Spatiotemporal Relationships Research Articles

Structural Transformer with Region Strip Attention for Video Object Segmentation

Memory-based methods in semi-supervised video object segmentation (VOS) achieve competitive performance by performing feature similarity between the current frame and memory frames. However, this operation involves two challenges: 1) instances of occlusion caused by interaction, and 2) interference from similar objects or clutters in the background. In this work, we propose a Structural Transformer with Region Strip Attention (STRSA) approach to address these challenges. Specifically, we build a Structural Transformer (ST) architecture to decompose the feature similarity between the long-term memory frames and the current frame into two aspects: a time–space part and an object significance part. This allows us to investigate the spatio-temporal relationship of pixels and capture the salient features of the objects. Therefore, the differences between pixels and the specificity of objects are fully exploited. In addition, we leverage the object location information from the long-term memory masks and the strip pooling to design a Region Strip Attention (RSA) module, which boosts the attention for the foreground regions and suppresses the background clutters. Extensive experiments on DAVIS, YouTube-VOS, and MOSE benchmarks prove that our method achieves satisfactory results and outperforms the retrained baseline model.

Trauma in Neonatal Acute Brain Slices Alters Calcium and Network Dynamics and Causes Calpain-Mediated Cell Death.

Preparing acute brain slices produces trauma that mimics severe penetrating brain injury. In neonatal acute brain slices, the spatiotemporal characteristics of trauma-induced calcium dynamics in neurons and its effect on network activity are relatively unknown. Using multiphoton laser scanning microscopy of the somatosensory neocortex in acute neonatal mouse brain slices (P8-12), we simultaneously imaged neuronal Ca2+ dynamics (GCaMP6s) and cytotoxicity (propidium iodide or PI) to determine the relationship between cytotoxic Ca2+ loaded neurons (GCaMP-filled) and cell viability at different depths and incubation times. PI+ cells and GCaMP-filled neurons were abundant at the surface of the slices, with an exponential decrease with depth. Regions with high PI+ cells correlated with elevated neuronal and neuropil Ca2+ The number of PI+ cells and GCaMP-filled neurons increased with prolonged incubation. GCaMP-filled neurons did not participate in stimulus-evoked or seizure-evoked network activity. Significantly, the superficial tissue, with a higher degree of trauma-induced injury, showed attenuated seizure-related neuronal Ca2+ responses. Calpain inhibition prevented the increase in PI+ cells and GCaMP-filled neurons in the deep tissue and during prolonged incubation times. Isoform-specific pharmacological inhibition implicated calpain-2 as a significant contributor to trauma-induced injury in acute slices. Our results show a calpain-mediated spatiotemporal relationship between cell death and aberrant neuronal Ca2+ load in acute neonatal brain slices. Also, we demonstrate that neurons in acute brain slices exhibit altered physiology depending on the degree of trauma-induced injury. Blocking calpains may be a therapeutic option to prevent acute neuronal death during traumatic brain injury in the young brain.

The critically endangered Chinese pangolin (Manis pentadactyla) surviving in the cracks under urbanization in Shenzhen

Understanding impacts of urbanization upon local wildlife is pivotal for fostering urban sustainability and biodiversity conservation. Urbanization can directly cause habitat loss and intensify interactions between wildlife due to habitat loss. Here we focus on a threatened flagship species, Chinese pangolin (Manis pentadactyla), in the eastern forests of Shenzhen, a megacity in China. We first conducted field surveys to uncover its fresh (used within a year) and abandoned burrows (unused > a year) and surrounding habitat characteries (elevation, slope, canopy closure, vegetation cover). We then deployed camera traps near the burrows to monitor the daily activities of the Chinese pangolin and sympatric species, and anthropogenic activities. These data were finally used to predict changes in habitat distributions/uses with species distribution models, and to statistically analyze the spatio-temporal relationships between the Chinese pangolin and its potential threats. We found the suitable habitats of pangolin burrows have been halved over the recent decade. In contrast to abandoned burrows, fresh burrows were more likely to be found at higher altitudes (200–500 m) and closer to reservoirs (300–500 m). Pangolins had strong spatial avoidance to presences of humans and domestic dogs, and reduced activities outside when potential predators were most active. We suggest that mitigating human disturbances in the eastern forests is crucial for the maintenance and recovery of the pangolin population in Shenzhen. Our study is among the first to shred a light on the impact of urban expansion on the Chinese pangolins and provides habitat maps and ecological knowledge of the Chinese pangolin for the forest restorations and conservation planning in Shenzhen.

