1. A system (100) for real-time environmental monitoring and response coordination, the system comprising: a user interface module (110) configured to receive environmental reports from a one or more users, each environmental report comprising geotagged data, textual input, visual evidence, and a severity level, and to transmit the environmental reports to a data processing module (120); a data storage module (130) configured to store the environmental reports including geospatial coordinates, timestamps, and associated metadata; the data processing module (120) operatively coupled to the user interface module (110) and the data storage module (130), and configured to: generate a textual embedding from the textual input using a language model (121) and a visual embedding from the visual evidence using a vision model (122); combine the textual embedding and the visual embedding using a multimodal fusion mechanism (123) that applies dynamically adjusted weighting factors based on confidence scores associated with each modality to generate a unified feature representation; classify and categorize the environmental reports based on the unified feature representation; perform spatiotemporal clustering (126) of the environmental reports based on geographic proximity and temporal proximity to identify environmental events; and compute an urgency score (127) for each environmental event based on a weighted function of report density, severity levels, and rate of report occurrence over time; a visualization engine module (140) operatively coupled to the data storage module (130) and the data processing module (120), and configured to generate geospatial visualizations including cluster-based heatmaps based on the environmental events and corresponding urgency scores; and a stakeholder interaction module (150) operatively coupled to the data processing module (120) and the visualization engine module (140), and configured to transmit notifications to stakeholders based on the environmental events and urgency scores and to track resolution status of the environmental events.

2. The system (100) as claimed in claim 1, wherein the confidence scores associated with each modality are determined based on one or more of model prediction confidence, data quality metrics, or completeness of the input data.

3. The system (100) as claimed in claim 1, wherein the multimodal fusion mechanism (123) comprises a weighted combination of the textual embedding and the visual embedding using dynamically computed weighting factors.

4. The system (100) as claimed in claim 1, wherein the spatiotemporal clustering (126) is performed using a density-based clustering algorithm that groups environmental reports based on a predefined spatial radius and temporal window.

5. The system (100) as claimed in claim 1, wherein the urgency score (127) is computed using a weighted model comprising a number of reports within a cluster, an average severity level, and a rate of incoming reports over time.

6. The system (100) as claimed in claim 1, wherein the stakeholder interaction module (150) comprises an event-driven alert system configured to generate notifications when the urgency score exceeds a predefined threshold within a defined geographic boundary.

7. The system (100) as claimed in claim 1, further comprising a data validation module (124) configured to perform one or more of duplicate report detection, image metadata verification, anomaly detection, and moderation of environmental reports.

8. A method for real-time environmental monitoring and response coordination, the method comprising: receiving environmental reports from a plurality of users, each environmental report comprising geotagged data, textual input, visual evidence, and a severity level; generating a textual embedding from the textual input and a visual embedding from the visual evidence; combining the textual embedding and the visual embedding using a multimodal fusion mechanism that applies dynamically adjusted weighting factors based on confidence scores associated with each modality to generate a unified feature representation; classifying the environmental reports based on the unified feature representation; performing spatiotemporal clustering of the environmental reports to identify environmental events; computing an urgency score for each environmental event based on report density, severity levels, and rate of report occurrence over time; and generating notifications and geospatial visualizations based on the environmental events and urgency scores.

9. The method as claimed in claim 8, wherein the textual embedding and the visual embedding are generated using models trained on a training dataset comprising historical environmental reports, annotated textual descriptions, labeled image data, and associated severity classifications.

10. The method as claimed in claim 8, wherein the dynamically adjusted weighting factors are computed in real time based on relative confidence scores of the textual embedding and the visual embedding.

