Spatio-temporal Aggregation Research Articles

STANet: A Novel Spatio-Temporal Aggregation Network for Depression Classification with Small and Unbalanced FMRI Data

STANet: A Novel Spatio-Temporal Aggregation Network for Depression Classification with Small and Unbalanced FMRI Data

MSA-GCN: Multistage Spatio-Temporal Aggregation Graph Convolutional Networks for Traffic Flow Prediction

Timely and accurate traffic flow prediction is crucial for stabilizing road conditions, reducing environmental pollution, and mitigating economic losses. While current graph convolution methods have achieved certain results, they do not fully leverage the true advantages of graph convolution. There is still room for improvement in simultaneously addressing multi-graph convolution, optimizing graphs, and simulating road conditions. Based on this, this paper proposes MSA-GCN: Multistage Spatio-Temporal Aggregation Graph Convolutional Networks for Traffic Flow Prediction. This method overcomes the aforementioned issues by dividing the process into different stages and achieves promising prediction results. In the first stage, we construct a latent similarity adjacency matrix and address the randomness interference features in similarity features through two optimizations using the proposed ConvGRU Attention Layer (CGAL module) and the Causal Similarity Capture Module (CSC module), which includes Granger causality tests. In the second stage, we mine the potential correlation between roads using the Correlation Completion Module (CC module) to create a global correlation adjacency matrix as a complement for potential correlations. In the third stage, we utilize the proposed Auto-LRU autoencoder to pre-train various weather features, encoding them into the model’s prediction process to enhance its ability to simulate the real world and improve interpretability. Finally, in the fourth stage, we fuse these features and use a Bidirectional Gated Recurrent Unit (BiGRU) to model time dependencies, outputting the prediction results through a linear layer. Our model demonstrates a performance improvement of 29.33%, 27.03%, and 23.07% on three real-world datasets (PEMSD8, LOSLOOP, and SZAREA) compared to advanced baseline methods, and various ablation experiments validate the effectiveness of each stage and module.

Spatio-temporal dynamics and recovery of commuting activities via bike-sharing around COVID-19: A case study of New York

The COVID-19 has led to significant changes in urban travel behaviors, with commuting being one of the most affected travel modes. Commuting cycling by bike-sharing systems (BSS) is regarded as a new transportation mode that is low-carbon and low-cost. However, its dynamic changes and spatiotemporal characteristics in different periods of COVID-19 still lack exploration. Therefore, this study adopts machine learning methods to identify commuter bike-sharing activities and develops a combined analysis method to analyze commuting cycling data via temporal, spatial, and spatiotemporal aggregation. Finally, we select the bike-sharing data in New York City from periods before, during, and after COVID-19 to conduct experiments. It has been found that commuting cycling experienced a “decrease-rebound” trend at the macroscopic level under the pandemic impact. However, at the micro level, urban mobility driven by this travel mode failed to fully recover, as evidenced by significant changes in spatial and temporal mobility patterns. The findings shall not only help traffic operators and managers discover the BSS commuting patterns but also reveal the pandemic impact on the travel behavior of urban residents, promoting the development of intelligent services for urban emergency management and traffic management.

Spatiotemporal interactions between jaguars (Panthera onca) and their potential prey in Amazonian islands

