Search

Ecological vulnerability assessment is crucial for environment protection, ecological restoration and resource utilization. However, many former studies have limitations in the indicator system of the assessment, which were not comparable for different types of ecologically vulnerable areas. It is difficult to apply directly to the ecological vulnerability assessment of different types and in various regions. Aiming to solve these problems, the study proposed a well-established and comprehensive indicator system for ecological vulnerability assessment and conducted ecological vulnerability assessment application of five types of typical ecologically vulnerable areas of China based on remote sensing, meteorological, geographic and other data. The results show that the average EVIs value of Zhangbei County ranging from 0.525 to 0.559 are the highest among the five research areas during the four periods, followed by Zoige region and Xiamen bay. However, the region with the lowest average EVI value varies. In 2005 and 2015, it was Taihe County, while in 2010 and 2020, it was the Sanjiangyuan region. The variation of average EVIs in the five typical areas presents slight fluctuation and remains generally stable from 2005 to 2020. It indicates that the environmental protection measures and projects undertaken by the Chinese government in recent years have had a striking effect, curbing the trend of ecological environment deterioration.

The secure and stable operation of power transmission lines is essential for electrical systems. Given that abnormal targets such as bird’s nests and defective insulators may lead to transmission failures, timely detection of these targets is imperative. This paper introduces the YOLO-PowerLite model, an advanced lightweight object detection model based on YOLOv8n, designed for efficient, real-time detection on resource-constrained unmanned aerial vehicles (UAVs) equipped with edge computing platforms. In the feature fusion module, YOLO-PowerLite incorporates the innovative C2f_AK module, significantly reducing the number of parameters and enhancing the adaptability and fusion capability of features at different scales. Meanwhile, the adoption of the Bidirectional Feature Pyramid Network (BiFPN) further optimizes the efficiency and effectiveness of feature processing. In addition, the newly designed lightweight detection head significantly reduces the number of parameters and computational requirements. The integration of the Coordinate Attention mechanism in the backbone network enhances the model’s ability to focus on and recognize abnormal targets in complex backgrounds. Experimental results show that YOLO-PowerLite achieves a mAP@0.5 of 94.2%, maintaining the accuracy of the original YOLOv8n while significantly reducing parameters, FLOPs, and model size by 42.3%, 30.9%, and 40.4%, respectively. Comparative analysis shows that YOLO-PowerLite surpasses other mainstream lightweight models in detection accuracy and computational efficiency. Deployment on the NVIDIA Jetson Xavier NX platform demonstrates an average processing time of 31.2 milliseconds per frame, highlighting its potential for real-time applications in monitoring transmission lines.

The severe wildfires that have ravaged Guangdong province, China, present a significant threat to the local ecosystem, socio-economics, and public health. Effective risk assessment is essential for early warning and timely prevention in wildfire management, thereby mitigating disaster losses. In this study, we compiled a dataset comprising 11,507 historical wildfire incidents in Guangdong Province spanning a decade (2011–2021) and developed a deep learning-based model to predict the likelihood of wildfire occurrence in the region. In addition to analyzing risk characteristics throughout the year, we also trained separate models for different seasons and analyzed the discrepancies in the contribution of driven factors to wildfire occurrence across seasons. Furthermore, the performance of our deep learning-based model was compared with that of traditional machine learning algorithms. The experimental results revealed that: (1) Factors such as relative humidity, temperature, NDVI, and precipitation exerted significant influence on wildfire occurrence. (2) The impact of wildfire driving factors varied across different seasons. (3) Our deep learning model outperformed traditional machine learning models, achieving a superior performance with an AUC of 0.962.

AbstractWildfires not only directly affect ecology but also have a huge impact on social and economic stability. Enhancing the understanding of wildfire influences can guide the design of targeted strategies in mitigation plans and adaptation strategies. Here, we explored the spatial-temporal characteristics of wildfires in the Beijing-Tianjin-Hebei region of China and the mechanisms of the effect of potential influencing factors on wildfire occurrence based on 2014–2020 wildfire data. The results show that wildfires mainly occur in spring and winter and decrease yearly. Among all the considered influencing factors, the occurrence of wildfire is most significantly affected by the temperature vegetation dry index (TVDI) and is also greatly affected by population and economy. Through interaction detection, it was found that the normalized difference vegetation index (NDVI) greatly interfered with wildfires. The high occurrence risk of wildfire was distributed in a blocky pattern in the middle and east of the study area. In addition, we also found a positive spatial correlation between the high wildland-urban interface (WUI) presence and high wildfire occurrence. These findings have improved our understanding of the impact of multiple influencing factors on wildfires and are supposed to facilitate the development of wildfire management strategies.

