Archive

  • 2021 Volume 23 Issue 11
    Published: 25 November 2021
      

  • Select all
    |
  • YIN Ling, LIU Kang, ZHANG Hao, XI Guikai, LI Xuan, LI Ziyin, XUE Jianzhang
    Download PDF ( ) HTML ( )   Knowledge map   Save

    The spread of infectious diseases is usually a highly nonlinear space-time diffusion process. Epidemiological models can not only be used to predict the epidemic trend, but also be used to systematically and scientifically study the transmission mechanism of the complex processes under different hypothetical intervention scenarios, which provide crucial analytical and planning tools for public health studies and policy-making. Since host behavior is one of the critical driven factors for the dynamics of infectious diseases, it is important to effectively integrate human spatiotemporal behavior into the epidemiological models for human-hosted infectious diseases. Due to the rapid development of human mobility research and applications aided by big trajectory data, many of the epidemiological models for Coronavirus Disease 2019 (COVID-19) have already coupled human mobility. By incorporating real trajectory data such as mobile phone location data at an individual or aggregated level, researchers are working towards the direction of accurately depicting the real world, so as to improve the effectiveness of the model in guiding actual epidemic prevention and control. The epidemic trend prediction, Non-pharmaceutical Interventions (NPIs) evaluation, vaccination strategy design, and transmission driven factors have been studied by the epidemiological models coupled with human mobility, which provides scientific decision-making aid for controlling epidemic in different countries and regions. In order to systematically understand this important progress of epidemiological models, this study collected and summarized relevant literatures. First, the interactions between the COVID-19 epidemic and human mobility were analyzed, which demonstrated the necessity of integrating the complex spatiotemporal behavior, such as population-based or individual-based mobility, activity, and contact interaction, into the epidemiological models. Then, according to the modeling purpose and mechanism, the models integrated with human mobility were discussed by two types: short-term epidemic prediction models and process simulation models. Among them, based on the coupling methods of human mobility, short-term epidemic prediction models can further be divided into models coupled with first-order and second-order human mobility, while process simulation models can be divided into models coupled with population-based mobility and individual-based mobility. Finally, we concluded that epidemiological models integrating human mobility should be developed towards more complex human spatiotemporal behaviors with a fine spatial granularity. Besides, it is in urgent need to improve the model capability to better understand the disease spread processes over space and time, break through the bottleneck of the huge computational cost of fine-grained models, cooperate cutting-edge artificial intelligence approaches, and develop more universal and accessible modeling data sets and tools for general users.

  • CUI Mingjie, YAO Xia, FANG Haoran, ZHANG Yangchengsi, YANG Degang, PEI Tao
    Download PDF ( ) HTML ( )   Knowledge map   Save

    The outbreaks of SARS and COVID-19 have had a serious impact on public health, social economy and so on in China, in order to reveal the common law and difference characteristics of space-time transmission of respiratory infectious diseases and the reasons behind them, using space-time statistical methods, systematically analyzed and compared the difference characteristics of space-time transmission between SARS and COVID-19, and combined with the transmission characteristics of the virus itself and temperature, traffic and other factors to analyze the causes. The study shows that, ① SARS experiences two stages, the rising period-flat phase, and the COVID-19 experiences three stages, the rising period-sharp rise-slow up period. ② In the mode of spatial transmission, the transmission intensity and range of COVID-19 is greater than that of SARS, and the overall connectivity of COVID-19 is greater and the provinces are more closely related to the outbreak of the virus. Both SARS and COVID-19 transmission have obvious spatial aggregation characteristics. They are based on proximity propagation and long-range leaps, and SARS has a secondary communication center, and COVID-19 diffusion center has not been relocated. ③ In the direction of space communication, SARS is centered in Beijing, Hong Kong and Guangdong, the direction of spatial communication is stronger, and COVID-19 is only spread outwards with Hubei as the center. ④ In terms of spatial transmission speed, the spread time of the first case in each province of SARS is relatively large, and the spread time of the first case in each province of COVID-19 is roughly divided by Hu Huanyong Line, showing a phenomenon of "fast in the east and slow in the west", and the spread time span is relatively short. ⑤ R0 is the main reason for the difference between the spatial transmission range of SARS and COVID-19 and the speed of spatial transmission. The temperature suitability of SARS and COVID-19 viruses is different, but spatial aggregation transmission and adjacent area transmission are occurring in areas with similar temperatures. Besides the virus transmission capacity and temperature impact, traffic is the main reason affecting SARS and COVID-19 space long-range leap transmission, and the spatial transmission speed of both is negatively related to the density of the road network.

