Search

Abstract Due to the powerful ability in capturing the global information, transformer has become an alternative architecture of CNNs for hyperspectral image classification. However, general transformer mainly considers the global spectral information while ignores the multiscale spatial information of the hyperspectral image. In this paper, we propose a multiscale spectral–spatial convolutional transformer (MultiFormer) for hyperspectral image classification. First, the developed method utilizes multiscale spatial patches as tokens to formulate the spatial transformer and generates multiscale spatial representation of each band in each pixel. Second, the spatial representation of all the bands in a given pixel are utilized as tokens to formulate the spectral transformer and generate the multiscale spectral–spatial representation of each pixel. Besides, a modified spectral–spatial CAF module is constructed in the MultiFormer to fuse cross‐layer spectral and spatial information. Therefore, the proposed MultiFormer can capture the multiscale spectral–spatial information and provide better performance than most of other architectures for hyperspectral image classification. Experiments are conducted over commonly used real‐world datasets and the comparison results show the superiority of the proposed method.

Based on the meteorological observational data over Qinghai-Xizang （Tibetan） Plateau and atmospheric reanalysis dataset during 1961 to 2020， the snowfall variation characteristics over the Qinghai-Xizang （Tibetan） Plateau and its related corresponding circulation system were analyzed in this paper.Main conclusions are drawn as following： the spatial distribution of snowfall shows a uniformity with less in the northwest and more in the southeast.The characteristics of interannual and interdecadal variability are very evident.The interannual variability of the snowfall over Qinghai-Xizang Plateau was strong.The first mode of Qinghai-Xizang （Tibetan） Plateau snowfall anomalies is regional uniformity.In terms of the key circulation systems that affect Qinghai-Xizang （Tibetan） Plateau snowfall， when the snowfall over Qinghai-Xizang （Tibetan） Tibetan Plateau is more， the upper troposphere corresponds to the positive phase of the southern Eurasian （SEA） teleconnection， characterized as positive anomalies over the southwestern Europe， the Arabian Sea， and the northeast Asia， and negative anomalies over the Middle East and the Qinghai-Xizang （Tibetan） Plateau， while the Middle East jet is stronger； the positive （negative） phase of North Atlantic Oscillation （NAO） is conducive to more （less） snowfall over Tibetan Plateau， via modulating the SEA teleconnection and key circulation systems such as the Middle East jet stream.

Abstract North China experienced an extreme precipitation event from July 29 to August 1, 2023 (i.e., the “23.7” event) causing severe floods, significant infrastructure damage and multiple fatalities. To enhance comprehension of the mechanism behind the extreme precipitation of the “23.7” event, water vapor transport paths and sources were determined, and water vapor contribution of each source was quantitatively evaluated based on Lagrangian methods. Results showed that the extreme precipitation of the “23.7” event was closely related to large‐scale water vapor transport and convergence from low‐latitude oceans. There were five main water vapor sources which corresponded to five transport pathways. Path 1 was derived from tropical West Pacific, containing the most trajectories (195), carrying the most water vapor (69.3%) and contributing the most to the extreme precipitation of the “23.7” event (45.7%). Path 2 was guided by the cross‐equatorial flow through South China Sea, contributing to 10.1% of the precipitation. Path 3 originating from eastern tropical Indian Ocean and Path 4 from the west source near the Caspian Sea contributed less to the precipitation. Last but not the least, water vapor evaporation from eastern China contributed more than 30% to the extreme precipitation, making this region another important water vapor source.

Abstract Unlike prior researches focusing on interannual or interdecadal changes of the Northwest Pacific subtropical high (NWPSH), this paper emphasizes the interdecadal changes in the relationship between May‐June and July‐August NWPSH. The correlation coefficients between them are 0.66 (passing the 99.9% confidence level) and 0.12 during 1979–2002 and 2003–2023. Therefore, the May‐June NWPSH has a strong connection with July‐August East Asian precipitation during 1979–2002, but not during 2003–2023. The interdecadal variations in the relationship of NWPSH are due to interdecadal changes of ENSO. During 1979–2002, the eastern‐type ENSO in prior winter, having longer durations, can influence May‐June and July‐August NWPSH and the North Indian Ocean. However, the central‐type ENSO in prior winter only persists it's influence until May‐June, and changes into a Pacific Dipole pattern in July‐August during 2003–2023. Despite the North Indian Ocean continually warms during 2003–2023, the tropical Pacific Dipole offsets its impacts on the NWPSH.

Abstract Background The relationship between the NHHR and kidney stone risk remains unknown. The purpose of this study was to evaluate the association between adult NHHR and kidney stone occurrence in USA. Methods This study used a variety of statistical techniques such as threshold effects, subgroup analysis, smooth curve fitting, multivariate logistic regression, and data from the National Health and Nutrition Examination Survey (NHANES) from 2007 to 2014. We aimed to clarify the relationship between the NHHR and kidney stone risk. Results The average age of the 21,058 individuals in this research was 49.70 ± 17.64 years. The mean NHHR was 3.00 ± 1.47, and the overall prevalence of kidney stone occurrence was 9.05%. The prevalence within the quartile ranges (Q1–Q4) was 7.01%, 8.71%, 9.98%, and 10.49%, respectively. The overall average recurrence rate of kidney stones was 3.05%, demonstrating a significant increase with increasing NHHR (Q1: 1.92%, Q2: 2.92%, Q3: 3.35%, Q4: 4.00%, P

Abstract In China, traditional medications for osteoporosis have significant side effects, low compliance, and high costs, making it urgent to explore new treatment options. Probiotics have demonstrated superiority in the treatment of various chronic diseases, and the reduction of bone mass in postmenopausal osteoporosis (PMOP) is closely related to the degradation and metabolism of intestinal probiotics. It is crucial to explore the role and molecular mechanisms of probiotics in alleviating PMOP through their metabolites, as well as their therapeutic effects. We aim to identify key probiotics and their metabolites that affect bone loss in PMOP through 16srDNA sequencing combined with non-targeted metabolomics sequencing, and explore the impact and possible mechanisms of key probiotics and their metabolites on the progression of PMOP in the context of osteoporosis caused by estrogen deficiency. The sequencing results showed a significant decrease in Lactobacillus acidophilus and butyrate in PMOP patients. In vivo experiments confirmed that the intervention of L. acidophilus and butyrate significantly inhibited osteoclast formation and bone resorption activity, improved intestinal barrier permeability, suppressed B cells, and the production of RANKL on B cells, effectively reduced systemic bone loss induced by oophorectomy, with butyric acid levels regulated by L. acidophilus. Consistently, in vitro experiments have confirmed that butyrate can directly inhibit the formation of osteoclasts and bone resorption activity. The above research results indicate that there are various pathways through which L. acidophilus inhibits osteoclast formation and bone resorption activity through butyrate. Intervention with L. acidophilus may be a safe and promising treatment strategy for osteoclast related bone diseases, such as PMOP.