Palaeolake Jeinimeni and ice-marginal geomorphology of the central Patagonian Lago General Carrera/Buenos Aires ice-lobe (46.5°S)

Robust geomorphological interpretations of ice-marginal glacial, meltwater and palaeolake landforms are needed to constrain models of landform evolution that can be used to underpin high-resolution geochronologies of deglaciation. The Lago General Carrera/Bueno Aires (LGCBA) ice lobe of the former Patagonian Ice Sheet is one of the most researched in Patagonia, yet there remain contrasting landform evolution models partly because the spatiotemporal relationships between landforms in the LGCBA basin and tributary valleys have not been fully investigated. To this end we present a detailed, high-resolution glacial geomorphological map of the Los Antiguos sector of the LGCBA basin, and the tributary Río Jeinimeni valley, alongside sediment-landform assemblages from a broad suite of ice-marginal, subglacial, glaciolacustrine, glaciofluvial and paraglacial landforms. We introduce evidence for a previously unmapped ice-dammed lake within the Jeinimeni valley, which may have existed during the Last Glacial Maximum. Multiple isolated, ice-marginal proglacial lakes, which formed during the recession of the LGCBA ice lobe, were also mapped. The landform evidence is synthesised into a refined landscape evolution model for the LGCBA basin and Jeinimeni valley during the Last Glacial Interglacial Transition. The data presented demonstrate that previously reconstructed high LGCBA lake-level stands above 500 m asl were likely glaciolacustrine landforms created by ice-marginal lakes dammed by the LGCBA ice lobe. The mapping in the Jeinimeni valley highlights the potential for renewed dating investigations of landforms in tributary valleys to provide maximum limiting ages for the retreat of their associated glaciers. Furthermore, the identification of glaciolacustrine varved sediments associated with palaeolake Jeinimeni and the Los Antiguos sector of the LGCBA palaeolake, highlights the potential of these annually-resolved archives to infer spatiotemporal dynamics of main ice-lobe/tributary glaciers to improve palaeoenvironmental and palaeoclimatic reconstructions in the region.

Drought and its ecological risk bundle from the perspective of watershed hydrological cycle

The mechanisms underlying the impacts of climate change and vegetation dynamics on hydrological drought in humid regions are still lacking. In this study, we connected the four components of meteorology-soil-vegetation-runoff to investigate the spatio-temporal response relationship between vegetation growth and different drought types. Based on the Variable Infiltration Capacity model and the Self-organizing Map Algorithm, we proposed ecological risk bundles at the grid scale to characterize the potential impacts of different types and levels of drought on vegetation. Furthermore, we quantified the driving impact of temporal and spatial changes in vegetation coverage on the propagation of meteorological-hydrological drought. The study found that the centers of gravity for the occurrence frequencies of extreme and mild drought shifted towards regions where vegetation growth was influenced by climate change. In certain regions of the watershed, vegetation exhibits significant spatial and temporal heterogeneity in its response to stress caused by different drought types. From 2004 to 2014, the stress on vegetation caused by moderate and mild meteorological droughts weakened, while soil moisture stress intensified after 2014. Simultaneously, the impacts of climate change and vegetation growth on runoff reached 48.25 % and 35.13 % respectively, and their synergistic effects triggered changes in the risk of co-concurrent return periods for hydrological drought events. Under the 100-year design return period, the co-occurrence return period of runoff shifted from its natural state of 162.9 years to 52.8 years, and the joint return period reversed its scenario, becoming shorter than the co-occurrence return period.

Whole-Genome Sequencing of Human and Porcine Escherichia coli Isolates on a Commercial Pig Farm in South Africa.

Escherichia coli is an indicator micro-organism in One Health antibiotic resistance surveillance programs. The purpose of the study was to describe and compare E. coli isolates obtained from pigs and human contacts from a commercial farm in South Africa using conventional methods and whole-genome sequencing (WGS). Porcine E. coli isolates were proportionally more resistant phenotypically and harbored a richer diversity of antibiotic resistance genes as compared to human E. coli isolates. Different pathovars, namely ExPEC (12.43%, 21/169), ETEC (4.14%, 7/169), EPEC (2.96%, 5/169), EAEC (2.96%, 5/169) and STEC (1.18%, 2/169), were detected at low frequencies. Sequence type complex (STc) 10 was the most prevalent (85.51%, 59/169) among human and porcine isolates. Six STcs (STc10, STc86, STc168, STc206, STc278 and STc469) were shared at the human-livestock interface according to multilocus sequence typing (MLST). Core-genome MLST and hierarchical clustering (HC) showed that human and porcine isolates were overall genetically diverse, but some clustering at HC2-HC200 was observed. In conclusion, even though the isolates shared a spatiotemporal relationship, there were still differences in the virulence potential, antibiotic resistance profiles and cgMLST and HC according to the source of isolation.