AbstractAlthough large carnivores usually prefer large prey, in some situations, they may shift their predation patterns towards smaller but abundant prey. The jaguar (Panthera onca) is a large carnivore capable of changing its diet according to prey and habitat availability. Here, we assessed the temporal and spatiotemporal interactions between jaguars and their prey in the Maracá‐Jipioca Islands (Northeastern Amazon, Brazil) through camera traps. We assessed overlapping activity patterns and tested for spatiotemporal segregation/avoidance between jaguars and nine potential prey species. We used a time‐to‐encounter approach, which consists in calculating the minimum time between prey and jaguar's detections, and vice versa, for each record of preys' species at a specific camera trap station, which translates into aggregation or avoidance behaviors. We found that these insular jaguars are more active in daylight periods when most of their prey are active and in locations used by species that cannot become nocturnal to avoid predators due to morphology constraints. Four prey species (great egret, white‐tailed deer, muscovy duck, and black‐and‐white tegu) presented moderate activity overlapping with jaguars. Agoutis and white‐tailed deer seek to spatiotemporally segregate from jaguars, although jaguars did not show spatiotemporal aggregation with any of the evaluated prey. Understanding the spatiotemporal dynamics is essential to establish the islands' trophic network composition and structure. This is fundamental information to efficiently allocate efforts for reducing costs and maximizing benefits in managing this population aiming to protect and conserve it, and consequently, the related ecosystems.

MAFormer: A cross-channel spatio-temporal feature aggregation method for human action recognition

Human action recognition has been widely used in fields such as human–computer interaction and virtual reality. Despite significant progress, existing approaches still struggle with effectively integrating hierarchical information and processing data beyond a certain frame count. To address these challenges, we introduce the Multi-AxisFormer (MAFormer) model, which is organized in terms of spatial, temporal, and channel dimensions of the action sequence, thereby enhancing the model’s understanding of correlations and intricate structures among and within features. Drawing on the Transformer architecture, we propose the Cross-channel Spatio-temporal Aggregation (CSA) structure for more refined feature extraction and the Multi-Axis Attention (MAA) module for more comprehensive feature aggregation. Moreover, the integration of Rotary Position Embedding (RoPE) boosts the model’s extrapolation and generalization abilities. MAFormer surpasses the known state-of-the-art on multiple skeleton-based action recognition benchmarks with the accuracy of 93.2% on NTU RGB+D 60 cross-subject split, 89.9% on NTU RGB+D 120 cross-subject split, and 97.2% on N-UCLA, offering a novel paradigm for hierarchical modeling in human action recognition.

Multimodal spatiotemporal aggregation for point cloud accumulation

Point cloud accumulation is a crucial technique in point cloud analysis, facilitating various downstream tasks like surface reconstruction. Current methods merely rely on raw LiDAR points, yielding unsatisfactory performance due to the limited geometric information, particularly in complex scenarios characterized by intricate motions, diverse objects, and an increased number of frames. In this paper, we introduce camera modality data, which is usually acquired alongside LiDAR data at minimal expense. To this end, we present the Multimodal Spatiotemporal Aggregation solution (termed MSA) to thoroughly explore and aggregate these two distinct modalities (sparse 3D points and multi-view 2D images). Concretely, we propose a multimodal spatial aggregation module to bridge the data gap between different modalities in the Bird’s-Eye-View (BEV) space and further fuse them by learnable adaptive channel-wise weights. By assembling their respective strengths, this module generates a reliable and consistent scene representation. Subsequently, we design a temporal aggregation module to capture continuous motion information across consecutive sequences, which is beneficial for identifying the motion state of the foreground scene and enabling the model to extend robustly to longer sequences. Experiments demonstrate MSA outperforms state-of-the-art (SoTA) point cloud accumulation methods across all evaluation metrics in the public benchmark, especially with more frames.

Ecological resilience in water-land transition zones: A case study of the Dongting Lake region, China

