Flood event construction method cooperatively driven by geographic scene and social data
By using a collaborative approach driven by geographic scenarios and social data, the problems of information timeliness and semantic ambiguity in flood disaster visualization have been solved, enabling accurate analysis and dynamic visualization of disaster information, and enhancing the public's awareness of disaster risks and emergency decision-making capabilities.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- SOUTHWEST JIAOTONG UNIV
- Filing Date
- 2026-01-16
- Publication Date
- 2026-05-08
AI Technical Summary
Existing technologies for flood disaster visualization suffer from several drawbacks, including missing disaster details, low public participation, poor information timeliness, inefficient social media data processing, and severe semantic ambiguity. These issues lead to inaccurate disaster information extraction, affecting the reliability and credibility of the visualization results.
By adopting a collaborative approach driven by geographic scenes and social data, this method achieves accurate semantic parsing and dynamic visualization of disaster information through social media data collection and preprocessing, geographic scene ontology structured modeling, flood disaster knowledge graph construction driven by large language models, and multimodal spatiotemporal narrative expression driven by semantic knowledge reasoning.
It has realized the entire chain path of flood disaster information from fragmented perception to systematic cognition, enhanced the public's real-time cognition and autonomous response capabilities, improved the scientific nature and social collaboration of disaster risk management, and reduced disaster losses.
Smart Images
Figure CN121998242A_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of flood control and flood prevention technology, specifically a flood event construction method driven by the collaboration of geographic scenes and social data. Background Technology
[0002] Floods, as one of the most frequent and destructive natural disasters globally, are characterized by their suddenness, widespread impact, and spatiotemporal uncertainty, leading to delays in the dissemination of disaster information and severely hindering public risk awareness and emergency decision-making efficiency. Traditional flood visualization methods largely rely on remote sensing imagery and numerical models, which, while capable of presenting the macroscopic extent of inundation, suffer from problems such as missing disaster details, low public participation, and poor information timeliness, failing to meet the public's urgent need for visualized disaster information.
[0003] The rise of social media data has provided a new dimension for dynamic perception of flood disasters. Platforms such as Twitter and Weibo generate millions of disaster-related tweets daily, offering advantages such as high real-time performance, wide coverage, and high public participation. However, existing research mostly uses traditional NLP techniques to process social media data, which faces problems such as low information extraction efficiency, severe semantic ambiguity, and significant noise interference. This leads to inaccurate extraction of disaster entities and relationships, thereby affecting the reliability and credibility of visualization results.
[0004] To this end, the present invention provides a flood event construction method driven by the collaborative integration of geographical scene and social data. Summary of the Invention
[0005] In order to overcome the shortcomings of the prior art, at least one technical problem raised in the background art is solved.
[0006] The technical solution adopted by this invention to solve its technical problem is: the flood event construction method driven by the collaboration of geographical scene and social data, as described in this invention, comprising the following steps: S1. Social media data collection and preprocessing; S2. Geographic ontology-based structural modeling for disaster-oriented geographic scenarios; S3, Construction of a flood disaster knowledge graph driven by a large language model; S4. Multimodal spatiotemporal narrative expression mechanism of flood disaster; S5. Visual representation of disaster spatiotemporal evolution driven by semantic knowledge reasoning.
[0007] Preferably, the specific implementation process of step S1 includes: Initial data collection from users across various platforms forms the raw social media dataset. : ; For the original social media dataset Preprocessing is performed, including at least noise filtering, format correction, timestamp unification, spatial alignment, and text normalization, to obtain a cleaned social media dataset. : ; Preferably, step S1 involves analyzing social media data. Before collecting the data, its quality should be scored: ;
[0008] When quality rating The social media data will only be considered if the following condition is met. Include the original social media dataset middle: ;
[0009] Preferably, the specific implementation process of step S2 includes: Constructing a geographic scene ontology model for flood disasters as a semantic framework: ;
[0010] Preferably, the specific implementation process of step S3 includes: In the geographical scene itself Under constraints, using large language models, from social media datasets Extracting structured knowledge : The extracted knowledge Represented as a triple:
[0011] Extraction results Mapping to Ontology In China, a knowledge graph of flood disasters is being constructed. :
[0012]
[0013] Preferably, the large language model mentioned in step S3 is an LLM, and the social media dataset is extracted using the following formula. Structured knowledge in:
[0014] Preferably, the flood disaster knowledge graph mentioned in step S3 Evolving over time:
[0015] Preferably, the specific implementation process of step S4 includes:
[0016] Introducing visual modalities and interaction modality This forms a flood narrative expression model:
[0017] At the semantic mapping layer, disaster entities are mapped according to the semantic hierarchy and spatial constraints in the ontology. By binding with geospatial location, semantic projection from the knowledge layer to the spatial visualization layer can be achieved:
[0018] At the visual layer, a multimodal expression strategy is adopted, combining semantic nodes, spatial trajectories and disaster images, and displaying the spread and impact of floods through color coding, dynamic symbolization and time-series animation.
