A comprehensive and up-to-date web-based interactive 3D emergency response and visualization system using Cesium Digital Earth: taking landslide disaster as an example

ABSTRACT As with the fast advances in the technologies of big Earth data and information communication, Web-based 3D GIS system has come a long way from a few years ago. These advances reflect in many aspects of 3D GIS such as higher real-time performance, enhanced interactivity, more realistic 3D visualization effect and improved user interface. This paper aims to present a comprehensive and up-to-date 3D Web GIS for Emergency Response using the current vue.js web application framework and the well-known Cesium API, taking landslide disaster as an example. Building upon recent advances in WebGL technology, we developed a suite of enhanced 3D spatial analysis functions, including interactive route planning, instant text/image/video messaging being incorporated into both 3D WebGL page and mobile GIS applications, and progressive 3D construction and AR visualization using LiDAR and camera over local emergency network or internet. Moreover, professional functions such as landslide susceptibility mapping, landslide monitoring, spatial temporal contingency plan management, landslide information management, personnel and equipment management, and communication are all implemented and integrated in the 3D GIS system. Most of the functions of the system are implemented using open-source projects, which is beneficial to the development of the 3D GIS research community.


Introduction
After the occurrence of natural disasters, decision-making based on spatial and nonspatial big data to achieve rapid emergency response and rescue is a key factor in reducing disaster losses.Geographic Information System (GIS) technology has been the key part of disaster emergency management system, which helps decision makers to quickly integrate, process and analyze disaster data (Abdalla & Li, 2010;Assilzadeh et al., 2010;Feng & Cui, 2021).The intuitive and interactive 3D visualization of multi-source geographic big data using modern 3D GIS technology (e.g.Cesium and Google Earth) can optimize realistic and dynamic expression of disaster scenes, making professional geological knowledge more easily understood by decision makers and the general public, and greatly improving the response speed and rescue capability of emergency services (Kang et al., 2018;Yu et al., 2016).
For example, landslide is one of the common natural disasters worldwide, which is widely distributed, sudden and easy to trigger secondary disasters, and can cause serious damage to human life, property, and natural environment (Petley, 2012;Pollock & Wartman, 2020).According to Emergency Event Database (EM-DAT), at least 5 million people from 65 countries have suffered from landslide disasters between 2000 and 2022, and 18,965 of them have lost their lives due to the disasters.The use of 3D GIS technology to provide a powerful display, analysis, and decision-making tool for decision and rescue personnel in landslide disaster scenarios is of great significance for disaster emergency response and disaster loss reduction.Liu et al. (2010) integrated multi-source geohazard data based on ArcGIS, GeoView, and other platforms to generate virtual 3D geohazard maps to provide data and functional support for early warning and decision-making systems.Traditional Client-Server (CS) architecture of the emergency response system has high development cost and is difficult to provide diversified services for disaster relief management as well as complicated system deployment in emergency situations, so the practical application is thus limited.
Web services-based WebGIS systems are increasingly used in emergency control and rescue systems because of the advantages of cross-platform, high scalability, and ease of integration (Li et al., 2015;Souza et al., 2017).Huang et al. (2016) integrated sensor data, geospatial data, and landslide forecasting models to develop a WebGIS system for landslide detection and forecasting.Chen et al. (2016) proposed a landslide multi-level management and emergency response platform using ArcGIS Flexviewer and Skyline ActiveX components, which jointly used 2D and 3D WebGIS to generate an integrated Web environment for decision makers for display and analysis.Hou et al. (2017) designed a geological hazard monitoring database which are used for storage and access management of spatial data, and developed an image analysis and processing model in ArcGIS Server, facilitating the monitoring of landslide phases and information management.
However, the Web end has its inherent limitations, one is that it relies on reliable and stable network communication conditions, which is difficult to guarantee at disaster sites where infrastructure is damaged; the other is that the limited computing resources of the Web side also limit the 3D visualization capability of large-scale geographic data and various interactive functions.The popularity and large number of applications of modern information technology, while bringing us rich big Earth data, also challenge the integration capability of different data services of the system (Guo et al., 2014;Li et al., 2022).And with the successive emergence of new concepts such as metaverse (Allam et al., 2022), virtual reality (VR) and augmented reality (AR) (Hu et al., 2018;Itamiya, 2021), the visualization details and realism of virtual scenes and rendering efficiency are also facing new requirements (Fu et al., 2022).Therefore, how to provide more comprehensive, realistic, and more effective functions and services for landslide management and emergency response while maintaining the characteristics of 3D WebGIS platform such as lightweight and easy deployment, and real-time and efficient performance, is the main problem faced by the current Web-based 3D emergency response system, and one of the problems we must face in moving from digital twin system to metaverse.
In this work, the up-to-date information technology is applied to the landslide disaster information management, and command and control system for emergency response, implementing real-time monitoring, information management, and visual display of the three-level command and control information (i.e.frontline rescuers, frontline command centers, and the headquarter).Our system is designed primarily for Search and Rescue (SAR) after landslides, as well as monitoring existing landslide sites to prevent secondary disasters.We use mobile communicating technology, Internet of Things (IoT) and sensor network to monitor the real-time condition of various physical entities in emergency response scenarios such as rescue workers, vehicles, and landslide sites, and then map them to digital twins of the disaster site for integrated visualization and management.
As shown in Table 1, the system provides corresponding Web services for different user groups to ensure the collaboration and consistency of user groups at all levels of personnel in all aspects of emergency response, thus serving more efficient and timely emergency response and rescue.Our system has the following highlights compared to other studies.(1) The use of the state-of-the art information technology and system framework seamlessly integrate multi-source spatial and non-spatial data service as the focus of system development, providing good scalability and practicality.(2) Communication hardware and software functions between different user groups at all command and control levels for emergency response are incorporated as an important part of the developed system, and we focus on interaction and information sharing of all user groups at different levels, solving the problem of lacking key rescue information at upper and lower levels in disaster scenarios.
(3) Building upon recent advances in WebGL technology and Cesium Digital Earth APIs, we developed a suite of enhanced 3D spatial analysis functions, including interactive and intelligent navigation, instant messaging being incorporated into both 3D WebGL page and mobile application, and progressive 3D construction using portable LiDAR and camera through transmission over local network or internet.Note that most of the functions of the system are implemented using open-source projects, which is beneficial to the development of the community.
In the following sections, we first describe the general framework of the system.Next, we illustrate the key technologies and workflows developed in each module of the system, showing the advantages of our system especially in the integration of emergency services.The paper then demonstrates the results of the application of the system in Wanzhou District, Chongqing province, China, to validate and test the features of the system.Finally, we summarize the characteristics of Web-based 3D interactive visualization technology for landslide disaster response and the outlook for future development.