Deep neural networks (DNNs) have attained remarkable success in aerial detection tasks, yet they remain susceptible to adversarial samples, posing a significant challenge for their practical applications. While numerous transferable attacks have been proposed, they frequently overlook the essential balance between attack effectiveness and the feasibility of physical implementation. In this article, we concentrate our efforts on adversarial attacks against aerial detection, crafting transferable adversarial patches that can be implemented in the physical world. To this end, we introduce localized pattern corruptions, such as light spots and shadows, around the target during the training phase. These corruptions could pull the image distributions closer to the decision boundaries of the surrogate model, thereby enhancing the transferability of patches. In addition, we avoid directly optimizing adversarial patterns with traditional gradient-based techniques. Instead, we opt to update the weights of a specialized generator, which employs multilayer perceptrons (MLPs) as its core component for mapping purposes. To verify the effectiveness of our method, we conduct experiments across both the digital and physical domains. The results reveal that our approach outperforms state-of-the-art methods in terms of attack performance.

Abstract Thermal layout optimization problems are common in integrated circuit design, where a large number of electronic components are placed on the layout, and a low temperature (i.e., high efficiency) is achieved by optimizing the positions of the electronic components. The operating temperature value of the layout is obtained by measuring the temperature field from the expensive simulation. Based on this, the thermal layout optimization problem can be viewed as an expensive combinatorial optimization problem. In order to reduce the evaluation cost, surrogate models have been widely used to replace the expensive simulations in the optimization process. However, facing the discrete decision space in thermal layout problems, generic surrogate models have large prediction errors, leading to a wrong guidance of the optimization direction. In this work, the layout scheme and its temperature field are represented by images whose relation can be well approximated by a deep neural network. Therefore, we propose an online deep surrogate-assisted optimization algorithm for thermal layout optimization. First, the iterative local search is developed to explore the discrete decision space to generate new layout schemes. Then, we design a deep neural network to build an image-to-image mapping model between the layout and the temperature field as the approximated evaluation. The operating temperature of the layout can be measured by the temperature field predicted by the mapping model. Finally, a segmented fusion model management strategy is proposed to online updates the parameters of the network. The experimental results on three kinds of layout datasets demonstrate the effectiveness of our proposed algorithm, especially when the required computational budget is limited.

The objective of this study was to optimize the extraction process of total flavonoids from Haloragis micrantha (Thunb.) R. Brown. and explore its antioxidant activity. With the yield of total flavonoids as an evaluation index, single factor tests were used to investigate the effects of ethanol volume fraction, solid-liquid ratio, ultrasonic temperature and extraction time on the total flavonoids content of Haloragis micrantha (Thunb.) R. Brown., and the Box-Behnken method was used to optimize the extraction process. The scavenging ability and total reducing power of total flavonoids of Haloragis micrantha (Thunb.) R. Brown. to DPPH and ABTS+ free radicals were detected. The results showed that optimum extraction process was ethanol volume fraction of 76.15%, liquid to material ratio of 1:39.66 g/mL, ultrasonic time of 41.14 min. The yield of total flavonoids from Haloragis micrantha (Thunb.) R. Brown. was 28.61±0.05 mg/g. The total flavonoids of Haloragis micrantha (Thunb.) R. Brown. could effectively scavenge DPPH and ABTS+ free radicals. Within a certain range, the scavenging ability increased with the increasing of mass concentration, and the total flavonoids of Haloragis micrantha (Thunb.) R. Brown. had high reducing power and better antioxidant activity. The total flavonoids of Haloragis micrantha (Thunb.) R. Brown. had the strongest ability to scavenge ABTS+ free radicals. When the concentration of total flavonoids was 3 mg/mL, the clearance rate of ABTS+ were 98.28%, which was equivalent to 3 mg/mL vitamin C (clearance rate was 99.81%). This process would be simple and feasible, and the extracted total flavonoids have good antioxidant activity.

Abstract Recovering the global accurate complex physics field from limited sensors is critical to the measurement and control of the engineering system. General reconstruction methods for recovering the field, especially the deep learning with more parameters and better representational ability, usually require large amounts of labeled data which is unaffordable in practice. To solve the problem, this paper proposes uncertainty guided ensemble self-training (UGE-ST), using plentiful unlabeled data to improve reconstruction performance and reduce the required labeled data. A novel self-training framework with the ensemble teacher and pre-training student designed to improve the accuracy of the pseudo-label and remedy the impact of noise is first proposed. On the other hand, uncertainty guided learning is proposed to encourage the model to focus on the highly confident regions of pseudo-labels and mitigate the effects of wrong pseudo-labeling in self-training, improving the performance of the reconstruction model. Experiments including the airfoil velocity and pressure field reconstruction and the electronic components’ temperature field reconstruction indicate that our UGE-ST can save up to 90% of the data with the same accuracy as supervised learning.

Abstract Deep Neural Networks have been successfully applied in hyperspectral image classification. However, most of prior works adopt general deep architectures while ignore the intrinsic structure of the hyperspectral image, such as the physical noise generation. This would make these deep models unable to generate discriminative features and provide impressive classification performance. To leverage such intrinsic information, this work develops a novel deep learning framework with the noise inclined module and denoise framework for hyperspectral image classification. First, the spectral signature of hyperspectral image is modeled with the physical noise model to describe the high intra‐class variance of each class and great overlapping between different classes in the image. Then, a noise inclined module is developed to capture the physical noise within each object and a denoise framework is then followed to remove such noise from the object. Finally, the CNN with noise inclined module and the denoise framework is developed to obtain discriminative features and provides good classification performance of hyperspectral image. Experiments are conducted over two commonly used real‐world datasets and the experimental results show the effectiveness of the proposed method. The implementation of the proposed method and other compared methods could be accessed at https://github.com/shendu‐sw/noise‐physical‐framework.

Objective To identify potential risk factors for cerebrospinal fluid (CSF) leakage after craniovertebral junction (CVJ) anomaly surgery and to provide a reference for clinical practice. Methods Sixty-six patients who underwent elective CVJ anomaly surgery during a 6-year period (April 2013 to September 2019) were retrospectively included. Research data were collected from the patients’ medical records and imaging systems. Patients were divided into CSF leak and no CSF leak groups. Univariate tests were performed to identify potential risk factors. For statistically significant variables in the univariate tests, a logistic regression test was used to identify independent risk factors for CSF leakage. Results The overall prevalence of CSF leakage was 13.64%. Univariate tests showed that a basion-dental interval (BDI) > 10 mm and occipitalized atlas had significant intergroup differences (p < 0.05). Multivariate analysis indicated that a BDI > 10 mm was an independent risk factor for CSF leakage, and patients with CVJ anomalies with a BDI > 10 mm were more likely to have postoperative CSF leaks (odds ratio, 14.67; 95% confidence interval, 1.48–30.88; p = 0.004). Conclusion It is necessary to maintain vigilance during CVJ anomaly surgery in patients with a preoperative BDI > 10 mm to avoid postoperative CSF leaks.