[0019] At the interaction layer, users can base their interactions on semantic relationships. Dynamic queries are performed to create a semantically driven, interactive narrative experience.
[0020] Preferably, the specific implementation process of step S5 includes: Calculating the strength of causal relationships between events based on semantic reasoning:
[0021] Each event Mapped to a triple This forms a set of spatiotemporal constraints:
[0022] The causal relationships between events are determined by semantic reasoning functions. Decision, when the event With the event If the following equation is satisfied, then the two are considered to have satisfied the conditions for forming a causal chain:
[0023] Preferably, the specific implementation process of step S5 further includes: At the visualization level, the reasoning results are mapped onto the narrative visualization framework, forming a semantically driven visualization expression mechanism, specifically including three mapping modes: Time evolution mapping: The formation, spread, and recovery process of a disaster are shown through a timeline and stage-by-stage animation; Spatial diffusion mapping: Based on a 3D terrain scene, it shows the propagation path and impact range of floods in geographic space; Semantic narrative mapping: The disaster event chain obtained through reasoning is embedded into the visual interface in the form of textual narrative and semantic nodes, realizing a closed-loop process from semantic reasoning to visual narrative.
[0024] The beneficial effects of this invention are as follows: 1. The flood event construction method driven by the collaboration of geographic scene and social data described in this invention utilizes the structured rules of geographic scene ontology and the collaboration of large language models to achieve accurate semantic parsing of disaster information on social media, transforming non-standard disaster descriptions into structured disaster scene elements; combining disaster semantic knowledge reasoning and spatiotemporal evolution narrative framework, it constructs a dynamic expression mechanism for the causal chain of disasters; relying on the collaborative architecture of knowledge graph and narrative visualization module, it realizes multi-level dynamic rendering and layered display of disaster information, and achieves hierarchical presentation of disaster information through time axis linkage and spatial scaling technology.
[0025] 2. The flood event construction method driven by the synergy of geographical scenes and social data described in this invention reconstructs the entire chain path of disaster information from fragmented perception to systematic cognition through a technical closed loop of "semantic parsing-knowledge generation-spatiotemporal extrapolation-visual narrative". It realizes the leap of disaster information expression from static display to dynamic narrative and from macro description to causal reasoning, strengthens the public's real-time cognition and autonomous response capabilities to disaster risks, significantly enhances the scientific nature and social collaboration of disaster risk management, and has important value for reducing disaster losses and maintaining social stability. Attached Figure Description
[0026] The invention will now be further described with reference to the accompanying drawings.
[0027] Figure 1 This is a flowchart of the method of the present invention. Detailed Implementation
[0028] To make the technical means, creative features, objectives and effects of this invention easier to understand, the invention will be further described below in conjunction with specific embodiments.
[0029] like Figure 1 As shown, the flood event construction method driven by the collaborative integration of geographical scenes and social data of the present invention includes the following steps: S1. Social media data collection and preprocessing; This paper presents an efficient social media data collection and cleaning process designed for platforms such as Twitter, Facebook, Weibo, and Douyin, aiming to automate the acquisition and structured processing of flood disaster-related tweets. The core of this process lies in a customized crawling strategy, which precisely sets key disaster keywords, geofencing boundary conditions, and time window parameters to achieve targeted crawling of disaster-related information. During the collection process, the crawler simultaneously parses and extracts the core text content and precise posting timestamps of each social media post. After initial collection, a raw social media dataset is formed. :
[0030] Each record This represents a single piece of social media data, containing information such as text, time, and location.