Description:FIELD OF INVENTION:
[001] The present disclosure relates generally to environmental monitoring and management systems. More particularly, it relates to systems and methods for real-time environmental reporting, response coordination, and spatial.
BACKGROUND OF THE INVENTION:
[002] The field of environmental monitoring and response coordination systems has seen significant technological advancements, traditionally leveraging various sensing modalities to gather information about environmental conditions. These include networks of fixed sensors for parameters such as air and water quality, satellite imaging for macro-level observations of land use changes and atmospheric conditions, and meteorological data for predicting environmental events. Such data is commonly integrated with Geographical Information Systems (GIS) for spatial analysis and visualization, primarily serving regulatory bodies and environmental agencies to inform policy and enforcement.
[003] Despite these advancements, significant limitations persist in effectively addressing the escalating environmental degradation in urban and semi-urban areas. These challenges are exacerbated by unchecked industrial emissions, widespread vehicular pollution, and ineffective waste management practices. While numerous non-governmental organizations (NGOs) and civic bodies are actively involved in environmental protection, their efforts are often isolated, poorly coordinated, and suffer from a lack of public visibility. Consequently, citizens are frequently unaware of how to effectively report environmental hazards or contribute meaningfully to ongoing initiatives.
[004] A substantial problem lies within existing government-operated pollution monitoring systems. These systems are typically costly to deploy and maintain, resulting in limited geographic coverage that often excludes rapidly developing or remote regions. Furthermore, they frequently suffer from delayed data distribution, creating a significant lag between environmental events and the availability of actionable information. Technologies such as satellite imaging and fixed sensors, while valuable, generally provide only macro-level data, which is often insufficient for the localized, fine-grained intervention required to address specific pollution sources or emerging environmental hazards effectively.
[005] From a technical perspective, a critical deficiency exists in current environmental monitoring platforms: they lack robust mechanisms to effectively integrate real-time crowd-sourced geo-tagged visual evidence with real-time severity classification and spatial aggregation. This technical gap results in fragmented datasets, where disparate pieces of information are not coherently linked or analyzed. Such fragmented data fails to provide an actionable, real-time visualization of pollution hotspots, making it challenging to identify areas of concentrated environmental concern accurately and promptly. This limitation, in turn, severely restricts the effectiveness of institutional response mechanisms, as decision-makers often lack the comprehensive and timely information required to deploy resources efficiently and mitigate environmental damage. Moreover, reliance on traditional reactive methods, such as manual inspections, periodic sampling, and unstructured public complaints, inherently suffers from delays in data collection, analysis, and dissemination, hindering a dynamic understanding of environmental conditions and proactive management.
[006] Therefore, there is a compelling need for an improved system and method that can overcome the aforementioned problems. Such a system must be capable of effectively receiving diverse user-generated environmental reports (comprising geotagged data, textual input, visual evidence, and a severity level), processing these reports using advanced multimodal analysis to generate unified feature representations, classifying and categorizing environmental issues, performing spatiotemporal clustering to identify environmental events, and computing dynamic urgency scores. Furthermore, the system needs to generate comprehensive geospatial visualizations, including cluster-based heatmaps, and provide a robust mechanism for stakeholder interaction and transparent resolution tracking, thereby fostering a more connected, responsive, and accountable environmental management ecosystem.
OBJECT OF THE INVENTION:
[007] The primary objective of the present disclosure is to provide a system and method for real-time, localized environmental monitoring by leveraging crowd-sourced geo-tagged visual data and citizen participation.
[008] Another objective of the present disclosure is to overcome the limitations of existing environmental monitoring systems that suffer from fragmented data, delayed response, limited geographic coverage, and insufficient integration of real-time citizen input.
[009] Yet another objective of the present disclosure is to enable advanced data processing, including artificial intelligence (AI) analysis, for classifying environmental issues, performing spatiotemporal clustering to identify environmental events, and computing dynamic urgency scores.
[010] Still another objective of the present disclosure is to generate intuitive and actionable geospatial visualizations, such as cluster-based heatmaps, for enhanced situational awareness and real-time identification of pollution hotspots.
[011] A further objective of the present disclosure is to facilitate improved communication, coordination, and rapid response among citizens, non-governmental organizations (NGOs), and administrative authorities regarding environmental issues.
[012] Another objective of the present disclosure is to enhance transparency, accountability, and civic engagement through comprehensive impact tracking and visible resolution statuses of reported environmental concerns.
[013] It is a further objective of the present disclosure to provide a scalable, decentralized, and user-friendly platform that encourages sustained public action and provides rich, actionable data for policy development and environmental management.
SUMMARY OF THE INVENTION:
[014] According to one aspect of the present disclosure, a system for real-time environmental monitoring and response coordination is provided, comprising a user interface module configured to receive environmental reports from users. Each environmental report includes geotagged data, textual input, visual evidence, and a severity level, which are transmitted to a data processing module. The system further includes a data storage module for storing these environmental reports, encompassing geospatial coordinates, timestamps, and associated metadata. The data processing module generates a textual embedding from the textual input using a language model and a visual embedding from the visual evidence using a vision model. The data processing module combines these embeddings using a multimodal fusion mechanism that applies dynamically adjusted weighting factors based on confidence scores associated with each modality to generate a unified feature representation. The system classifies and categorizes the environmental reports based on the unified feature representation, performs spatiotemporal clustering of the environmental reports based on geographic and temporal proximity to identify environmental events, and computes an urgency score for each environmental event based on a weighted function of report density, severity levels, and rate of report occurrence over time. A visualization engine module generates geospatial visualizations, including cluster-based heatmaps, based on the environmental events and corresponding urgency scores. A stakeholder interaction module transmits notifications to stakeholders based on the environmental events and urgency scores and tracks the resolution status of the environmental events.
[015] According to another aspect of the present disclosure, a method for real-time environmental monitoring and response coordination is provided, comprising receiving environmental reports from a plurality of users, where each environmental report includes geotagged data, textual input, visual evidence, and a severity level. The method generates a textual embedding from the textual input and a visual embedding from the visual evidence. The method combines the textual embedding and the visual embedding using a multimodal fusion mechanism that applies dynamically adjusted weighting factors based on confidence scores associated with each modality to generate a unified feature representation, and classifies the environmental reports based on the unified feature representation. The method performs spatiotemporal clustering of the environmental reports to identify environmental events, computes an urgency score for each environmental event based on report density, severity levels, and the rate of report occurrence over time, and generates notifications and geospatial visualizations based on the environmental events and urgency scores.
[016] The confidence scores associated with each modality are determined based on one or more of model prediction confidence, data quality metrics, or the completeness of the input data. The multimodal fusion mechanism comprises a weighted combination of the textual embedding and the visual embedding using dynamically computed weighting factors. The spatiotemporal clustering is performed using a density-based clustering algorithm that groups environmental reports based on a predefined spatial radius and temporal window. The urgency score is computed using a weighted model comprising a number of reports within a cluster, an average severity level, and a rate of incoming reports over time. The stakeholder interaction module comprises an event-driven alert system configured to generate notifications when the urgency score exceeds a predefined threshold within a defined geographic boundary. The system further comprises a data validation module configured to perform one or more of duplicate report detection, image metadata verification, anomaly detection, and moderation of environmental reports. The textual embedding and the visual embedding are generated using models trained on a training dataset comprising historical environmental reports, annotated textual descriptions, labeled image data, and associated severity classifications. The dynamically adjusted weighting factors are computed in real time based on relative confidence scores of the textual embedding and the visual embedding.
[017] This disclosure improves environmental monitoring capabilities, enhances stakeholder coordination, and promotes community involvement in environmental stewardship. The foregoing paragraphs have been provided by way of general introduction and are not intended to limit the scope of the following claims. The described embodiments, together with further advantages, will be best understood by reference to the following detailed description taken in conjunction with the accompanying drawings.
BRIEF DESCRIPTION OF DRAWINGS:
[018] FIG. 1 is a block diagram illustrating an exemplary system architecture with interconnected modules, according to an embodiment of the present invention.
[019] FIG. 2 is a block diagram illustrating the detailed components of an exemplary data processing module, according to an embodiment of the present invention.
[020] FIG. 3 is a flow diagram illustrating an exemplary overall operational process and user journey of the system, according to an embodiment of the present invention.
[021] FIG. 4 is a flow diagram illustrating an exemplary workflow for citizen task resolution within the system, according to an embodiment of the present invention.