Abstract Compared to interannual and interdecadal variations, inadequate attention has been given to intraseasonal changes in the summer northwest Pacific subtropical high (NWPSH). During 1981–2005, the north Indian Ocean (NIO) and NWPSH exhibit significant relationships from June to August, while the influences of the Boreal Summer Intraseasonal Oscillation 1 (BSISO1) in June and August differ from its impacts in July, resulting in a significant correlation between the NWPSH in June and August. The models capture NIO's roles but underestimate BSISO1's influences, leading to an underestimation (overestimation) of the relationship between NWPSH in June and August (adjacent two months). A dynamical‐statistical approach based on the close relationship between June and August NWPSH is proposed to improve models' prediction capabilities. Improved results show the mean correlation coefficients (root mean square errors) between prediction and observation for August precipitation and geopotential height increase by 57% and 21% (decrease by 0.27 mm/day and 1.61 m).

Abstract Convolutional neural networks (CNNs) have successfully demonstrated their powerful predictive performance in a variety of tasks. However, it remains a challenge to estimate the uncertainty of these predictions simply and accurately. Deep Ensemble is widely considered the state-of-the-art method which can estimate the uncertainty accurately, but it is expensive to train and test. MC-Dropout is another popular method that is less costly but lacks the diversity of predictions resulting in less accurate uncertainty estimates. To combine the benefits of both, we introduce a ReLU-Based Uncertainty Estimation (RBUE) method. Instead of using the randomness of the Dropout module during the test phase (MC-Dropout) or using the randomness of the initial weights of CNNs (Deep Ensemble), RBUE uses the randomness of activation function to obtain diverse outputs in the testing phase to estimate uncertainty. Under the method, we propose strategy MC-DropReLU and develop strategy MC-RReLU. The uniform distribution of the activation function’s position in CNNs allows the randomness to be well transferred to the output results and gives a more diverse output, thus improving the accuracy of the uncertainty estimation. Moreover, our method is simple to implement and does not need to modify the existing model. We experimentally validate the RBUE on three widely used datasets, CIFAR10, CIFAR100, and TinyImageNet. The experiments demonstrate that our method has competitive performance but is more favorable in training time.

In order to evaluate the effect of RRE on reducing blood lipid and protecting liver, a model of nonalcoholic fatty liver disease (NAFLD) mice induced by high-fat diet was used, and then the body weight, liver weight and liver coefficient of each group were calculated, and the TC、TG、HDL-C、LDL-C、AST、ALT、AKP were checked by sample test. The liver tissue was stained with HE and oil Red O for pathological observation and fat particles observation; Animal experiment results showed that RRE intervention could effectively improve the process of liver pathological changes, alleviate inflammatory injury and reduce the accumulation of liver fat particles in NAFLD mice; both high and medium doses of RRE could significantly reduce the levels of TC, TG and LDL-C in serum of NAFLD mice (P0.05); Through network pharmacology, the intersection of Rhinacanthus nasutus and NAFLD disease-related targets was carried out, and the effective active substances and action mechanism of RRE were predicted combined with molecular docking technology, which showed that rhinacanthin-C had good binding potential with the intersection targets. The experimental results showed that the RRE could reduce blood lipid, improved liver lipid deposition and inflammatory injury in NAFLD mice, so as to protect the liver. Its mechanism may be that the active substance rhinacanthin-C can improve liver lipid deposition and metabolic disorder through AMPK-SREBP1c-SIRT1 signal pathway. The experimental results can provide a scientific basis for the development of Zhuang medicine Rhinacanthus nasutus as a national drug for reducing blood lipid and improving fatty liver.

Background: Spinal cord injury (SCI) has a high disability rate and mortality rate. Recently, LncRNA XIST has been found to be involved in the regulation of inflammatory responses. Therefore, we aimed to investigate the role of XIST in the occurrence and development of SCI and the specific regulation mechanism. Methods: 100 ng/mL lipopolysaccharide (LPS) was used to treat mouse microglia BV2 cells. Hitting spinal cord was performed to C57BL/6 mice for establishing SCI model. Real-time reverse transcriptase-polymerase chain reaction (RT-qPCR), Western blot, Immunofluorescence (IF) and Enzyme linked immunosorbent assay (ELISA) experiments were used to explore the function of XIST, miR-124–3p and IRF1 in LPS-induced BV2 cells. RT-qPCR, Nissl staining, IF, Western blot and ELISA experiment were performed to study the function of XIST in SCI mice. Dual-luciferase reporter assay, RNA immunoprecipitation (RIP), RT-qPCR and Western blot assays were utilized to identify the interaction among XIST, miR-124–3p and IRF1. Results: XIST was upregulated in LPS-induced BV2 cells and spinal cord tissues of SCI mice. Overexpression of XIST promoted the M1 microphages polarization and cytokines concentration in LPS-stimulated BV2 cells, aggravated SCI of mice. Downregulated XIST promoted M1-to-M2 conversion of microglial and relieved the injury of SCI mice. Mechanism verification indicated that XIST acted as a molecular sponge of miR-124–3p and regulated IRF1 expression. Increased miR-124–3p or reduced IRF1 inhibited M1 polarization of microglial and decreased the production of inflammatory cytokines in LPS-induced BV2 cells. Increased XIST or decreased miR-124–3p had an opposite of on LPS-induced BV2 cells. Conclusion: Overexpression of XIST enhanced M1 polarization of microglia and promoted the level of inflammatory cytokines through sponging miR-124–3p and regulating IRF1 expression.

Xinjiang is a typical arid and semi-arid Mountain basin system, which make the regional ecosystem extremely fragile. Studying the influence of climate on vegetation is conducive to qualitatively analyze the change trend of vegetation coverage in this region. Therefore, utilizing vegetation coverage and main meteorological elements (temperature, precipitation, relative humidity, sunshine hours) data in Xinjiang province, this paper carried out the influence of multiple meteorological elements on vegetation coverage changes, and constructed a model of the impact of multiple meteorological elements on the growing season vegetation coverage based on random forest. The model can better simulate the vegetation coverage in 2017 and 2018, with an average error of 0.027, in consequence it can well forecast whether the vegetation is high-density or low-density in this area. Correlation analysis and variable importance show that the critical meteorological factors affecting vegetation cover change are relative humidity and sunshine hours, accounting for 73% of the vegetation coverage area. The results are helpful to understand how meteorological factors affect the vegetation coverage, and then provide a theoretical reference for the construction of ecological security in Xinjiang.