Wildfire is one of the main hazards affecting large areas and causes great damage all over the world, and the rapid development of the wildland-urban interface (WUI) increases the threat of wildfires that have ecological, social, and economic consequences. As one of the most widely used methods for tracking fire, remote sensing can provide valuable information about fires, but it is not always available, and needs to be supplemented by data from other sources. Social media is an emerging but underutilized data source for emergency management, contains a wealth of disaster information, and reflects the public’s real-time witness and feedback to fires. In this paper, we propose a fusion framework of multi-source data analysis, including social media data and remote sensing data, cellphone signaling data, terrain data, and meteorological data to track WUI fires. Using semantic web technology, the framework has been implemented as a Knowledge Base Service and runs on top of WUIFire ontology. WUIFire ontology represents WUI fire–related knowledge and consists of three modules: system, monitoring, and spread, and tracks wildfires happening in WUIs. It provides a basis for tracking and analyzing a WUI fire by fusing multi-source data. To showcase the utility of our approach in a real-world scenario, we take the fire in the Yaji Mountain Scenic Area, Beijing, China, in 2019 as a case study. With object information identified from remote sensing, fire situation information extracted from Weibo, and fire perimeters constructed through fire spread simulation, a knowledge graph is constructed and an analysis using a semantic query is carried out to realize situational awareness and determine countermeasures. The experimental results demonstrate the benefits of using a semantically improved multi-source data fusion framework for tracking WUI fire.

In this article, the trends in global vegetative drought were investigated using MODIS- and AVHRR-based NDVI products. A set of selected methods were employed to perform trend analysis including trend test, trend location detection, and trend estimates. Accounting for the effect of the global geographical heterogeneity on trend analysis, the NDVI dataset was aggregated on designated divisions in view of latitude ranges and climate zones. From the results, it was concluded that: AVHRR has longer-term records that provide a critical historical perspective on vegetation activities necessary for global change research, and continuity and correctness is achievable from AVHRR VCI given the systematic offset between the NDVI values derived from the two sensors and the characteristics of the VCI algorithm. From a pixel level global trend analysis map, the proportion of pixels with rising trends is 54.7% in the world, 67.6% and 47.5% in the northern hemisphere and the southern hemisphere, respectively, which means there is an overall rising trends in the global VCI values, especially in the northern hemisphere. The North Temperate and the South Tropical have overall increases in the VCI values while all climate zones have overall increases in the VCI values. The piecewise trends basically adhere to the results of overall trend identification although there are some local variations. There are obvious rising trends during the latest years for all the climate zones. Dominant down trends were identified in A, B, Cw, Ds, and E while the piecewise trends for both Cs and Df are dominantly rising before 2000. An average of about four breakpoints were detected from both the climate zone- and latitude range-aggregated divisions Thus, the mean duration for a piecewise trend is 7-9 years.

Air temperature (Ta) is an essential parameter for science research and engineering practice. While the traditional site-based approach is only able to obtain observations in limited and discrete locations, satellite remote sensing is promising to retrieve some environmental variables with spatially continuous coverage. Nowadays, land surface temperature (Ts) measurements can be obtained from some satellite sensors (e.g., MODIS), further enabling us to estimate Ta in view of the relationship between Ta and Ts. In this article, we proposed a two-phase integrated framework to estimate daily mean Ta nationwide. In the first phase, multivariate linear regression models were fitted between site-based observations of daily mean air temperature (Ta-mean) and MODIS land surface temperature products (including Terra day: TMOD-day, Terra night: TMOD-night, Aqua day: TMYD-day, and Aqua night: TMYD-night) conditional on some covariates of environmental factors. The fitted models were then used to predict Ta-mean from those covariates at unobserved locations. The predicted Ta-mean were looked on as stochastic variables, and their distributions were also obtained. In the second phase, Bayesian maximum entropy (BME) methods were used to produce spatially continuous maps of Ta-mean taking the meteorological station observations as hard data and the predicted Ta-mean in the first phase as soft data. It is shown that the proposed approach is promising to improve the interpolation accuracy significantly, comprehensively considering the prior knowledge and the context of space variability and correlation, which will enable it to compile spatially continuous air temperature products with higher accuracy.