  • LI Zhao, GAO Huiying, DAI Xiaoyi, SUN Hai
    Download PDF ( ) HTML ( )   Knowledge map   Save

    The COVID-19 epidemic poses a great threat to public health and people's lives, which has initiated new challenges to the prevention and control system of the epidemic in China. In all efforts for epidemic control and prevention, predicting the risk of epidemic spread is of great practical importance for scientific prevention and control, and precise strategies. To predict the risk of an epidemic rapidly and quantitatively, this paper fused multi-source spatiotemporal data and established a risk prediction model for epidemic transmission by coupling LSTM algorithm and cloud model. Firstly, a simulation model of the spatiotemporal spread of infectious diseases was built based on GIS and LSTM algorithm, which simulated the infectious disease's spatiotemporal transmission process by learning rules in historical epidemic data. At the same time, to improve the simulation accuracy, this paper took 1 km × 1 km for the spatial scale, and days for the temporal scale as the study scale. Secondly, this paper applied the simulated data of infectious cases and the spatiotemporal influence factors on the spread of the epidemic to construct risk evaluation indicators. Finally, the cloud model and adaptive strategies were applied to construct an epidemic risk assessment model. In this way, the epidemic risk assessment at multiple spatial scales was achieved. In the empirical study phase, based on the Beijing COVID-19 epidemic data from 11 June 2020 to 25 June 2020, this paper simulated the process of the spatial evolution of the epidemic from 26 June 2020 to 1 July 2020. To test the advantage of the LSTM model applied to simulate spatiotemporal spread of infectious diseases, four machine learning models were introduced for comparison, including GA-BP Neural Network, Decision Regression Tree, Random Forest, and Support Vector Machine. The results were as follows: ① Compared with other conventional machine learning models, the LSTM model with time-series relationship had higher simulation accuracy (MAE=0.002 61) and better fitting degree (R-Square=0.9455). This showed that the LSTM model considering the temporal relationship between epidemic data was more suitable for epidemic spatial evolution simulation. ② The application results showed that the coupled model can not only fully consider the influence of infection source factors, weather factors, epidemic spread factors and epidemic prevention factors on the spread of transmission risk and reflect the trend of risk evolution, but also quickly quantify regional risk levels. Therefore, the coupled model based on LSTM algorithm and cloud model can effectively predict the transmission risk of epidemic, and also provide a method reference for establishing spatial-temporal transmission models and assessing epidemic risk.

  • ZHANG Hao, YIN Ling, LIU Kang, MAO Liang, FENG Shengzhong, CHEN Jie, MEI Shujiang
    Download PDF ( ) HTML ( )   Knowledge map   Save

    Many cities in China have adopted a series of Non-Pharmaceutical Interventions (NPIs) and rapidly suppressed the 1st wave of COVID-19 epidemic in 2020. It is critical to evaluate the effectiveness of these NPIs for future epidemic control. However, as a variety of NPIs were applied together in practice, it is difficult to evaluate the effectiveness of a single type of intervention by epidemiological observation. Taking Shenzhen city as an example, this study used a spatially explicit agent-based model by integrating mobile phone location data, travel survey data, building survey data and other multi-source spatiotemporal big data to evaluate the effectiveness of different types of NPIs in the suppression of the 1st wave of COVID-19 epidemic in Shenzhen. The simulation results show that the peak of the epidemic would have appeared on the 127th day since Jan 1st of 2020, resulting in an average of 72.26% of the population to be infected without any interventions. In the 1st wave of Shenzhen epidemic, except for the hospitalization of confirmed cases and intercity traffic restrictions, the stay-at-home order was the most effective one, followed by comprehensive isolation and quarantine measures (for close contacts, imported population and suspected cases), mask wearing, and orderly resumption of work. The stay-at-home order and comprehensive isolation and quarantine measures can effectively control the large-scale outbreak of the COVID-19, which are identified as the core measures; Mask wearing and orderly resumption of work can only reduce the overall infection size and delay the epidemic peak, which are identified as secondary measures. Considering the socioeconomic costs and the receding compliance to interventions in the post-epidemic period, this study suggests that the core measures and secondary measures should be combined to control the sporadic cases. Specifically, the local government can give the highest priority to isolation and quarantine measures for confirmed cases and high-risk individuals, complemented by mask wearing. In addition, our model can reveal the high-risk infection areas at a community level, which can help deploy control measures within an urban environment. In summary, this study demonstrated the advantages of integrating spatiotemporal big data and agent-based models to simulate the spread processes of infectious diseases in an urban environment: it can not only simulate the evolving processes of an epidemic at a fine-grained scale, but also evaluate the effectiveness of the NPIs at an individual level and for activity-travel behaviors, which can be useful for precise intervention.