In recent years, transformers have shown great potential in hyperspectral image processing and have also been gradually applied in hyperspectral target detection (HTD). Nonetheless, applying a typical transformer to HTD remains challenging. The heavy computation burden of the multi-head self-attention (MSA) in transformers limits its efficient HTD, while the limited ability to extract local spectral features can reduce the discrimination of the learned spectral features. To further explore the potential of transformers for HTD, for balance of representation ability and computational efficiency, we propose a dual-branch Fourier-mixing transformer network for hyperspectral target detection (DBFTTD). First, this work explores a dual-branch Fourier-mixing transformer network. The transformer-style network replaces the MSA sublayer in the transformer with a Fourier-mixing sublayer, which shows advantages in improving computational efficiency and learning valuable spectral information effectively for HTD. Second, this work proposes learnable filter ensembles in the Fourier domain that are inspired by ensemble learning to improve detection performance. Third, a simple but efficient dropout strategy is proposed for data augmentation. Sufficient and balanced training samples are constructed for training the dual-branch network, and training samples for balanced learning can further improve detection performance. Experiments on four data sets indicate that our proposed detector is superior to the state-of-the-art detectors.

Abstract World cultural landscape heritage (WCLH), recognized as a combination of cultural relics and natural landscapes with outstanding significance and universal value, is unique in terms of its differentiation from the deliberate human creativity of general cultural heritage and the “deartificialization” of natural heritage. To date, the management of WCLH has become increasingly standardized. However, with the prevalence of heritage resource development activities such as heritage tourism, the phenomenon of “urbanization”, “commercialization”, “artificialization” and other issues have gradually emerged in WCLH sites. Thus, the management issues of WCLH have become increasingly serious, leading to intense concerns about the unsustainable development of WCLH. Drawing inspirations from management effectiveness (ME) evaluation research of protected areas and acknowledging the uniqueness of WCLH, this study constructs a WCLH ME evaluation system consisting of four general criteria (management foundation, management system, management measures and management performance), 16 factors and 34 indicators. The evaluation system is applied and verified through an empirical study of five existing WCLH sites in China. The empirical results show that the ME of Chinese WCLH is at a “good” level. Specifically, the management of Chinese WCLH is overall impressive in indicators of management planning, heritage protection performance and economic performance but shows deficiency in indicators of protection fees, management infrastructure, management assessment, management institution, social performance, etc. Finally, this study discusses the management issues of respective heritage sites to provide suggestions and inspirations for the development, protection and management of the sites and other cultural landscapes in China and even the world.

Abstract Background Large bone defects have always been a great challenge for orthopedic surgeons. The use of a good bone substitute obtained by bone tissue engineering (BTE) may be an effective treatment method. Artificial hydroxyapatite, a commonly used bone defect filler, is the main inorganic component of bones. Because of its high brittleness, fragility, and lack of osteogenic active elements, its application is limited. Therefore, its fragility should be reduced, its osteogenic activity should be improved, and a more suitable scaffold should be constructed. Methods In this study, a microhydroxyapatite whisker (mHAw) was developed, which was doped with the essential trace active elements Mg2+ and Sr2+ through a low-temperature sintering technique. After being formulated into a slurry, a bionic porous scaffold was manufactured by extrusion molding and freeze drying, and then SiO2 was used to improve the mechanical properties of the scaffold. The hydrophilicity, pore size, surface morphology, surface roughness, mechanical properties, and release rate of the osteogenic elements of the prepared scaffold were detected and analyzed. In in vitro experiments, Sprague–Dawley (SD) rat bone marrow mesenchymal stem cells (rBMSCs) were cultured on the scaffold to evaluate cytotoxicity, cell proliferation, spreading, and osteogenic differentiation. Results Four types of scaffolds were obtained: mHAw-SiO2 (SHA), Mg-doped mHAw-SiO2 (SMHA), Sr-doped mHAw-SiO2 (SSHA), and Mg-Sr codoped mHAw-SiO2 (SMSHA). SHA was the most hydrophilic (WCA 5°), while SMHA was the least (WCA 8°); SMHA had the smallest pore size (247.40 ± 23.66 μm), while SSHA had the largest (286.20 ± 19.04 μm); SHA had the smallest Young's modulus (122.43 ± 28.79 MPa), while SSHA had the largest (188.44 ± 47.89 MPa); and SHA had the smallest compressive strength (1.72 ± 0.29 MPa), while SMHA had the largest (2.47 ± 0.25 MPa). The osteogenic active elements Si, Mg, and Sr were evenly distributed and could be sustainably released from the scaffolds. None of the scaffolds had cytotoxicity. SMSHA had the highest supporting cell proliferation and spreading rate, and its ability to promote osteogenic differentiation of rBMSCs was also the strongest. Conclusions These composite porous scaffolds not only have acceptable physical and chemical properties suitable for BTE but also have higher osteogenic bioactivity and can possibly serve as potential bone repair materials.

Generating accurate estimation of water inherent optical properties (IOPs) from hyperspectral images plays a significant role in marine exploration. Traditional methods mainly adopt bathymetric models and numerical optimization algorithms to deal with this problem. However, these methods usually tend to simplify the bathymetric models and lack the capability of describing the discrepancy between reference spectrum and estimation spectrum, resulting in a limited estimation performance. To get a more precise result, in this work, we propose a novel network based on deep learning to retrieve the IOPs. The proposed network, named as IOPs estimation network (IOPE-Net), explores a hybrid sequence structure to establish IOPs estimation module that acquires high-dimensional nonlinear features of water body spectrums for water IOPs estimation. Moreover, considering the insufficiency of labeled training samples, we design a spectrum reconstruction module combined with classical bathymetric model to train the proposed network in an unsupervised manner. Then, aiming at further promoting the estimation performance, a multicriterion loss is developed as the objective function of IOPE-Net. In particular, we construct a hierarchical multiscale sequence loss as the key component to retain the details of original spectral information. Thus, the discrepancy between different spectrums can be better described by the obtained learning model. Experimental results on both simulated and real datasets demonstrate the effectiveness and efficiency of our method in comparison with the state-of-the-art water IOPs retrieving methods.