DETAILED DESCRIPTION OF THE INVENTION:
[022] Aspects of the present disclosure are best understood by reference to the description set forth herein. All the aspects described herein will be better appreciated and understood when considered in conjunction with the following descriptions. It should be understood, however, that the following descriptions, while indicating preferred aspects and numerous specific details thereof, are given by way of illustration only and should not be treated as limitations. Changes and modifications may be made within the scope herein without departing from the spirit and scope thereof, and the present disclosure herein includes all such modifications.
[023] The present invention addresses the need for a comprehensive, real-time solution to escalating environmental degradation by uniquely leveraging citizen participation, advanced artificial intelligence (AI), and geospatial technology. The present invention transforms raw, crowd-sourced data into actionable intelligence, fostering improved coordination and accountability among citizens, non-governmental organizations (NGOs), and administrative authorities. The system operates as a decentralized platform designed for scalability and seamless user interaction, providing an interactive, data-driven interface for environmental issue reporting, monitoring, and response coordination.
[024] The system receives diverse environmental reports from one and more users, each comprising geo-tagged observational data, visual evidence, textual input, and/or a user-assigned severity level. These reports are then processed by a data processing module. The data processing module comprising multimodal artificial intelligence (AI), generating textual and visual embeddings that are intelligently fused into a unified feature representation using dynamically adjusted weighting factors based on confidence scores. This robust AI framework enables precise classification and categorization of environmental issues, spatiotemporal clustering to identify distinct environmental events, and the computation of dynamic urgency scores based on report density, severity, and occurrence rate.
[025] Further, the system facilitates dynamic NGO tracking, transparently indicating which organizations are actively handling particular issues and their current resolution status. These processed insights are then presented via an intuitive visualization engine, which generates real-time, cluster-based heatmaps. These heatmaps not only depict the density and severity of environmental concerns but also dynamically update their color, such that an area's color transitions to green upon successful resolution of an issue by a responsible NGO, thereby providing immediate visual feedback of impact and promoting transparency. Beyond visualization, a dedicated stakeholder interaction module ensures efficient communication and coordination. This module is configured to transmit timely notifications to relevant stakeholders including NGOs and authorities and to citizens via geo-radius alerts, informing them of new environmental concerns within a defined proximity to their location. Through this integrated approach, the present invention significantly enhances situational awareness, accelerates response times, and empowers communities to actively participate in environmental stewardship, moving beyond passive reporting to active, data-driven change.
[026] The detailed architecture and functionality of the system are further elaborated with reference to the accompanying figures.
[027] The modules described herein, including the user interface module, data processing module, data storage module, visualization engine module, stakeholder interaction module, and associated sub-modules, are functional components and are not limited to any specific implementation. Each module may be implemented in hardware, software, firmware, or a combination thereof, such as dedicated integrated circuits, embedded systems, software executed on general-purpose processors, or cloud-based virtualized resources. The allocation of functionality between hardware and software may vary based on design, performance, and cost considerations without departing from the scope of the invention.
[028] The system may include one or more processors configured to execute instructions and perform computations for module operations. Such processors may include CPUs, GPUs, FPGAs, ASICs, microcontrollers, or combinations thereof, and may be implemented as single-core, multi-core, or distributed systems, including cloud or edge computing environments. The selection of processors may vary depending on computational requirements, power constraints, and cost considerations.
[029] The system may further utilize various types of memory for storing data, instructions, and intermediate results, including volatile memory such as RAM and cache, and non-volatile memory such as ROM, flash memory, SSDs, and HDDs. Memory may be integrated with processors, externally connected, or distributed across networked or cloud-based systems. The choice of memory configuration may vary based on access speed, storage capacity, and system architecture.
[030] The language, vision, and classification models employed by the data processing module are not limited to any specific artificial intelligence architecture. These models may include deep neural networks, such as convolutional, recurrent, or transformer-based models, as well as ensemble methods, support vector machines, decision trees, or rule-based approaches. The models may be pre-trained, fine-tuned using domain-specific environmental data, or trained from scratch, and may be updated or replaced over time to improve performance without departing from the scope of the invention.
[031] FIG. 1 depicts a system 100 for real-time environmental monitoring and response coordination as per the exemplary embodiment of the present invention, which integrates various functional modules to facilitate community-driven environmental reporting and stakeholder engagement. The system 100 comprises a user interface module 110 configured to receive environmental reports from one or more users. These environmental reports include geotagged data, textual input, visual evidence, and a severity level. The user interface module 110 transmits these reports to a data processing module 120. A data storage module 130 is operatively coupled to store the environmental reports, including geospatial coordinates, timestamps, and associated metadata. The data processing module 120 is operatively coupled to both the user interface module 110 and the data storage module 130, processing the received data to generate valuable insights. Furthermore, a visualization engine module 140 is operatively coupled to the data storage module 130 and the data processing module 120, enabling the generation of geospatial visualizations. Lastly, a stakeholder interaction module 150 is operatively coupled to the data processing module 120 and the visualization engine module 140, designed to transmit notifications to stakeholders and track the resolution status of environmental events. The interconnected nature of these modules supports a comprehensive and responsive environmental monitoring platform, enabling diverse data inputs to translate into actionable intelligence and coordinated responses.
[032] The user interface module 110 functions as the primary point of interaction for users to submit environmental reports. This module is a lightweight, accessible web platform that enables immediate submission of environmental concerns, incorporating location data, descriptions, severity assessments, and images. The user interface module 110 can be implemented using frontend technologies such as React.js, providing a responsive and intuitive web interface optimized for both desktop and mobile users, allowing them to easily report environmental issues, track complaint statuses, and visualize pollution data. Alternative embodiments for the user interface module 110 may include native mobile applications developed for iOS and Android platforms, or integration with social media platforms for direct reporting. Additional embodiments may include voice-activated reporting mechanisms or even smart device integrations for automated data capture, further simplifying the reporting process for citizens.
[033] The data processing module 120 is the core analytical component of the system 100 and is configured to transform raw environmental reports into structured, standardized, and actionable information. The data processing module 120 performs a series of operations including data validation, preprocessing, feature extraction, classification, and prioritization of reported environmental events. In some embodiments, the module analyzes incoming data such as text descriptions, images, geolocation coordinates, timestamps, and user inputs to generate enriched datasets and corresponding urgency scores. The processed outputs are utilized by downstream modules, including clustering, visualization, and notification components, to enable efficient decision-making and response actions. Further details of the data processing workflow are illustrated in FIG. 2.
[034] The data storage module 130 is configured to securely store all environmental reports and associated metadata. This includes critical information such as geospatial coordinates (latitude and longitude), timestamps, issue descriptions, image links, report categories, and severity levels. The data storage module 130 utilizes a PostgreSQL database, often hosted on a platform like SupaBase, which allows for structured storage and efficient querying of user-submitted data. The robust nature of PostgreSQL supports the relational integrity required for detailed report tracking and analytical operations. Alternative embodiments for the data storage module 130 may include NoSQL databases such as MongoDB or Cassandra for greater flexibility with unstructured data, particularly if the scope of reported data expands significantly, or cloud-based data warehouses like Amazon Redshift for large-scale analytical processing. An additional embodiment could involve blockchain technology for immutable record-keeping of reports, enhancing trust and transparency in the data.
