Search Results (13)

Sign language recognition technology serves as a crucial bridge, fostering meaningful connections between deaf individuals and hearing individuals. This technological innovation plays a substantial role in promoting social inclusivity. Conventional sign language recognition methodologies that rely on static images are inadequate for capturing the dynamic characteristics and temporal information inherent in sign language. This limitation restricts their practical applicability in real-world scenarios. The proposed framework, called SSTA-ResT, integrates ResNet, soft spatiotemporal attention, and Transformer encoders to achieve this objective. The framework utilizes ResNet to extract robust spatial feature representations, employs the lightweight SSTA module for dual-path complementary representation enhancement to strengthen spatiotemporal associations, and leverages the Transformer encoder to capture long-range temporal dependencies. Experimental results on the LSA64 Argentine Sign Language (ASL) dataset demonstrate that the proposed method achieves an accuracy of 96.25%, a precision of 97.18%, and an F1 score of 0.9671. These results surpass the performance of existing methods across all metrics while maintaining a relatively low model parameter count of 11.66 M. This demonstrates the framework’s effectiveness and practicality for sign language video recognition tasks. Full article

Seeking to advance mutual clustering of the tourism economy and cultural industries while safeguarding cultural sustainability in tourism, this paper delves into the patterns of co-development and the contributing forces across spatial and temporal dimensions in the Yellow River Basin. Using a combined spatial and temporal analytical lens, along with spatial autocorrelation testing and a spatial Durbin model embedded in a synergetic systems approach, the present study analyzes the evolutionary characteristics of the spatiotemporal pattern of the collaborative agglomeration of the Yellow River Basin’s tourism and cultural industries in 2011 and 2021 and the internal mechanism of its influencing factors. We then propose countermeasures and suggestions to boost the quality–efficiency synergy agglomeration of the basin’s tourism and cultural industries. The results showed the following: ① From 2011 to 2021, a positive overall spatial autocorrelation was noted in the basin’s tourism and cultural industries. Temporally, it presented a variation trend of “rise–fall–rise”, and spatially, it presented a distribution characteristic of “higher in the central and eastern regions versus in its western parts”. ② From 2011 to 2021, the local spatial autocorrelation (LSA) of the basin’s tourism and cultural industries remained at a low level. Moreover, significant differences were noted in the LSA among different regions. In spatial terms, the clustering intensity of tourism and cultural industries was stronger in the central and eastern parts of the basin versus in its western parts. ③ Influencing variables for tourism–culture collaborative agglomeration across the basin involve both temporal superposition effects and spatial radiation driving effects. The industrial economy, policies, and innovation exert enduring effects on the development and cross-regional spillover outcomes of the two collaborative agglomerations. Serving as a theoretical reference and policy resource, this study addresses how to promote the quality–efficiency synergy in the Yellow River Basin’s tourism and cultural industries while enhancing cultural sustainability in the tourism industry. Moreover, it can also provide experiences and references for other similar regions. Full article