Detecting underwater targets from hyperspectral imagery makes a profound impact on marine exploration. Available methods mainly tackle this problem by modifying the land-based detection algorithms with classical bathymetric models, which usually fail to remove the interference of background and ignore the effect of depth information, leading to a poor detection performance. To achieve a more precise result, in this work we propose a novel network based on hyperspectral unmixing (HU) methodology and bathymetric models to detect the desired underwater targets. The proposed network, called underwater target detection network (UTD-Net), first develops a novel joint anomaly detector with classical HU methods to separate out target-water mixed pixels, which is devoted to eliminate the adverse influence of background. Then, we explore a bathymetric model-based autoencoder to unmix the target-water mixed pixels for acquiring the target-associated abundance values and maps. One dimension convolutional neural network is exploited to construct the encoder part of above autoencoder for the sake of addressing spectral variability problem. Moreover, considering the physical meaningless endmembers issue, a particular spectral constraint is imposed on the objective function as a training guidance. In this way, the autoencoder would be capable of generating specific endmembers and their corresponding abundance maps. Finally, according to the physical essence of abundance maps, we figure out the detection result by fusing the outcomes of autoencoder with weight coefficients determined by abundance values. Qualitative and quantitative illustrations demonstrate the effectiveness and efficiency of UTD-Net in comparison with the state-of-the-art underwater target detection methods.

In this study, the northward-moving tropical cyclones (TCs) that influence Northeast China (NEC) in July–September are identified based on the distance between the TC center and the boundaries of NEC. Then, based on a machine learning algorithm named hierarchical agglomerative clustering, the tracks of northward-moving TCs that influence NEC are classified into the eastern-track type and the turning type. In NEC, the precipitation induced by eastern-track type TCs gradually decreases from east to west, and the precipitation induced by turning-type TCs gradually decreases from south to north. For eastern-track type TCs, negative Niño3 sea surface temperature (SST) anomalies in preceding January–March can induce cyclonic circulation anomalies and positive vorticity anomalies over the Philippine Sea during subsequent July–September, which favors the genesis of TCs. Moreover, the westerly anomaly in the subtropical western north Pacific and the strengthening of cyclonic steering flow over the Philippines jointly steer the TCs to move northward along the northerly airflow on the west of the western Pacific subtropical high, which favors the genesis of eastern-track type TCs. For turning-type TCs, the positive SST anomalies in the West Wind Drift area during preceding May–July cause positive vorticity anomalies from the northern Philippines to Taiwan from July to September. The cyclonic steering flow over the Philippines and the anticyclonic steering flow over the Sea of Japan lead the TCs to move northwestwards and then turn to northeast, which is conducive to the genesis of turning type northward-moving TCs. Finally, the results of numerical experiments have confirmed these findings.

Abstract Background Intervertebral disc degeneration (IDD) is characterized by the loss of nucleus pulposus cells (NPCs) and phenotypic abnormalities. Accumulating evidence suggests that long noncoding RNAs (lncRNAs) are involved in the pathogenesis of IDD. In this study, we aimed to investigate the functional effects of lncRNA MALAT1 on NPCs in IDD and the possible mechanism governing these effects. Results We validated the decreased expression of MALAT1 in the IDD tissues, which was associated with decreased Collagen II and Aggrecan expression. In vitro, overexpressed MALAT1 could attenuate the effect of IL-1β on NPC proliferation, apoptosis, and Aggrecan degradation. In vivo, MALAT1 overexpression attenuated the severity of disc degeneration in IDD model rats. Our molecular study further demonstrated that MALAT1 could sponge miR-503, modulate the expression of miR-503, and activate downstream MAPK signaling pathways. The effects of MALAT1 on NPCs were partially reversed/aggregated by miR-503 mimics/inhibitor treatment. Conclusion Our data suggested that the MALAT1-miR-503-MAPK pathway plays a critical role in NPCs, which may be a potential strategy for alleviating IDD.

Intensifying global climate change has a significant influence on the vegetation, which is the basis of most of Earth’s ecosystems. It is urgent to identify the critical meteorological elements of vegetation coverage changes to address the problems induced by climate change. Many studies, ranging from theoretical advances to data-driven analyses, have been devoted to investigating meteorological elements’ roles in changing vegetation coverage. However, little has been considered in the aspect of the meteorological elements’ seasonal scale in data-driven studies. Herein, taking China as an example, we collected satellite-derived vegetation coverage data from 2000 to 2020. We then analyzed the meteorological elements, on a seasonal scale, that affect the vegetation coverage change in terms of temperature, precipitation, and solar radiation. We revealed that the critical meteorological elements facilitating vegetation coverage area change differ in both time and space and gave a detailed analysis in line with such findings. Moreover, an apparent seasonal delay effect of meteorological elements on the vegetation coverage change is uncovered.

Background: Long non-coding RNA (lncRNA) hypoxia inducible factor 1α-antisense RNA 1 (HIF1A-AS1) serves critical roles in cardiovascular diseases (CVDs). Vascular endothelial cells (VECs) are vulnerable to stimuli. Our previous study revealed that knockdown of HIF1A-AS1 reduces palmitic acid-induced apoptosis and promotes the proliferation of human VECs (HUVECs); however, the underlying mechanism remains unclear. Material and Methods: Cell Counting Kit-8, flow cytometry, transwell invasion, and wound healing were applied to detect the function of HUVECs. Moreover, miRNA sequencing (miRNA-seq) and RNA sequencing (RNA-seq) were conducted to uncover its underlying mechanism. Quantitative Polymerase Chain Reaction (qPCR) was implemented to assess the accuracy of miRNA-seq. A co-expression network was generated to determine the relationship between differentially expressed miRNAs (DEmiRNAs) and differentially expressed genes (DEGs). Results: Knockdown of HIF1A-AS1 promoted the proliferation, migration, and invasion but reduced the apoptosis of HUVECs, and the overexpression of this lncRNA had the opposite effect. Numerous DEmiRNAs and DEGs were identified, which might contribute to this phenomenon. Multiple target genes of DEmiRNAs were associated with cell proliferation and apoptosis, and overlapped with DEGs identified from RNA-seq. Finally, the network manifested that lncRNA HIF1A-AS1 moderated the function of HUVECs by not only regulating the expression of some genes directly but also by influencing a few miRNAs to indirectly mediate the expression of mRNAs. Conclusions: The results suggested that HIF1A-AS1 might regulate HUVEC function by not only regulating the expression of some genes directly but also by influencing some miRNAs to indirectly mediate the expression level of mRNA.