[0031] In addition, regarding social media data Before collecting the data, its quality should be scored:
[0032] in, Rate the accuracy of the time; Scoring for spatial accuracy; Rate the credibility of the content; , , These are the weighting coefficients; When quality rating The social media data will only be considered if the following condition is met. Include the original social media dataset middle:
[0033] in, This is the scoring threshold.
[0034] Raw social media data suffers from severe semantic ambiguity, non-standardized expression, and fragmented spatiotemporal relationships. Therefore, raw social media datasets... The process involves systematic preprocessing. First, regular expressions are used for noise filtering, removing advertising text, duplicate content, and non-disaster-related entries. Data with abnormal posting times or formats is automatically corrected or removed. Second, posting timestamps are uniformly converted to the ISO 8601 standard format, and geocoding services (such as the GeoNames API) are used to map the original coordinates to standardized administrative regions (province / city / county). Finally, text standardization is performed using word segmentation, stop word removal, and spell correction techniques. Colloquialisms, abbreviations, and other non-standard expressions are grammatically and semantically standardized. The result is the preprocessed social media dataset. :
[0035] S2. Geographic ontology-based structural modeling for disaster-oriented geographic scenarios;
[0036] To achieve semantic organization and knowledge representation of social media information related to flood disasters, this invention constructs a disaster-oriented geographic scene ontology model. As a formal knowledge representation method, geographic ontology provides semantic constraints and logical rules for entity recognition and relation extraction of social media data by defining geographic concepts, attributes and their relationships, enabling the information extraction process to shift from text matching to semantic understanding, and ensuring the consistency and interpretability of the results.
[0037] The semantic expression of geographic information follows the "six elements of geography" framework, namely time, place, people, events, phenomena, and things. Time depicts the temporal characteristics of disaster occurrence and development, and is the basis for dynamic evolution analysis. Place reflects the spatial location and geographical distribution of events, and is the key to spatial correlation of disasters. People represent the actors or affected groups of events. Events describe the dynamic process of flood occurrence, spread, and response. Phenomena reflect the observable manifestations of disasters, such as road flooding, building collapses, and traffic disruptions. Things refer to specific objects affected by disasters or involved in emergency response, such as buildings, bridges, rivers, or drainage systems. The coupled expression of the six elements realizes the complete modeling of disaster information at the temporal, spatial, and semantic levels, providing a structured semantic framework for flood knowledge extraction. The geographic scene ontology model constructed based on the six elements of geography can be formally represented as follows:
[0038] in, It is a collection of entities, corresponding to the core semantic objects among the six elements; It is a set of attributes used to describe the characteristic information of each entity, such as time format, spatial coordinates or disaster level; For a set of relations, define semantic connections between entities, such as "occurring at", "located in", "cause", "affect", etc. It is a hierarchical classification system used to organize concepts and relationships at different semantic levels; This is a set of constraint rules that defines the logical dependencies and extraction boundaries between entities and relationships.
[0039] By constructing this geographic scene ontology, a unified semantic template can be formed during the knowledge extraction stage of social media data, enabling structured mapping of time, space, and semantic elements in flood disaster tweets.
[0040] S3, Construction of a flood disaster knowledge graph driven by a large language model; Social media texts are characterized by high unstructuredness, colloquialism, and semantic ambiguity. Traditional rule-based or shallow machine learning-based text information extraction methods struggle to accurately identify complex semantics and contextual logic. To achieve a deep understanding and automatic extraction of implicit knowledge from flood disaster-related tweets, this invention introduces a knowledge extraction mechanism driven by a Large Language Model (LLM). Leveraging its powerful capabilities in natural language understanding and contextual reasoning, the LLM can identify semantic relationships from non-standardized expressions and achieve high-precision identification of disaster entities, attributes, and relationships even with few or zero samples, thereby significantly improving the intelligence and generalization performance of disaster knowledge extraction.