Basketball technique action recognition using 3D convolutional neural networks

This research investigates the recognition of basketball techniques actions through the implementation of three-dimensional (3D) Convolutional Neural Networks (CNNs), aiming to enhance the accurate and automated identification of various actions in basketball games. Initially, basketball action sequences are extracted from publicly available basketball action datasets, followed by data preprocessing, including image sampling, data augmentation, and label processing. Subsequently, a novel action recognition model is proposed, combining 3D convolutions and Long Short-Term Memory (LSTM) networks to model temporal features and capture the spatiotemporal relationships and temporal information of actions. This facilitates the facilitating automatic learning of the spatiotemporal features associated with basketball actions. The model’s performance and robustness are further improved through the adoption of optimization algorithms, such as adaptive learning rate adjustment and regularization. The efficacy of the proposed method is verified through experiments conducted on three publicly available basketball action datasets: NTURGB + D, Basketball-Action-Dataset, and B3D Dataset. The results indicate that this approach achieves outstanding performance in basketball technique action recognition tasks across different datasets compared to two common traditional methods. Specifically, when compared to the frame difference-based method, this model exhibits a significant accuracy improvement of 15.1%. When compared to the optical flow-based method, this model demonstrates a substantial accuracy improvement of 12.4%. Moreover, this method showcases strong robustness, accurately recognizing actions under diverse lighting conditions and scenes, achieving an average accuracy of 93.1%. The research demonstrates that the method reported here effectively captures the spatiotemporal relationships of basketball actions, thereby providing reliable technical assessment tools for basketball coaches and players.

Water Flow Prediction Based on Improved Spatiotemporal Attention Mechanism of Long Short-Term Memory Network

The prediction of water plant flow should establish relationships between upstream and downstream hydrological stations, which is crucial for the early detection of flow anomalies. Long Short-Term Memory Networks (LSTMs) have been widely applied in hydrological time series forecasting. However, due to the highly nonlinear and dynamic nature of hydrological time series, as well as the intertwined coupling of data between multiple hydrological stations, the original LSTM models fail to simultaneously consider the spatiotemporal correlations among input sequences for flow prediction. To address this issue, we propose a novel flow prediction method based on the Spatiotemporal Attention LSTM (STA-LSTM) model. This model, based on an encoder–decoder architecture, integrates spatial attention mechanisms in the encoder to adaptively capture hydrological variables relevant to prediction. The decoder combines temporal attention mechanisms to better propagate gradient information and dynamically discover key encoder hidden states from all time steps within a window. Additionally, we construct an extended dataset, which preprocesses meteorological data with forward filling and rainfall encoding, and combines hydrological data from multiple neighboring pumping stations with external meteorological data to enhance the modeling capability of spatiotemporal relationships. In this paper, the actual production data of pumping stations and water plants along the East-to-West Water Diversion Project are taken as examples to verify the effectiveness of the model. Experimental results demonstrate that our STA-LSTM model can better capture spatiotemporal relationships, yielding improved prediction performance with a mean absolute error (MAE) of 3.57, a root mean square error (RMSE) of 4.61, and a mean absolute percentage error (MAPE) of 0.001. Additionally, our model achieved a 3.96% increase in R2 compared to the baseline model.

