. Experimental results based on three popular hyperspectral datasets demonstrate that the proposed method outperforms other related counterparts in terms of classification accuracy and generalization performance, with OA improvements of 0.97%-16.83%, 1.25%-6.89%, and 0.85%-6.67% for Indian Pines, Pavia University, and Salinas, respectively.

Convolutional neural network (CNN) has exhibited enormous potentials in hyperspectral image (HSI) classification owing to excellent locally modeling ability. Although excellent performance of CNN-based methods has been witnessed, there still have some limitations of their internal network backbone. On the one hand, modeling long-distance context dependencies is an inborn defect, which leads to receptive field limitation and insufficient feature capture in HSI. On the other hand, CNN-based methods usually need various sample distribution to train and cannot infer dynamically, which may not capture the inherent changes of HSI data well. To overcome the abovementioned issues, we propose a novel local transformer with spatial partition restore network (SPRLT-Net) for HSI classification. First, local transformer is introduced to obtain the spatial attention weights dynamically by measuring the similarity between related pixel pairs. Second, a spatial partition restore (SPR) module is designed to split the input patch into several overlapping continuous subpatches as sequential. With the obtained attention weights at hand, the SPR module restores the sequential to the original patch. Finally, a fully connected layer is used for classification. SPRLT-Net can capture global context dependencies, and the dynamical attention weights can adapt the inherent changes of HSI spatial pixels. Experimental results based on spatially disjoint samples and randomly selected samples of five benchmark datasets demonstrate that SPRLT-Net outperforms the other state-of-the-art methods in terms of classification accuracy, generalization performance, and computational complexity.

Deep learning (DL) has exhibited huge potentials for hyperspectral image (HSI) classification due to its powerful nonlinear modeling and end-to-end optimization characteristics. Although the superior performance of DL-based methods has been witnessed, some limitations can still be found. On the one hand, existing DL frameworks usually resorted to first-order statistical features, whereas they rarely considered second-order or higher order statistical features. On the other hand, the optimization of complex hyperparameters (e.g., the layer number and convolutional kernel size) is time-consuming and a very tough task, making the designed DL framework unexplainable. To overcome these challenges, we propose a novel attention-based second-order pooling network (A-SPN). First, a first-order feature operator is designed to model the spectral–spatial information of HSI. Second, an attention-based second-order pooling (A-SOP) operator is designed to model discriminative and representative features. Finally, a fully connected layer with softmax loss is used for classification. The proposed framework can obtain second-order statistical features in an end-to-end manner. In addition, A-SPN is free of complex hyperparameters tuning, making it more explainable and easily equipped for classification tasks. Experimental results based on three common hyperspectral data sets demonstrate that A-SPN outperforms other traditional and state-of-the-art DL-based HSI classification methods in terms of generalization performance with limited training samples, classification accuracy, convergence rate, and computational complexity.