  • LIAO Jiaxin, WU Qiyong, LAN Xiaoji, ZHANG Hongqing
    Download PDF ( ) HTML ( )   Knowledge map   Save

    In order to extract the travel trajectory of urban residents more conveniently, analyze the daily spatial behavior of individuals, and provide data support for the decision-making of urban management measures, this paper proposes an urban travel trajectory extraction method based on WiFi probe data, which mainly solves the problem of map matching and lost trajectory reconstruction of WiFi probe data. First, extract the track record sequence by sorting the terminal MAC code and timestamp in multiple columns, and use the RSSI value to extract the candidate point set located on the road network for each record. Secondly, an algorithm based on local evaluation is designed: for each candidate point, the spatio-temporal relationship between the candidate point set extracted from the adjacent records is used to evaluate its temporal consistency and spatial consistency, and then the final score is obtained by combining with the weight function dynamically constructed in inverse time ratio, then the highest score point in each candidate point set is selected as the best matching point. Finally, a depth-first-based path search algorithm is used to search for all feasible paths between the upper and lower points of the lost trajectory, and then the optimal reconstruction path is determined based on the TOPSIS method. In this paper, the WiFi probe data collected in the central area of Dongguan City is used as the experimental data to test, and more than 60 000 tracks can be extracted every day on average. Compared with the GPS data, the feasibility of the method is verified, which provides a new solution for urban travel trajectory mining.

  • QI Lin, QIN Kun, LUO Ping, YAO Borui, ZHU Zhaoyuan
    Download PDF ( ) HTML ( )   Knowledge map   Save

    Conflicts occur frequently at any time and any place in the world. Conflicts often erupt between two or more parties. Analyzing the relation between various conflicts and monitoring the development and evolution of conflicts can help provide measures to intervene in conflicts and provide humanitarian assistance in the embryonic stage of conflicts, which can further help avoid the escalation of conflicts. Various conflict has attracted lots of attention from the public. The occurrence of various conflicts is usually reported by the news media in a timely manner, and each event information can be automatically collected by computers and recorded in news databases. The conflict news database contains a wealth of information. It provides a feasible way to extract the information of conflict events from the new data, quantify the conflict intensity, and analyze the change of national conflict intensity. The GDELT is such an excellent event database which monitors news from different sources around the world in real time, automatically extracts events and event attribute information in news, and classifies the event into conflict events and cooperation events. This paper uses GDELT event database as the data source and comprehensively obtains the number of events, the impact of the events, and the degree of attention to conflict events. We propose a method to quantitatively express the intensity of conflicts by using the global conflict index and the local conflict index for different spatial scales. At the global scale, we calculate the global conflict index of countries around the world to measure the intensity of national conflicts and analyze the spatial distribution of the intensity of global national conflicts. At the country level, the local conflict index is calculated to measure the change of conflict intensity in a country. Based on the quantitative expression of conflict intensity, a distance-based time series conflict detection method is employed to detect the occurrence of conflict events. The results show that: 1) Countries with high conflict intensity are mainly concentrated in Africa and the Middle East, and there is obvious spatial agglomeration of global conflict intensity; 2) The sudden increase in the national conflict index usually corresponds to the occurrence of some conflict events. The method of conflict detection in this paper can effectively detect the sudden increase in time and provide support for the early warning of conflicts. The research results of this paper can provide references for the analysis of international conflict relations and the decision-making of international rescue organizations.