System design and framework
We designed and developed a landslide emergency response, command and control system using the Vue.js 1 web application framework (i.e.JavaScript framework for building user interfaces).Figure 1 shows the main structure of our system and the key technologies used.The system is divided into three modules for all levels of users, including landslide monitoring and visualization, emergency rescue and control, and disaster scenario integration and decision support.Most of the functions and services of the system are based on the state-of the art open-source projects as shown in Table 1.The combined use and application of these technologies aims to improve the modernity and usefulness of our disaster response system and to improve natural disaster emergency response in virtual and augmented reality, and the digital twin environments of the disaster site.

Disaster rescue command center
Disaster rescue command center uses Web-based disaster response and visualization system.Each function of the system is developed as a separate component of Vue, thus increasing the reusability of the functions and efficiently organizing the many task functions of the system.Another advantage of using Vue framework is its bi-directional data binding and virtual DOM, the former can ensure the synchronization of the data layer and view layer real-time updates, while the latter significantly improves the operational efficiency of the Web platform (js, 2022).In addition, the Cesium Digital Earth library is used to implement the map visualization at the core of the system.Cesium has been rapidly developing in the past decade to support more types of 3D geospatial data (e.g.glTF models and 3D Tiles) and uses WebGL for hardware-accelerated graphics to render each geographic element into virtual terrain scenes efficiently with high quality (Cesium, 2022;Zheng et al., 2019).Based on the above technologies and their new features, we designed the service layer and front-end layer of each module.Note that the joint use of Vue and Cesium makes the system have rich 3D interactive visualization effects, and greatly reduces the difficulty of integrating different data services.The real-world information communication and integration within the disaster rescue command center is synchronized and visualized on the system, enabling interactive 3D perception of the disaster-related information and providing a more powerful analysis and display environment for emergency command and control as well as decision-making.Applications and Web system in front-end layer access resources and invoke services in application services layer through unified and standardized interfaces based on RESTful architecture.