Fragilized Ecologies and Multilayer Violence in Two Oil Films About the Niger Delta

ABSTRACT: Nigeria's oil-rich Niger Delta region has continuously been conceived as a context of terror and political violence. However, this representation of the Niger Delta is deemed too schematic and misleading since it does not account for oil production's complicated local and human issues. In fact, violent mining geographies like the Niger Delta are always permeated by various forms of subtle and imperceptible violence that, if overlooked, may lead to misrepresentations. Rob Nixon has theorized such invisibilized forms of violence as "slow violence." Building on this concept of slow violence as long emergencies that are hard to harness and turn into dramatic stories, this article examines how Black November : Struggle for the Niger Delta (Jeta Amata, 2012) and Oloibiri (Curtis Graham and Samantha Iwowo, 2016), two films set in the vulnerable ecosystem of the Niger Delta oil-producing enclaves, take on the representational challenge posed by the slow violence of ecological destruction. It studies how the filmmakers conceptualize slow violence and explores their strategies to make visible subtle forms of harm that resist representation. More specifically, it analyzes how spatiotemporal relations, images, and narrative sequences are combined to witness the slow violence of oil production and ecological destruction.

Spatio-temporal variation and relationship between agricultural efficiency and irrigation intensity in a semi-arid region of India

A surging population in Karnataka State, a semi-arid region in India, poses a threat to both food security and livelihood sustainability, necessitating a concentrated effort to bolster agricultural efficiency and achieve United Naton’s Sustainable Development Goal 2 (zero hunger). Therefore, in order to address the pressing issue of food scarcity in Karnataka, this study meticulously examined the spatio-temporal variation of agricultural efficiency and irrigation intensity in Karnataka, uncovering its significant dependence of agricultural efficiency on irrigation intensity. Specifically, this study used a one-way analysis of variance (ANOVA) to ascertain significant differences in the means of agricultural efficiency and irrigation intensity during 2004–2005 and 2018–2019. This study showed that the maximum improvement in agricultural efficiency index was recorded in Belgaum (40.24), Gulbarga (24.77), and Yadgir districts (22.92) between 2004–2005 and 2018–2019, which indicated the progressing trend and better scope for agriculture extension. On the contrary, some districts expressed threat (a decline of above 20.00 of agricultural efficiency index) and needed special care for the improvement of agricultural efficiency in four northern districts (Bagalkot, Bidar, Raichur, and Bijapur), three southern districts (Chitradurga, Chikballapur and Hassan), and two southern districts (Koppal and Gadag) in Karnataka. During 2004–2005, irrigation intensity varied from 3.19% to 56.39%, with the lowest irrigation intensity in Kodagu District and the highest irrigation intensity in Shimoga District. During 2018–2019, irrigation intensity changed from 0.77% to 72.77%, with the lowest irrigation intensity in Kodagu District and the highest in Dakshin Kannad District. Moreover, the research scrutinized the complex relationship between agricultural efficiency and irrigation intensity, with the correlation coefficient increased from 0.162 during 2004–2005 to 0.255 during 2018–2019. It implies that in both periods, a low positive correlation existed between these two variables. Over time, several factors (high-yield seeds and chemical fertilizers) other than irrigation intensity gradually became essential for agricultural efficiency. This research offers a wealth of valuable insights for regional planners and policy-makers contending with comparable challenges in various regions of India and other developing countries.

Precipitation prediction based on variational mode decomposition combined with the crested porcupine optimization algorithm for long short-term memory model

Accurate precipitation prediction is very important for meteorological disaster prevention, water resources management, and agricultural decision making. To improve the accuracy of precipitation prediction, a hybrid model based on variational mode decomposition (VMD), crested porcupine optimization algorithm (CPO), and long short-term memory model (LSTM) is proposed in this paper. The model first uses VMD to decompose the precipitation time series into intrinsic mode functions of different frequencies to capture the multi-scale characteristics of precipitation data. Then, the CPO algorithm is used to optimize LSTM adaptive parameters to improve the global search ability and robustness of the model. Finally, the decomposed precipitation component is input into the LSTM network to learn the spatiotemporal dependence relationship and improve the ability of long-term prediction. The experimental results show that compared with the traditional LSTM model, CPO-LSTM model, and VMD-LSTM model, the hybrid model achieves better performance in many evaluation indices and effectively improves the accuracy of precipitation prediction. The application of the model can provide an effective tool for the fields of meteorology and water resources management, as well as provide new ideas for related research.

M2 receptors are required for spatiotemporal sequence learning in mouse primary visual cortex.