[035] The visualization engine module 140 is responsible for generating intuitive and informative geospatial visualizations based on the processed environmental events and their corresponding urgency scores. This module creates cluster-based heatmaps that visually represent the density and severity of reported issues across geographical areas. The visualization engine module 140 employs MapBox API for accurate geotagging and dynamic, interactive map rendering. In some embodiments, the visualization engine module 140 comprising heatmap engine to generate the color coding. The heatmaps typically color-code regions, for instance, grey for no activity, yellow for low, orange for moderate, and red for high activity, providing immediate situational awareness. In some embodiments, the map or corresponding heatmap region turns green when an NGO resolves a reported issue, thereby visually indicating successful remediation or closure of the environmental event. This enables NGOs and authorities to identify environmental hotspots efficiently, monitor issue severity, and maintain situational awareness. The heatmap engine updates dynamically as new reports are received and as issue statuses change, thereby reflecting current environmental conditions and resolution progress in real time. Alternative embodiments for the visualization engine module 140 could include using Google Maps API or OpenStreetMap for mapping services, or employing different visualization libraries such as D3.js or Leaflet.js for custom and more complex interactive elements. Additional embodiments might feature 3D visualizations for urban areas or augmented reality (AR) overlays for on-site environmental assessment, offering a more immersive data exploration experience.
[036] The stakeholder interaction module 150 facilitates communication and coordination among multiple stakeholders, including citizens, non-governmental organizations (NGOs), and administrative authorities. The stakeholder interaction module 150 is configured to transmit notifications, alerts, and updates to relevant stakeholders based on detected environmental events and their associated urgency scores, thereby ensuring timely awareness and enabling prompt response actions. In some embodiments, the notifications may include real-time alerts, status updates, escalation messages, or resolution confirmations delivered via one or more communication channels, including mobile applications, web interfaces, SMS, or push notification services. The stakeholder interaction module 150 further enables stakeholders to access and interact with geospatial visualizations, including dynamic heatmaps generated by the visualization engine module, thereby allowing users to monitor environmental conditions, identify hotspots, and track issue resolution status. In alternative embodiments, the module may support bidirectional communication, enabling stakeholders to provide feedback, update event statuses, or coordinate response efforts within the system. Furthermore, the stakeholder interaction module 150 tracks the resolution status of environmental events, allowing users to see NGO involvement and monitor problem resolutions, which contributes to transparency and accountability. The module can send automated alerts when the urgency score exceeds a predefined threshold within a defined geographic boundary. Alternative embodiments for the stakeholder interaction module 150 may include integrated chat functionalities, in-app messaging systems, or direct email/SMS gateways for broader notification reach. Additional embodiments could incorporate a feedback mechanism for stakeholders to report on their actions and resolution outcomes, further enriching the data on impact tracking and closure.
[037] FIG. 2 illustrates a detailed view of the data processing module 120, which is operatively coupled to the user interface module and the data storage module, showing its internal components and their interconnections for advanced environmental report analysis. The data processing module 120 encompasses a language model 121, a vision model 122, a multimodal fusion sub-module 123, a data validation sub-module 124, a classification and categorization sub-module 125, a spatiotemporal clustering sub-module 126, an urgency scoring sub-module 127, and a report generation module 128. These sub-modules collaboratively process environmental reports, extracting features, validating data, categorizing issues, identifying event clusters, assessing urgency, and preparing comprehensive reports for stakeholders. This modular design within the data processing module 120 allows for specialized processing steps, ensuring thorough analysis and accurate insights derived from diverse user submissions. In various embodiments, the data processing module 120 functions as the core artificial intelligence (AI) engine of the system, wherein its operations are driven by machine learning and/or deep learning models, including the language model 121, vision model 122, and multimodal fusion sub-module 123, enabling adaptive analysis, pattern recognition, and data-driven decision-making across environmental reports.
[038] The language model 121 within the data processing module 120 is configured to generate a textual embedding from the textual input provided in environmental reports. This involves employing natural language processing (NLP) techniques to understand and represent the semantic meaning of the text. The language model 121 converts textual descriptions into numerical vectors that capture contextual information, which can then be used for classification and further analysis. For instance, if a user describes "smog in the air," the language model 121 processes this to identify it as related to air pollution. Alternative embodiments for the language model 121 include utilizing transformer-based models like BERT, RoBERTa, or GPT variants for enhanced contextual understanding, or smaller, more efficient models for real-time inference on resource-constrained environments. Additional embodiments might incorporate domain-specific lexical databases or ontologies to improve the accuracy of environmental term recognition and classification.
[039] The vision model 122, also a component of the data processing module 120, is designed to generate a visual embedding from the visual evidence, such as images, submitted with environmental reports. The vision model 122 employs computer vision techniques to analyze the content of photographs, detecting relevant features and patterns that indicate environmental issues. For example, the vision model 122 can identify images depicting deforestation, waste accumulation, or water pollution. Alternative embodiments for the vision model 122 include convolutional neural networks (CNNs) like ResNet, VGG, or EfficientNet, or object detection models like YOLO or Faster R-CNN for identifying specific elements within images such as plastic bottles or smoke plumes. Additional embodiments may involve integrating image metadata verification within the vision model 122, such as checking GPS tags or timestamps, to augment the reliability of the visual evidence.
[040] The multimodal fusion sub-module 123, integrated within the data processing module 120, combines the textual embedding from the language model 121 and the visual embedding from the vision model 122. This sub-module applies dynamically adjusted weighting factors based on confidence scores associated with each modality to generate a unified feature representation. The fusion mechanism ensures that information from both text and images is leveraged effectively, especially when one modality might be more informative or reliable than the other. For instance, if an image is blurry but the text description is precise, the textual embedding might receive a higher weight. Alternative embodiments for the multimodal fusion sub-module 123 could involve attention mechanisms to selectively focus on relevant parts of each modality, or more complex neural network architectures specifically designed for multimodal learning, such as early fusion, late fusion, or hybrid fusion strategies. Additional embodiments might incorporate additional modalities, such as audio data for noise pollution or sensor readings, further enriching the unified feature representation.
[041] The data validation sub-module 124 is configured to perform various measures to enhance the reliability and quality of incoming environmental reports. This sub-module can execute tasks such as duplicate report detection, image metadata verification, anomaly detection, and moderation workflows for environmental reports. Duplicate report detection prevents redundancy and ensures that multiple submissions of the same incident do not skew statistics or response efforts. Image metadata verification checks for inconsistencies in geotags or timestamps, while anomaly detection flags unusual reporting patterns. Alternative embodiments for the data validation sub-module 124 could include machine learning models trained to identify suspicious patterns in user submissions, or integration with external geographical data to verify the reported location against known landmarks or features. Additional embodiments might involve a human-in-the-loop moderation system, where flagged reports are reviewed by trained personnel to ensure accuracy before processing.
[042] The classification and categorization sub-module 125, part of the data processing module 120, is responsible for classifying and categorizing environmental reports based on the unified feature representation generated by the multimodal fusion sub-module 123. This sub-module assigns reports to predefined environmental categories, such as air pollution, sanitation issues, deforestation, or urban infrastructure degradation. The categorization is automated, employing AI-based techniques, keyword logic, or a combination thereof, ensuring consistent and efficient sorting of diverse reports. Alternative embodiments for the classification and categorization sub-module 125 may include hierarchical classification models that can categorize issues at different levels of granularity, or transfer learning approaches where pre-trained models are fine-tuned for environmental datasets. Additional embodiments could incorporate user-defined categories or dynamic category generation based on emergent environmental concerns, providing greater flexibility to the system.