  • ZHONG Yutong, WEI Jing, ZHENG Yueming, YAN Fuli
    Download PDF ( ) HTML ( )   Knowledge map   Save

    The quiescent period after favorable meteorological conditions greatly reduces the influence of PM2.5 transmission across regions, thus it can reveal emissions of local sources. In this paper, the concept of the pollutant distribution in the quiescent period is introduced to reveal characteristics of local source emissions. A spatial refinement method of the PM2.5 emission inventory based on remote sensing data is also proposed. Firstly, the high spatial and temporal resolution PM2.5 data fusion method was constructed using the ChinaHighPM2.5 data retrieved by MODIS MCD19A2. Then, the selection method of pollutants for the quiescent period after favorable meteorological conditions in Tangshan was established. The favorable meteorological condition is east wind, with a wind speed above 3 m/s at 10-m height. For other wind directions, it should be sustained strong wind, with a wind speed between 5~10 m/s. For quiescent period, the wind speed should be below 1.5~2.0 m/s. Furthermore, the total PM2.5 emissions from the MEIC Inventory were allocated based on the PM2.5 data of the selected quiescent period. At the same time, referring to the traditional interpolation methods, based on the data of GDP, population density, road network, and land use type, the spatial distribution of PM2.5 in each pollution sources of the inventory was allocated into 1 km×1 km grids. Finally, the simulation data of WRF-CMAQ and the measured data of ground stations were used in the validation. The results show that, firstly, the method of PM2.5 data fusion can effectively improve the temporal and spatial resolution of PM2.5 observational data, and it is significantly correlated with the observational data on the ground (R2=0.94, RMSE=4.64 µg/m 3, NMB=2%, NME=7%). Secondly, the concept of the quiescent period after favorable meteorological conditions is introduced, and the selection method of the quiescent period is established. The cross-region transmission of PM2.5 is effectively reduced, thus better reflecting the spatial distribution characteristics of local source emissions. Thirdly, the accuracy verification results based on WRF-CMAQ simulation method show that compared with the traditional area interpolation method, NME decreased 7%, NMB decreased 10%, RMSE decreased 1.54 µg/m 3, and R 2 increased 11%. This method provides a new idea for the spatial refinement of the emission inventory.

  • YANG Fei, HUA Yixin, LI Xiang, LI Po, YANG Zhenkai, CAO Yibing
    Download PDF ( ) HTML ( )   Knowledge map   Save

    The construction of new smart cities has been proposed as an important strategic goal in the 14th Five Year Plan of China. Information management of urban facilities lays a solid foundation for the construction of new smart cities, which relies on the standard, uniform, and detailed modeling and management of such facilities. Current solutions for urban facility modeling are mainly based on three models, i.e., spatial data model using maps as templates, traditional object-oriented spatial data model, and real-time GIS data model. However, limited by the traditional visualization approaches of computers, it is hard for spatial data model using maps as templates to describe and express the hierarchical structure, behavioral interaction, and other necessary information of urban facilities with the data organization method of “geometry & attributes”; traditional object-oriented spatial data model can be used to describe the geometry, attributes, and relationships of urban facilities, but the hierarchical structure and behavioral interaction are still not considered in this model; the object state of real-time GIS data model mainly records traditional spatial and thematic attributes of urban facilities, while lacking the description of the composition structure and behavioral capabilities of temporal and spatial objects. Therefore, all the three models are unable to support the comprehensive and microscopical expression of facility objects. Current related research of urban facility management can be seen mainly in the domains of computer, Internet of Things (IoT) or Geographic Information System (GIS). Research in the domains of computer and IoT mainly focuses on the management of the data generated by urban facilities, ignoring the management of the facilities' own information; real-time GIS in the domain of GIS realizes the management of the information of facility spatiotemporal entities by managing facility data instead of facilities themselves. For example, it expresses taxis and cameras with the real-time data stream of them. Therefore, it still cannot support the management of the facilities' own information. To solve the problems above, firstly the classification and coding of urban facilities was completed by extending the GBT30428.2-2013 standard; then an urban facility object description model based on the Multi-granularity Spatiotemporal Object Data Model (MGSTODM) was proposed, which supports comprehensive description and expression of urban facilities' information; furthermore, an urban facilities management method was proposed, with the design of the corresponding technical architecture and the implementation of the prototype system for the mega level urban facilities management based on cloud storage afterwards. Finally, the proposed object description model and management method of urban infrastructure was proved to be feasible, effective, and efficient.

