A graph-based evaluation framework for smart disaster response systems using IoT design quality metrics Authors: Iqra Qayyum, Tahir Alyas, Qaiser Abbas, Turki Alghamdi, Nasser Aljohani, Saad Said Alqahtany, Muhammad Shoaib, Aidarus Mohamed Ibrahim Published: 2026 ABSTRACT AbstractDisaster-response systems have been progressively installed in smart cities through IoT-enabled systems that assist in early warning, situational awareness, and coordination of emergencies. Nevertheless, the majority of existing studies measure such systems by means of data analytics, accuracy of prediction, or anomaly detection by AI, and the quality of the system-level design is not quantified in most cases. Specifically, no standardized framework exists to evaluate the complexity of architecture, resilience, recoverability, and reusability of IoT disaster-response systems before their implementation. The paper presents a graph-based design evaluation model, which represents the IoT disaster-response architectures as layered graphs and suggests a collection of quantitative design quality measures. In contrast to the learning-based approaches, the suggested framework is also algorithm-free and assesses the structure of the system, which can be compared objectively among the various IoT system setups. The framework is validated through a flood-response case study conducted in Pakistan where changes in sensor density, gateway connectivity, and interactions on decision-layers are examined. Evidence indicates that network redundancy is a validated variable that minimizes the complexity of detection and increases system recoverability, whereas sparse connectivity enhances the severity propagation and response latency. The developed framework offers a viable pre-deployment assessment tool, which augments AI-based monitoring systems and that will aid in robustly designed disaster-response systems.Keywords: Internet of Things, Design quality metrics, Smart city, Disaster response frameworkSubject terms: Environmental sciences, Engineering Introduction Disasters have the potential to cause widespread destruction and loss of life; therefore, it is essential that disaster response systems be able to respond to these events quickly and effectively. However, the unpredictable and complex nature of disasters can make this challenging. The design and evaluation of disaster response systems involves integrating multiple components, such as sensors, communication networks, and decision-making algorithms1. In recent years, there has been a growing focus on developing smart disaster response systems that leverage advanced technologies, including artificial intelligence, machine learning, and the Internet of Things (IoT), to enhance the effectiveness and efficiency of disaster response. However, despite these efforts, there is a lack of a comprehensive evaluation framework for disaster response systems that can provide a systematic and consistent assessment of the system’s ability to respond effectively to disasters. A disaster management framework encompasses essential stages, including preparedness, response, recovery, mitigation, and adaptation2,3. Research on the influence of sealing ratio on the combustion process and temperature field of pool fires has been conducted extensively, but it remains insufficient for emergency rescue after the original fire extinguishing system fails4. Measuring design quality according to the defined rules of any disaster response framework can be useful in resolving problems efficiently5. This research addresses the challenge of designing effective and efficient disaster response systems for smart cities. Disasters can cause widespread destruction and loss of life; therefore, response systems must be able to respond to these events quickly and effectively6. However, the unpredictable and complex nature of disasters can make this a challenging task. Metrics are designed to check and compare a system efficiency, maintainability, performance, understandability, and complexity7. By defining design quality metrics for smart disaster response systems and incorporating them into the design process, it is possible to ensure that the smart city systems are optimized for effective disaster response. Rockfalls are a significant geohazard in mountainous regions, with a catastrophic impact potential that could lead to the failure of dangerous rock masses within their travel distance. In the context of slope risk assessment, a comprehensive understanding is needed of the impact of failure processes and the dynamic behaviour of rockfall movements. The objective of this work is to examine how rockfall failure behaviours and disaster processes affect the hazardous rock mass in advance by using three-dimensional discontinuous deformation analysis (3D DDA)8. Abstract graph schema representations offer a flexible and intuitive approach to modelling the various components and relationships within a disaster response system, facilitating the identification of areas for improvement and supporting effective decision-making during a disaster. Overall, the goal is to improve the quality of disaster response systems and ensure that they can effectively address the complex and unpredictable nature of disaster events. By defining design quality metrics for smart disaster response systems, it is possible to ensure the development of high-quality systems optimised for effective disaster response. While efforts have been made to develop disaster response systems, a systematic and consistent approach is needed to evaluate these systems and identify areas for improvement. Current evaluation frameworks for disaster response systems often focus on individual components or specific aspects of the system, such as the accuracy of sensor data or the speed of alert dissemination. However, there is a need for a holistic evaluation framework that considers the complexity and interrelated nature of the disaster response system as a whole. Despite the increasing importance of disaster response systems, there is a lack of a comprehensive evaluation framework that considers the complex and interrelated nature of disaster response. By incorporating these metrics into the design process, it is possible to ensure the development of high-quality disaster response systems that are optimised for effective and efficient operation in the event of disasters2,5,6. Metrics for measuring design quality are crucial because if an issue develops after the system is implemented and deployed, it could result in system failure, financial losses, or even risk to human life in the case of a health and hazards management system9. These metrics can evaluate the existing framework at its design level. Thus, designing quality metrics for a smart disaster response framework can help choose a robust framework to respond to in any disastrous situation. The main contributions of this study are as follows. To establish a design phase acceptance standard for the quality of services provided by disaster response systems, design quality metrics are proposed for the IoT ecosystem of smart cities. To establish a design phase acceptance standard for the quality of services provided by disaster response systems, design quality metrics are proposed for the IoT ecosystem of smart cities. Using an innovative flood response Internet of Things system as a case study, we evaluate the proposed metrics. The outcomes demonstrated that the proposed metrics help determine the quality level of the IoT design model in disaster response activities. Using an innovative flood response Internet of Things system as a case study, we evaluate the proposed metrics. The outcomes demonstrated that the proposed metrics help determine the quality level of the IoT design model in disaster response activities. The proposed metrics are useful in evaluating the design quality of the Internet of Things for disaster response. These metrics can be used to compare different smart IoT disaster response frameworks for improved quality. The proposed metrics are useful in evaluating the design quality of the Internet of Things for disaster response. These metrics can be used to compare different smart IoT disaster response frameworks for improved quality. When studying disaster management using IoT, most prior research has focused on individual system factors, such as sensor accuracy or alert