[0041] Under the semantic constraints of the geographic scene ontology, a prompt word template oriented towards disaster semantics is designed to guide the model to perform structured understanding of social media text according to the "six elements of geography" framework. The model completes the automatic extraction of disaster semantics through multiple rounds of prompts, mainly including: ① Entity recognition stage, extracting semantic units such as time, location, events, people, phenomena, and things related to floods; ② Attribute extraction stage, identifying the feature information of entities, such as disaster level, water depth, affected area, or time precision; ③ Relationship extraction stage, mining the semantic associations between entities, such as "occurred at," "caused," "affected," "located in," etc. The extraction results are represented in structured form as follows:
[0042] in, It is a collection of disaster-related entities, including semantic objects such as time, location, event, phenomenon, person, and thing identified in the tweet; It is a set of attributes used to describe the characteristic information of various entities, such as disaster level, time precision or spatial range; This set of semantic relations defines the logical connections between disaster elements, such as "occurred at", "caused", and "affected". This knowledge set constitutes the core structure of the semantic expression of flood disasters, providing basic data support for subsequent semantic mapping and knowledge graph construction.
[0043] To ensure the consistency and interpretability of the extracted results with the geographic scene ontology, a semantic mapping process was further carried out. By calculating semantic similarity, applying concept alignment rules, and using ontology hierarchical constraints, the extracted disaster entities were mapped to concept nodes in the ontology, achieving synonym unification and semantic fusion. Finally, the extracted disaster knowledge was fully mapped into the geographic scene ontology structure, forming a flood disaster knowledge graph with unified semantics and logical consistency. :
[0044] in, A collection of nodes, corresponding to entity instances. ; For the set of edges, correspondence instances ; A collection of node attributes, corresponding to attribute instances. .
[0045] Flood Disaster Knowledge Graph Evolving over time:
[0046] in, Update the graph operator; For map increments; This is newly added data; This is the time decay factor.
[0047] S4. Multimodal spatiotemporal narrative expression mechanism of flood disaster; Traditional disaster information visualization is often limited to two-dimensional maps or static charts, lacking a unified narrative mechanism that integrates semantic, visual, and interactive layers, making it difficult to express the spatiotemporal evolution logic and semantic connotations of flood disasters. In particular, for social media data, its unstructured characteristics and semantic complexity lead to fragmented and disjointed information presentation. Therefore, this invention constructs a multimodal narrative framework for flood disasters based on geographic scene ontology and knowledge extraction results, in order to achieve semantic integration from structured knowledge to dynamic visualization.
[0048] The framework is based on knowledge sets As the core input, a visual modality is introduced to achieve multimodal narrative expression. (Including 3D spatial scenes, images, charts, etc.) and interactive modalities (Including time control, semantic query, etc.), forming a flood narrative expression model:
[0049] in, For the purpose of expressing the narrative result; The fusion function describes the fusion relationship between semantic knowledge, visual mapping, and interaction mechanisms; the core idea of the model is to integrate the semantic structure of flood disasters. By transforming visual and interactive sensory channels into dynamic stories that users can understand, we can support cognitive analysis and process reconstruction of disasters.
[0050] At the semantic mapping layer, the system maps disaster entities according to the semantic hierarchy and spatial constraints in the ontology. By binding with geospatial location, semantic projection from the knowledge layer to the spatial visualization layer can be achieved:
[0051] in, Location of the disaster; For the corresponding time segment; This is a semantic attribute vector, such as disaster level or damage extent.
[0052] At the visual layer, a multimodal expression strategy is adopted, combining semantic nodes, spatial trajectories, and disaster images, and using color coding, dynamic symbolization, and time-series animation to display the spread and impact of the flood.
[0053] At the interaction layer, users can base their interactions on semantic relationships. Dynamic queries, such as "showing affected cities" and "tracing the path of flood propagation," create a semantically driven interactive narrative experience. Through a multimodal narrative framework that integrates semantics, visuals, and interaction, the complex spatiotemporal evolution of flood disasters can be presented in a more intuitive and interpretable way.
[0054] S5. Visual representation of disaster spatiotemporal evolution driven by semantic knowledge reasoning; Supported by a multimodal narrative framework, this invention constructs a disaster spatiotemporal evolution visualization model driven by semantic knowledge reasoning to achieve intelligent expression of the entire process of flood disaster from occurrence to spread to end. This model mines the event logic and causal chain hidden in the knowledge graph through semantic reasoning mechanism, and drives the spatiotemporal visualization narrative to realize the dynamic expression and cognitive reconstruction of the disaster evolution process.