Rescue workers and equipment
Rescue workers at disaster sites can also use the Web platform to learn about regional disaster situation and integrated disaster scenarios through mobile phones or laptops.In SAR, we developed an instant messaging Android app and adopt short-wave radio stations to ensure smooth and efficient communication between the units at all levels of personnel as well as sharing rescue task information.Smart bracelets and other sensors such as UAV, LiDAR and camera are used to monitor individual health condition and collect on-site disaster data.In addition to the web platform and mobile applications, AR devices are also used to provide augmented landslide information for rescue workers assisting frontline rescue tasks.

Geographic data
Our system supports multi-source and heterogeneous geographic data which includes: (1) road network data and Point of Interest (POI) data about local infrastructures; (2) geographic data related to SAR such as temporary shelter distribution, evacuation routes, and affected areas; (3) remote-sensing images, terrain models, and disaster thematic maps.
Images and terrain models are stored through commercial Object Storage Service (OSS).OSS supports low-cost, secure, and reliable storage of various types of data and can be accessed efficiently through RESTful interfaces.Other geographic data are stored in the server's PostgreSQL database and we use PostGIS extension to support geographic objects and spatial analysis.

Emergency response data
This includes all data that are generated during the emergency response such as rescue workers condition information, contingency plans, and sensor data at disaster sites.Most of them are stored in MySQL databases which are designed and tuned to meet the needs of different data types.The data that takes up more storage space is stored on server disk (e.g.image, video and LiDAR point cloud).Temporary dynamic data generated by users while using the system or application services are stored in memory.

Instant messaging data
Instant messaging data include user data, ordinary text messages, and media messages.The first two are stored in MySQL database while media messages, which are usually large, are stored on server disk as files and the related file paths are stored in database.

Evaluation of landslide hazard susceptibility
Landslide disasters are highly destructive and prone to secondary disasters, so estimating the probability of landslides at different locations in the region is extremely necessary for both pre-disaster warning and post-disaster rescue.In practical applications, sensors are usually deployed in landslide-prone areas to monitor the displacement, stress, groundwater, and other parameters in real time to predict the probability of landslide occurrence (Pecoraro et al., 2019;Qiang, 2020).This method can provide accurate prediction results, but it is difficult to be applied to large-scale study areas (Tengtrairat et al., 2021).
We use a deep learning-based landslide hazard susceptibility evaluation model that can predict the probability of landslides occurring at a point in a region from multiple sources of spatio-temporal big data such as digital elevation models (DEMs), hydrometeorology, satellite remote sensing images, and human engineering activities in the study area.The model provides more accurate long-term prediction values using easily accessible geographic data, which can be easily deployed as a Web service to provide decision support for disaster management and emergency relief.Since the deep learning model produces only point prediction results, kriging interpolation is used to interpolate all points in a certain area to generate a raster map in geotiff format, and then use GeoServer to publish Web Map Tile Service (WMTS) after selecting a suitable map rendering scheme to provide a web visualization platform for landslide hazard susceptibility evaluation.The Web Map Tile Service (WMTS) is published using GeoServer to provide the Web visualization system with landslide hazard susceptibility evaluation service.