dissemination mechanisms. This study addresses this gap by proposing a comprehensive evaluation framework consisting of nine design quality metrics that assess disaster-response systems at the architectural level. Among these, three metrics are novel contributions: (i) Disaster Sensor Health Monitoring (DSHM), which evaluates sensor reliability during disaster scenarios; (ii) Alert Dissemination Efficiency (ADE), which assesses the effectiveness of alert propagation across communication channels; and (iii) Disaster Component Reusability (DCR), which examines the reuse potential of system components across multiple disaster scenarios. Furthermore, existing metrics such as Disaster Detection Complexity (DDC) and Disaster Severity Identification (DSI) are extended using graph-based formulations to enable system-level analysis beyond single-dimensional indicators. The proposed framework is validated through a simulated flood-response case study based on the 2022 Pakistan floods, demonstrating its applicability in complex multi-region disaster scenarios.Within the framework proposed in the study, a multidimensional set of nine metrics on design quality are introduced, which together assess the disaster detection, severity estimation, disaster-type identification, sensing efficiency, data integrity, sensor-health stability, alert dissemination, component reusability and system recoverability. This breadth enables the evaluation of the lifecycle of disaster-response in a holistic manner instead of considering separate sub systems, a weakness of previous studies. Although artificial intelligence and machine learning are rapidly adopted in disaster-response systems, the current assessment practices are largely algorithm-focused, and they focus on the accuracy of prediction, the classification performance, or the ability to detect anomalies. These methods provide a little information about the quality of the underlying IoT architecture, especially on complexity, robustness, recoverability, and reusability of components. Therefore, the system designers do not have objective guidelines to determine the resiliency and scales of a disaster-response framework prior to deployment. Various important performance metrices are employed in this research to determine the quality of design of the IoT-based disaster-response systems. The complexity of Disaster Detection (DDC) indicates the degree of interaction and communication overhead in a system, and the efficiency of Alert Dissemination (ADE) is a measure of the efficiency of alert dissemination among stakeholders. Disaster System Recoverability (DSR) is the capacity of the system to recover its functionality following failures and Fault Tolerance (FT) is the capacity of the system to service incomplete component failures. Early Warning System Accuracy (EWSA) is a measure of how accurate alerts generated by the system are, and Cost Efficiency (CE) is a measure of the tradeoff between the performance of the system and the cost of its operation. The combined metrics give a holistic assessment of the system performance, resilience and the operational efficiency. In order to fill this gap, this paper presents a graph-based, design-oriented assessment scheme that models IoT disaster-response systems in layered graph forms. Nine quality metrics of quantitative design are suggested to assess complexity of detection, spread of disaster severity, sensing efficiency performance, data integrity, sensor health, alert propagation, recoverability and component reusability. The proposed framework is also not based on any learning models like AI-driven disaster-response models, which rely on available training data and predictive algorithms to provide the optimal design of IoT. This approach involving metrics and structuring is the main innovation of the work. In addition, by defining the whole smart-disaster-response architecture as an abstract graph schema we are structurally representing the dependencies between sensors, gateways, controllers and response-layer actuators. The graph-based view offers a formal framework to locate the bottlenecks, study the interactions between components, and calculate the quality measures, which are based on metrics. This method enhances interpretability and provides a common mathematical support to assess the performance of SDR in different situations. The rest of the paper is organized as follows. Section Literature review provides background on disaster response systems and the challenges associated with designing effective and efficient systems. Section 3 presents the Abstract graph schema representation of a smart disaster response framework. The proposed evaluation framework and design quality metrics for smart disaster response systems are presented in Section Methods. Section Results and analysis presents the results and analysis of the proposed framework. Finally, Sect. Conclusion concludes the paper. Literature review It is important to be able to predict landslide displacement for reliable landslide warning and forecasting. AI models have been employed for forecasting the displacement of step-like landslides. Nevertheless, a quantitative evaluation of the impact of rainfall and reservoir water levels on landslide displacement is challenging, and the interpretability of models for typical step-like landslides can be enhanced10. Metrics help eliminate the uncertainty of defects when assessing software quality and provide a factual basis for making informed judgments regarding quality. Any software can be tested in two ways, as proposed by Belachew et al.11, either manually (a formal method) or through metrics (a systematic measurement technique). However, the manual method is so costly that developers and analysts moved to software metrics to measure the system’s quality. This work demonstrates that a measurement-based modelling approach based on adaptive modelling can be used to recreate system behaviour when complete structural information is not known- analogous to the incomplete or uncertain IoT conditions our SDR framework is used to assess system performance. Its method of comparing substructures and mapping of latent responses is in line with our graph-based metrics that predict the quality of disaster-response even where some of the components of the response or interaction cannot be directly measured12. Efficient data collection from resource-constrained sensors is critical to the success of the next-generation Internet of Things (IoT). In this paper, we consider the problem of minimising energy consumption in data collection from a generic wirelessly powered sensor13. Moreover, these are designed for software quality, cost estimation, defects analysis, software size, testing, and maintainability. The three most important quality metrics — complexity, maintainability, and functionality — are covered in this paper by Ergasheva and Kruglov9. It is still unclear about the disturbance and damage laws of the surrounding rock during such tunnel construction, resulting in frequent occurrences of roadway collapse, settlement, and other engineering hazards. The current theories and methods for predicting ground settlement are still not accurate enough14. Different cities have implemented the smart city framework to enhance citizens’ lives by effectively handling natural disasters. Kumar et al.15 proposed a peer-vetted creative crowdsourcing process as part of India’s smart city transformation framework, which indirectly contributes to disaster preparedness and response. Four decisive areas illustrate the city transformation: Information and Communication Technology (ICT) infrastructure, physical infrastructure, planning, and deploying smart city solutions. However, the framework does not directly support disaster response solutions. The recent advancement and development in technology, combination, and information demand a resilient framework for handling disasters in coastal cities. Tonmoy et al.16 developed a disaster resilience framework utilising IoT for flood management. The framework utilised a four-layered architecture to enhance the natural disaster resilience of coastal communities. However, there is a lack of decision support systems specifically designed for smart cities to enhance disaster management. To address the lack of decision-making in disastrous situations, Jung et al.17 propose an Intelligent