[043] The spatiotemporal clustering sub-module 126 performs spatiotemporal clustering of environmental reports based on geographic proximity and temporal proximity to identify environmental events. This sub-module groups similar reports that occur close to each other geographically and within a specific time window, allowing the system to identify distinct environmental incidents rather than treating each report in isolation. For example, multiple reports of smoke in the same area over a short period would be clustered into a single "air pollution event." Alternative embodiments for the spatiotemporal clustering sub-module 126 include density-based clustering algorithms such as DBSCAN or OPTICS for identifying arbitrary-shaped clusters, or grid-based clustering methods for large-scale datasets. Additional embodiments may integrate weather data or topographic information to refine clustering accuracy, recognizing that environmental events can be influenced by such factors.
[044] The urgency scoring sub-module 127 is configured to compute an urgency score for each environmental event identified by the spatiotemporal clustering sub-module 126. This score is based on a weighted function of report density, severity levels provided by users, and the rate of report occurrence over time. A high density of reports, severe user ratings, and a rapid increase in new reports would result in a higher urgency score, signaling a more pressing environmental concern. This allows stakeholders to prioritize their response efforts effectively. Alternative embodiments for the urgency scoring sub-module 127 could involve machine learning models trained on historical response times and impact data to predict optimal urgency levels, or dynamically adjustable weighting factors based on specific policy priorities. Additional embodiments might integrate external data such as population density or proximity to vulnerable ecosystems to further refine the urgency score.
[045] The report generation module 128, a final component within the data processing module 120, is responsible for compiling synthesized information from the processed environmental events. This module prepares structured data for the visualization engine module 140 and the stakeholder interaction module 150. The report generation module 128 ensures that all relevant details, including classified categories, unified feature representations, cluster data, and urgency scores, are coherently assembled. This compilation forms the basis for the heatmaps, notifications, and dashboards presented to various users. Additional embodiments might incorporate predictive analytics within the report generation module 128, forecasting potential future environmental issues based on current trends and historical data.
[046] In some embodiments, the language model 121, vision model 122, classification and categorization sub-module 125, urgency scoring sub-module 127, or combinations thereof, are trained using supervised, weakly supervised, semi-supervised, or self-supervised learning techniques. Training data may include historical environmental reports comprising textual descriptions, captured images, geographic coordinates, timestamps, user-provided severity indicators, moderation outcomes, and ground-truth category labels. Representative categories may include air pollution, waste accumulation, water contamination, deforestation, noise pollution, sanitation issues, and urban infrastructure degradation.
[047] In some embodiments, textual training samples may include statements such as “smog near the school at 7 a.m.,” “garbage piling up beside the canal,” and “oil-like residue floating on the river surface,” each associated with one or more target labels. Visual training samples may include images depicting smoke plumes, illegal dumping, stagnant wastewater, tree-cutting activity, potholes, or damaged drainage systems. Multimodal training samples may pair text and image data from a common report and further include metadata such as location and time of submission.
[048] In some embodiments, the training data is curated by removing corrupted files, normalizing text, resizing images, deduplicating substantially similar reports, validating metadata where available, and assigning labels through expert review, crowd annotation, heuristic labeling, or combinations thereof. The training data may be partitioned into training, validation, and test sets.
[049] In some embodiments, the language model 121 is fine-tuned on environmental-report text to generate contextual embeddings, and the vision model 122 is trained or fine-tuned on environmental-report images to generate visual embeddings. In some embodiments, the multimodal fusion sub-module is trained using weighted averaging mechanisms.
[050] The classification and categorization sub-module 125 may be trained using one or more output heads configured for single-label classification, multi-label classification, hierarchical classification, or severity estimation. The training objective may include cross-entropy loss, binary cross-entropy loss, focal loss, contrastive loss, triplet loss, ranking loss, regression loss, or combinations thereof. Model parameters may be updated by gradient-based optimization, including stochastic gradient descent, Adam, AdamW, RMSProp, or related optimizers.
[051] In some embodiments, the spatiotemporal clustering sub-module 126 is configured using clustering parameters such as geographic radius, temporal window, minimum sample count, and density thresholds. The urgency scoring sub-module 127 may be trained using historical event outcomes, response times, verified severity levels, escalation records, or environmental impact data to learn a mapping from event features to urgency scores.
[052] By way of example, a training sample may include: (i) text input stating “thick smoke and burning smell near Sector 9 industrial area,” (ii) an associated image depicting a visible smoke plume, (iii) metadata indicating latitude, longitude, and a timestamp, and (iv) a target label of air pollution with a severity label of high. The language model 121 converts the text into a textual embedding vector, the vision model 122 converts the image into a visual embedding vector, and the multimodal fusion sub-module 123 combines the vectors into a unified feature representation.
[053] The unified feature representation is then processed by the classification and categorization sub-module 125 to produce class probabilities. If the predicted probability for air pollution is lower than the target label indicates, a classification loss is computed. The model parameters are then updated using backpropagation and an optimizer such as AdamW. This process is repeated across batches of training samples for multiple epochs until one or more stopping criteria are satisfied, such as convergence of validation loss, attainment of a target score, or early stopping based on validation performance.
[054] In an illustrative embodiment, a classifier head receives a fused feature vector and computes output logits for categories including air pollution, waste accumulation, and water contamination. For a particular sample, the model may initially output probabilities while the target label corresponds to air pollution. A loss value is computed, for example using cross-entropy loss, and gradients with respect to the classifier parameters are determined. Using a learning rate of 0.00001, the optimizer updates the classifier weights and bias terms in a direction that increases the likelihood of the target class for similar samples in subsequent iterations.
[055] In another illustrative embodiment, the urgency scoring sub-module 127 receives event-level features including report count, average user severity, rate of report arrival, and proximity to a school or hospital. For an event having features [12 reports, average severity 4.6/5, 9 reports within 30 minutes, proximity score 0.8], the sub-module may initially predict an urgency score of 0.63 while the target urgency score is 0.89. A regression loss is computed and parameters are updated so that later iterations produce a score closer to the target.
[056] Representative multimodal training samples may include:
(1) Text: “Black smoke is coming from the factory chimney.” Image: smoke plume. Label: air pollution. Severity: high.
(2) Text: “Plastic waste has accumulated beside the lake.” Image: littered shoreline. Label: waste accumulation. Severity: medium.
(3) Text: “Brown foamy water is entering the canal.” Image: discolored outflow. Label: water contamination. Severity: high.
(4) Text: “Several trees were cut overnight near the hillside road.” Image: freshly cut stumps. Label: deforestation. Severity: medium.
(5) Text: “The road drain is broken and sewage is overflowing.” Image: damaged drain with overflow. Label: sanitation issue and infrastructure degradation. Severity: high.
[057] In some embodiments, such samples are manually labeled, machine-assisted and human-verified, or derived from historical enforcement, moderation, or inspection records.
[058] In some embodiments, the trainable parameters of one or more models include embedding-layer weights, attention weights, convolutional kernel weights, projection weights, classifier-layer weights, bias terms, normalization parameters, or combinations thereof. The parameter values are initialized randomly, pre-trained on external corpora or image datasets, or transferred from prior environmental-report models, and are iteratively updated during training based on one or more loss functions. The exact dimensionality and numerical values of the parameters may vary depending on the selected architecture, available compute resources, and target deployment constraints.
[059] FIG. 3 illustrates the user journey and system architecture for the environmental monitoring and response coordination platform 200. This comprehensive workflow demonstrates how user input is transformed into actionable intelligence and transparently communicated to stakeholders and the community.
[060] At step 202, the “citizen opening the system” stage represents the initial interaction point where a user accesses the environmental monitoring platform via a web browser or mobile application. This step reflects the user’s intent to engage with the platform, whether to report an issue, review existing reports, or view environmental data. The system provides an accessible and user-friendly interface to facilitate this interaction.