  • FENG Yehan, CHEN Liang, HE Xiaodong
    Download PDF ( ) HTML ( )   Knowledge map   Save

    The Sky View Factor (SVF) is one of the most important indicators to characterize urban radiation fluxes and urban thermal environment. Therefore, it is a key morphological parameter to study the Urban Heat Island (UHI) effect. Studies have shown that SVF has a strong relationship with UHI intensity. Nevertheless, the relationships found can be contradictory. This is primarily due to the fact that the cases studied are often in different regions with different climatic conditions. In addition, the influences of trees are sometimes ignored due to the lack of vegetation data or the limitation of calculating methods. How to calculate SVF quickly and accurately is important to urban climate research. SVF is typically calculated by four types of methods: fisheye photo methods, 3D GIS methods, GPS methods, and street view image methods. Compared with the other types of methods, calculating SVF using street view images has many advantages, such as widely available data, low cost, high efficiency, and the ability to consider the influences of trees and other obstacles. On the one hand, street view images provide the possibility for fast and accurate calculation of SVF in large-scale areas. On the other hand, the street view image method is still at its developing stage and more work needs to be done to verify its application in various urban environments. In this study, we proposed an automatic SVF calculation method using street view images and deep learning algorithms, and then applied the method to the UHI study in the city center of Shanghai. Baidu static panoramas and Deeplabv3+ were used to detect sky range while MATLAB code was written to calculate SVF. A Landsat-8 OLI / TIRS image was also used to retrieve land surface temperature at street level in the study area. Based on the Local Climate Zones (LCZ) scheme, we combined large-scale SVF value with the land use and building morphology to examine the relationship between SVF and UHI intensity. The results showed that Deeplabv3+ can detect the sky and non-sky range effectively in different scenarios (MIOU=91.64%). The SVF calculated using the proposed method was in good agreement with that calculated using fish-eye photos (R2=0.8869). The LCZ scheme provides new insights for the relationship between SVF and UHI. For LCZ5 and LCZ1, the highest correlation coefficients were 0.68 and -0.79, respectively. The proposed method was shown to be applicable in high-density and complex urban environments. In addition, the calculation of large-scale continuous SVF provides the possibility for zonal understandings of the UHI effect based on the LCZ scheme.

  • HE Bin, WU Wenzhou, KANG Lu, SU Fenzhen
    Download PDF ( ) HTML ( )   Knowledge map   Save

    In recent years, with the continuous exploitation and utilization of marine resources, marine spatial planning has become more and more important, among which fishery resources account for the main proportion. In order to provide auxiliary information for the monitoring and planning of fishery resources, this paper obtained the 2018 Automatic Identification System data of the South China Sea and surrounding countries, extracted the activity intensity of fishing vessels and carried out preprocessing, sampling processing, and GIS spatial analysis, and then mathematically analyzed the spatial and temporal characteristics. The results showed that, firstly, in 2018, fishing vessels in the South China Sea and surrounding countries were mainly distributed regionally, concentrated in areas within 100 km of the coast of China and Vietnam. Fishing activities were frequent in autumn and November. The average activity intensity of fishing vessels was higher during the day than at night, with the maximum activity intensity at 16:00 PM; Secondly, the intensity of fishing activities in main ports of Guangdong, Guangxi, and Hainan provinces is clustered as dots, with the intensity of fishing activities bigger than 100. The sea area near some ports is striped, and the intensity of activities of other fishing vessels in the South China Sea is larger than that of other islands in the Paracel Islands. The activity intensity of fishing vessels is smaller than 2; Thirdly, the regional distribution of Vietnamese fishing activities is obvious, showing a stable mass clustering distribution in Ho Chi Minh Port, with little change in activity intensity throughout the year. The activity intensity of nearshore fishing vessels remains at 50~100. Vietnamese fishing vessels are banned in the South China Sea. There are two areas with strong activity in the southwestern part of Hainan Province within the fishing line. In 2018, days with fishing activity accounted for 87.71% of the total sampling days, with on average 7~10 fishing boats every hour in the area. During the moratorium period, the average number of boats every hour is bigger than 5, which poses a great threat to China's south China sea fishery resources. In this paper, AIS data research and analysis of fishing vessel activities can provide data support for marine spatial planning and relevant government departments.