Water-land transition zones (WLTZs) are quintessential, complex human-land systems where aquatic and terrestrial environments intermingle, playing a pivotal role in bolstering global ecological resilience (ER). Human terrestrial production activities have encroached upon ecologically sensitive aquatic areas, weakening the inherent ER patterns in WLTZs. To address these issues, we developed an ER assessment process chain across the disturbance-carrying-recovery (D-C-R) dimensions. Applied it to the Dongting Lake region, this model diagnosed the spatiotemporal variations in ER from 2000 to 2020 and revealed the stress mechanisms of various driving factors. The results indicate: (1) The ER spatial pattern in the Dongting Lake region featured a radiating enhancement structure originating from a low-value core in the central plain’s lake network area, displaying distinct water-land gradient variations. Areas with declining ER levels were mainly distributed around the main body of the lake, water system runoff areas, and near governmental centers. (2) The spatiotemporal differentiation effects of ER manifested as an east–west axial development trend, tending toward polarization. The spatiotemporal aggregation effect appeared as L-L clustering and cold spots in the north, gradually expanding toward the peripheral forest areas. (3) Factors related to land use were the dominant drivers of spatiotemporal changes in ER. Simultaneously, the nonlinear interaction intensity among these driving factors showed a significant annual increase. This research framework can be extended to assess ER in similar regions, potentially offering policy recommendations for the ecological protection and sustainable management of WLTZs.

Epidemiological Characteristics of Varicella in Anhui Province, China, 2012-2021: Surveillance Study.

Varicella is a mild, self-limited disease caused by varicella-zoster virus (VZV) infection. Recently, the disease burden of varicella has been gradually increasing in China; however, the epidemiological characteristics of varicella have not been reported for Anhui Province. The aim of this study was to analyze the epidemiology of varicella in Anhui from 2012 to 2021, which can provide a basis for the future study and formulation of varicella prevention and control policies in the province. Surveillance data were used to characterize the epidemiology of varicella in Anhui from 2012 to 2021 in terms of population, time, and space. Spatial autocorrelation of varicella was explored using the Moran index (Moran I). The Kulldorff space-time scan statistic was used to analyze the spatiotemporal aggregation of varicella. A total of 276,115 cases of varicella were reported from 2012 to 2021 in Anhui, with an average annual incidence of 44.8 per 100,000, and the highest incidence was 81.2 per 100,000 in 2019. The male-to-female ratio of cases was approximately 1.26, which has been gradually decreasing in recent years. The population aged 5-14 years comprised the high-incidence group, although the incidence in the population 30 years and older has gradually increased. Students accounted for the majority of cases, and the proportion of cases in both home-reared children (aged 0-7 years who are not sent to nurseries, daycare centers, or school) and kindergarten children (aged 3-6 years) has changed slightly in recent years. There were two peaks of varicella incidence annually, except for 2020, and the incidence was typically higher in the winter peak than in summer. The incidence of varicella in southern Anhui was higher than that in northern Anhui. The average annual incidence at the county level ranged from 6.61 to 152.14 per 100,000, and the varicella epidemics in 2018-2021 were relatively severe. The spatial and temporal distribution of varicella in Anhui was not random, with a positive spatial autocorrelation found at the county level (Moran I=0.412). There were 11 districts or counties with high-high clusters, mainly distributed in the south of Anhui, and 3 districts or counties with high-low or low-high clusters. Space-time scan analysis identified five possible clusters of areas, and the most likely cluster was distributed in the southeastern region of Anhui. This study comprehensively describes the epidemiology and changing trend of varicella in Anhui from 2012 to 2021. In the future, preventive and control measures should be strengthened for the key populations and regions of varicella.

BusEye: A multi-stream approach for driver behavior analysis on public bus driver cameras

In the contemporary era, modern public buses are equipped with multiple built-in cameras, offering real-time or recorded monitoring capabilities through centralized systems. This research endeavors to introduce a novel approach for classifying driver behaviors at the signalized road intersections with traffic lights by leveraging a public bus vehicular camera that encompasses the road and thus traffic lights. The proposed method involves a 2-step multi-head self-attention network that employs the extracted traffic light state, and optic flow information besides the raw vehicular camera video. Subsequently, a bi-directional long short-term memory based spatio-temporal aggregation strategy is applied to guide the learning final discriminative representation. To evaluate the proposed approach, we introduce a benchmark dataset called BusEye, which covers the challenging real-life scenarios that are captured on actual vehicular cameras mounted on public busses. We provide extensive analysis and ablation studies on the BusEye dataset that shows our proposed solution outperforms the state-of-the-art video classification methods both on a per-class basis and overall. Based on the outcomes derived from the validation and test data, the proposed method demonstrated remarkable efficacy in classifying driver behaviors at signalized intersections.