Near-land surface air temperature (NLSAT) is an important meteorological and climatic parameter widely used in climate change, urban heat island and environmental science, in addition to being an important input parameter for various earth system simulation models. However, the spatial distribution and the limited number of ground-based meteorological stations make it difficult to obtain a large range of high-precision NLSAT values. This paper constructs neural network, long short-term memory, bi-directional long short-term memory, support vector machine, random forest, and Gaussian process regression models by combining MODIS data, DEM data, and meteorological station data to estimate the NLSAT in China’s mainland and compare them with actual NLSAT observations. The results show that there is a significant correlation between the model estimates and the actual temperature observations. Among the tested models, the random forest performed the best, followed by the support vector machine and the Gaussian process regression, then the neural network, the long short-term memory, and the bi-directional long short-term memory models. Overall, for estimates in different seasons, the best results were obtained in winter, followed by spring, autumn, and summer successively. According to different geographic areas, random forest was the best model for Northeast, Northwest, North, Southwest, and Central China, and the support vector machine was the best model for South and East China.

Drought is one of the billion-dollar natural disasters and hard to trace and measure. In recent years drought monitoring becomes much easier with remote sensing. However, it is still difficult to pin vegetation variances on drought because of the delay of the caused vegetation stress. To assess vegetative drought, it is important to first understand the relationship between meteorological condition and vegetation condition, and measure the vegetation responses to meteorological drought. It would be very helpful for effective early warning about agricultural drought. This study uses the CONUS as the study area, and utilizes remote sensing products such as NDVI/VCI (normalized difference vegetation index/vegetation condition index) and SPI/SPEI (standardized precipitation index/standardized precipitation evapotranspiration index) to give a comparative evaluation to the vegetation’s drought response. The used vegetation products and meteorological data are ensured to be consistent. The scale and lag of vegetation response to drought for various vegetation types and aridity levels were thoroughly investigated. The results show that: The AVHRR and MODIS NDVI series pairs and the meteorological drought index series pairs (SPEI and SPI) have fairly good consistencies. Among them, 69.5% and 84% have rho (correlation coefficient) values greater than 0.8 respectively. For the NDVI series pairs, the maximum rho value is 0.98, the minimum rho value is −0.47, and the mean rho value is 0.79, which are 0.97, 0.68, and 0.87 respectively for the meteorological index series pairs. Compared to rho values of the meteorological index series pairs, the rho values of the NDVI series pairs have more outliers indicating instability. The correlation between SPEI and VCI significantly relies on time lags and has high spatial inhomogeneity. 1- and 2-month lags of SPEI have more significant positive correlation with VCI in Arid, Semi-Arid and Dry sub-humid areas of central west CONUS with less precipitation and lower temperature. For various time scales (time scale is SPEI reference range), the most significant positive correlation between SPEI and VCI happens in the time scales of 6- to 12-month in summer, the time scales of 3- and 6-month in spring and autumn, and the time scales of 2- and 3-month in winter regardless of time lags. Despite of the different vegetation types and aridity levels, the maximum correlations between SPEI and VCI are observed in Hyper arid regions in January, Arid regions in April, Semi-arid regions in July, and Dry sub-humid regions in October. Shrub has prominent responses in January and April, grass responses appear in July and October, and Evergreen forest shows least responses in all seasons. The results add more insights of the connection between vegetation and climate, and guide the development of new technology leveraging remote sensing data for vegetation drought monitoring and early-warning. The results are also helpful to provide important references for studying large-scale physiological and phenological properties of the vegetation under different climate conditions.

Indoor 3D reconstruction and navigation element extraction with point cloud data has become a research focus in recent years, which has important application in community refinement management, emergency rescue and evacuation, etc. Aiming at the problem that the complete wall surfaces cannot be obtained in the indoor space affected by the occluded objects and the existing methods of navigation element extraction are over-segmented or under-segmented, we propose a method to automatically reconstruct indoor navigation elements from unstructured 3D point cloud of buildings with occlusions and openings. First, the outline and occupancy information provided by the horizontal projection of the point cloud was used to guide the wall segment restoration. Second, we simulate the scanning process of a laser scanner for segmentation. Third, we use projection statistical graphs and given rules to identify missing wall surfaces and “hidden doors”. The method is tested on several building datasets with complex structures. The results show that the method can detect and reconstruct indoor navigation elements without viewpoint information. The means of deviation in the reconstructed models is between 0–5 cm, and the completeness and correction are greater than 80%. However, the proposed method also has some limitations for the extraction of “thick doors” with a large number of occluded, non-planar components.