Being an essential issue in global climate warming, the response of urban green spaces to air pollution and climate variability because of rapid urbanization has become an increasing concern at both the local and global levels. This study explored the response of urban vegetation to air pollution and climate variability in the Bucharest metropolis in Romania from a spatiotemporal perspective during 2000–2024, with a focus on the 2020–2024 period. Through the synergy of time series in situ air pollution and climate data, and derived vegetation biophysical variables from MODIS Terra/Aqua satellite data, this study applied statistical regression, correlation, and linear trend analysis to assess linear relationships between variables and their pairwise associations. Green spaces were measured with the MODIS normalized difference vegetation index (NDVI), leaf area index (LAI), photosynthetically active radiation (FPAR), evapotranspiration (ET), and net primary production (NPP), which capture the complex characteristics of urban vegetation systems (gardens, street trees, parks, and forests), periurban forests, and agricultural areas. For both the Bucharest center (6.5 km × 6.5 km) and metropolitan (40.5 km × 40.5 km) test areas, during the five-year investigated period, this study found negative correlations of the NDVI with ground-level concentrations of particulate matter in two size fractions, PM2.5 (city center r = −0.29; p < 0.01, and metropolitan r = −0.39; p < 0.01) and PM10 (city center r = −0.58; p < 0.01, and metropolitan r = −0.56; p < 0.01), as well as between the NDVI and gaseous air pollutants (nitrogen dioxide—NO₂, sulfur dioxide—SO₂, and carbon monoxide—CO. Also, negative correlations between NDVI and climate parameters, air relative humidity (RH), and land surface albedo (LSA) were observed. These results show the potential of urban green to improve air quality through air pollutant deposition, retention, and alteration of vegetation health, particularly during dry seasons and hot summers. For the same period of analysis, positive correlations between the NDVI and solar surface irradiance (SI) and planetary boundary layer height (PBL) were recorded. Because of the summer season’s (June–August) increase in ground-level ozone, significant negative correlations with the NDVI (r = −0.51, p < 0.01) were found for Bucharest city center and (r = −76; p < 0.01) for the metropolitan area, which may explain the degraded or devitalized vegetation under high ozone levels. Also, during hot summer seasons in the 2020–2024 period, this research reported negative correlations between air temperature at 2 m height (TA) and the NDVI for both the Bucharest city center (r = −0.84; p < 0.01) and metropolitan scale (r = −0.90; p < 0.01), as well as negative correlations between the land surface temperature (LST) and the NDVI for Bucharest (city center r = −0.29; p< 0.01) and the metropolitan area (r = −0.68, p < 0.01). During summer seasons, positive correlations between ET and climate parameters TA (r = 0.91; p < 0.01), SI (r = 0.91; p < 0.01), relative humidity RH (r = 0.65; p < 0.01), and NDVI (r = 0.83; p < 0.01) are associated with the cooling effects of urban vegetation, showing that a higher vegetation density is associated with lower air and land surface temperatures. The negative correlation between ET and LST (r = −0.92; p < 0.01) explains the imprint of evapotranspiration in the diurnal variations of LST in contrast with TA. The decreasing trend of NPP over 24 years highlighted the feedback response of vegetation to air pollution and climate warming. For future green cities, the results of this study contribute to the development of advanced strategies for urban vegetation protection and better mitigation of air quality under an increased frequency of extreme climate events. Full article

Sentinel-3 synergy (SYN) VEGETATION (VGT) products were designed to provide continuity to the SPOT/VEGETATION (SPOT VGT) base products archive. Since the PROBA-V mission acted as a gap filler between SPOT VGT and Sentinel-3, and in principle, a continuous series of data products from the combined data archives of SPOT VGT (1998–2014), PROBA-V (2013–2020) and Sentinel-3 SYN VGT (from 2018 onwards) are available to users, the consistency of Sentinel-3 SYN VGT with both the latest SPOT VGT (VGT-C3) and PROBA-V (PV-C2) archives is highly relevant. In past years, important changes have been implemented in the SYN VGT processing baseline. The archive of SYN VGT products is therefore intrinsically inconsistent, leading to different consistency levels with SPOT VGT and PROBA-V throughout the years. A spatio-temporal intercomparison of the combined time series of VGT-C3, PV-C2 and Sentinel-3 SYN VGT 10-day NDVI composite products with an external reference from LSA-SAF, and an intercomparison of Sentinel-3 SYN V10 products with a climatology of VGT-C3 resp. PV-C2 for three distinct periods with different levels of product quality have shown that the subsequent processing baseline updates have indeed resulted in better-quality products. It is therefore essential to reprocess the entire Sentinel-3 SYN VGT archive; a uniform data record of standard SPOT VGT, PROBA-V and Sentinel-3 SYN VGT products, spanning over 25 years, would provide valuable input for a wide range of applications. Full article

Efficient operation of urban metro systems depends on accurate passenger flow predictions, a task complicated by intricate spatiotemporal correlations. This paper introduces a novel spatiotemporal graph neural network (STGNN) designed explicitly for predicting multistep passenger flow within metro stations. In the spatial dimension, previous research primarily focuses on local spatial dependencies, struggling to capture implicit global information. We propose a spatial modeling module that leverages a dynamic global attention network (DGAN) to capture dynamic global information from all-pair interactions, intricately fusing prior knowledge from the input graph with a graph convolutional network. In the temporal dimension, we design a temporal modeling module tailored to navigate the challenges of both long-term and recent-term temporal passenger flow patterns. This module consists of series decomposition blocks and locality-aware sparse attention (LSA) blocks to incorporate multiple local contexts and reduce computational complexities in long sequence modeling. Experiments conducted on both simulated and real-world datasets validate the exceptional predictive performance of our proposed model. Full article