[0055] The semantic reasoning layer operates based on a joint mechanism of "knowledge graph + large language model"; the knowledge graph provides computable logical rules and entity relationship networks, while the large language model provides semantic understanding and completion capabilities; the two work together to form a hybrid reasoning system: the former is responsible for rule-based reasoning, and the latter is responsible for semantically enhanced reasoning; therefore, the formula for calculating the strength of causal relationships between events is:
[0056] in, and Representing knowledge graphs The two event entities in the text reflect event nodes at different stages or links in the disaster process; Indicates an event and Semantic relationships between them, such as "cause", "affect", or "alleviate"; The contextual relationships between events are extracted from textual semantics or social media data by a large language model to characterize the semantic relevance between events; Representation of knowledge graph The global structural constraints are used to reflect the consistency between the ontological logic and the structure contained in the graph; It is a semantic reasoning function based on a large language model, which realizes the identification of potential semantic connections between events through semantic understanding and completion; This is a logical reasoning function based on ontology rules, used to ensure that the reasoning results conform to knowledge constraints and causal logic; the final result is... It represents the semantic strength and causal direction between events, thereby supporting the construction and reasoning of disaster evolution chains.
[0057] When generating event chains, time and space are two core dimensions; to characterize the spatiotemporal evolution of flood disasters, each event... Mapped to a triple This forms a set of spatiotemporal constraints:
[0058] in, Indicates the time when the event occurred, ensuring that the time is non-negative and conforms to the actual order of occurrence; This set represents the spatial location of an event, which can be latitude and longitude or projected plane coordinates, used to visualize its distribution in geographic space; It describes the dynamic trajectory of the entire flood event in time and space, providing basic data for subsequent visualization.
[0059] The causal relationships between events in an event chain are determined by a semantic reasoning function. Decision, when the event With the event If the following equation is satisfied, then the two are considered to have satisfied the conditions for forming a causal chain:
[0060] in, The filtering threshold is used to filter out event pairs with weak semantic reasoning strength, retaining only credible causal relationships. This condition ensures that the generated event chain is both temporally logical and semantically credible, thus forming a dynamic evolutionary semantic chain from static knowledge.
[0061] At the visualization level, the reasoning results are mapped into a narrative visualization framework, forming a semantically driven visualization expression mechanism. Specifically, this includes three mapping modes: ① Temporal evolution mapping: using a timeline and stage-by-stage animation to show the formation, spread, and recovery process of a disaster; ② Spatial diffusion mapping: based on a three-dimensional terrain scene, showing the propagation path and impact range of floods in geographic space; ③ Semantic narrative mapping: embedding the disaster event chain obtained through reasoning into the visual interface in the form of textual narrative and semantic nodes, achieving an integrated expression that is "readable, viewable, and understandable," and ultimately realizing a closed-loop process from semantic reasoning to visual narrative.
[0062] The terms "front," "back," "left," "right," "top," and "bottom" all refer to the figures in the accompanying drawings. Figure 1 Based on the perspective of the observer, the side of the device facing the observer is defined as the front, the left side of the observer is defined as the left, and so on.
[0063] In the description of this invention, it should be understood that the terms "center", "longitudinal", "lateral", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer", etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limiting the scope of protection of this invention.
[0064] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of the present invention is defined by the appended claims and their equivalents.
Claims
1. A flood event construction method driven by the collaborative integration of geographic scene and social data, characterized by: Includes the following steps: S1. Social media data collection and preprocessing; S2. Geographic ontology-based structural modeling for disaster-oriented geographic scenarios; S3, Construction of a flood disaster knowledge graph driven by a large language model; S4. Multimodal spatiotemporal narrative expression mechanism of flood disaster; S5. Visual representation of disaster spatiotemporal evolution driven by semantic knowledge reasoning; The specific implementation process of step S1 includes: Initial data collection from users of various platforms, including social media data. The quality is rated as follows: ; in, Rate the time accuracy; Scoring for spatial accuracy; Rate the credibility of the content; , , These are the weighting coefficients; When quality rating When the following formula is satisfied, the social media data will be... Include the original social media dataset middle: ; in, The scoring threshold; Forming the original social media dataset : ; Each record This represents a single piece of social media data, including text, time, and location information.
