Band selection (BS) is a vital technique for improving efficiency of hyperspectral image (HSI) processing. This letter proposes a dual-view structural similarity subspace clustering model (DVS3C) for BS. Traditional low-rank subspace clustering (LRSC) methods rely solely on single-view data (e.g., original HSI), potentially leading to the loss of critical information (e.g., spatial structures) and insufficient exploitation of the multi-dimensional features of HSI for optimal BS. To do so, DVS3C constructs a spatial view alongside the spectral view, leveraging global spectral-spatial information through subspace clustering to achieve complementary advantages between views. Besides, to overcome LRSC’s limitations in capturing band local structure, DVS3C introduces a structural similarity matrix to deeply exploit intraview neighborhood relationships of bands, further reducing band redundancy. Ultimately, an adaptive dual-view fusion strategy that iteratively optimizes a consensus matrix while dynamically adjusting the contribution of each view is designed to ensure view consistency. Experimental results on four public datasets demonstrate its remarkable stability and superiority. The source code is available athttps://github.com/ydk0912/DVS3C.

Contrastive learning has recently demonstrated great potential in hyperspectral image few-shot classification. However, conventional methods mainly emphasize instance-level similarity while neglecting class structure, often leading to the separation of intra-class samples. Moreover, the lack of explicit class prototype modeling often leads to ambiguous decision boundaries. To address these issues, this paper proposes a prototype-guided supervised momentum contrastive learning (PGSMC) for hyperspectral cross-domain few-shot classification. PGSMC first designs an asymmetric augmentation module to enhance sample diversity by applying distinct augmentation strategies and encoders to the query and key views. Subsequently, a momentum queue is employed to store historical key features and their corresponding labels. This mechanism enables smooth updates of class-level momentum prototypes and mitigates the limitations imposed by mini-batch training. Finally, a momentum prototype contrastive loss is formulated to guide the model toward class-level feature representations, thereby promoting more discriminative decision boundaries. Overall, PGSMC enables query samples to contrast with more representative momentum prototypes, enhancing inter-class separability and promoting well-defined class boundaries. Extensive experiments on five hyperspectral image datasets demonstrate that PGSMC significantly outperforms existing few-shot learning methods.

Compound drought and heat events (CDHEs) have been intensifying under global climate change, posing significant threats to arid and semi-arid regions and underscoring the imperative of precise regional monitoring. However, most existing studies focus on single hazards; even in those addressing CDHEs, the compound indices employed are typically constrained by coarse spatial resolution (exceeding 0.1°), which hinders fine-scale evaluations. To address the above issue, this study proposes a novel framework for developing the High-Resolution Compound Drought and Heat Index (HCDHI), an indicator designed to finely reflect CDHEs. Specifically, three machine learning models (XGBoost, SVR, and ANN) are employed to fit station-based Standardized Precipitation Evapotranspiration Index (SPEI) and Standardized Temperature Index (STI) values, using features extracted from remote sensing and reanalysis data. The best-performing model is interpreted via the SHapley Additive exPlanations (SHAP) method for transparency and then applied to generate 1 km-resolution SPEI and STI datasets. The HCDHI is subsequently obtained by integrating the SPEI and STI datasets using the copula function. When applied to Ningxia, a typical arid and semi-arid region of China, during 2001–2022, the XGBoost model demonstrated superior performance (R²=0.70/0.72, RMSE=0.56/0.51), closely aligning with observed data. SHAP analysis identified the precipitation (27.2% of total importance) as the primary contributor to the drought monitoring model and the land surface temperature (22.1%) as the key factor to the heat monitoring models, respectively; potential evapotranspiration (PET), vapor pressure deficit (VPD), and land surface temperature (each exceeding 10%) also contributed substantially to drought–heat interactions. Monitoring results indicated that the CDHEs affect a larger area and exhibit greater severity than single hazards. The proposed high-resolution HCDHI significantly enhances the capacity for CDHEs monitoring and facilitates fine-scale assessment, particularly in data-scarce regions.

In recent years, convolutional neural networks (CNNs) have been extensively used in hyperspectral image (HSI) classification tasks and achieved desirable performance. However, CNNs expand the receptive field by stacking convolutional layers and pooling layers, but the actual receptive field is still insufficient, so it is hard to capture the global representations of HSIs. In addition, the existing CNN models used for HSI classification have low-computational efficiency, and cannot effectively fuse spectral and spatial information. To alleviate the above limitations, this article proposes a novel efficient lightweight spectral–spatial transformer (ELS2T) specially designed for HSI classification. The transformer model can model the long-distance feature dependencies in the HSI cube. First, we design a global multiscale attention module (GMAM) to effectively highlight the useful information and weaken the useless information. Second, considering the different importance of spectral and spatial information for classification tasks, an adaptive feature fusion module (AFFM) is proposed to adaptively fuse the acquired spectral and spatial information. Finally, to improve the computational efficiency, we design the lightweight separable spatial–spectral self-attention ($\text{S}^{3}\text{A}$) module to replace the multihead self-attention (MHSA) module in the transformer encoder. Experimental results on the four well-known hyperspectral datasets show that our model is superior to the other state-of-the-art deep learning (DL) methods in both computational efficiency and classification performance.