Decision Support System (IDSS) conceptual framework for disaster management in South Korea. Using AI and big data for Intelligent Decision Support, this framework leverages big data and Convolutional Neural Network (CNN) to mitigate and respond to natural disasters, such as heat/cold fires, as well as wildfires. The disaster layer was related to admin and used decision support analysis algorithms. However, the framework has mitigation risks and a lack of system security, resulting in less accuracy in precisely detecting disasters. A smart city spatial disaster management framework by Arepalli et al.18 has been designed to overcome Amravati city’s precision and accuracy issue. This framework used remote sensing, IoT, and Geospatial tools for preparedness, response, and recovery from natural disasters. However, this disaster framework is specific to a river (water-based disaster) and lacks the versatility and adaptability found in existing frameworks. Every year, Indonesia faces countless disasters, whether natural or man-made. Therefore, research was conducted on a smart system that handled multiple types of disasters to analyse all possible disaster management technologies. A smart disaster management system was proposed by Sukmaningsih et al.19 to respond to natural disasters by applying thematic tweet classification. The disaster layer that handled the response was related to the internet. However, the social media approach lacks intelligent disaster technologies to make the system more resilient. The lack of intelligent disaster technologies was cover by implementing a data-driven approach using Artificial Intelligence. Integrating a smart, resilient system with critical infrastructure made real-time predictions for earthquakes and fires in Lisbon (Portugal). Smart City 4Forces by Elvas et al.20 proposed a framework for decision-making in disaster response that enables timely action. However, more work is needed to utilize infrastructure development to mitigate risks, and the framework is limited to just one city. West Java (a province of Indonesia) meets high disaster rates yearly. West Java resilience culture province framework proposed in Janitra21 to handle natural disasters. The researchers sought to adopt digitisation as a means to mitigate post-disaster losses. The major components of the proposed framework are resilient knowledge, citizen, infrastructure, resilience centre, ecology, index, and financing. Despite this, there remains a need for a comprehensive response framework to manage multiple disasters in a smart city. Another theoretical IoT-based framework, presented by Sood et al.5, was proposed to mitigate disaster losses. The proposed technique was based on an IoT cyber-physical framework that handled both natural disasters and panic-oriented disasters. A fog-cloud-centric IoT- based system utilized three layers: the IoT sense layer, the cloud layer, and the fog layer. However, the proposed system lacks efficiency by ignoring the sensor’s energy and offloading the fog layer. The world is undergoing urbanisation at an accelerated rate, resulting in increasingly common and pronounced urban ecological problems. It is relevant to assess the quality in ecological sense of an urban area for the control of environment-oriented policies22. The framework has a four-tier architecture (sensor, database, decision support, and alert). However, there is no dynamic information monitoring for decision-makers. A theoretical smart city framework based on GIS-mobile IT, proposed by Hartama et al.23, is proposed for North Sumatra for natural disaster management. The proposed system and an internet-based disaster layer are based on mitigation and disaster recovery. However, the system lacks the main disaster response stage with no practical implementation of disaster risk reduction. A flooding resilience disaster information framework, proposed by Kumar et al.24, is designed for smart communities to enhance preparedness and facilitate the response phase. The proposed system was designed to predict disasters and get knowledge-based recommendations using sensors and physical and demographic data. However, there is a need to evaluate and validate the smart city framework that collaborates with the local emergency managers. The Zaman et al.25 study proposed the smart city approach for disaster management in Pakistan. The Decomposed Theory of Planned Behaviour (DTPB) is the hypothesis proposed for disaster management. An integrated evaluation Multi-Criteria Decision-Making (MCDM) pathogenic framework by Abdel-Basset et al.26 is proposed to measure the performance of smart disaster response systems under uncertain situations. The integrated smart disaster frameworks could provide a high level of consideration in decision-making. However, researchers need to modify the framework techniques to align with modern technology, thereby achieving better accuracy. Recent research has noted the drawback of the fixed threshold choice in dynamic and heterogeneous IoT settings. Alserhani et al.27 showed that fixed cut-off values may worsen the reliability of systems when sensor behaviour and network characteristics are changing, and suggested adaptive threshold mechanisms as a means of higher adaptability in the assessment of integrity in IoT. On the same note, Leong28 has pointed out that though threshold-based triggers would still be critical in the IoT-based disaster and emergency monitoring system, threshold-based triggers would have to be empirically tuned to the environmental and operational conditions to prevent false alarms or slow reaction to disasters or emergencies. Moreover, it was demonstrated by Gad et al.29 that optimizing the threshold based on data is significantly better in terms of the reliability of decisions in computational system, which supports the idea that arbitrarily chosen cut-offs can have an adverse effect on the system-level evaluation. All these findings are encouraging reasons and indicate the necessity of an empirically supported and flexible threshold policy in the framework of disaster-response appraisal of IoT. Hartama et al.6 utilised IT resources to respond to natural disasters promptly under the umbrella of the smart city concept in North Sumatra and Aceh. The proposed study integrated the mobile IT and Geographic Information Systems to gather geospatial information. However, the system must build such strategies to predict future disaster management. The worst and most worrying type of drought is hydrological drought, which has numerous harmful long-term effects on natural ecosystems, agricultural potential, and the water balance in a given area30. In another paper, Boukerche and Coutinho2 discussed disaster response and smart detection for smart cities to address the lack of connectivity issues during or after a disaster. However, the researchers pointed out some future challenges for deploying smart system components, including sensing, processing, and response. The above in-depth literature survey describes the major implementation of a smart city framework for disaster management, aiming to make people’s lives more secure in a city. Different smart city frameworks were proposed for famous world cities, such as West Java, Lisbon, Taiwan, South Korea, Amaravati, North Sumatra, and Aceh. Most frameworks are based on multiple advanced technological layers such as ICT, IoT, and AI technologies for implement tion and deployment purposes. System reliability, fault propagation, and risk assessment of engineered systems have been common in the traditional system evaluation methods of Fault Tree Analysis (FTA), Reliability Block Diagrams (RBD), and Failure Mode and Effects Analysis (FMEA). FTA pays attention to determination of possible failure paths based on hierarchical logical structures, whereas RBD models rely on system reliability based on component interactions and dependencies on failures. On the same note, FMEA critically assesses failure modes and their effects on the performance of the system. Nevertheless, the techniques are mainly aimed at determining the static reliability and fail to explicitly represent dynamic data flow, multi-layer interactions, and real-time patterns of communication that are inherent to the disaster-response system that uses IoT. By comparison, the proposed graph-based frameworks represent the dynamics of structural connectivity