A Study on the Spatiotemporal Aggregation and Corridor Distribution Characteristics of Cultural Heritage: The Case of Fuzhou, China

Fuzhou is an important city on the southeast coast of China with rich cultural heritage resources, and it is a national historical and cultural city in China. The main purposes of this study are the following: 1. To study the current situation and spatial distribution of cultural heritage; 2. To study the construction of cultural heritage corridors; 3. To utilize the holistic characteristics of heritage corridors to integrate the cultural heritage elements along the corridors and achieve the optimal allocation of resources. This study collects data from 605 cultural heritage sites in Fuzhou City and establishes a GIS database. The spatial distribution characteristics of cultural heritage in Fuzhou City are studied using aggregation analysis, nearest neighbor analysis, and buffer analysis in GIS technology. The following results were obtained: there is a cultural heritage corridor in Fuzhou City with the Minjiang River serving as the main land and water transportation line, and the closer one is to the water transportation line, the more intensive the distribution of cultural heritage points; and there are three cultural heritage node areas and one special cultural heritage node area. The study of Fuzhou’s heritage corridor system has important reference significance for the protection of cultural heritage in coastal historical cities in China and even in East Asia and has exploratory value for the holistic protection of cultural heritage and for the planning and development of tourism.

Spatio-temporal dynamics and distributional trend analysis of African swine fever outbreaks (2020-2021) in North-East India.

African swine fever (ASF) is a highly contagious, notifiable, and fatal hemorrhagic viral disease affecting domestic and wild pigs. The disease was reported for the first time in India during 2020, resulted in serious outbreaks and economic loss in North-Eastern(NE) parts, since 47% of the Indian pig population is distributed in the NEregion. The present study focused on analyzing the spatial autocorrelation, spatio-temporal patterns, and directional trend of the disease in NE India during 2020-2021. The ASF outbreak data (2020-2021) were collected from the offices of the Department of Animal Husbandry and Veterinary Services in seven NE states of India to identify thepotential clusters, spatio-temporal aggregation, temporal distribution, disease spread, densitymaps, and risk zones. Between 2020 and 2021, a total of 321 ASF outbreaks were recorded, resulting in 59,377 deaths. The spatial pattern analysis of the outbreak data (2020-2021) revealed that ASF outbreaks were clustered in 2020 (z score = 2.20, p < .01) and 2021 (z score = 4.89, p < .01). Spatial autocorrelation and Moran's I value (0.05-0.06 in 2020 and 2021) revealed the spatial clustering and spatial relationship between the outbreaks. The hotspot analysis identified districts of Arunachal Pradesh, Assam and districts of Mizoram, Tripura as significant hotspots in 2020 and 2021, respectively. The spatial-scan statistics with a purely spatial and purely temporal analysis revealed six and one significant clusters, respectively. Retrospective unadjusted, temporal, and spatially adjusted space-time analysis detected five, five, and two statistically significant (p < .01) clusters, respectively. The directional trend analysis identified the direction of disease distribution as northeast-southwest (2020) and north-south (2021),indicatethe possibility of ASF introduction to India from China. The high-risk zones and spatio-temporal pattern of ASF outbreaks identified in the present study can be used as a guide for deploying proper prevention, optimizing resource allocation and diseasecontrol measures in NE Indian states.

An Intelligent Framework for Resilience Recovery of FANETs with Spatio-Temporal Aggregation and Multi-Head Attention Mechanism

An Intelligent Framework for Resilience Recovery of FANETs with Spatio-Temporal Aggregation and Multi-Head Attention Mechanism

Epidemiological characteristics and spatio-temporal aggregation of severe fever with thrombocytopenia syndrome in Jinan City, China, 2018-2022.