Progressive 3D modeling through emergency local IP network
It is often difficult for experts and decision makers to truly grasp the actual situation of the disaster site only through two-dimensional information such as photos and remote sensing images, and the construction of virtual geographic disaster environment can provide a three-dimensional interactive real perceptual space, which can better support disaster acquisition and decision research and judgment (Denolle et al., 2014).However, the dynamic changes of landslide disaster are complex, and the scene animation simulation and visualization are difficult and inefficient (Fu et al., 2022).Using Simultaneous Localization and Mapping (SLAM) technology for 3D reconstruction of large landslide scenes not only achieve accurate mapping of landslides on digital earth systems but also improve the automation of modeling work.Figure 2 shows the workflow of our progressive front-end to back-end 3D modeling of landslides.
3D printing technology helps us design a low-cost, lightweight, and portable handheld LiDAR and camera-integrated acquisition device for emergency rescue workers, as shown in Figure 2. The rescue workers are often the first to reach the disaster site after a disaster, and they can use this device to achieve fast and convenient 3D data acquisition.Since the limited computing power of the data acquisition equipment is difficult to support the 3D construction of large scenes, the multi-machine communication technology of the Linux ROS system is applied to guarantee the consistency of data acquisition and 3D map construction at the front and back ends.The RGB images and LiDAR point cloud data are transmitted to the front-line command vehicle and command center synchronously over a dedicated IP network being composed of short-wave communication radio stations, and the back-end server computing resources.
We collect GNSS coordinates of ground control points, and use geographic alignment methods to georeference the landslide model.After converting the model data format to 3DTiles format, the new features of Cesium library can be used for 3D interactive visualization of the landslide on the Web.The oblique imagery collected by the UAVs can be triangulated to obtain an oblique photographic model.Based on this data, we carry out the coarse alignment of the aerial model and ground model by Iterative Closest Point (ICP) algorithm, and then manually select the control point pairs to determine the coordinate transformation relationship between the two models and complete the fine alignment of the aerial and ground model.The above process can achieve 3D reconstruction of large landslides using air-ground integrated approach, which greatly enriches the visualization effect of landslide model from multiple perspectives.

Emergency communication network
Landslides are often accompanied by other natural disasters such as earthquakes, floods, and mudslides.Thus, regional power and communication facilities can be severely damaged, seriously affecting the collaborative efficiency of disaster command and rescue (Luo et al., 2020).Therefore, we design an emergency communication network as shown in Figure 3  communication equipment for the system based on short-wave radio communication is used to support the rapid setup of a temporary LAN between the front-line rescue personnel and the intermediate center.The front-line center can transmit data to the headquarter through communication satellites or temporary base stations, thus guaranteeing smooth network communication between all levels of personnel in an emergency.
Our system is equipped with instant messaging (IM) module called WildFireChat 2 (WildFireChat, 2022), and is deployed in C/S and B/S architectures on portable devices including rescue worker's cell phones or watches, and the frontline command and control center.The frontline emergency command and control 3D web system can transmit audio, video, and text data to the rescue worker's App on the mobile device through private network communication, and the rescue team is also able to use the group chat function for efficient collaboration in the field.On the central platform, we use the locations of frontline personnel and vehicles as the hotspots, and use the form of a status card to integrate communication data, individual health monitoring data and 3D geographic information data for visualization.The tight coupling of modern IM technology with 3D GIS can provide the command center with a more fine-grained and accurate reference basis for disaster emergency response and decision-making.
As shown in Figure 3, the system also supports real-time transmission and live streaming of video from the UAVs and other devices.Nginx proxy server for video stream pulling, conversion, and distribution is used to support a variety of video protocols such as RTMP and HttpFlv.The Web system uses components developed based on the Video.jslibrary to load real-time video and play it.Both front and back end system only need to go through a simple operation to make the command center synchronously view the camera video images transmitted by the UAV, command vehicles, and other devices.

Interactive route planning
The system also provides a solution for 3D interactive path planning for the needs of personnel mobilization and resource distribution in emergency disasters (Mantoro et al., 2021).An open-source project Graphhopper routing engine 3 is used to provide Brower/ Server-based path planning and navigation services (GraphHopper, 2022).Our experimental road network data come from the open map OpenStreetMap.Graphhopper can perform fast and efficient path planning using heuristic algorithms such as Contraction Hierarchies (CH) and A* algorithms (Geisberger et al., 2008;Hart et al., 1968).We deploy Graphhopper and road network data on the cloud server, and the front end encapsulates the interactive visualization function of path planning as a Vue component, which requests the service in real time by calling the API interface.The component also makes the calling interface accessible to other modules to support the saving and exporting of route paths in order to link with the functions of contingency plan management.
By drawing point-line surface elements on the Cesium layer to express the start and end points and via points, planning routes and areas to avoid, rescue workers can visually find the optimal path to the disaster site on the 3D map, and easily view the terrain along the way and interact in real time.Compared with traditional 2D map-based path planning, our method improves the dynamics, interactivity, and visualization of geolocation services in disaster emergency response.The third-party map data and path planning services are well coupled together in the Vue Web framework, which can provide more professional geolocation services for emergency response and is more in line with the development trend of big Earth data information system.