and interaction between sensing, networking, decision, and response layers. This allows a more comprehensive analysis of system complexity, timeliness and information propagation. Thus, the suggested framework is a complement to the traditional methods of reliability since it can be extended to the current setting of IoT-based systems, where the reliability of the individual systems is not the sole aspect of the system performance but also the efficiency of the communications, data integrity, and adaptability in response. This work revealed that most frameworks lack the measurement aspects at their design level. Moreover, there is a need for quality metrics to make smart disaster response frameworks robust, ideal, and heterogeneous. The metrics can enhance the quality of disaster response strategies to save lives in the face of upcoming disasters. A research matrix for existing framework analysis is presented in Table 1. Table 1Research matrix for existing frameworks analysis.Author & YearSmart region (case study)Proposed frameworkTechnology usedDisaster typeDisaster response layerDesign metrics for evaluationKumar et al.15IndiaSmart city transformation frameworkPeer-vetted crowdsourcing-✓×Tonmoy et al.16Coastal areasDisaster resilience frameworkIoTFlood××Jung et al.17South KoreaIDSS conceptual frameworkCNNHeat/cold fires✓×Arepalli et al.18Amravati citySmart city spatial disaster management frameworkGeospatial tools & IoTFlood××Sukmaningsih et al.19IndonesiaSmart disaster management systemTweet thematic classificationFlood✓×Elvas et al.20Lisbon (Portugal)Smart city 4 forcesAIEarthquake/fires✓×Janitra21West JavaWest Java resilience culture province frameworkICT basedNatural✓×Sood et al.5-IoT cyber-physical frameworkFog, cloud, IoT-basedNatural✓×Hong et al.22TaiwanDisaster alerting framework based on smart citysensor-basedEarthquake✓×Hartama et al.23North Sumateratheoretical smart city frameworkGIS-mobile ITNatural××Kumar et al.24FloodFlood disaster resilience informa- tion frameworkSensor-basedNatural✓×Zaman et al.25PakistanDecomposed Theory for Planned BehaviorBig dataNatural××Khalifeh et al.27-IoT cyber-physical frameworkWSNsNatural & man-made××Hayajneh et al.28-Drone Empowered Small Cellular NetworksDroneNatural××Alsamhi et al.29--IoT and droneNatural××Hartama et al.6North Sumatra & AcehIT Resource FrameworkIT and geospatialNatural✓×This PaperPakistan (Dam, Hills)Smart City Disaster Response frameworkIoT, AI, CloudAll Natural✓✓ Research matrix for existing frameworks analysis. Abstract graph schema representation of a smart disaster response framework This study employs the disaster response framework proposed by Hong et al.22, which integrates various resources within a smart city Internet of Things network to enhance its resilience to natural disasters. The abstract graph schema representation for this smart disaster response framework utilises graph data structures to model its various components and relationships, as illustrated in Fig. 1. The collaborative smart city disaster response framework addresses heterogeneity and supports an Artificial Intelligence-based decision-making process, making it resilient to on-time disaster response. The framework is based on four major layers (i.e., sensor layer, networking layer, decision layer, and response layer) that are essential for providing evacuation plans during disaster response. Abstract schema representation for the disaster response framework. The base layer (named sensor layer) consists of disaster-specific sensors, actuators, and a gateway. These sensor devices monitor real-time environment data and pass it to the IoT disaster-detecting gateway node. On the other hand, actuators perform various actions in disastrous situations. Sensors, devices, and actuators communicate via the gateway. Data coming from sensors is further filtered and formatted by the gateway. When the sensor value approaches/exceeds the threshold value (set by decision-makers), it indicates that disaster will happen and generates alert signals. Above the IoT disaster-detecting gateway (sensor layer), the cloud data collection platform (network layer) logs the sensor’s data. The cloud platform is responsible for extracting meaningful disaster data and then integrating and storing it on the cloud. The data can be updated and retrieved between the platform and the gateway. In addition, this layer keeps examining and updating sensor data as per controller demand (the layer also processes/extracts data using advanced artificial intelligence techniques). Furthermore, decision-makers and experts interact with the decision layers that are above the network layer. In this layer, the AI decision-making controller is dedicated to processing high-level disaster data and making decisions based on the current disaster situation. The system converts sensor data into useful information for precise decision-making in disaster response. The controller addresses disastrous situations through risk area analysis and informs the response layer about potential disaster situations. Furthermore, the decision-making controller helps alert-generating devices (response layer) generate alerts and disseminate them with the aid of actuators. The controller identifies the area under threat as well. In the response layer, IoT alert-generating devices handle various disaster responses, including alerts, areas, information, and resources. However, this sends information to government bodies to report damage, provides alternative routes to rescue teams, and alerts emergency operation centres to the resources needed. Also, these IoT devices disseminate alert messages to citizens via various channels (i.e., phone, TV, radio, and social media). Module representation of framework For the disaster response framework in a smart city Internet of Things network (Fig. 1), we define each node of the graph with the following terms. Sensing Devices: St — Sensor and actuators for disaster gateway devices: Gs — IoT disaster-detecting gateway low-end devices: Pc — Cloud data collection platform middleware devices: Cd — AI decision-making controller high-end devices: Ar — IoT alert-generating devices. The integer numbers t, s, c, d, and r represent the number of devices integrated with each framework layer. However, there can be more than one of these devices, depending on the network integration. The underlying bidirectional relationship of edges and vertices show some meaningful interaction that connects each framework layer. The proposed framework, though represented with the help of the flood-response scenario, is based on a graph-based abstraction that can be customized to other disaster types, including earthquakes, wildfires, and landslides. The layered representation (sensor, network, decision, and response) is the same, but the particular elements of each layer may vary depending on the disaster type. An example of this is that the flood sensors could be substituted by seismic sensors, accelerometers and vibration sensors during earthquake conditions and thermal sensors and smoke sensors could be used during wildfire monitoring. Likewise, there may be variations in communication needs and decision-making timeframes based on disaster types. Ultra-low latency and quick alert dissemination is needed to respond to time-critical events like earthquakes, whereas slower-onset events like floods or landslides may have sensing coverage and data aggregation as priority. Thresholds, weights, and component definitions can be modified to refine the proposed metrics to these domain-specific needs. Table 2 shows how the suggested graph-based IoT disaster-response model could be tailored to various disaster types by changing the type of sensors, focusing on particular design quality indicators, and changing communication needs and decision-making timeframes. It underlines the adaptability of the framework to different working conditions without changing the evaluation framework. Table 2Adaptability of proposed framework across different disaster scenarios.Disaster typeSensor typesKey priority metricsCommunication requirementsDecision timelineFloodWater level sensors, rain gauges, ultrasonic sensorsDSI, SC, DSRModerate bandwidth, continuous monitoringMedium (minutes to hours)EarthquakeSeismic sensors, accelerometers, vibration sensorsDTI, ADE, DSRUltra-low latency, real-time alertsVery fast (seconds)WildfireTemperature sensors, smoke detectors, gas sensorsDSI, ADE, SCWide-area coverage, rapid broadcastFast (seconds to minutes)LandslideSoil moisture sensors, tilt sensors, geophonesSC, DSI, DSRModerate bandwidth, periodic updatesMedium (minutes)Urban FireHeat sensors, flame detectors, CO sensorsDTI, ADEHigh-speed alert disseminationVery fast (seconds) Adaptability of proposed framework across different disaster scenarios. Methods This section defines several metrics to measure the quality and make our smart disaster response IoT framework responsive to disastrous situations Disaster Detection Complexity Metrics (DDC). Table 3 contains a summary of the entire list of symbols and parameters that were used in the suggested graph-based assessment model, and it contains similar definitions and measurement units, which are intended to help interpret and reproduce the analytical models in a transparent manner. Table 3Unified notation and symbol definitions.SymbolDescriptionUnit/domain (St) Number of disaster-specific sensorsCount(Gs)Number of IoT disaster-detecting gatewaysCount(Pc)Cloud/middleware data-collection platformsCount(Cd)AI decision-making controllersCount(Ar)Alert-generating/response devicesCount(R)Disaster-affected regionIndex(R_n)Total number of regionsCount(T)Total number of disaster typesCount(E)Total number of system entitiesCount(C)Communication channelsCount(DDC)Disaster detection complexityEdge count(DSI)Disaster severity identification metricRatio [0,1](DTI)Disaster timeliness index metricRatio [0,1](SC)Sensing complexityCount(DDI)Decision data integrity metricRatio [0,1](DSHM)Disaster sensor health monitoringRatio [0,1](ADE)Alert dissemination efficiencyRatio [0,1](DSR)Disaster system recoverabilityBinary / Ratio(DCR)Disaster component reusabilityRatio(\phi, \chi, \psi, \omega)Inter-layer edge mappingsGraph edges Unified notation and symbol definitions. Disaster detection complexity metrics (DDC) The Disaster Detection Complexity (DDC) metric is a measure of the structural complexity of the IoT-based disaster-response system, which counts the number of interactions that the system will need to operate to identify and react to a disaster event. The proposed framework represents the system as a graph G (V), E), with V representing the system components (sensors, gateways, controllers and actuators) and E representing communication connections among these system components. Formally, the disaster detection complexity is defined as the total number of edges involved in the detection and response process as define in Eq. (1): where E is the set of all interaction edges in the system and is the weight of edge e. In the basic formulation, all edges are assumed to have a uniform weight , representing a single interaction (request or response). For a system consisting of n interaction paths, the total disaster detection complexity can also be expressed in Eq. (2) : where represents the number of interactions in the detection-response path. The average disaster detection complexity across multiple scenarios is computed in Eq. (3): where denotes the complexity of the scenario. that represents the total number of interactions within the system. A higher DDC value indicates increased system interaction complexity, which may result in greater communication overhead and reduced efficiency, whereas lower values reflect simpler and more efficient system designs. Based on this, the complexity is categorized into three levels: Worst Case (< 5), Average Case (5–10), and Best Case (> 10), representing increasing levels of system interaction and operational capability. Design level disaster detection complexity metrics The Design-Level Disaster Detection Complexity (DDC design) is an index of the complexity of the IoT disaster-response architecture at the architectural design level. This metric contrasts with the operational version of the DDC metric, which assesses runtime interactions, and represents the general connectivity and integration of system components, at all levels. The IoT disaster-response system is represented as a graph , where V denotes system components and E denotes communication links. At the design level, the complexity is defined as the total number of possible interconnections among system components as defind in Eq. (4): where represents the total number of edges in the system graph, assuming uniform interaction weight. Let t,s,c,d,r denote the number of devices in the sensor, gateway, cloud/platform, decision, and response layers, respectively. The total number of interconnected devices is given by Eq. (5): The design-level complexity can also be interpreted in terms of connectivity per node as defined in Eq. (6): where represents the number of connections associated with node . The average design-level complexity is computed as defined in Eq. (7): The design-level disaster detection complexity lies in the range , where higher values indicate a well-connected and efficient architecture, while lower values reflect sparse connectivity and potential bottlenecks. The complexity is categorized as Worst (< 0.5), Average (0.5–0.8), and Best (> 0.8), representing increasing levels of system robustness and communication efficiency. Disaster severity identification metric (DSI) The Disaster Severity Identification (DSI) metric quantifies the intensity of a disaster within a specific region based on sensor observations. In the proposed framework, disaster-specific sensors deployed across different areas continuously monitor environmental conditions, and their readings are used to estimate the severity level of the disaster. Formally, the DSI metric is defined as the ratio of active or threshold-exceeding sensor readings to the total number of deployed sensors in a given region as defined in Eq. (8): where represents the total number of sensors in the region, and denotes the normalized status of the sensor (active = 1, inactive = 0). A region may consist of multiple areas or gateways, represented by , which collectively contribute to the overall severity estimation. The DSI value falls in the range [0,1] with higher values indicating a higher level of disaster. The severity levels are defined as low (< 0.5), medium (0.50–0.8), and high (> 0.8) based on the predefined thresholds. The Smart Decision-Making Agent (SDMA) defines these thresholds and aids in making decisions in time to respond to an emergency and allocate resources. Disaster timeliness index metrics (DTI) The Disaster Timeliness Index (DTI) is used to assess how responsive the system is to disaster events, how well it responds to them within an acceptable time frame, according to the activation and processing of disaster-specific sensors. In this model, it takes only one type of sensor as the one that can detect a given disaster in a certain area at a particular time. Formally, for a single region, the DTI metric is defined in Eq. (9): where represents the total number of relevant disaster-detection sensors in region R, and denotes the activation status of the sensor (active = 1, inactive = 0). The decision layer component processes these sensor inputs to identify the disaster type. For multiple regions and disaster types, the DTI metric is defined in Eq. (10) : where N is the number of regions, and represents the timeliness index for region j. The DTI value lies within , where higher values indicate faster and more reliable disaster identification, while lower values reflect delayed or insufficient response. A threshold of DTI ⩾ 0.5indicates successful identification of disaster type within the required time window. Sensing complexity metrics (SC) Sensing Complexity (SC) measure measures the effectiveness and reliability of disaster sensing; it is used to measure the number of active sensors that detect hazards in various locations. Effective sensing is very important in minimizing risks and enhancing response time by availing proper and timely information during the occurrence of a disaster. Formally, the sensing complexity for a region is defined in Eq. (11): where represents the number of active disaster-detection sensors in region , and is the total number of regions in the IoT system. The average sensing complexity across all regions is computed by Eq. (12): Similarly, the sensing complexity with respect to disaster types is defined in Eq. (13): where T represents the total number of disaster