Severe fever with thrombocytopenia syndrome (SFTS) has become a significant public health issue in Jinan City. However, the analysis of epidemiological characteristics and spatio-temporal clustering of SFTS in Jinan has not been studied yet. SFTS data from 2018-2022 in Jinan City were obtained from the China Information System for Disease Control and Prevention. Global spatial autocorrelation and local spatial autocorrelation analyses were performed using ArcGIS 10.2 software, and spatiotemporal hotspot area detection was carried out using SatScan 9.6 software. Between 2018 and 2022, 680 SFTS cases were reported in Jinan City, resulting in 53 deaths and an average case fatality rate of 7.8%. 99.0% of cases occurred between April and October, 91.9% individuals were over 50 years old, and 87.79% were primarily farmers. A positive spatial correlation of SFTS in Jinan was observed (Moran's I value between 0.135-0.197, P<0.001), indicating spatial aggregation, primarily in Licheng, Zhangqiu, Laiwu, and Gangcheng districts in southeastern Jinan. Spatiotemporal scanning detected one class I and two class II aggregation areas, with the class I aggregation area (RR = 5.66, LLR = 192.547, P<0.001) locating in southeastern Jinan City, comprising 31 towns/streets, and an aggregation time from 13 May 2020 to 13 October 2022. Spatial and temporal aggregation of SFTS is evident in Jinan. Based on the spatial and temporal distribution and epidemiological characteristics, prevention and control measures such as public education, monitoring, and training should target key populations in high-incidence epidemic areas.

Histogram cube: towards lightweight interactive spatiotemporal aggregation of big earth observation data

ABSTRACTIn the era of Earth Observation (EO) big data, interactive spatiotemporal aggregation analysis is a critical tool for exploring geographic patterns. However, existing methods are inefficient and complex. Their interactive performance greatly depends on large-scale computing resources, especially data cube infrastructure. In this study, from a green computing perspective, we propose a lightweight data cube model based on the preaggregation concept, in which the frequency histogram of EO data is employed as a specific measure. The cube space was divided into lattice pyramids by the Google S2 grid system, and histogram statistics of the EO data were injected into in-memory cuboids. Therefore, exploratory aggregation analysis of EO datasets could be rapidly converted into multidimensional-view query processes. We implemented the prototype system on a local PC and conducted a case study of global vegetation index aggregation. The experiments showed that the proposed model is smaller, faster and consumes less energy than ArcGIS Pro and XCube, and facilitates green computing strategies involving a cube infrastructure. Due to the standalone mode, larger dataset will result in longer cube building time with indexing latency. The efficiency of the approach comes at the expense of accuracy, and the inherent uncertainties were examined in this paper.

Evaluating Driving Safety of Road Alignment Conditions by Predicted Driver’s Psychophysiological Workload Using Vehicle Maneuvering Data

Evaluating driving safety according to road alignment conditions is a fundamental for safe road design and traffic operations to prevent traffic crashes. A significant intrinsic factor affecting driving safety is the driver’s psychophysiological workload generated in response to various external environmental changes during road driving. Electroencephalogram (EEG) is a well-known measure to evaluate the driver workload resulting from the response to the internal and external stimuli experienced while driving vehicles. However, the use of EEG in research of the transportation engineering domain requires not only high-cost experiments but also great efforts in data analysis. To address these issues, this study developed a promising method to predict the psychophysiological workload affecting driving safety using vehicle maneuvering data. A total of 30 drivers participated in driving simulation experiments using a high-fidelity motion-based driving simulator to obtain both psychophysiological workload data and vehicle maneuvering data to be used for developing the methodology. The proposed methodology consists of two parts: the prediction of the level of psychophysiological workload using a K-nearest neighbor (KNN) algorithm and the derivation of the alertness risk index (ARI) based on the spatiotemporal aggregation of the prediction results. Additionally, evaluating driving safety by different road conditions is conducted based on the proposed methodology.