  • ZHAO Quanhua, WANG Xiao, WANG Xuefeng, LI Yu
    Download PDF ( ) HTML ( )   Knowledge map   Save

    Under the attack of strong cold wave in winter, a large-scale freezing phenomenon appears in Liaodong Bay. In order to analyze the change rules of sea ice condition and the environmental influencing factors in Liaodong Bay from 2015 to 2020, Sentinel-1A/B data are selected to carry out sea ice monitoring in Liaodong Bay. First, the Gray Level Co-occurrence Matrix is used to count the texture features. Then the optimal feature combination is selected based on the Bhattacharyya Distance, and the Maximum Likelihood method is adopted to classify sea ice. Then, according to the classification results of sea ice, the sea ice condition levels of Liaodong Bay in recent 5 years are determined. The change rules of ice condition characters such as the outer edge of sea ice, area, type, and freezing probability are analyzed. Finally, the influence of sea water depth on sea ice condition is discussed, the relationship between sea ice condition and sea temperature, air temperature, and wind speed is studied through correlation analysis. The main conclusions are as follows. Firstly, the Sentinel-1B image on February 1, 2020 is tested using the proposed method and different texture feature combination methods. The results show that the overall classification accuracy and Kappa coefficient of the proposed method are 93.16% and 0.85, respectively, which is the highest classification accuracy among all methods. The overall classification accuracy and Kappa coefficient of all images from 2015 to 2020 are above 85% and 0.80, respectively, which meet the accuracy requirements for sea ice monitoring. Secondly, in last November and December, the sea ice types are mainly primary ice, with gray ice in between. Gray ice is the main ice in January and early February, with primary and white ice in between. Gray and primary ice is dominant in late February and early March. There are differences in the sea ice freezing probability in Liaodong Bay. The sea ice freezing probability in the north coast is higher than that in the south, and the probability in the east is higher than that in the west. Sea depth in Liaodong Bay has different effects on sea ice development in different ranges of [-10, 0], [-20, -10], and [-30, -20]. The Pearson correlation coefficients of sea ice condition with sea temperature, air temperature, and wind speed in Liaodong Bay are -0.55 (P<0.01)、-0.59 (P<0.01)、and -0.22 (P=0.19), indicating that the sea ice condition is negatively correlated with sea temperature and air temperature, and has a low correlation with wind speed.

  • MAI Jianfeng, XIAN Yuyang, LIU Guilin
    Download PDF ( ) HTML ( )   Knowledge map   Save