The abnormal characteristics of extremely warm early summer (EWES) in North China under different decadal backgrounds were contrastively analyzed. Their relationships with upper- and lower-level atmospheric circulation and global sea surface temperature anomalies (SSTAs) are also discussed. Results show that temperature anomalies of EWES in North China are overall higher than normal in both cold (1961–1993) and warm (1994–2019) periods, but the anomalies of the latter are much higher than that of the former. EWES in North China is directly related to the circulation lying between 40° and 50°N in the middle troposphere, which leads to positive temperature anomalies occurring from the bottom to the upper level of the troposphere together with a high anomaly trend tilting northward. The persistent and strong Eurasian continental high (ECH) and weak Northeast China cold vortex (NECV) activity, together with the strong western Pacific subtropical high (WPSH) are major factors that directly lead to EWES in North China. ECH and WPSH are stronger and larger, and NECV are weaker and more northward in the warm period than in the cold period. In addition, the positive SSTAs in the tropical Indian Ocean and the Kuroshio area are favorable for the stronger and larger ECH and WPSH as well as the weaker and more northward NECV, causing strong anticyclonic and downward circulation system controlling North China, resulting in the extremely warm temperatures there. The joint impact of the positive tropical Indian Ocean SSTAs and the Kuroshio region SSTAs is more significant during warm than cold periods, resulting in much stronger EWES in North China during warm periods.

The interannual variation of snowfall over the Tibetan Plateau (TP) in early winter (November–December) and its related atmospheric attribution are clarified. Meanwhile, the influence of tropical sea surface temperatures (SSTs) on TP snowfall is investigated by diagnostic analyses and Community Atmosphere Model (CAM5) simulations. The leading mode of TP snowfall in early winter features a spatially uniform pattern with remarkable interannual variability. It is found that the Indian Ocean Dipole (IOD) and El Niño Southern Oscillation (ENSO) are main external forcing factors for TP snowfall. Positive IOD with positive ENSO and positive IOD with neutral ENSO cases both have remote impact on motivating Southern Eurasia (SEA) pattern, which can induce an anomalous cyclone around the TP. The corresponding anomalous ascending motion and cold air in the mid-upper troposphere provide the dynamical and thermal conditions for heavy snowfall. The low-level southwesterly winds are enhanced over the Arabian Sea and Bay of Bengal, bringing abundant water vapor into the TP for excessive snowfall. Furthermore, CAM5 simulation experiments forced by IOD- and ENSO-related SST anomalies are performed to verify their combined and independent effects on TP snowfall in early winter. It is confirmed that either positive IOD or El Niño has certain impacts on motivating circulation anomalies favorable for snowfall over the TP. However, IOD plays a leading role in producing the excessive snowfall-related atmospheric conditions, and there is an asymmetric influence of ENSO and IOD on the TP snowfall.

Cultural landscapes are joint masterpieces of man and nature with outstanding universal value. Adequate knowledge of their evolutionary process and mechanism is crucial to their development, protection, and management. However, theoretical understanding about such has been limited as existing studies tend to focus on the descriptive and interpretative analysis of the evolutionary process and pay less attention to the underlying mechanism of the process. Integrating the traditional perspective of landscape ecology in cultural landscape research and theories of path dependence and path creation in evolutionary economic geography, this paper constructs a triple-layered integrated analytical framework of cultural landscape evolution and applies the framework to empirically examine the cultural landscape evolution of Mount Lushan. To grasp an accurate and full picture of the process, field observation and historical data collection were carried out, and a combination of thematic analysis and chronological organization was conducted. The research finds that the cultural landscape evolution of Mount Lushan has experienced three stages, i.e., coexistence and mutual influence of multiple cultures, conflict and integration of Chinese and Western cultures, as well as landscape transformation, revival, and expansion. Such evolution is a non-linear, dynamic, and complex process across which the elements, functions, and patterns of landscapes were constantly constructed and reconstructed. Fundamentally, it is the result of the synergistic effect of path dependence and path creation, and is driven by the interplay of the behavior of associated actors and the change of contextuality. The findings of this study can provide some strategic references for the management practice of cultural landscape heritage sites.

It is urgent to improve the electrochemical performance of anode for supercapacitors. Herein, we successfully prepare Fe2O3/porous carbon composite materials (FPC) through hydrothermal strategies by using oily sludge waste. The hierarchical porous carbon (HPC) substrate and fine loading of Fe2O3 nanorods are all important for the electrochemical performance. The HPC substrate could not only promote the surface capacitance effect but also improve the utilization efficiency of Fe2O3 to enhance the pseudo-capacitance. The smaller and uniform Fe2O3 loading is also beneficial to optimize the pore structure of the electrode and enlarge the interface for faradaic reactions. The as-prepared FPC shows a high specific capacitance of 465 F g−1 at 0.5 A g−1, good rate capability of 66.5% retention at 20 A g−1, and long cycling stability of 88.4% retention at 5 A g−1 after 4000 cycles. In addition, an asymmetric supercapacitor device (ASC) constructed with FPC as the anode and MnO2/porous carbon composite (MPC) as the cathode shows an excellent power density of 72.3 W h kg−1 at the corresponding power density of 500 W kg−1 with long-term cycling stability. Owing to the outstanding electrochemical characteristics and cycling performance, the associated materials’ design concept from oily sludge waste has large potential in energy storage applications and environmental protection.

Snow cover over the Tibetan Plateau plays a vital role in the regional and global climate system because it affects not only the climate but also the hydrological cycle and ecosystem. However, high-quality snow data are hindered due to the sparsity of observation networks and complex terrain in the region. In this study, a nonlinear time series analysis method called phase space reconstruction was used to obtain the Tibetan Plateau snow depth by combining the FY-3C satellite data and in situ data for the period 2014–2017. This method features making a time delay reconstruction of a phase space to view the dynamics. Both of the grids and their nearby in situ snow depth time series were reconstructed with two appropriate parameters called time delay and embedding dimension. The values of the snow depth for grids were averaged over the in situ observations and retrieval of the satellite if their two parameters were the same. That implies that the two trajectories of the time series had the same evolution trend. Otherwise, the snow depth values for grids were averaged over the in situ observation. If there were no in situ sites within the grids, the retrieval of the satellite remained. The results show that the integrated Tibetan Plateau snow depth (ITPSD) had an average bias of –1.35 cm and 1.14 cm, standard deviation of the bias of 3.96 cm and 5.67 cm, and root mean square error of 4.18 cm and 5.79 cm compared with the in situ data and FY-3C satellite data, respectively. ITPSD expressed the issue that snow depth is usually overestimated in mountain regions by satellites. This is due to the introduction of more station observations using a dynamical statistical method to correct the biases in the satellite data.