Acetylcholine is a neurotransmitter that plays a variety of roles in the central nervous system. It was previously shown that blocking muscarinic receptors with a nonselective antagonist prevents a form of experience-dependent plasticity termed "spatiotemporal sequence learning" in the mouse primary visual cortex (V1). Muscarinic signaling is a complex process involving the combined activities of five different G protein-coupled receptors, M1-M5, all of which are expressed in the murine brain but differ from each other functionally and in anatomical localization. Here we present electrophysiological evidence that M2, but not M1, receptors are required for spatiotemporal sequence learning in mouse V1. We show in male mice that M2 is highly expressed in the neuropil in V1, especially in thalamorecipient layer 4, and colocalizes with the soma in a subset of somatostatin-expressing neurons in deep layers. We also show that expression of M2 receptors is higher in the monocular region of V1 than it is in the binocular region but that the amount of experience-dependent sequence potentiation is similar in both regions and that blocking muscarinic signaling after visual stimulation does not prevent plasticity. This work establishes a new functional role for M2-type receptors in processing temporal information and demonstrates that monocular circuits are modified by experience in a manner similar to binocular circuits.NEW & NOTEWORTHY Muscarinic acetylcholine receptors are required for multiple forms of plasticity in the brain and support perceptual functions, but the precise role of the five subtypes (M1-M5) are unclear. Here we show that the M2 receptor is specifically required to encode experience-dependent representations of spatiotemporal relationships in both monocular and binocular regions of mouse V1. This work identifies a novel functional role for M2 receptors in coding temporal information into cortical circuits.

An entity alignment approach coupling NGBoost and SHAP for constructing spatio-temporal evolution knowledge graph from historical atlases

The historical atlases provide a wealth of information about the evolution of geography over time and space. The alignment of geographical entities across varying time periods is a crucial aspect of extracting meaningful insights into the spatio-temporal dynamics of geography. This paper proposes a geographic entity alignment approach coupling Natural Gradient Boosting (NGBoost) with SHapley Additive exPlanation (SHAP). Taking the historical atlas of China as a case study, a geographic entity alignment model based on NGBoost is constructed considering the different kinds of similarity features of geographic entities, including semantic, distance, shape, size and topology. The contribution of similarity features in the NGBoost model is analyzed using the SHAP framework so as to improve the explanatory capacity of the model. The spatio-temporal evolution relationships of geographic entities are generated by association rules depending on alignment types and represented as quadruples, for constructing geographic knowledge graphs. The proposed NGBoost method was found a superior accuracy by comparing with BP neural networks, random forests, and other alternative methods for aligning geographic entities. The constructed geographic spatio-temporal evolution knowledge graphs offer valuable support for the queries of evolutionary knowledge.

Research on the spatio-temporal coupling relationship between agricultural green development efficiency and food security system in China

Food security and agricultural green development are fundamental issues concerning the survival of mankind. To realize the coordinated development of them is of great significance to promote the green transformation of agriculture and ensure national food security. However, few studies have analyzed the coupling relationship between agricultural green development and food security. Therefore, this study complements the research on the coupling relationship between them. First, we use the agricultural panel data of 31 provinces in China from 2010 to 2021 to measure the agricultural green development efficiency and food security level by Super-SBM model and entropy weight method. Then, the coupling relationship and spatiotemporal evolution of the two are analyzed by coupling coordination degree model. The results show that, first of all, China's agricultural green development efficiency and food security level have improved overall, but there are regional differences. Secondly, the degree of coupling and coordination between the two is significantly improved, and most provinces develop from the break-in stage to the coordination stage, and the regional differences are relatively reduced. Finally, Beijing, Shanghai and other developed areas are mostly of the lagging food security type, Liaoning, Shandong and other major grain producing areas have reached a high degree of coordination, while Tibet, Qinghai and other western regions are still in the break-in stage. According to the development situation of different regions, corresponding suggestions are put forward. For regions with lagging food security type, the government should give full play to the application of science and technology in agriculture and promote green and low-carbon planting technologies. For regions with low coupling and coordination degree, the government should improve support policies and build a collaborative operation system, adjust planting structure to improve land utilization rate and food security level. Finally, the government should work together to build agricultural industrial parks and give full play to the leading role of competitive provinces to achieve common development.