Landslide AR visualization
The 3D LiDAR sensor can obtain dense point cloud for interpreting, processing, and analyzing the landslide model, the geology experts can then guide the front-line emergency rescue and on-site rescue work in a targeted manner based on their professional knowledge, such as interpreting the area and scope of the landslide and directing the deployment of monitor sensors.The more intuitive AR technology is used in this part to make up for the shortage of traditional text or audio-video communication in front-line operation guidance through the method of virtual-real integration and mixed reality visualization (Dini & Mura, 2015;Li et al., 2022).
As shown in Figure 2, after collecting the data from the LiDAR and camera, we use the method described in (Shi et al., 2022) to build up a visual feature database, while recording the positional information of the visual feature points in the LiDAR point cloud map.The analysis decision results from the geology experts will be edited on the landslide 3D model directly through the well-known game engine Unity software to produce AR visualization content.With the above processing procedure, frontline rescue workers are able to use AR glasses and edge computing devices to accurately match the real images captured by the camera with virtual landslide information and AR content overlaid in the real scene.Due to the deep-learning based feature extraction method, which are lightweight, real-time, and robust at night (Shi et al., 2022), our AR function is able to overcome the complex disaster environments and thus operate under extreme lighting conditions, such as day and night, greatly expanding the utility of this function in disaster emergency scenarios.

Disaster scenarios integration and visualization
Our system mainly integrates the following scenarios: dispatch and management of relief forces and supplies, realistic, and real-time visualization of the disaster situation, monitoring of the disaster sites and health condition of rescue workers, management, and direction of SAR tasks.In this way, rapidly summarizing and displaying various disaster information in the disaster area helps managers and decision makers to accurately control the disaster situation in order to efficiently direct the relief work.Using the Cesium JS library, the system supports 3D mapping and visualization of various types of geospatial data on maps or remote sensing images.For example, mapping the extent and affected areas of the landslide and the distribution of rescue efforts on a 3D map will help the headquarter gain a more complete understanding of the overall situation.Other information such as meteorological data and casualties, after data processing and synthesis, can be displayed on the system interface in the form of statistical charts, tables, live video, etc.By organizing and presenting the above two types of information in a reasonable manner, our system can achieve a multi-source data and information integration of the disaster scenarios.
Our WebGL-based system is built from the latest information technology, so it can effectively utilize hardware computing resources.Compared with the previous results, the system can achieve faster response and real-time loading and display of massive spatial big data.To solve the problem that the traditional system does not have a high degree of integration between two and three dimensions (Chen et al., 2016), we display 2D maps and 3D scenes simultaneously in the same screen partition.Synchronized user interaction operations in the two scenes enable flexible switching and linkage of different dimensions at multiple scales in a unified interface.

Contingency plan and management
Based on the comprehensive service functions and powerful data organization and visualization capabilities, we can build up a flexible and comprehensive disaster emergency plan within the system.The emergency response plan is a set of pre-defined disaster response tasks agreed in advance by the command center and experts in consultation before a disaster occurs.The system will be able to call Cesium map and Vue application components for operational linkage of the task content of each action node preset for specific rescue actions, thus facilitating rapid emergency response and decision-making.The contingency plan records the rescue plan, data visualization information, system component data, etc.The system directly loads the relevant modules after starting the contingency plan.Through tight coupling at the system level, our system can achieve faster response and easier disaster management.In addition, a timeline management function is designed for the contingency plan, which allows to view specific information of each task in the plan execution process in real time, thus effectively strengthening the disaster process management capability.