Sign language recognition technology can help people with hearing impairments to communicate with non-hearing-impaired people. At present, with the rapid development of society, deep learning also provides certain technical support for sign language recognition work. In sign language recognition tasks, traditional convolutional neural networks used to extract spatio-temporal features from sign language videos suffer from insufficient feature extraction, resulting in low recognition rates. Nevertheless, a large number of video-based sign language datasets require a significant amount of computing resources for training while ensuring the generalization of the network, which poses a challenge for recognition. In this paper, we present a video-based sign language recognition method based on Residual Network (ResNet) and Long Short-Term Memory (LSTM). As the number of network layers increases, the ResNet network can effectively solve the granularity explosion problem and obtain better time series features. We use the ResNet convolutional network as the backbone model. LSTM utilizes the concept of gates to control unit states and update the output feature values of sequences. ResNet extracts the sign language features. Then, the learned feature space is used as the input of the LSTM network to obtain long sequence features. It can effectively extract the spatio-temporal features in sign language videos and improve the recognition rate of sign language actions. An extensive experimental evaluation demonstrates the effectiveness and superior performance of the proposed method, with an accuracy of 85.26%, F1-score of 84.98%, and precision of 87.77% on Argentine Sign Language (LSA64). Full article

Land surface temperature (LST) and land surface albedo (LSA) are the two key regional and global climate-controlling parameters; assessing their behavior would likely result in a better understanding of the appropriate adaptation strategies to mitigate the consequences of climate change. This study was conducted to explore the spatiotemporal variability in LST and LSA across different land use/cover (LULC) classes in northwest Iran. To do so, we first applied an object-oriented algorithm to the 10 m resolution Sentinel-2 images of summer 2019 to generate a LULC map of a 3284 km² region in northwest Iran. Then, we computed the LST and LSA of each LULC class using the SEBAL algorithm, which was applied to the Landsat-8 images from the summer of 2019 and winter of 2020. The results showed that during the summer season, the maximum and minimum LSA values were associated with barren land (0.33) and water bodies (0.11), respectively; during the winter season, the maximum LSA value was observed for farmland and snow cover, and the minimum value was observed in forest areas (0.21). The maximum and minimum LST values in summer were acquired from rangeland (37 °C) and water bodies (24 °C), respectively; the maximum and minimum values of winter values were detected in forests (4.14 °C) and snow cover (−21.36 °C), respectively. Our results revealed that barren land and residential areas, having the maximum LSA in summer, were able to reduce the heating effects to some extent. Forest areas, due to their low LSA and high LST, particularly in winter, had a greater effect on regional warming compared with other LULC classes. Our study suggests that forests might not always mitigate the effects of global warming as much as we expect. Full article

Land surface albedo (LSA) directly affects the radiation balance and the surface heat budget. LSA is a key variable for local and global climate research. The complexity of LSA variations and the driving factors highlight the importance of continuous spatial and temporal monitoring. Snow, vegetation and soil are the main underlying surface factors affecting LSA dynamics. In this study, we combined Global Land Surface Satellite (GLASS) products and ERA5 reanalysis products to analyze the spatiotemporal variation and drivers of annual mean blue-sky albedo for stable land cover types in the middle-high latitudes of the Northern Hemisphere (30~90°N) from 1982 to 2015. Snow cover (SC) exhibited a decreasing trend in 99.59% of all pixels (23.73% significant), with a rate of −0.0813. Soil moisture (SM) exhibited a decreasing trend in 85.66% of all pixels (22.27% significant), with a rate of −0.0002. The leaf area index (LAI) exhibited a greening trend in 74.38% of all pixels (25.23% significant), with a rate of 0.0014. Blue-sky albedo exhibited a decreasing trend in 98.97% of all pixels (65.12% significant), with a rate of −0.0008 (OLS slope). Approximately 98.16% of all pixels (57.01% significant) exhibited a positive correlation between blue-sky albedo and SC. Approximately 47.78% and 67.38% of all pixels (17.13% and 25.3% significant, respectively) exhibited a negative correlation between blue-sky albedo and SM and LAI, respectively. Approximately 10.31%, 20.81% and 68.88% of the pixel blue-sky albedo reduction was mainly controlled by SC, SM and LAI, respectively. The decrease in blue-sky albedo north of 40°N was mainly caused by the decrease in SC. The decrease in blue-sky albedo south of 40°N was mainly caused by SM reduction and vegetation greening. The decrease in blue-sky albedo in the western Tibetan Plateau was caused by vegetation greening, SM increase and SC reduction. The results have important scientific significance for the study of surface processes and global climate change. Full article