Drought is one of the most serious natural disasters in the world and causes great economic losses in China every year, especially in its southwest region. Yet, few studies have reported the quantitative comprehensive risk of drought in the Yunnan, Guizhou, and Guangxi provinces of China. Taking these three provinces as the study area, we obtained annual precipitation, disaster loss, and agricultural planting data during 1964–2013. Following an optimal estimation of annual precipitation by the Bayesian maximum entropy method, we mapped the annual Standardized Precipitation Index. Based on the theory of information diffusion and exceeding probability, the hazard of drought was evaluated. We also fit the vulnerability curves using the drought loss data. As a basis, we constructed a multiplicative formula to calculate the comprehensive risk of drought, which integrates the hazard and the vulnerability and produces drought loss rate (DLR) maps. We found that the DLR caused by mild drought was about 3%, moderate drought 10%, severe drought 25%, and extreme drought 50%. We also created a risk zoning map to provide practical information, such as a scientific basis for optimization of regional allocation of resources for drought preparedness and response.

Spatiotemporal patterns are one of the greatest interests and provide valuable insights into chronological events occurring in space. A tropical cyclone (TC) track is defined as a sequence of successive points, and several different types of analyses are performed to explore the temporal and spatial patterns of the TCs in the Northwest Pacific and along the coastline of China during 1949–2014. Results show that (1) the number of TCs is getting more frequent from April to August and less frequent from August to October with the peak occurring in August almost every year; (2) the mean of the sizes of the annual temporal clusters during 1949–2014 is 52.5 (days), the standard deviation is 17.0 (days), and the average starting point is the 210.5th day; (3) the spatial clusters are located in two areas: the boundary of Guangxi and Guangdong provinces and the boundary of Fujian and Zhejiang provinces; and (4) the within-strata variance is less than the between-strata variance, which implies the locational and seasonal factors are the potential determinants of the heterogeneity of the TCs. Furthermore, several maps representing the hazards of TCs are produced. According to the resultant maps, 12 coastal prefectures (Zhanjiang, Maoming, Fuzhou, Huizhou, Yangjiang, Qinzhou, Ningde, Quanzhou, Jiangmen, Nanning, Zhangzhou, and Hangzhou) have return periods of less than two years, and the two island provinces of Hainan and Taiwan are visited by TCs the most. Guangdong, Guangxi, Fujian, and Zhejiang provinces in particular suffered severely from the destructive TCs.

Gathering very accurate spatially explicit data related to the distribution of mean annual precipitation is required when laying the groundwork for the prevention and mitigation of water-related disasters. In this study, four Bayesian maximum entropy (BME) models were compared to estimate the spatial distribution of mean annual precipitation of the selected areas. Meteorological data from 48 meteorological stations were used, and spatial correlations between three meteorological factors and two topological factors were analyzed to improve the mapping results including annual precipitation, average temperature, average water vapor pressure, elevation, and distance to coastline. Some missing annual precipitation data were estimated based on their historical probability distribution and were assimilated as soft data in the BME method. Based on this, the univariate BME, multivariate BME, univariate BME with soft data, and multivariate BME with soft data analysis methods were compared. The estimation accuracy was assessed by cross-validation with the mean error (ME), mean absolute error (MAE), and root mean square error (RMSE). The results showed that multivariate BME with soft data outperformed the other methods, indicating that adding the spatial correlations between multivariate factors and soft data can help improve the estimation performance.

With the continuous progress of human production and life, air quality has become the focus of attention. In this paper, Beijing, Tianjin, Hebei, Shanxi, Shandong and Henan provinces were taken as the study area, where there are 58 air quality monitoring stations from which daily and monthly data are obtained. Firstly, the temporal characteristics of the air quality index (AQI) are explored. Then, the spatial distribution of the AQI is mapped by the inverse distance weighted (IDW) method, the ordinary kriging (OK) method and the Bayesian maximum entropy (BME) method. Additionally, cross-validation is utilized to evaluate the mapping results of these methods with two indexes: mean absolute error and root mean square interpolation error. Furthermore, the correlation analysis of meteorological factors, including precipitation anomaly percentage, precipitation, mean wind speed, average temperature, average water vapor pressure and average relative humidity, potentially affecting the AQI was carried out on both daily and monthly scales. In the study area and period, AQI shows a clear periodicity, although overall, it has a downward trend. The peak of AQI appeared in November, December and January. BME interpolation has a higher accuracy than OK. IDW has the maximum error. Overall, the AQI of winter (November), spring (February) is much worse than summer (May) and autumn (August). Additionally, the air quality has improved during the study period. The most polluted areas of air quality are concentrated in Beijing, the southern part of Tianjin, the central-southern part of Hebei, the central-northern part of Henan and the western part of Shandong. The average wind speed and average relative humidity have real correlation with AQI. The effect of meteorological factors such as wind, precipitation and humidity on AQI is putative to have temporal lag to different extents. AQI of cities with poor air quality will fluctuate greater than that of others when weather changes and has higher correlation with meteorological factors.