Analyzing the impact of ocean-atmosphere teleconnections on rainfall variability in the Brazilian Legal Amazon via the Rainfall Anomaly Index (RAI)

Drought events are mainly associated with alterations in ocean-atmosphere patterns, leading to abnormally dry conditions due to reduced precipitation. This study aims to enhance the understanding of the spatiotemporal relationship of atmospheric teleconnections associated with droughts in the Brazilian Legal Amazon (BLA), based on the Rainfall Anomaly Index (RAI), utilizing 1675 precipitation grid points from the Global Precipitation Climatology Centre (GPCC) for the period 1970–2019. The results revealed that Principal Component Analysis (PCA) and cluster analysis of RAI demonstrated significant climatic variability, with 15 components explaining 68% of the total variance and identifying two homogeneous regions closely aligning with observed conditions. Atmospheric teleconnections showed substantial variation in spatial and temporal extent in extreme drought events in the years 2005, 2010, 2015–2016, and 2019 in the BLA. The findings indicate that on an interannual scale, the dominant teleconnections in the BLA are the Atlantic Multidecadal Oscillation (AMO), North Atlantic Oscillation (NAO), Tropical North Atlantic (TNA), Pacific North American (PNA) pattern, and Atlantic Meridional Overturning Circulation (AMOC). In contrast, on a decadal scale, the Pacific Decadal Oscillation (PDO), Southern Oscillation Index (SOI), Oceanic Niño Index (ONI), and Tropical South Atlantic (TSA) prevail. Our results provide valuable insights into the nature of drought events in the Amazon, including strategies for the early detection of their intensity and affected areas, demonstrating that atmospheric teleconnections significantly influence both interannual and decadal periods. These influences are concentrated on specific temporal scales and during particular years. Conclusively, this study underscores the critical role of atmospheric teleconnections in shaping drought dynamics in the BLA, offering a foundation for predictive models and mitigation strategies to combat climate-induced challenges in the region.

Historical Insights into Sustainable Development: Analyzing the Spatiotemporal Dynamics of Ancient Trade and Settlements

Ancient trade routes significantly influenced political structures, cultural exchanges, economic development, and landscape transformations across diverse regions. The Ancient Tea Horse Road, a prominent historical commercial pathway in China, is crucial for ecosystem and cultural heritage conservations. This study uses content and comparative analysis, Geographic Information System (GIS) technology, and spatial analysis to examine the spatiotemporal relationships between 13,026 settlement points identified via Google Earth and 445 recorded settlement names within the Dali Bai Autonomous Prefecture, Yunnan Province. The findings indicate that the Ancient Tea Horse Road evolved from a simple linear form to a complex radial network, reflecting its growth from local to broad regional connectivity. Settlements related to administrative, postal, and commercial functions emerged in its vicinity, underscoring the route’s impact on social structures and economic activities. The development of settlements, mainly within a 10 km radius of the route, was closely linked to its expansion, alteration, and decline, offering insights into contemporary settlement distribution patterns. This research not only enhances the understanding of the Ancient Tea Horse Road and its surrounding settlements but also supports the protection of ancient trade routes and their cultural heritage. It contributes to sustainable development by integrating ecological preservation with the safeguarding of ancient trade routes and their cultural legacies.

ADDGCN: A Novel Approach with Down-Sampling Dynamic Graph Convolution and Multi-Head Attention for Traffic Flow Forecasting

An essential component of autonomous transportation system management and decision-making is precise and real-time traffic flow forecast. Predicting future traffic conditionsis a difficult undertaking because of the intricate spatio-temporal relationships involved. Existing techniques often employ separate modules to model spatio-temporal features independently, thereby neglecting the temporally and spatially heterogeneous features among nodes. Simultaneously, many existing methods overlook the long-term relationships included in traffic data, subsequently impacting prediction accuracy. We introduce a novel method to traffic flow forecasting based on the combination of the feature-augmented down-sampling dynamic graph convolutional network and multi-head attention mechanism. Our method presents a feature augmentation mechanism to integrate traffic data features at different scales. The subsampled convolutional network enhances information interaction in spatio-temporal data, and the dynamic graph convolutional network utilizes the generated graph structure to better simulate the dynamic relationships between nodes, enhancing the model’s capacity for capturing spatial heterogeneity. Through the feature-enhanced subsampled dynamic graph convolutional network, the model can simultaneously capture spatio-temporal dependencies, and coupled with the process of multi-head temporal attention, it achieves long-term traffic flow forecasting. The findings demonstrate that the ADDGCN model demonstrates superior prediction capabilities on two real datasets (PEMS04 and PEMS08). Notably, for the PEMS04 dataset, compared to the best baseline, the performance of ADDGCN is improved by 2.46% in MAE and 2.90% in RMSE; for the PEMS08 dataset, compared to the best baseline, the ADDGCN performance is improved by 1.50% in RMSE, 3.46% in MAE, and 0.21% in MAPE, indicating our method’s superior performance.