Landslide represents an increasing menace causing huge casualties and economic losses, and rainfall is a predominant factor inducing landslides. Landslide susceptibility assessment (LSA) is a commonly used and effective method to prevent landslide risk, however, the LSA does not analyze the impact of the rainfall on landslides which is significant and non-negligible. Therefore, the spatiotemporal LSA considering the inducing effect of rainfall is proposed to improve accuracy and applicability. In this study, the influencing factors are selected using the chi-square test, out-of-bag error and multicollinearity test. The spatial LSA are thus obtained using the random forest (RF) model, deep belief networks model and support vector machine, and compared using receiver operating characteristic curve and seed cell area index to determine the optimal assessment result. According to the heavy rainfall characteristics in the study area, the rainfall period is divided into four stages, and the effective rainfall model is employed to generate the rainfall impact (RI) maps of the four stages. The spatiotemporal LSAs are obtained by coupling the optimal spatial LSA and various RI maps and verified using the landslide warning map. The results demonstrate that the optimal spatiotemporal LSA is obtained using the spatial LSA of the RF model and temporal LSA of the rainfall data in the peak stage. It can predict the area where rainfall-induced landslides are likely to occur and prevent landslide risk. Full article

Satellite altimeters can be used to derive long-term and large-scale sea ice thickness changes. Sea ice thickness retrieval is based on measurements of freeboard, and the conversion of freeboard to thickness requires knowledge of the snow depth and snow, sea ice, and sea water densities. However, these parameters are difficult to be observed concurrently with altimeter measurements. The uncertainties in these parameters inevitably cause uncertainties in sea ice thickness estimations. This paper introduces a new method based on least squares adjustment (LSA) to estimate Arctic sea ice thickness with CryoSat-2 measurements. A model between the sea ice freeboard and thickness is established within a 5 km × 5 km grid, and the model coefficients and sea ice thickness are calculated using the LSA method. Based on the newly developed method, we are able to derive estimates of the Arctic sea ice thickness for 2010 through 2019 using CryoSat-2 altimetry data. Spatial and temporal variations of the Arctic sea ice thickness are analyzed, and comparisons between sea ice thickness estimates using the LSA method and three CryoSat-2 sea ice thickness products (Alfred Wegener Institute (AWI), Centre for Polar Observation and Modelling (CPOM), and NASA Goddard Space Flight Centre (GSFC)) are performed for the 2018–2019 Arctic sea ice growth season. The overall differences of sea ice thickness estimated in this study between AWI, CPOM, and GSFC are 0.025 ± 0.640 m, 0.143 ± 0.640 m, and −0.274 ± 0.628 m, respectively. Large differences between the LSA and three products tend to appear in areas covered with thin ice due to the limited accuracy of CryoSat-2 over thin ice. Spatiotemporally coincident Operation IceBridge (OIB) thickness values are also used for validation. Good agreement with a difference of 0.065 ± 0.187 m is found between our estimates and the OIB results. Full article

The evaporation of water from land into the atmosphere is a key component of the hydrological cycle. Accurate estimates of this flux are essential for proper water management and irrigation scheduling. However, continuous and qualitative information on land evaporation is currently not available at the required spatio-temporal scales for agricultural applications and regional-scale water management. Here, we apply the Global Land Evaporation Amsterdam Model (GLEAM) at 100 m spatial resolution and daily time steps to provide estimates of land evaporation over The Netherlands, Flanders, and western Germany for the period 2013–2017. By making extensive use of microwave-based geophysical observations, we are able to provide data under all weather conditions. The soil moisture estimates from GLEAM at high resolution compare well with in situ measurements of surface soil moisture, resulting in a median temporal correlation coefficient of 0.76 across 29 sites. Estimates of terrestrial evaporation are also evaluated using in situ eddy-covariance measurements from five sites, and compared to estimates from the coarse-scale GLEAM v3.2b, land evaporation from the Satellite Application Facility on Land Surface Analysis (LSA-SAF), and reference grass evaporation based on Makkink’s equation. All datasets compare similarly with in situ measurements and differences in the temporal statistics are small, with correlation coefficients against in situ data ranging from 0.65 to 0.95, depending on the site. Evaporation estimates from GLEAM-HR are typically bounded by the high values of the Makkink evaporation and the low values from LSA-SAF. While GLEAM-HR and LSA-SAF show the highest spatial detail, their geographical patterns diverge strongly due to differences in model assumptions, model parameterizations, and forcing data. The separate consideration of rainfall interception loss by tall vegetation in GLEAM-HR is a key cause of this divergence: while LSA-SAF reports maximum annual evaporation volumes in the Green Heart of The Netherlands, an area dominated by shrubs and grasses, GLEAM-HR shows its maximum in the national parks of the Veluwe and Heuvelrug, both densely-forested regions where rainfall interception loss is a dominant process. The pioneering dataset presented here is unique in that it provides observational-based estimates at high resolution under all weather conditions, and represents a viable alternative to traditional visible and infrared models to retrieve evaporation at field scales. Full article