[061] In alternative embodiments, step 202 may include access through QR codes placed in public areas, integration with smart city kiosks, or other rapid-access mechanisms. Additional embodiments may support multi-language interfaces to accommodate diverse user populations, thereby enhancing accessibility and inclusivity of the platform.
[062] At step 204, a citizen selecting the “report pollution” option indicates the user’s action to initiate the environmental reporting process. This step typically involves navigating through the user interface module 110 to locate a prominent button or link designated for reporting environmental concerns. This action leads to the display of a report submission form.
[063] In alternative embodiments, the “report pollution” action may be initiated through voice commands, which can be particularly useful for accessibility or hands-free operation. Other embodiments may include a quick-capture feature that directly opens a camera interface for immediate submission of visual evidence. Additional embodiments may provide customizable shortcut options for frequent users to streamline the reporting process.
[064] At step 206, a citizen filling the report form involves the user inputting the necessary details for an environmental report. This includes providing textual input, selecting a severity level such as high, moderate, or low, and uploading visual evidence such as photographs. The form also captures geotagged data, either automatically via GPS or manually through map selection, ensuring precise tagging of the incident location. In alternative embodiments, predefined templates for common issues may be provided to expedite data entry, or speech-to-text functionality may be used for detailed descriptions. Additional embodiments may incorporate AI-assisted suggestions for categorization based on initial keywords or images, guiding users to provide more accurate information.
[065] At step 208, reporting an incident represents the culmination of the user’s submission, wherein the completed environmental report is transmitted from the user interface module 110 to the data storage module or cloud backend service at step 218. This action sends all collected data, including geotagged observations, photographs, textual descriptions, and severity levels, for processing and storage. The transmission may be secured to ensure that the data reaches the system reliably for further analysis. In alternative embodiments, offline reporting capabilities may be supported, wherein data is stored locally and transmitted once an internet connection is available. Additional embodiments may include cryptographic hashing or similar techniques to ensure data integrity during transmission.
[066] At step 210, initiation of multimodal AI processing signifies the commencement of advanced data processing within the data processing module 120, involving the language model 121 and the vision model 122. At this stage, textual input is processed to generate textual embeddings, and visual evidence is analyzed to produce visual embeddings. These embeddings are combined by the multimodal fusion sub-module 123 to form a unified representation. This analysis supports accurate categorization and interpretation of the environmental issue. In alternative embodiments, real-time sentiment analysis of textual input may be performed to assess public perception, or generative models may be used to produce summarized descriptions of visual evidence. Additional embodiments may incorporate edge computing to perform initial processing on the user device, thereby reducing latency and bandwidth usage.
[067] At step 212, report generation involves processing performed by the report generation module 128, wherein the categorized environmental report is prepared for visualization and stakeholder notification. This may include aggregating similar regional reports, evaluating urgency, and compiling relevant data into a structured format for real-time display and analysis. In alternative embodiments, reports may be generated in compliance with regulatory standards or automatically translated into multiple languages for broader accessibility. Additional embodiments may include generating visual summaries such as infographics to improve interpretability.
[068] At step 214 and step 216, a frontend interface is generated to provide stakeholder access to the system through a web application and a mobile application. The frontend interface is configured to enable interaction between the stakeholders and the system by presenting environmental event data, urgency indicators, geospatial visualizations, alerts, and remediation status information. The frontend interface further enables stakeholders to submit incident reports, upload supporting content, access geotagged map views, receive notifications, and monitor the progress of issue resolution. The frontend interface represents the user-facing component of the system, which may be implemented using technologies such as React.js to provide a responsive and intuitive experience. This interface enables users to submit reports, track complaint status, and visualize environmental data through maps and heatmaps. The frontend interface may be optimized for both desktop and mobile platforms to ensure accessibility. In alternative embodiments, the frontend may be implemented as a progressive web application with offline capabilities or as a dedicated desktop application. Additional embodiments may include customizable themes and accessibility features to accommodate diverse user needs.
[069] At step 218, the data storage module or cloud backend service leverages platforms such as Supabase, an open-source alternative to Firebase, to manage user authentication, secure API services, and serverless backend logic. This service validates and processes reports in real time, serving as a central hub for data management and operational logic. The cloud backend service ensures scalability, real-time performance, and secure handling of user data. In alternative embodiments, other serverless platforms such as AWS Lambda, Google Cloud Functions, or Azure Functions may be utilized, providing flexibility in infrastructure selection. Additional embodiments may incorporate a microservices architecture, enabling individual backend functions to be independently developed, deployed, and scaled, thereby enhancing system resilience and maintainability.
[070] At step 220, a geospatial mapping API, such as the Mapbox API, is integrated into the system to enable users to accurately geotag their submissions and visualize environmental issues on a dynamic, interactive map. This API provides foundational mapping capabilities, allowing precise plotting of report locations and rendering of interactive map layers. The geospatial mapping API supports the development of location-aware features throughout the platform. In alternative embodiments, mapping services such as Google Maps API, HERE Technologies API, or OpenStreetMap API may be utilized, each offering different features and licensing models. Additional embodiments may integrate satellite imagery or remote sensing data to enhance geographical context and support advanced environmental analysis.
[071] At step 222, a relational database, such as a PostgreSQL database hosted on Supabase, is employed for structured storage of user-submitted data. This includes geolocation data, timestamps, issue descriptions, image references, and report categories. The relational database ensures data integrity, consistency, and efficient querying, which are essential for real-time analytics and visualization. In alternative embodiments, other SQL-based systems such as MySQL or Microsoft SQL Server may be used, or cloud-managed database services such as Amazon RDS or Google Cloud SQL may be employed. Additional embodiments may incorporate data warehousing solutions for historical analysis to derive long-term environmental insights.
[072] At step 224, a real-time heatmap engine, driven by the visualization engine module 140, generates dynamic heatmaps that visually represent the density and severity of reported environmental issues. The heatmaps may be color-coded to provide immediate visual interpretation of activity levels and resolution status across different geographic areas. For example, regions with no activity may be displayed in grey, regions with low activity may be displayed in yellow, regions with moderate activity may be displayed in orange, and regions with high activity may be displayed in red. In some embodiments, the map or corresponding heatmap region turns green when an NGO resolves a reported issue, thereby visually indicating successful remediation or closure of the environmental event. This enables NGOs and authorities to identify environmental hotspots efficiently, monitor issue severity, and maintain situational awareness. The heatmap engine updates dynamically as new reports are received and as issue statuses change, thereby reflecting current environmental conditions and resolution progress in real time. In alternative embodiments, different heatmap generation techniques may be used, such as kernel density estimation with variable bandwidths or alternative visual encodings such as cluster markers. Additional embodiments may allow users to customize visualization parameters, including spatial radius, temporal aggregation windows, and color-scale thresholds.
[073] At step 226, the NGO dashboard refers to a specialized interface within the stakeholder interaction module 150 that enables non-governmental organizations to monitor and manage environmental reports. The dashboard allows filtering of reports based on category, location, and time, providing a centralized view of environmental issues to support coordinated response and resource allocation. In alternative embodiments, the dashboard may include customizable widgets, integration with task management systems, or collaborative tools for multiple users. Additional embodiments may incorporate predictive analytics to estimate resource requirements based on incoming report patterns.