Machine learning methods have achieved good performance and been widely applied in various real-world applications. They can learn the model adaptively and be better fit for special requirements of different tasks. Generally, a good machine learning system is composed of plentiful training data, a good model training process, and an accurate inference. Many factors can affect the performance of the machine learning process, among which the diversity of the machine learning process is an important one. The diversity can help each procedure to guarantee a totally good machine learning: diversity of the training data ensures that the training data can provide more discriminative information for the model, diversity of the learned model (diversity in parameters of each model or diversity among different base models) makes each parameter/model capture unique or complement information and the diversity in inference can provide multiple choices each of which corresponds to a specific plausible local optimal result. Even though diversity plays an important role in the machine learning process, there is no systematical analysis of the diversification in the machine learning system. In this paper, we systematically summarize the methods to make data diversification, model diversification, and inference diversification in the machine learning process. In addition, the typical applications where the diversity technology improved the machine learning performance have been surveyed including the remote sensing imaging tasks, machine translation, camera relocalization, image segmentation, object detection, topic modeling, and others. Finally, we discuss some challenges of the diversity technology in machine learning and point out some directions in future work. Our analysis provides a deeper understanding of the diversity technology in machine learning tasks and hence can help design and learn more effective models for real-world applications.

Recently, more and more attention has been focused on the remote sensing scenes since they contain plentiful spectral and spatial information. In order to obtain good performance for scene representation, a proper model for feature extraction and large amounts of labeled training samples are required. However, in real-world applications, it usually cannot provide enough labeled samples since labeling is always time-consuming. To overcome this problem, this work develops a novel unsupervised deep feature learning framework with iteratively refined pseudo-classes for remote sensing scene representation. First, we introduce the center points to construct the pseudo-classes and assign the pseudo labels to the training samples. Then, a pseudo-center loss is developed by decreasing the intra-class variance between the learned features of the samples and the corresponding center points to iteratively refine the pseudo classes with the training samples in the training process. Moreover, to increase the inter-class variance between different pseudo classes and further improve the performance of the unsupervised learning, this work imposes the diversity-promoting priors over the center points. Finally, the unsupervised learning framework is developed by joint learning of the diversified pseudo-center loss and pseudo-class-based softmax loss where the pseudo-class-based softmax loss is to update the convolutional neural network (CNN) with the pseudo-classes and the diversified pseudo-center loss is to iteratively refine the pseudo-classes with the features learned from the CNN. Experiments are conducted over three real-world remote sensing scene datasets to validate the effectiveness of the proposed method and the experimental results show the superiority of the method when compared with other state-of-the-art methods.

Xianglian pill (XLP) is a typical traditional Chinese herbal medicine prescription composed of Coptidis Rhizoma and Aucklandiae Radix. It has been used to treat gastrointestinal disease for centuries. In the present study, the potential mechanisms of XLP in the treatment of ulcerative colitis (UC) were predicted by integrative pharmacology-based approach. Then, the main compounds of XLP were detected by liquid chromatography-mass spectrometry (LC-MS/MS). Finally, we verified the mechanism of XLP in the treatment of UC in a dextran sulfate sodium (DSS) model. C57BL/6 mice were randomly divided into the control group, DSS group, 5-aminosalicylic acid (5-ASA) group which was used as the positive drug control, XLP low, medium, and high dose group, with 10 mice per group. Except for the control group, acute colitis model was induced in the other mice by administering 3% DSS for consecutive 7 days. Mice in 5-ASA and XLP groups were administered with 5-ASA (50 mg/kg) or XLP (0.8, 1.6, 3.2 g/kg) via oral gavage once per day respectively. Body wight and disease activity index were assay during drug intervention. On day 8, all animals in this experiment were sacrificed and colon tissues were collected for analysis after measurement of the length. The results showed that XLP alleviate DSS -induced acute colitis in mice, including inhibition the secretion of pro-inflammatory cytokines, repairing the dysfunction of intestinal epithelial barrier, enhanced autophagy, and blocked the activation of PI3K/Akt/mTOR pathway. Furthermore, inhibiting autophagy by 3-methyladenine attenuated the protective effects of XLP on colitis. The underlying mechanism may be that Xianglian pill promote autophagy by blocking the activation of PI3K/Akt/mTOR signaling pathway.

China is the world’s largest consumer of fertilizer, with fertilizer plants widely distributed throughout the country. With the removal and closing of fertilizer factories in recent years, pollutant surveys and risk assessments (human health risks) for these sites have become increasingly necessary. However, there has been little research on contaminated fertilizer factory sites. This study aimed to characterize the distribution of pollutants, assess the health risk of the site, and calculate the remediation area and volume in a typical fertilizer plant site in North China. A total of 443 samples were collected in 2019; they indicated that the study site had high concentrations of copper (Cu), ammonia-nitrogen (NH3-N), total petroleum hydrocarbons (TPH), and fluoride at maximum ratios (the ratio of the highest value of all test data for a particular pollutant to the standard value of the pollutant) of 3.30, 2.55, 19.69, and 1.10, respectively. The health risk assessment results suggested that some hazard quotients exceeded the threshold safe level (>1 established by environmental regulations). The risk control values of soil were 2000 mg/g (Cu), 826 mg/g (TPH), and 1549 mg/g (NH3-N), and the total remediation soil volume was 72860.71 m3. The results provided basic information on soil pollution control and environmental management in a contaminated fertilizer plant site.

Based on remote sensing data of vegetation coverage, observation data of basic meteorological elements, and support vector machine (SVM) method, this study develops an analysis model of meteorological elements influence on vegetation coverage (MEVC). The variations for the vegetation coverage changes are identified utilizing five meteorological elements (temperature, precipitation, relative humidity, sunshine hour, and ground temperature) in the SVM model. The performance of the SVM model is also evaluated on simulating vegetation coverage anomaly change by comparing with statistical model multiple linear regression (MLR) and partial least squares (PLS)-based models. The symbol agreement rates (SAR) of simulations produced by MLR, PLS, and SVM models are 55%, 57%, and 66%, respectively. The SVM model shows obviously better performance than PLS and MLR models in simulating meteorological elements-related interannual variation of vegetation coverage in North China. Therefore, the introduction of the intelligent analysis method in term of SVM in model development has certain advantages in studying the internal impact of meteorological elements on regional vegetation coverage. It can also be further applied to predict the future vegetation anomaly change.

In the aim to go beyond the performance tradeoffs of classic electric double-layer capacitance and pseudo-capacitance, composites made out of carbon and pseudo-capacitive materials have been a hot-spot strategy. In this paper, a nest-like MnO2 nanowire/hierarchical porous carbon (HPC) composite (MPC) was successfully fabricated by a controllable in situ chemical co-precipitation method from oily sludge waste. Due to the advantages of high surface area and fast charge transfer for HPC as well as the large pseudo-capacitance for MnO2 nanowires, the as-prepared MPC has good capacitance performance with a specific capacitance of 437.9 F g−1 at 0.5 A g−1, favorable rate capability of 79.2% retention at 20 A g−1, and long-term cycle stability of 78.5% retention after 5000 cycles at 5 A g−1. Meanwhile, an asymmetric supercapacitor (ASC) was assembled using MPC as the cathode while HPC was the anode, which exhibits a superior energy density of 58.67 W h kg−1 at the corresponding power density of 498.8 W kg−1. These extraordinary electrochemical properties highlight the prospect of our waste-derived composites electrode material to replace conventional electrode materials for a high-performance supercapacitor.