2. The flood event construction method driven by the collaborative integration of geographic scene and social data according to claim 1, characterized in that: The specific implementation process of step S1 also includes: For the original social media dataset Preprocessing is performed, including at least noise filtering, format correction, timestamp unification, spatial alignment, and text normalization, to obtain a cleaned social media dataset. : 。 3. The flood event construction method driven by the collaborative integration of geographic scene and social data according to claim 2, characterized in that: The specific implementation process of step S2 includes: Constructing a geographic scene ontology model for flood disasters as a semantic framework: ; in, For a collection of entities; A collection of attributes; For a set of relations; A hierarchical classification system; This is a set of constraint rules.
4. The flood event construction method driven by the collaborative integration of geographical scene and social data according to claim 3, characterized in that: The specific implementation process of step S3 includes: In the geographical scene itself Under the constraints of large language models, from social media datasets Extracting structured knowledge : The extracted knowledge Represented as a triple: ; in, A collection of disaster-related entities; A collection of attributes; It is a set of semantic relations; Extraction results Mapping to Ontology In China, a knowledge graph of flood disasters is being constructed. : ; ; in, A collection of nodes, corresponding to entity instances. ; For the set of edges, correspondence instances ; A collection of node attributes, corresponding to attribute instances. .
5. The flood event construction method driven by the collaborative integration of geographical scene and social data according to claim 4, characterized in that: The large language model mentioned in step S3 is an LLM, and the social media dataset is extracted using the following formula. Structured knowledge in: ; in, This is the extraction function.
6. The flood event construction method driven by the collaborative integration of geographic scene and social data according to claim 4, characterized in that: The flood disaster knowledge graph mentioned in step S3 Evolving over time: ； in, Update the graph operator; For map increments; This is newly added data; This is the time decay factor.
7. The flood event construction method driven by the collaborative integration of geographic scene and social data according to claim 4, characterized in that: The specific implementation process of step S4 includes: Introducing visual modalities and interaction modality This forms a flood narrative expression model: ; in, For the purpose of expressing the narrative result; This is the fusion function; At the semantic mapping layer, disaster entities are mapped according to the semantic hierarchy and spatial constraints in the ontology. By binding with geospatial location, semantic projection from the knowledge layer to the spatial visualization layer can be achieved: ； in, Location of the disaster; For the corresponding time segment; It is a semantic attribute vector; At the visual layer, a multimodal expression strategy is adopted, combining semantic nodes, spatial trajectories and disaster images, and displaying the spread and impact of floods through color coding, dynamic symbolization and time-series animation. At the interaction layer, users can base their interactions on semantic relationships. Dynamic queries are performed to create a semantically driven, interactive narrative experience.
8. The flood event construction method driven by the collaborative integration of geographic scene and social data according to claim 7, characterized in that: The specific implementation process of step S5 includes: Calculating the strength of causal relationships between events based on semantic reasoning: ; in, and Representing knowledge graphs The two event entities in; Indicates an event and The semantic relationship between them; For the contextual association between events; Representation of knowledge graph Global structural constraints; This is a semantic reasoning function based on a large language model; It is a logical reasoning function based on ontology rules; Indicates the semantic strength and causal direction between events; Each event Mapped to a triple This forms a set of spatiotemporal constraints: ; in, The time when the event occurred; The spatial location where the event occurred; The causal relationships between events are determined by semantic reasoning functions. Decision, when the event With the event If the following equation is satisfied, then the two are considered to satisfy the condition for forming a causal chain: ; in, This is the filtering threshold.
9. The flood event construction method driven by the collaborative integration of geographic scene and social data according to claim 8, characterized in that: The specific implementation process of step S5 also includes: At the visualization level, the reasoning results are mapped onto the narrative visualization framework, forming a semantically driven visualization expression mechanism, specifically including three mapping modes: Time evolution mapping: The formation, spread, and recovery process of a disaster are shown through a timeline and stage-by-stage animation; Spatial diffusion mapping: Based on a 3D terrain scene, it shows the propagation path and impact range of floods in geographic space; Semantic narrative mapping: The disaster event chain obtained through reasoning is embedded into the visual interface in the form of textual narrative and semantic nodes, realizing a closed-loop process from semantic reasoning to visual narrative.