Study introduction
We selected Wanzhou district in Chongqing, China, as the study area for system deployment and also conducted experiments on 3D modeling techniques at the Huangyukou landslide in Beijing, as shown in Figure 4. Wanzhou district is the second largest district in Chongqing, located in the hinterland of the Three Gorges reservoir area, with a complex geological environment and frequent geological disasters in the region, among which landslide disasters occur most frequently and have the greatest impact (Wang et al., 2020).
The Huangyukou landslide in Yanqing District, Beijing, is caused by human engineering activities and rainfall that lead to slope instability.The landslide topography is undulating and the volume of the landslide body is large, which is suitable for 3D modeling and AR visualization testing (Zhao, 2020).
In this case, we conducted a simulation experiment in the areas according to the emergency response process after the occurrence of landslide disaster.In practice, rescue workers at disaster sites need to collect disaster data and practice rescue work such as cleaning up landslide body, laying sensors, searching for survivors and setting up camps.Based on the disaster situation, the headquarter commands and dispatches all rescue  forces and resources as well as making decisions and guiding the frontline relief work to ensure efficient and coordinated SAR activities.Frontline command centers have to implement the decisions made by the headquarter and transmit all types of disaster data between the frontline and the headquarter in an emergency.Most of the data in the system are generated or collected during the experiment such as IM data, health condition of the workers and meteorological data while thematic data and other static data (e.g.road network data, contingency plan and disaster situation) are prepared in advance.

Disaster scenario integration and visualization
Figure 5 shows the 3D comprehensive and integrated disaster scenarios map produced by our system after the disaster based on the data of the disaster area in the database and aggregating the data from other multiple sources.Figure 5(a) represents comprehensive information including the distribution of affected people in the disaster area, crowd gathering distribution, building information, road information, real-time distribution locations of rescue parties, etc. Figure 5(a,b) demonstrates the support and visualization effect of the system for various multi-source geospatial data.Remote sensing images, raster, vector geometry, 3D model, layer annotation, tilt photography and other disaster data can be efficiently visualized and organized for visualization and management.Scenarios integration and visualization module.(Continued).
Therefore, through the overlay display of multi-source spatial-temporal geographic data on the Cesium map, a thematic map of the disaster area can be quickly produced.
The system summarizes various disaster information in the region and can be updated in real time according to the new incoming data, thus assisting crowd evacuation, resources allocation, emergency rescue and other actions, and guaranteeing decision support for decision makers, and command and control management.The contingency plan function shown in Figure 5(d) can load the disaster rescue plan in advance.Once the plan is started, it can quickly start a series of rescue work and display the corresponding key information on the map.Dynamic adjustment of the timeline allows direct viewing of the current work progress and the status of the disaster at different stages, thus realizing the intuitive perception of the disaster situation.The system also supports the linkage of 2D and 3D visualization display as shown in Figure 5(e), through screen partition, displaying 2D and 3D scenes, respectively, in one screen, and can interactively control the perspective, scale and other parameters.This function is useful for more realistic visualization and analysis of multi-source geo-temporal big data.

Emergency rescue command and control management
We integrate and visualize the disaster monitoring data, communication network information, and geographic location data of the rescue worker, as shown in Figure 6(a).The system can monitor in real time and in the form of cards to a single person's network status, life status, geographic location, camera real-time video, and other information for the command center to coordinate management and rescue.Emergency rescue and command management module.(Contiuned).
Figure 6(e) shows the interactive path planning service, by setting the starting place, destination can be then intelligently calculated and recommended to produce multiple optimal routes, and evaluate the estimated time spent.In the actual rescue operation, personnel mobilization and resource distribution need to bypass the disaster area, and through a series of necessary passing points on the way.The route planning tool enables a simple click shown in Figure 6(e) to set the avoidance area and passing points to meet the need of path planning under a real disaster.Utilizing advanced technology development such as Vue and Cesium, not only allows us to integrate more data services on the system but also to support us to develop a real-time and interactive visualization functions.
The system can manage task information according to different units and stages of rescue work, while tracking and recording task execution and field feedback information, thus implementing continuous command and control and realizing interactive sensing and information sharing of all users at different levels.As shown in Figure 6(b-d), after the command center starts the task, the WebGIS map dynamically displays the scene image as well as human and vehicle trajectory, and the stakeholders can click on the terminal icon to conveniently view the scene video and trajectory information, and use the instant  messaging module to communicate the disaster situation, assign rescue tasks, and guide front-line operations.