    As a natural disaster, rainfall-triggered landslide causes tremendous losses to mankind and then seriously affects living environments of mankind, especially in highly economically developed urban agglomeration areas. Many scholars have carried out data collection and related researches from the perspectives of real-time monitoring, process mechanism analysis, risk assessment, and prediction. However, the abovementioned studies lack the prediction of the potential distribution of landslides from the perspective of climate change. Therefore, we employed the Maximum Entropy (MaxEnt) model to simulate the current distribution of potential rainfall-triggered landslides combined with a series of indicators related to precipitation, including geology, topography, vegetation cover, the precipitation of April, precipitation of May, precipitation of July, precipitation of wettest quarter, Enhanced Vegetation Index (EVI), elevation, slope, lithology, and distance to fault. Then we revealed different effects of those influencing factors on the spatial distribution of landslides in Guangdong Province. Finally, we predicted the future potential distribution of rainfall-triggered landslides under the Shared Socio-economic Pathways (SSPs) scenarios from 2021 to 2060. The results show that the average AUC (Area under the receiving operator curve) value of the model simulation was 0.820, exceeding the standard of "very accurate" simulation effect, and the Kappa coefficient was also 0.823 after 10 repeated simulations. We found that precipitation of wettest quarter, precipitation of July, precipitation of April, and elevation significantly affected the distribution of landslides, in which the wettest season rainfall was the most contributing factor. Specifically, the precipitation of wettest quarter between 593~742 mm, precipitation of July between 139~223 mm, precipitation of April between 154~186 mm, and elevation between 81~397 m were highly correlated to the occurrence of rainfall-triggered landslides. Currently, the area of high risk of rainfall-triggered landslides in Guangdong Province was 1.28×104 km2, accounting for 7.59% of Guangdong Province. Spatially, it is mainly distributed in the eastern part of Guangdong Province. However, under three future SSPs scenarios (SSP1-2.6, SSP3-7.0, and SSP5-8.5) of two periods (i.e., 2021—2040, 2041—2060), the potential distribution range and harm of the areas above the risk in Guangdong Province have shown an expansion trend. The area increased the most under the SSP1-2.6 scenario during the period of 2041—2060. The simulated future high-risk distribution of landslide had potential harm to Guangdong-Hong Kong-Macao Greater Bay Area and Eastern Guangdong urban agglomeration. The findings can provide scientific evidence for the future smart sustainable territory development plan from the perspective of prediction of landslide distribution.

  • WEN Chao, ZHAN Qingming, FAN Zhiyu, ZHAN De, ZHAO Huang, WU Kai
    Download PDF ( ) HTML ( )   Knowledge map   Save

    Urban development has a great impact on water bodies in many aspects, especially for cities rich in water resources. Thus, it is urgent to carry out relevant monitoring and research. This study takes Wuhan as study area. Based on relevant studies at home and abroad, as well as Wuhan water protection policies, this study put forward a technical route, integrating water area, water quality, water landscape, and waterfront ecology, to reflect the spatiotemporal characteristics of urban scale water more comprehensively. Specifically, the random forest model was used to obtain water information of 67 key water bodies from remote sensing images from 1979 to 2019. Based on the information, the changes of water area and water landscape were analyzed. The evolution characteristics of water quality and waterfront ecology were analyzed based on water quality monitoring data and Remote Sensing Ecological Index (RSEI), respectively. Then, the influencing factors of the changes in water body area were analyzed using the Multi-scale Geographical Weighted Regression model (MGWR). This study aims to provide scientific support for the government to formulate differentiated water protection policies, and provide useful reference for the multi-factor analysis of water bodies in other regions. The results showed that the total water areas of Wuhan and 67 key water bodies had decreased by 10.75% and 13.12%, respectively. There were significant differences in the changes of water bodies in the downtown area and the suburban area. The water landscape showed a degradation trend. The Perimeter Area Fractal Dimension (PAFRAC), Edge Density(ED), Mean Patch Area(MPA), Aggregation Index(AI), and Cohesion Index (COHESION) decreased by 6.43%, 79.35%, 1.55%, and 10.94%, respectively. The water quality of Wuhan was deteriorating. Most of the rivers and reservoirs are of class III or above all the year round. Most of the lakes in the downtown area are of class V or below. Most of the lakes in the suburbs are of class IV or V. Changes of RSEI of the waterfront in the downtown area and the suburb area showed that the eco-environment of the waterfront zone was recovering. The average RSEI of waterfront in the downtown area had increased by 14.29%. MGWR analysis showed that among the natural factors, the increase of relative humidity had more effects on the recovery of lakes in Jiangxia, while the increase of precipitation had a more significant impact on the recovery of lakes with smaller water area. Among the socioeconomic factors, the increase of GDP in each administrative region was helpful to water restoration, especially for water bodies in the downtown area, Huangpi, and Xinzhou. The increase of Impervious Surface (IS) proportion in waterfront area had led to the shrinkage of most water bodies. For a few key recovering water bodies, the growth of IS due to relevant protection policies had a positive effect on the recovery of these bodies.