[074] At step 228, the citizen dashboard provides users with a personalized interface displaying their submitted reports and broader environmental insights. Users can track report status, review historical submissions, and observe regional trends. The citizen dashboard enhances transparency and encourages civic engagement by demonstrating how individual contributions support environmental action. In alternative embodiments, the dashboard may include gamification elements such as badges or leaderboards to incentivize participation. Additional embodiments may provide subscription-based alerts for specific locations or issue categories.
[075] At step 230, map and heatmap updates refer to the continuous refreshing of visual data across both the NGO dashboard and the citizen dashboard. This real-time updating mechanism ensures that stakeholders have access to the most current environmental information. In alternative embodiments, predictive models may be integrated to forecast potential hotspot development or environmental changes. Additional embodiments may allow configurable refresh intervals or manual update controls.
[076] At step 232, NGO response and issue resolution represent the operational phase in which non-governmental organizations act upon identified environmental concerns. Based on insights derived from the NGO dashboard and system-generated alerts, NGOs deploy resources and implement corrective measures. This stage reflects the transition from data analysis to real-world intervention. In alternative embodiments, integration with government agencies or emergency services may be provided to enable coordinated responses. Additional embodiments may include structured reporting protocols for documenting resolution activities and outcomes.
[077] At step 234, impact tracking and visualization represent the final stage, wherein the outcomes of interventions are monitored and presented through visual indicators. Following resolution, system updates may reflect status changes, such as modifications in heatmap representation to indicate mitigation. This functionality enhances transparency and accountability by allowing users to observe the effects of reported issues. In alternative embodiments, detailed environmental metrics, such as air quality improvements or waste reduction levels, may be presented. Additional embodiments may include trend analysis over defined time intervals to demonstrate long-term environmental progress and cumulative impact.
[078] FIG. 4 illustrates the workflow 300 for citizen task resolution in the system, demonstrating how a newly submitted report progresses through categorization, urgency assessment, report generation, stakeholder notification, and eventual resolution with visual feedback.
[079] At step 302, a new report submitted represents the initiation of the resolution workflow, corresponding to report incident 208 described in FIG. 3. At this stage, a citizen’s geotagged observations, photographs, textual descriptions, and severity levels are transmitted to the cloud-based infrastructure for processing. The receipt of this report triggers subsequent analytical operations within the system. In alternative embodiments, report submission may include bulk uploads by organizations or programmatic submission via API endpoints. Additional embodiments may incorporate immediate validation mechanisms, such as spam detection, duplicate detection, or format verification, to ensure data quality at the point of entry.
[080] At step 304, categorization by keyword or AI logic is performed by the data processing module 120, particularly the classification and categorization sub-module 125. This step involves analyzing the report using keyword-based logic, machine learning models, or a hybrid approach to assign one or more environmental categories, such as air pollution, sanitation issues, or infrastructure degradation. This ensures that reports are properly classified for downstream processing. In alternative embodiments, active learning systems may be employed, wherein user or expert feedback is used to iteratively improve classification accuracy. Additional embodiments may utilize ensemble models or support multi-label classification, allowing a single report to be associated with multiple categories. In this step further, an urgency assessment is performed by the urgency scoring sub-module 127 within the data processing module 120. The system computes an urgency score based on factors such as report density within a geographic area, user-provided severity levels, and the rate of incoming reports over time. This score determines prioritization for stakeholder notification and response. In alternative embodiments, the weighting of these factors may be dynamically adjusted based on contextual parameters such as seasonal variations or policy priorities. Additional embodiments may integrate external real-time data sources, including weather conditions or public health advisories, to enhance the contextual accuracy of urgency determination.
[081] At step 306, AI-based report generation is performed by the report generation module 128, which compiles a structured and comprehensive report based on categorized data and computed urgency. The generated report includes details such as location, category, severity, and associated visual and textual evidence, and is formatted for stakeholder consumption. In alternative embodiments, the system may generate customized reports tailored to different stakeholder roles, or produce reports compliant with regulatory standards. Additional embodiments may include interactive report formats that allow stakeholders to explore underlying data or historical trends.
[082] At step 308, NGO or authority notification is performed by the stakeholder interaction module 150, which transmits automated alerts to relevant entities based on issue type, geographic location, and urgency level. This ensures timely and targeted communication to appropriate stakeholders. In alternative embodiments, notifications may be delivered through multiple channels, including email, SMS, push notifications, or integration with enterprise systems. Additional embodiments may include escalation mechanisms, wherein unresolved issues trigger notifications to higher authorities after predefined thresholds.
[083] At step 310, NGO action represents the execution phase in which the notified NGO or authority undertakes measures to address the reported environmental issue. Such actions may include cleanup operations, field investigations, awareness campaigns, or coordination with other agencies. This stage reflects the transition from system-driven analysis to real-world intervention. In alternative embodiments, the platform may provide integrated task management tools enabling NGOs to assign, monitor, and update task status. Additional embodiments may include resource coordination features, such as requesting equipment, personnel, or volunteer support through the system.
[084] At step 312, heatmap visualization update represents the feedback mechanism provided by the visualization engine module 140, indicating the resolution status of environmental issues. Upon action or resolution, the system updates visual indicators, such as transitioning from warning colors (e.g., red, orange, or yellow) to green, on an interactive heatmap. This enhances transparency and allows users to observe the impact of their reports. In alternative embodiments, the visualization may include gradient-based indicators reflecting partial resolution, or temporal overlays illustrating the progression of issue resolution over time. Additional embodiments may incorporate animated transitions or historical playback features to provide deeper insight into environmental improvements.
[085] The embodiments of the present disclosure as disclosed herein are intended to be illustrative and not limiting. Other embodiments are possible and modifications may be made to the embodiments without departing from the spirit and scope of the disclosure. As such, these embodiments are only illustrative of the inventive concepts contained herein.
, Claims:1. A system (100) for real-time environmental monitoring and response coordination, the system comprising:
a user interface module (110) configured to receive environmental reports from a one or more users, each environmental report comprising geotagged data, textual input, visual evidence, and a severity level, and to transmit the environmental reports to a data processing module (120);
a data storage module (130) configured to store the environmental reports including geospatial coordinates, timestamps, and associated metadata;
the data processing module (120) operatively coupled to the user interface module (110) and the data storage module (130), and configured to:
generate a textual embedding from the textual input using a language model (121) and a visual embedding from the visual evidence using a vision model (122);
combine the textual embedding and the visual embedding using a multimodal fusion mechanism (123) that applies dynamically adjusted weighting factors based on confidence scores associated with each modality to generate a unified feature representation;
classify and categorize the environmental reports based on the unified feature representation;
perform spatiotemporal clustering (126) of the environmental reports based on geographic proximity and temporal proximity to identify environmental events; and
compute an urgency score (127) for each environmental event based on a weighted function of report density, severity levels, and rate of report occurrence over time;
a visualization engine module (140) operatively coupled to the data storage module (130) and the data processing module (120), and configured to generate geospatial visualizations including cluster-based heatmaps based on the environmental events and corresponding urgency scores; and
a stakeholder interaction module (150) operatively coupled to the data processing module (120) and the visualization engine module (140), and configured to transmit notifications to stakeholders based on the environmental events and urgency scores and to track resolution status of the environmental events.
2. The system (100) as claimed in claim 1, wherein the confidence scores associated with each modality are determined based on one or more of model prediction confidence, data quality metrics, or completeness of the input data.