African landscape fires are widespread, recurrent and temporally dynamic. They burn large areas of the continent, modifying land surface properties and significantly affect the atmosphere. Satellite Earth Observation (EO) data play a pivotal role in capturing the spatial and temporal variability of African biomass burning, and provide the key data required to develop fire emissions inventories. Active fire observations of fire radiative power (FRP, MW) have been shown to be linearly related to rates of biomass combustion (kg s⁻¹). The Meteosat FRP-PIXEL product, delivered in near real-time by the EUMETSAT Land Surface Analysis Satellite Applications Facility (LSA SAF), maps FRP at 3 km resolution and 15-min intervals and these data extend back to 2004. Here we use this information to assess spatio-temporal variations in fire activity across sub-Saharan Africa, and identify an overall trend of decreasing annual fire activity and fuel consumption, agreeing with the widely-used Global Fire Emissions Database (GFEDv4) based on burned area measures. We provide the first comprehensive assessment of relationships between per-fire FRE-derived fuel consumption (Tg dry matter, DM) and temporally integrated Moderate Resolution Imaging Spectroradiometer (MODIS) net photosynthesis (PSN) (Tg, which can be converted into pre-fire fuel load estimates). We find very strong linear relationships over southern hemisphere Africa (mean r = 0.96) that are partly biome dependent, though the FRE-derived fuel consumptions are far lower than those derived from the accumulated PSN, with mean fuel consumptions per unit area calculated as 0.14 kg DM m⁻². In the northern hemisphere, FRE-derived fuel consumption is also far lower and characterized by a weaker linear relationship (mean r = 0.76). Differences in the parameterization of the biome look up table (BLUT) used by the MOD17 product over Northern Africa may be responsible but further research is required to reconcile these differences. The strong relationship between fire FRE and pre-fire fuel load in southern hemisphere Africa is encouraging and highlights the value of geostationary FRP retrievals in providing a metric that relates very well to fuel consumption and fire emission variations. The fact that the estimated fuel consumed is only a small fraction of the fuel available suggests underestimation of FRE by Spinning Enhanced Visible and Infrared Imager (SEVIRI) and/or that the FRE-to-fuel consumption conversion factor of 0.37 MJ kg⁻¹ needs to be adjusted for application to SEVIRI. Future geostationary imaging sensors, such as on the forthcoming Meteosat Third Generation (MTG), will reduce the impact of this underestimation through its ability to detect even smaller and shorter-lived fires than can the current second generation Meteosat. Full article

The Land Surface Albedo (LSA) was estimated in the broader Mediterranean area, on an 8-day basis, for the period 2001–2012. MODIS (Moderate Resolution Imaging Spectroradiometer) albedo product parameters, at 1 km × 1 km spatial resolution, were used. LSA changes during the above study period were also estimated, based on annual average values. Results revealed increasing LSA trends dominating in the Levant region and decreasing in NW Africa, of the order of 3.3% and −6.6%, respectively, while mixed signs were observed in southern Europe. Three factors that can determine the LSA changes were investigated: land cover changes, rainfall changes and Aerosol Optical Thickness (AOT) spatio-temporal variability. The analysis made clear that land cover and rainfall changes affect LSA at local and regional scales, while the effect of AOT was not important. Land cover changes revealed deforestation hot spots, where LSA was increased by 13%–14%, while an increase in rainfall over many areas in NW Africa appears to have caused a corresponding decrease in LSA by over 5%. These findings highlight the importance of a global and continuous LSA monitoring at both regional and local scales, which is necessary for both climate monitoring and modeling studies. Full article