3D modeling and AR visualization
The test experiment of the AR functional module is carried out at the Beijing Huangyukou landslide.We walked about 250 m around the landslide body with portable LiDAR and camera data acquisition equipment, and completed the 3D reconstruction in the command vehicle with the help of LAN communication of radio network, as shown in Figure 7(a).Figure 7(b) displays air-ground integrated 3D modeling of the landslide, where the landslide model constructed by in-field portable equipment is precisely seamless merged with the UAV oblique photographic model after geographic alignment.This helps experts and decision makers to get a more complete and immersive view of the actual disaster situation.
The constructed landslide model was decoded and a visual feature library was constructed in the indoor operation, and information such as observation stations and landslide areas were manually marked.Finally, AR tests were conducted using HoloLens 2 AR glasses.After the evaluation of GNSS ground control points, the accuracy of our map construction and visual positioning accuracy are shown in Table 2. Figure 7(c,d) show our AR visualization results, and you can see that our method has been able to be maturely applied to the disaster site.Under poorly lit and twilight conditions, our system is still able to accurately and robustly localize and visualize landslide AR content.

Conclusion
After disaster, the fusion and 3D interactive visualization of various types of big Earth data can enable decision makers and the general public to understand the expert knowledge intuitively, thus improving the response speed and decision-making ability of emergency response, rescue command and control.Currently, Web GIS systems with Web services are widely used, but with the fast development of big Earth data and modern mobile communication technologies, it is difficult to meet the demand for large-scale integration of big data and function services.Our emergency response system based on Cesium and Vue integrates a variety of new information technologies, which can provide more comprehensive, faster and more convenient Web services and 2-3D visualization effects in disaster emergency situation.In addition, most of the services we developed are based on open-source projects and easy to reproduce.Secondly, the system coupled the communication and location data in rescue with other multimedia and disaster information to improve the collaboration between various rescue groups in emergency situations.
Our experiments in Wanzhou District and Huangyukou landslide demonstrate that our system has stronger interactive visualization ability and richer functionality than traditional systems, which greatly improves in practicality, modernity, and realism.
The system has made a further step towards the application of natural disaster response and emergency management in the era of VR, AR and meta-universe.However, our 3D WebGIS system encounters low frame rate and high hardware resources consumption when loading and visualizing complex large scenes and 3D models.And data acquisition and processing are not fully automated and only partially integrated into the system and thus often require manual operations.In the future, we will continue to explore related state-of-the-art technologies and try to apply WebGPU to our system, progressing towards the next-generation VR/AR or meta-universe system for emergency response.

Notes
Jianhua Gong is a professor at the Aerospace Information Research Institute, Chinese Academy of Sciences, and Director of the Division of Virtual Geographic Environments.His research areas include cartography and geovisualization, virtual geographical environment, health GIS, and computable man-earth relationship.He originally proposed and developed the concept and framework of virtual geographic environment (VGE), which has become an important frontier research field of geographic information science.He has published three monographs and more than 160 articles (over 60 included in the SCI) in national and international journals and conferences.The Virtual Geographic Environments Division he leads is dedicated to 3-D GIS and geovisualization, visual analytics of spatial-temporal big data, and virtual and augmented reality.

Figure 1 .
Figure 1.Landslide disaster emergency response, command and control system framework.
. Rescue workers and monitor equipment at the landslide sites acquire hazard information which are then transmitted to the intermediate data center through Local Area Network (LAN).The intermediate center is usually composed of a command vehicle or a temporary camp with communication equipment.A set of emergency

Figure 3 .
Figure 3. Construction of emergency communication network.
Figure 5. Scenarios integration and visualization module.
Figure 6.Emergency rescue and command management module.

Banghui
Yang received the M.S. degrees in Geographic Information System from Wuhan University, China, in 2005 and the Ph.D. degrees in cartography and geographic information systems from Graduate University of Chinese Academy of Sciences, China, in 2011 respectively.He is currently an associated professor at the Aerospace Information Research Institute, Chinese Academy of Sciences.His research activities focus on geographic information systems and remote sensing applications.