Study on the Expansion Potential of Artificial Oases in Xinjiang by Coupling Geomorphic Features and Hierarchical Clustering

The study of the expansion potential of artificial oases based on remote sensing data is of great significance for the rational allocation of water resources and urban planning in arid areas. Based on the spatio-temporal relationship between morphogenetic landform types and the development of artificial oases in Xinjiang, this study explored the development pattern of artificial oases in the past 30 years by using trend analysis and centroid migration analysis, constructing a series of landform–artificial oasis change indices, and investigating the suitability of different landforms for the development of artificial oases based on geomorphological location by adopting a hierarchical clustering method. The following conclusions are drawn: (1) From 1990 to 2020, the area of artificial oases in the whole territory continued to increase, with significant expansion to the south from 2005 to 2010. (2) Six categories of landform types for artificial oasis development were created based on the clustering results. Of these, 7.39% and 6.15% of the area’s geomorphological types belonged to the first and second suitability classes, respectively. (3) The optimal scale for analyzing the suitability of landforms for the development of artificial oases over the past 30 years in the whole area was 8 km, which could explain more than 96% of the changes in the growth of artificial oases. The distribution of landforms of first- and second-class suitability within the 8 km buffer zone of an artificial oasis in the year 2020 was 10.55% and 9.90%, respectively, and landforms of first-class suitability were mainly concentrated in the near plain side of the urban agglomerations located on the northern and southern slopes of the Tianshan Mountains, and the urban agglomerations at the southern edge of Altai Mountains. This study quantified the potential of different geomorphological types for the development of artificial oases and provided a basis for site selection in future artificial oasis planning and urban construction.

Spatiotemporal coupling analysis of surface water and groundwater in Tibet based on multi-source sensors

Abstract. Lakes and groundwater are two crucial components of the global terrestrial water cycle, collectively forming a vital network for Earth's water resources. However, in the Tibetan region, the spatiotemporal relationship between lakes and groundwater and their impact on the local hydrological cycle remain inadequately understood. Satellite remote sensing serves as an effective observational tool, enabling comprehensive investigation and analysis of surface lakes and groundwater in Tibet with high spatial resolution. Therefore, this study integrates Landsat and Cryosat-2 satellite data to examine the spatiotemporal patterns of surface lake extents and water levels in Tibet. Additionally, combining Gravity Recovery and Climate Experiment (GRACE) satellite observations with the Global Land Data Assimilation System (GLDAS) model data, we quantitatively analyze the spatiotemporal variations in groundwater storage and its correlation with lake water. The results indicate that: 1) Rivers and lakes in Tibet are mainly located in the central and northwest regions, displaying noticeable intra-annual variations; 2) Substantial lagged relationships exist between groundwater storage and lake water levels and areas, revealing that lakes contribute significantly to groundwater replenishment, especially in the Ngari prefecture and Lhokha prefecture. This study comprehensively utilizes multi-source remote sensing data to dynamically monitor surface and groundwater in Tibet, providing robust support for a better understanding of the interaction between groundwater and surface water.

Deep knowledge distillation: A self-mutual learning framework for traffic prediction

Traffic flow prediction in spatio-temporal networks is a crucial aspect of Intelligent Transportation Systems (ITS). Existing traffic flow forecasting methods, particularly those utilizing graph neural networks, encounter limitations. When processing large-scale graph data, the depth of these models can restrict their ability to effectively capture complex relationships and patterns. Additionally, these methods often focus mainly on local neighborhood information, which can limit their capability to recognize and analyze global relationships and patterns within the graph data. Therefore, we proposed a deep knowledge distillation model, tailored to effectively capture spatio-temporal patterns in traffic flow prediction. This model incorporates a bidirectional random walk process on a directed graph, enabling it to effectively capture both spatial and temporal dependencies. Utilizing a blend of mutual learning and self-distillation, our approach enhances the detection of spatio-temporal relationships within traffic data and improves the feature perception ability at both local and global levels. We tested our model on two real-world datasets, achieving notable improvements in prediction accuracy, especially for predictions within a one-hour timeframe. In comparison to the baseline model, our proposed model achieved accuracy improvements of 0.19 and 0.18 on the respective datasets. These results highlight the success of using mutual learning and self-distillation to transfer knowledge effectively within and between models and to improve the model’s capability in identifying and extracting features.