  • CHU Guozhong, LI Mengmeng, WANG Xiaoqin
    Download PDF ( ) HTML ( )   Knowledge map   Save

    Urban building type information is crucial to many urban applications such as the identification of urban functional areas and estimation of urban environmental variables. This paper presents a new method to extract urban building types using multi-scale features and integrating height features derived from high resolution remote sensing images. We first conduct an image semantic segmentation to extract building and shadow objects from remote sensing images, and then estimate the height of buildings based upon the directional relationship of a building object and its shadow information. Following multi-scale image analysis concept, we extract a series of multi-scale features regarding the height, geometry, and spatial structure of building objects. Last, we use a machine learning method based upon random forest to classify building types. We also analyze the impact of different spatial units of building types on classification results. Experiments were conducted in Fuzhou, Fujian province, China, using a Chinese GF-2 satellite images acquired on February 18, 2020. Our results show that: (1) The overall accuracy of building type classification combined with multi-scale features reached 82.98%, and the kappa coefficient was 0.77, which was better than other conventional methods, namely a Multi-scale Classification Without Height Features (MCNH), a Single-scale Classification Incorporating Height Features (SC), and a Single-scale Classification Without Height Features (SCNH) in this paper; (2) The classification accuracy of middle-low residential buildings and high-rise commercial and residential buildings was improved by adding height features. Compared with classification results without using height features, the overall accuracy was improved by 11.28%; (3) The fusion of image features at multiple scales can reduce the misclassification of adjacent buildings into dense buildings. Compared with a single-scale classification method, the proposed method improved overall accuracy by 2.77%. We conclude that the use of high-resolution remote sensing images provides an effective strategy to estimate building heights based upon shadow information and improves the classification accuracy of urban building types, particularly when detailed digital surface model data are absent. In addition, the fusion of multi-scale image features can improve the characterization of complex building types in urban areas and the subsequent classification accuracy accordingly. Nevertheless, we also observed that the results of classified building types were affected by the initial extraction of building information from high resolution remote sensing images, implying that a further improvement of building type classification can be done by improving the extraction methods, e.g., using a more advanced semantic segmentation model.

  • LU Dajin, LI Dong, ZHU Xiaoxiao, NIE Sheng, ZHOU Guoqing, ZHANG Xingyi, YANG Chao
    Download PDF ( ) HTML ( )   Knowledge map   Save

    ICESat-2 (Ice, Cloud, and land Elevation Satellite-2) launched by NASA (National Aeronautics and Space Administration) in 2018 is a laser altitude measurement satellite. The advanced topographic laser altimeter system (ATLAS) instrument on-board ICESat-2 employs a micro-pulse and multi-beam photon counting laser altimeter system with low energy consumption, high detection sensitivity, and high repetition rates, and thus greatly improves the sampling density in the along-track distance. However, it introduces a significant number of solar noise photons in the raw data. How to effectively remove the noise photons and classify the signal photons into ground photons and canopy photons is critical for subsequent applications such as the estimation of terrain elevation and forest height, and it has been a hot and challenging topic in the current research. In this paper, a denoising and classification algorithm based on convolutional neural network was proposed. The convolutional neural network has made a series of breakthrough research results in the fields of image classification, object detection, semantic segmentation, and so on. To remove obvious noise photons, the photons were first divided into grids in the along-track distance and elevation direction, and the rough signal photons were gridded into pictures. Then, the convolutional neural network was employed to perform the final denoising and classification. Finally, the proposed algorithm was tested with the airborne LiDAR datasets, including DSM (Digital Surface Model) and DTM (Digital Terrain Model), and was further compared with ATL08 (land and vegetation height) products. Experimental results show that our proposed algorithm can remove noise photons effectively in bare land and forest areas. Moreover, this algorithm can simultaneously remove noise photons and classify signal photons into ground photons and canopy photons in forest areas. The R 2 and RMSE values of the retrieved ground surface in the bare land areas were 1.0 and 0.72 m, respectively. In the forest areas, the R 2 of the estimated ground surface and canopy surface were 1.0 and 0.70 with the RMSE values of 1.11 m and 4.99 m, respectively. The reason for this result may be that it is difficult for photons to penetrate the forest canopy and reach the ground surface in forest areas, which causes the RMSE value of the forest area to be larger than that of the bare land area. In this paper, the deep learning algorithm was used to realize the denoising and classification of photon counting data, and good results were achieved in bare land and forest areas, which provides a reference for subsequent photon counting LiDAR data processing.