An Attentive Spatio-Temporal Learning-Based Network for Cardiovascular Disease Diagnosis

Automated diagnosis of cardiovascular diseases (CVDs) has become an imperative need for remote or in-hospital heart monitoring. This is a challenging task because of the tenuous morphological variation of the electrocardiogram (ECG) signal across different cardiac diseases. Existing works have attempted to learn the diagnostic representation by capturing the lead-specific morphological variation of a multilead ECG signal. In this work, we have developed an attentive spatio-temporal learning network (ASTLNet) that can learn better diagnostic representation by exploiting the concurrent spatio-temporal variation of a multilead ECG signal. The ASTLNet consists of two modules, i.e., spatio-temporal representation learning (STRL) module and attentive spatio-temporal aggregation (ASTA) module. The STRL module is designed to learn the multiscale spatio-temporal representation, and the ASTA module is designed to aggregate the learned representation. Experiments on the three publicly available datasets, i.e., PTB, PTB-XL, and CPSC-2018, demonstrate that the proposed model can effectively learn the spatio-temporal variation of the ECG signal and gives superior performance compared to the state-of-the-art methods.

Temporal partitioning and the potential for avoidance behaviour within South African carnivore communities.

Carnivora occupy many ecological niches fundamental to ecosystem functioning. Within this diverse order, carnivore species compete to establish dominance, ensure survival and maintain fitness. Subordinate carnivores must, therefore, adapt their behaviour to coexist with dominant species. One such strategy is the partitioning of temporal activity patterns. We aim to determine interspecific avoidance patterns among sympatric carnivores by examining coexistence along a temporal axis. We compared the temporal activity patterns of 13 carnivore species using multi-seasonal camera trapping data from four protected areas across South Africa: Associated Private Nature Reserves, Madikwe Game Reserve, Mountain Zebra National Park and Tswalu Kalahari Reserve. Interspecific coefficients of overlap in diel and core activity periods were calculated over the study period and during the wet and dry seasons. Furthermore, interspecific spatiotemporal behaviour was examined using time-to-event analyses. Our results showed that complete avoidance of diel activity patterns was rare among South African carnivore species. Most species were predominantly nocturnal and, therefore, diel activity overlap was high, whereas core activity overlap was significantly lower (p < .001). Diel activity overlap was significantly lower during the dry than wet seasons (p = .045). Lastly, evidence of spatiotemporal aggregation revolved around scavenging species. We show the importance of seasonality in the temporal avoidance behaviours of South African carnivores while highlighting the need for fine-scaled behavioural analyses. Overall, we show that the daily activity patterns of most subordinate South African carnivore species are not influenced by top-down forces in the form of competitional suppression and risk exerted by dominant species. If avoidance is required, it is more likely to manifest as fine-scaled avoidance of core activity periods. We suggest that the focus on core activity periods might be a more suitable tool for interspecific temporal partitioning research.

Aircraft and road traffic noise, insulin resistance, and diabetes: The role of neighborhood socioeconomic status in San Diego County

Evidence linking traffic noise to insulin resistance and diabetes is limited and unanswered questions remain regarding the potential effect modification by neighborhood socioeconomic status (nSES). We aimed to assess socioeconomic inequalities in noise exposure, whether road and aircraft noise exposures were associated with insulin resistance or diabetes, and whether nSES modified these relationships. Among the Community of Mine Study in San Diego County, road and aircraft noise exposure at enrollment was calculated based on the static (participant's administrative boundary, and circular buffer around participant homes), and dynamic (mobility data by global positioning system, GPS) spatio-temporal aggregation methods. Associations of noise with insulin resistance (HOMA-IR) or type 2 diabetes (T2DM) were quantified using generalized estimating equation models adjusted for sex, age, ethnicity, individual income, and air pollution (nitrogen dioxide) exposure. Additive interaction between noise and nSES was assessed. Among 573 participants (mean age 58.7 y), participants living in low nSES were exposed to higher levels of aircraft and road noise using noise level at the census tract, circular buffer, or Kernel Density Estimation (KDE) of GPS data. Participants exposed to road noise greater or equal to the median (53 dB(A)) at the census tract and living in low nSES had an increased level of insulin resistance (β = 0.15, 95%CI: −0.04, 0.34) and higher odds of T2DM (Odds Ratio = 2.34, 95%CI: 1.12, 4.90). A positive additive interaction was found as participants living in low nSES had higher odds of T2DM. The impact of noise exposure on insulin resistance and T2DM differs substantially by nSES. Public health benefits of reducing exposure to road or aircraft noise would be larger in individuals living in low nSES.