types observed across regions. Sensing complexity is defined in positive real numbers where larger numbers correspond to good sensing coverage and better system reliability and smaller numbers correspond to inadequate sensing and increased likelihood of late detection. The complexity is divided into Worst (< 5.0), Average (5.0–8.0) and Best (> 8.0) that are the growing sensing effectiveness and system responsiveness. Decision data integrity metric (DDI) The Decision Data Integrity (DDI) measure assesses the accuracy, consistency, and reliability of data transferred between IoT disaster-response system network layer and the decision layer. The integrity of data is necessary to make the right decisions since the controller of the decision-making would base the process on the proper sensor information to evaluate the state of the disaster and produce the right responses. Formally, the DDI metric is defined in Eq. (14): where DAT represents the number of data access attempts made by the network layer, denotes the total number of authenticated data access attempts, and IE~ represents the number of data integrity errors encountered during data transmission or retrieval. DDI value is in the range [0,1], the closer to 1 the higher the data integrity and the few errors in data, which means that there is reliable decision-making, and the lower it is, the more data inconsistencies and the chances of system failure. Disaster sensor health monitoring metric (DSHM) Disaster Sensor Health Monitoring (DSHM) measure is used to assess the stability and functionality of sensors used within the sensor layer of the IoT disaster-response system. As sensor performance can deteriorate or even malfunction in case of disaster conditions, sensor health monitoring is important to make sure that the data are collected correctly and that the entire system operates properly. Formally, the DSHM metric is defined in Eq. (15): where represents the total number of sensors in a region, and denotes the status of the sensor (active = 1, inactive = 0). The DSHM value is in [0,1], so values near 1 imply that the majority of sensors are operating correctly, and the system will be reliable, whereas values near 0 imply a greater rate of sensor failures and lower reliability of the system. To determine how well the sensor is performing, the threshold is set at 0.5, which represents bad sensor health (worst case) and good sensor health (best case). Alert dissemination efficiency metric (ADE) The Alert Dissemination Efficiency (ADE) measure is used to assess the efficacy of the transmission of alerts by the response layer to the concerned stakeholders such as government agencies, NGOs, and the masses, via several communication mediums. The dissemination should be efficient so that the response and coordination during disasters are timely. Formally, the ADE metric is defined in Eq. (16): where represents the total number of communication channels, denotes the total number of entities involved in receiving alerts, indicates the status of the channel, and represents the status of the entity (active = 1, inactive = 0). The ADE value is in [0,1], where the higher the value, the more efficient the notification of the alert in various channels and organizations and the timely response, and the lower the value, the more communication gaps and slow delivery of information. The value of 0 is the worst-case scenario, and 1 is the optimal dissemination efficiency. Disaster system recoverability metric (DSR) Disaster System Recoverability (DSR) measures how well the IoT disaster-response system can recover the failed components and resume the normal operation within a reasonable time. It is a measure of the resilience of the system by determining the effectiveness of the failed components in being detected, repaired and activated back into service in the event of a disaster. Formally, the DSR metric is defined in Eq. (17): where represents the number of system components successfully restored to operational status, and denotes the total number of components in the system. The value of DSR is in the range of [0 1], the higher it is closer to 1, the higher is the system recoverability and system resilience, and the closer it is to 0, the lower is the system recoverability and system failure. The metric is defined as Worst Case (< 0.5), which means that response activities were not undertaken; Average Case (0.5–0.8), which means that the recovery of the system was not efficient and reliable; and Best Case (> 0.8), which means that the recovery of the system was efficient and reliable. Disaster component reusability metrics (DCR) The Disaster Component Reusability (DCR) is a metric that measures the reusability of system components in various disaster events within the IoT disaster-response framework. Reusability increases flexibility in the system, lowers the cost of deployment, and improves the scalability of systems as components can be used in multiple functions under different conditions. Reuse factors of the individual parts are determined using reuse factors that are defined on the different layers of the system: Sensor Reuse Factor (SRF), Low-End Device Reuse Factor (LRF), Middleware Reuse Factor (MRF), and High-End Device Reuse Factor (HRF) which are the number of successful transactions or accesses of the respective parts. Formally, the component reusability metric is defined in Eq. (18): where c represents the total number of component categories considered in the system. The overall system reusability factor is given in Eq. (19): where denotes the reuse factor of the component category. The value of DCR is in [01], bigger values are more component reusability and better system adaptability, whereas smaller values are less reuse and less efficiency. Good reusability facilitates the economic and scalable design of disaster-response systems. Metric weighting and aggregation strategy To facilitate the possibility of a single evaluation and ranking of alternative IoT disaster-response architectures, the proposed design quality measures are combined into a composite performance score. As there are various application situations that can be interested in specific characteristics of the system (e.g., time-related, reliability, or reusability), the framework allows both equal-weight and domain-specific weighted aggregation schemes. Formally, the overall system performance score is computed as in Eq. (20): where represents the normalized value of the metric, denotes the corresponding weight assigned to that metric, and m is the total number of metrics. The weights satisfy the condition as shown in Eq. 21: In the baseline evaluation, all metrics are assigned equal weights (i.e., ) to ensure an unbiased comparison across different system configurations. The framework however permits weights to flexibly adjust according to domain needs. As an example, greater weights can be given to timeliness (DTI) and alert dissemination efficiency (ADE) in time-critical disaster situations, but to system recoverability (DSR) and component reusability (DCR) in long-term infrastructure planning. Such a flexible weighting mechanism makes the framework more applicable as it allows decision-makers to fit the evaluation process to certain operational priorities. Table 4 provides the extended design quality measures based on the underlying structure of evaluation to enable a higher-level evaluation of the robust, efficient, scalable, and cost-effective disaster-response systems based on IoT capabilities. The metrics are a complement to the central measures and make it possible to carry out a quantitative analysis of various operational settings in a comprehensive manner. Table 4Extended design quality metrics.MetricSymbolMathematical definitionVariable definitions (concise)DescriptionFault toleranceFT DSR: system recoverability; DSHM: sensor healthIndicates the ability of the IoT disaster-response system to remain operational despite partial failures of sensors, gateways, or servicesData throughput efficiencyDTE : processed data; : generated dataMeasures how efficiently the system processes generated sensor data under disaster-induced load conditionsEnergy consumption optimizationECO : consumed energy; : max energy budgetEvaluates the efficiency of energy utilization across sensing, processing, and response layers during disaster operationsSecurity & privacy assuranceSPA SPA=1 - IE IE: integrity error during data accessQuantifies the degree to which data integrity, secure access, and privacy preservation are maintained during system operationScalability & adaptability (extended)SAE SC: sensing complexity; DSR: recoverabilityAssesses the system’s capability to scale and adapt dynamically as disaster intensity, data volume, or device count increasesInteroperability of IoT devicesIOD : successful communications; : total attemptsMeasures the effectiveness of communication and coordination among heterogeneous IoT devices and platformsAccuracy of early warning systemsEWSA TP/TN: correct alerts; FP/FN: incorrect alertsEvaluates the correctness of early warning alerts generated by the system in identifying disaster eventsTimeliness of emergency alertsTEA : alert delay; : max allowed delayMeasures the ability of the system to disseminate emergency alerts within acceptable response time constraintsCost efficiency of proposed designCE performance score; : total costAssesses the balance between achieved system performance and the total deployment and operational cost Extended design quality metrics. Table 5 gives a summarized overview of all the suggested design quality metrics which will be used to assess the IoT-based disaster-response framework. It shows each metric and its mathematical formulation, the range of values and their interpretation to achieve clarity, consistency and reproducibility of the evaluation process. The Table 6helps to easily compare various system architectures by emphasizing some important performance characteristics including complexity, severity detection, timeliness, sensing efficiency, data integrity, system resilience, and component reusability. On the whole, it is a coherent source that can be used to comprehend the contributions made by each metric to the measurement of the quality of disaster-response systems in terms of their structure and functioning. Table 5Summary of design quality metrics.Metric nameSymbolMathematical definitionRangeInterpretationDisaster detection complexityDDC NMeasures total system interactions; higher values indicate increased complexityDesign-level detection complexity E{N}Disaster severity identificationDSI [0,1]Indicates disaster intensity based on sensor activationDisaster timeliness indexDTI [0,1]Measures responsiveness in identifying disaster eventsSensing complexitySC ℝ⁺Represents sensing coverage and detection capabilityDecision data integrityDDI [0,1]Measures accuracy and reliability of transmitted dataSensor health monitoringDSHM [0,1]Indicates operational status of sensorsAlert dissemination efficiencyADE [0,1]Measures effectiveness of alert propagationSystem recoverabilityDSR [0,1]Reflects system resilience and recovery capabilityComponent reusabilityDCR [0,1]Measures reuse potential of system components Summary of design quality metrics. Threshold justification and sensitivity considerations Some of the measures suggested use threshold values to determine the level of quality including low/high severity, acceptable levels of complexity and system recoverability. These thresholds are based on the existing ideas of IoT operational practices, binary-classification cutoffs, and disaster-management severity levels. As an example, the 0.5 value in the Disaster Sensor Health Monitoring (DSHM) and DSR measurement is based on the traditional fault threshold of sensor networks decision. On the same note, the Disaster Severity Identification (DSI) levels (< 0.5, 0.5–0.8, > 0.8) are the typical hydrological severity levels of flood-early-warning systems. Sensitivity analysis demonstrates that minor changes in threshold boundaries (e.g. changing DSI to 0.85) do not affect the overall trend of the metric, although they do affect the severity labels, and thus maintain the consistency of interpretation. Similarly, changing complexity levels of DDC or SC leads to proportional changes with no change in system ranking. Each metric output is connected to practical design implications, so high DDC values suggest the necessity of more gateways or load-balancing nodes, low DSHM values imply sensor redundancy or maintenance issues, sub-optimal ADE values imply the presence of communication bottlenecks, and lower DSR values imply the need to implement strategies that can help make the system resilient. This guarantees that the threshold based interpretations can be directly converted into design choices in intelligent disaster-response systems. Threshold selection and sensitivity analysis The threshold values used in the proposed evaluation metrics (e.g., 0.5 and 0.8) have been chosen through the analysis of simulated data of IoT disaster-response sensors at different sensor density, communication delay, and noise levels. Statistical analysis of normalized metric distributions showed that the values near 0.5 are always the values which are consistent between low level of confidence and moderate level of confidence of the operational state, and the values which are equal to or greater 0.8 are those which are high severity or high confidence of the system. Such empirical threshold choice matches with the more recent IoT literature that documents that fixed cut-offs can not operate successfully in heterogeneous settings unless they are based on an analysis that is data-driven or that they are supplemented by adaptive behavior. Based on this, the suggested framework also promotes adaptive thresholding approach, wherein cut-off values are dynamically calculated based on percentile based statistic (e.g. median or upper-quartile values) based on real-time sensor readings. A sensitivity analysis also confirmed that the relative ranking of alternative IoT system configurations is not influenced (statistically) by moderate threshold variations (± 10–15%), which implies that the suggested framework of evaluation has a high resilience against the plausible threshold perturbation. It should be mentioned that the suggested measures are intended to be universal and flexible, and the scales of those measures are not to be perceived as absolute universal norms but as comparative ones. This enables the framework to be utilized in various IoT setups and disaster cases. In contrast to the traditional reliability models, FTA, RBD, and FMEA, which aim to study failure analysis and component-wise risk, the suggested metrics offer a system-level analysis of IoT architectures through the interaction complexity, sensing efficiency, and real-time responsiveness. This difference allows the framework to assess the structural resilience and the performance of operations during disaster-response situations. Results and analysis Mapping of QoS factors, disaster response framework services, and design quality metrics To demonstrate the significance of newly defined quality metrics, we mapped the suite of metrics to disaster response framework services and QoS (Quality of Service) factors, as shown in Fig. 2. QoS factors are derived from two different software quality models: ISO/IEC 912632 and ISO/IEC 25010:201133. Although this study focuses on seismic risk analysis, its emphasis on selecting hazard-consistent data across varying intensity levels aligns with our framework’s goal of evaluating disaster-response systems under diverse and realistic hazard conditions. Its approach to improving accuracy through multi-period hazard modelling parallels our use of metric-driven, graph-based evaluation to ensure reliability and consistency across different disaster scenarios34. Disaster response framework services. The ISO/IEC 9126 quality model comprises six quality characteristics (i.e., functionality, reliability, efficiency, usability, maintainability, and portability) that are mapped to the ISO/IEC 25010:2011 characteristics. ISO/IEC 25010:2011 is the updated version of ISO/IEC 912635, which has eight quality characteristics (i.e., functional suitability, security, performance efficiency, reliability, usability, maintainability, portability, and compatibility) that are mapped to disaster response framework services. Our framework services include disaster detection & monitoring, measuring disaster severity levels, analysing disasters & types, regional data sensing for disaster response, ensuring decision integrity, device health monitoring, generating response alterations, recovering system components, and reusing system