Recently, convolutional neural networks (CNNs) have been widely used in hyperspectral image (HSI) classification with appreciable performance. However, the current CNN-based HSI classification methods have limitations in exploiting the multiscale features and extracting sufficiently discriminative features, and usually adopted dimensionality reduction method such as PCA leads to some or all of the physical information of the original band may be lost. To address the above problems, in this letter, we propose an adaptive multiscale feature attention network (AMFAN) for HSI classification. First, we use a band selection algorithm to perform data dimensionality reduction, which helps maintain the original characteristics of the image. Second, different from existing multiscale feature extraction methods that give features of different scales the same degree of importance, we propose an adaptive multiscale feature residual module (AMFRM) to give multiscale features different importance. Finally, due to the input of the HSI classification model based on deep learning being the patch cube, the only available initial information is the category of the center pixel. However, the patch often contains pixels different from the center pixel category, and existing attention mechanisms do not consider the impact of such pixels on HSI classification, so we design a novel position attention module (PAM) to calculate the similarity between the center (target) pixel and surrounding pixels, and then pay more attention to the pixels with high similarity to the center pixel. Besides, we also use a spectral attention module (SAM) to obtain more discriminative spectral features. Experimental results show that the proposed AMFAN effectively improves the classification accuracy, and outperforms the state-of-the-art CNNs.

The current regional-scale process-based photosynthesis models use biome-specified values of maximum carboxylation rate at 25 °C (Vm25) in simulating ecosystem gross primary productivity (GPP). These models ignore the variations inVm25over time and space, resulting in substantial errors in regional estimates of cropland GPP. Thus, to resolve this problem, we used the ensemble Kalman filter (EnKF) to assimilate tower-based GPP from five maize flux sites into a process-based mode to obtain the “apparent” value ofVm25and then modeled this parameter using machine learning (ML) algorithms. The results showed thatVm25increased during the early growing season and then decreased after reaching a peak value in the middle of the growing season. The coefficient of determination (R2) (root mean square error, RMSE) for SCOPES-Crop with EnKF-derived variedVm25in simulating daily GPP across all site-days increased (decreased) by 0.17 (5.63 μmol m-2s-1) on average compared to that for the model with fixedVm25. We used four ML algorithms, namely artificial neural network, random forest, extreme gradient enhancement, and convolutional neural network (CNN), to model theVm25of maize. The CNN algorithm yielded the best results. The average of theR2(RMSE) values of simulated GPP using CNN-basedVm25over the three flux sites is 0.93 (1.95 μmol m-2s-1), higher (smaller) than that using fixedVm25. This study implies that representing the seasonal variations inVm25can facilitate improved estimates of GPP and the ML methods are useful tools for modeling the variation inVm25.

The tropical cyclone (TC) convective structure evolution during rapid intensification (RI) process were explored by using the differenced infrared and water vapor imagery (IRWV) from Himwari-8 satellite in the western North Pacific. The radial and azimuthal profiles and the morphological features were extracted from 1-hour interval images and several key patterns and the rules considering the location, shapes, and magnitude of the IRWV were identified through the whole RI process. The results show that the development and maintenance of a good spin structure of the negative IRWV were crucial for TC intensification. The RI-onset periods were normally connected with the sudden change of IRWV and the inward movement to the inner-core area. The pinhole eye features were normally a sign of continue RI, while the sudden decrease of IRWV area or the appearance of symmetric central overcast features indicates the ending of RI process.

Soil moisture content is one of the most important variables for monitoring and assessing the drought condition. In this letter, a modified perpendicular drought index (MPDI) derived from the Moderate Resolution Imaging Spectroradiometer satellite data was validated using in situ soil moisture measurements in Henan province of North China. The soil moisture at different depths of a soil layer and time lag on usefulness of the MPDI for estimating soil moisture content was analyzed; the effectiveness of the MPDI was evaluated with the observed soil moisture content and the comprehensive drought index K for different soil types. The results showed that the MPDI was significantly correlated with soil moisture content for the top soil layer with 10 cm depth. For a time lag ranging from 0 to 4 days, the correlation of MPDI to soil moisture was significant with no time lag in the depth of top 10 cm (r = -0.867). For the stations of the same soil texture type of loam, the correlation coefficient between MPDI and soil moisture is higher than that of all soil texture types. In a regional scale, the MPDI reflected an obvious spatial pattern of drought under different growing stages in the drought years over the study area.