Preferential aggregation areas of Seriola dumerili: Estimation of environmental influences and insights towards sustainable exploitation

A key aspect of the sustainable management of commercially important coastal pelagic fish stocks is the identification of the spatio-temporal dynamics of aggregation behaviour. An exemplary case species is the greater amberjack, Seriola dumerili. This top-predator is an important commercial resource exploited by recreational and commercial fisheries worldwide. Greater amberjack is known to form aggregations during important stages of its life-history, making it highly vulnerable to overfishing with potential consequences to its conservation. To improve the limited knowledge on the spatio-temporal aggregation areas of this important resource, this study combined fishing fleet dynamics (Vessel Monitoring System records), environmental and landing data into a species distribution model to predict such vulnerable events in space and time around the Balearic Islands. Our results indicated that S. dumerili aggregations were generally encountered in coastal areas at shallow depths (<100 m), low salinities, low chlorophyll concentrations, and weak meridional and zonal current velocities around the Balearic Islands. The preference for these environmental variables and processes is probably related to searching for an optimal scenario that maximises the success of larval development and survival of its offspring. Moreover, the distribution model suggested the possibility of a longer spawning period than previously predicted. This research demonstrates the use of a perceptive statistical tool that efficiently combines fisheries-dependent data and environmental data to highlight essential factors that influence the aggregation behaviour of S. dumerili and thus provides essential information for the conservation and sustainable exploitation not only of the greater amberjack, but also for other gregarious species highly-vulnerable to fishing by their aggregation behaviour.

Spatiotemporal Analysis of Forest Fires in China from 2012 to 2021 Based on Visible Infrared Imaging Radiometer Suite (VIIRS) Active Fires

Forest fire regimes are changing as a function of increasing global weather extremes, socioeconomic development, and land use change. It is appropriate to use long-term time series satellite observations to better understand forest fire regimes. However, many studies that have analyzed the spatiotemporal characteristics of forest fires based on fire frequency have been inadequate. In this study, a set of metrics was derived from the VIIRS active fire data in China, from 2012 to 2021, through spatial extraction, spatiotemporal clustering, and spread reconstruction to obtain the frequency of forest fire spots (FFS), the frequency of forest fire events (FFE), the frequency of large forest fire events (LFFE), duration, burned area, and spread rate; these metrics were compared to explore the characteristics of forest fires at different spatiotemporal scales. The experimental results include 72.41 × 104 forest fire spots, 7728 forest fire events, 1118 large forest fire events, and a burned area of 58.4 × 104 ha. Forest fires present a significant spatiotemporal aggregation, with the most FFS and FFE in the Southern Region and the most severe LFFE and burned area in the Southwest Region. The FFS, FFE, and LFFE show a general decreasing trend on an annual scale, with occasional minor rebounds. However, the burned area had substantial rebounds in 2020. The high incidence of forest fires was concentrated from March to May. Additionally, 74.7% of the forest fire events had a duration of less than 5 days, while 25.3% of the forest fire events lasted more than 5 days. This helps us to understand the characteristics of more serious or higher risk forest fires. This study can provide more perspectives for exploring the characteristics of forest fires, and more data underpinning for forest fire prevention and management. This will contribute towards reasonable forest protection policies and a sustainable environment.