3. The system (100) as claimed in claim 1, wherein the multimodal fusion mechanism (123) comprises a weighted combination of the textual embedding and the visual embedding using dynamically computed weighting factors.

4. The system (100) as claimed in claim 1, wherein the spatiotemporal clustering (126) is performed using a density-based clustering algorithm that groups environmental reports based on a predefined spatial radius and temporal window.

5. The system (100) as claimed in claim 1, wherein the urgency score (127) is computed using a weighted model comprising a number of reports within a cluster, an average severity level, and a rate of incoming reports over time.

6. The system (100) as claimed in claim 1, wherein the stakeholder interaction module (150) comprises an event-driven alert system configured to generate notifications when the urgency score exceeds a predefined threshold within a defined geographic boundary.

7. The system (100) as claimed in claim 1, further comprising a data validation module (124) configured to perform one or more of duplicate report detection, image metadata verification, anomaly detection, and moderation of environmental reports.

8. A method for real-time environmental monitoring and response coordination, the method comprising:
receiving environmental reports from a plurality of users, each environmental report comprising geotagged data, textual input, visual evidence, and a severity level;
generating a textual embedding from the textual input and a visual embedding from the visual evidence;
combining the textual embedding and the visual embedding using a multimodal fusion mechanism that applies dynamically adjusted weighting factors based on confidence scores associated with each modality to generate a unified feature representation;
classifying the environmental reports based on the unified feature representation;
performing spatiotemporal clustering of the environmental reports to identify environmental events;
computing an urgency score for each environmental event based on report density, severity levels, and rate of report occurrence over time; and
generating notifications and geospatial visualizations based on the environmental events and urgency scores.

9. The method as claimed in claim 8, wherein the textual embedding and the visual embedding are generated using models trained on a training dataset comprising historical environmental reports, annotated textual descriptions, labeled image data, and associated severity classifications.

10. The method as claimed in claim 8, wherein the dynamically adjusted weighting factors are computed in real time based on relative confidence scores of the textual embedding and the visual embedding.