Band selection has imposed great impacts on hyperspectral image processing in recent years. Unfortunately, few existing methods are proposed for hyperspectral underwater target detection (HUTD). In this paper, a novel unsupervised band selection method is proposed for HUTD by embedding the bathymetric model into the band selection process. Considering the dependence between targets and background, a bathymetric latent spectral representation learning scheme is designed to investigate a physically meaningful subspace where the desired targets are the most distinguishable from the background. This calculated subspace is exploited as a reference to select out desired bands based on the spectral distance metric. Then, we propose an iteration-based band subset generation strategy for the sake of promoting the diversity of the band selection results and taking full advantage of the ample spectral information. Moreover, a representative band selection approach based on sparse representation is also conducted to eliminate the redundant information among adjacent bands. The band selection result is eventually achievable by connecting the representative bands of all the band subsets. Qualitative and quantitative evaluations demonstrate the effectiveness and efficiency of the proposed method in comparison with state-of-the-art band selection methods.

The WMO-IOC Joint Technical Commission for Oceanography and Marine Meteorology (JCOMM) has devised a coordination mechanism for the fit-for-purpose delivery of an end-to-end system, from ocean observations to met-ocean operational services. This paper offers a complete overview of the activities carried out by JCOMM and the status of the achievements up to 2018. The JCOMM stakeholders consist of the research and operational institutions of WMO members and the IOC member states, which mandated JCOMM to devise an international strategy to move toward the achievement of the United Nations Sustainable Development Goals. The three areas of activity are the Observation Program Area (OPA), the Data Management Program Area (DMPA) and the Services and Forecasting Services Program Area (SFSPA), and several expert teams have been established to contribute to the international coordination efforts. OPA is organized into observing networks connected by different observing technologies, DMPA organizes the overall near-real time and delayed mode data assembly, and the delivery methodology and architecture, and the SFSPA coordinates the met-ocean services resulting from the observations and data management. Future developments should enhance coordination in these three program areas by considering the inclusion of new and emergent observing technologies, the interoperability of met-ocean data assembly centers and the establishment of efficient research to operations protocols, in addition to better fit-for-purpose customized services in both the public and private sectors.

Underwater target detection (UTD) is one of the most attractive research topics in hyperspectral imagery (HSI) processing. Most of the existing methods are presented to predict the signatures of desired targets in an underwater context but ignore the depth information which is position-sensitive and contributes significantly to distinguishing the background and target pixels. So as to take full advantage of the depth information, in this paper a self-improving framework is proposed to perform joint depth estimation and underwater target detection, which exploits the depth information and detection results to alternately boost the final detection performance. However, it is difficult to calculate depth information under the interference of a water environment. To address this dilemma, the proposed framework, named self-improving underwater target detection framework (SUTDF), employs the spectral and spatial contextual information to pick out target-associated pixels as the guidance dataset for depth estimation work. Considering the incompleteness of the guidance dataset, an expectation-maximum liked updating scheme has also been developed to iteratively excavate the statistical and structural information from input HSI for further improving the diversity of the guidance dataset. During each updating epoch, the calculated depth information is used to yield a more diversified dataset for the target detection network, leading to a more accurate detection result. Meanwhile, the detection result will in turn contribute in detecting more target-associated pixels as the supplement for the guidance dataset, eventually promoting the capacity of the depth estimation network. With this specific self-improving framework, we can provide a more precise detection result for a hyperspectral UTD task. Qualitative and quantitative illustrations verify the effectiveness and efficiency of SUTDF in comparison with state-of-the-art underwater target detection methods.

Droughts and floods have frequently occurred in Southwest China (SWC) during the past several decades. Yet, the understanding of the mechanism of precipitation in SWC is still a challenge, since the East Asian monsoon and Indian monsoon potentially influence the rainfall in this region. Thus, the prediction of precipitation in SWC has become a difficult and critical topic in climatology. We develop a novel multi-variable network-based method to delineate the relations between the global sea surface temperature anomalies (SSTA) and the precipitation anomalies (PA) in SWC. Our results show that the out-degree patterns in the Pacific, Atlantic and Indian Ocean significantly influence the PA in SWC. In particular, we find that such patterns dominated by extreme precipitation change with the seasons. Furthermore, we uncover that the teleconnections between the global SSTA and rainfall can be described by the in-degree patterns, which dominated by several vital nodes within SWC. Based on the characteristics of these nodes, we find that the key SSTA areas affect the pattern of the nodes in SWC with some specific time delays that could be helpful to improve the long-term prediction of precipitation in SWC.

Deep learning methods, especially convolutional neural networks (CNNs), have shown remarkable ability for remote sensing scene classification. However, the traditional training process of standard CNNs only takes the point-wise penalization of the training samples into consideration, which usually makes the learned CNNs sub-optimal especially for remote sensing scenes with large intra-class variance and low inter-class variance. To address this problem, deep metric learning, which incorporates the metric learning into the deep model, is used to maximize the inter-class variance and minimize the intra-class variance for better representation. This work introduces structured metric learning for remote sensing scene representation, a special deep metric learning which can take full advantage of the training batch. However, the deep metrics only consider the pairwise correlation between the training samples, and ignores the classwise correlation from the class view. To take the classwise penalization into consideration, this work defines the center points of the learned features of each class in the training process to represent the class. Through increasing the variance between different center points and decreasing the variance between the learned features from each class and the corresponding center point, the representational ability can be further improved. Therefore, this work develops a novel center-based structured metric learning to take advantage of both the deep metrics and the center points. Finally, joint supervision of the cross-entropy loss and the center-based structured metric learning is developed for the land-use classification in remote sensing. It can joint learn the center points and the deep metrics to take advantage of the point-wise, the pairwise, and the classwise correlation. Experiments are conducted over three real-world remote sensing scene datasets, namely UC Merced Land-Use dataset, Brazilian Coffee Scene dataset, and Google dataset. The classification performance can achieve 97.30%, 91.24%, and 92.04% with the proposed method over the three datasets which are better than other state-of-the-art methods under the same experimental setups. The results demonstrate that the proposed method can improve the representational ability for the remote sensing scenes.