In this letter, an improved algorithm for aerosol retrieval is presented by employing the non-Lambertian forward model (forward model) (NL_FM) in the Moderate Resolution Imaging Spectroradiometer (MODIS) dark target (DT) algorithm to reduce the uncertainties induced when using the Lambertian FM (L_FM). This new algorithm was applied to MODIS measurements of the whole year of 2008 over Eastern China. By comparing the results with that of AERONET, we found that the accuracy of the aerosol optical depth (AOD) retrieval was improved with the regression plots concentrating around the 1 : 1 line and two-thirds falling within the expected error (EE) envelope EE = ±0.05±0.1τ (from 53.6% with L_FM to 68.7% with NL_FM at band 0.55 μm). Surprisingly, more accurate retrieval of the AOD demonstrated significantly improved the Ångstrom exponent (AE) retrieval, which is related to particle size parameters. The regression plots tended to concentrate around the 1 : 1 line, and many more fell within the EE = ±0.4 from 53.6% with L_FM to 80.9% with NL_FM. These results demonstrate that including the NL_FM in the MODIS DT algorithm has the potential to significantly improve both AOD and AE retrievals with respect to AERONET in comparison to the L_FM used in the current MODIS operational retrievals.

Drought is the first place in all the natural disasters in the world. It is especially serious in North China Plain. In this paper, different soil water content control levels at winter wheat growth stages are performed on Gucheng Ecological Meteorological Integrated Observation Experiment Station of CAMS, China. Some canopy parameters, including growth conditions, dry weight, physiological parameters and hyperspectral reflectance, are measured from erecting stage to milk stage for winter wheat in 2009. The relationship between canopy parameters and soil relative moisture, canopy water content and water indices of winter wheat are established. The results show that some parameters, such as SPAD and dry weight of leaves, decrease with the increasing of soil relative moisture, while other parameters, including dry weight of caudexes, above ground dry weight, height, photosynthesis rate, intercellular CO2concentration, stomatal conductance and transpiration rate, increase corresponding to the soil relative moisture. Obvious linear relationship between stomatal conductance and transpiration rate is established with 45 samples, which R reaches to 0.6152. Finally, the fitting equations between canopy water content and water indices are regressed with b5, b6 and b7 of MODIS bands. The equations are best with b7 and worst with b5. So the fitting equations with b7 can be used to inverse the canopy water content of winter wheat using MODIS or other remote sensing images with similar bands range to MODIS in Hebei Province.

Qinghai-Tibetan Plateau plays an important role in estimating net primary productivity of grassland ecosystem for global carbon cycling research. In this paper, boreal ecosystem productivity simulator (BEPS) model was modified according to the characteristics of grassland canopy. A hypothesis of horizontal homogeneity and vertical layer was put forward for grassland canopy and BEPS was modified to GEPS (grassland ecosystem productivity simulator) to simulate the NPP of grassland ecosystem. With MODIS products (MOD15A2 and MOD12Q1) and routine meteorological data, net primary productivity of grassland ecosystem was simulated in Qinghai-Tibetan Plateau in 2006 based on GEPS. The result shows that NPP of grassland ecosystem in Qinghai-Tibetan Plateau is between 20 and 500 gC/m2·a, which is close to the other studies. The spatial distribution of NPP of grassland ecosystem in Qinghai-Tibetan Plateau has the trend of decreasing from east to west. Finally the seasonal change of NPP was investigated based on monthly NPP, which has good coherence with the seasonal changes of temperature. This study suggests that the process model - GEPS - is suitable to simulate NPP of grassland ecosystem in Qinghai-Tibetan Plateau.

The accurate prediction of crop yield is of great help for grain policy making as the importance of food in human life. By assuming a homogeneous and vertical laminar structure and introducing a multilayer-two-big-leaf model, we developed a radiative transfer equation for winter wheat canopy and a model named RS-P-YEC (Remote Sensing — Photosynthesis — Yield Estimation for Crop) for winter wheat yield estimation. In this model, we converted the net primary productivity to winter wheat yield using harvest index. In this study, we estimated yield of winter wheat in North China Plain using the RS-P-YEC model. The simulated yield agrees well with observations from agro-meteorological stations and the R2reaches to 0.817. This study demonstrates that RS-P-YEC model is useful in the yield estimation of winter wheat in North China Plain with widely available remotely sensed images.