components. In addition, these services are directly mapped with newly defined quality metrics for smart IoT-based disaster response framework. These metrics (see Sect. 4) include disaster detection complexity metrics, severity identification, type identification, sensing complexity metrics, decision data integrity, sensor health monitoring, alert dissemination efficiency, system recoverability, and disaster components reusability metrics. A mapping of QoS Factors32,33,35, Disaster Response Framework Services, and Design Quality Metrics is given in Fig. 2. In the middle of summer 2022, a severe flood struck Pakistan. The flood covered one-fifth of the total land, which increased the probability of severe alert levels across the country. We used the smart city flood response system for Pakistan as a case study to evaluate our proposed metrics. We consider a region with a specific number of cities. Each city is embedded with various disaster-specific sensors for monitoring. We consider that numerous flood-sensing devices send data to the regional gateway (e.g., Sindh, KPK, Baluchistan, and Punjab). We assumed that a flood disaster in multiple regions of Pakistan would trigger the activation of different sensors, decision-making, and the sending of a response to save the victims, as illustrated in Fig. 3. A list of disaster-specific sensing. Flow diagram of smart flood response IoT system. Devices are given in Table 6. The actuators that perform various actions to reduce human effort in the system are shown in Table 7. In this smart flood response setup, various flood sensors, including WRS, ULS, SSG, RLS, and AWS, are deployed under the gateways of the dam and the river. Sensors are providing real-time disaster data to gateways. Actuators like DHM perform disaster response activities as needed. All these IoT devices work under the sensor, network, decision, and response layers. The IoT gateways/regions share data with the cloud to process and store it in the network layer. Furthermore, data is sent to the flood decision-maker controller to perform artificial intelligence processes, enabling decisions to be made based on the severity level. According to the disaster status from sensors, the decision-making controller performs risk area analysis and develops strategies. The controller keeps analyzing the gathered data to generate emergency plans. The IoT alert system causes sensors to update the flood responder, acting as an “Emergency Action Planning Agent,” which sends responses to find a route and raise awareness among victims, the government, and non-governmental organisations. Moreover, various other actuators help these devices spread alerts through multiple communication channels to the victims. Table 6Disaster-specific sensor devices used in the flood-response case study.SensorsUsageValue generateWSR-88D radar sensor (WRS)Detect rainfall location, duration & track storm systemmax 80 nm26 g radar LevSel sensor (RLS)Tide measurement in water reservoirmax 70 mAutomated rain gauges (ARG)Flood warning systemsmax 500 mm/hrAnemometer wind sensor (AWS)Measure wind speed and directionmax 20 mAHygrometer sensor (HS)Measure humidity & air temperaturemax 90%WaterLOG water level sensor (WLS)Water Level Monitoringmax 20 mAWL705 ultrasonic level sensor (ULS)Measure flow & water surface distance from sensormax 20 mALM35 temperature sensor (LTS)Measure temperaturemax 150 °CVibration sensor (VS)Used to trigger the effect of earthquake alarmmax 15 mASeismic sensors (SS)Monitor earth core motionmax 100 HzAccelerometer sensor (AS)Detects the acceleration caused by an earthquakemax 20 mA Disaster-specific sensor devices used in the flood-response case study. Table 7List of actuator devices.ActuatorsUsageDevice health manager (DHM)Check & resolve device failures Set sensors threshold values Make & handle emergency plansSmart decision making agent (SDMA) emergency action planning agent (EAPA) disaster alert spread agent (DASA) smart spillway gates (SSG)Spread alert to victim & action taken bodies Opening & closing dam spillway gates automaticallyJDS earthquake alarm (JEA)Early warning by detecting earthquake soundwave List of actuator devices. In order to make the proposed framework more practical, the case study of Pakistan flood-response was further simulated by evaluating various IoT architectural configurations. There are three representative designs that were modeled and this represents low, medium, and high density sensor-gateway deployments that are consistent with modern disaster-response systems in the IoT. The proposed graph-based metrics were assessed on each configuration allowing quantitative comparison of alternative system design, as opposed to computing metrics individually. The findings reveal that the architectures that are more sensor redundancy and have equal connectivity to the gateway have lower disaster detection complexity, higher alert dissemination efficiency, and faster recovery of the system. Conversely, sparse deployments have propagation and delay alert generation that increases. These quantitative comparisons demonstrates the idea that the suggested metrics help to rank (objectively) the IoT architecture options and offer the practical guidance on the design optimization. The further analysis is consistent with the modern IoT disaster-response practice as it is based on real-time sensor surveillance, aggregation of data at the gates, and decision layers that are centralized and common in modern early warning systems. Although the AI-based anomaly detector and predictive models work on sensor streams of data, their performance is determined by the strength and performance of the underlying IoT framework. The suggested metrics are intended to complement AI-based methods offering the opportunity to test the resilience, efficiency of data flow, and recoverability of the system before its deployment to make informed design decisions that align with the trends in the current IoT technology. In order to confirm the feasible practicality of the suggested assessment model other than representative analysis, the flood-response case study was expanded to cover three separate IoT disaster-response architectures. The configurations are the typical design decisions in real world: (i) low-density centralized architecture, (ii) a middle-level redundancy balanced architecture, and (iii) a high-redundancy distributed architecture. The presented design quality measures were applied to each architecture, and it was possible to quantitatively compare and rank different system designs. The findings demonstrate that the architectures with increased sensor redundancy and uniform distribution of gateways always reach the lowest detection complexity, maximum alert disseminating efficiency, and recoverability of the system. Conversely, sparse centralized designs have greater severity propagation and sluggish response behavior. The aggregated performance score is based on a proposed weighting scheme and used to rank the various IoT architectures. In this research, all metrics are given equal weight with an attempt to make a fair comparison but other weighting plans can be implemented according to priorities within the application. The suggested measurements allow the objective comparison and ranking of the alternative IoT disaster-response framework, where Design C reveals to be more robust, efficient in terms of alerts, and recoverable, confirming the diagnostic and decision-support nature of the proposed framework as shown in Table 8. Despite the fact that the given framework provides a complete range of design quality metrics, the comparative validation revolves around a representative set that explicitly reflects the architectural complexity, response efficiency, system robustness, and decision accuracy. The metrics to be chosen included Disaster Detection Complexity (DDC), Alert Dissemination Efficiency (ADE), Disaster System Recoverability (DSR), Fault Tolerance (FT), Early Warning System Accuracy (EWSA), and Cost Efficiency (CE) since all of them are indicative of the most vital dimensions that affect the disaster-response effectiveness. Table 8Comparative evaluation and ranking of IoT disaster-response architectures.Design configurationSensor densityGateway strategyDDC ↓ADE ↑DSR ↑FT ↑EWSA ↑CE ↑Overall rankDesign A (centralized–low density)LowCentralized0.420.460.410.440.430.563 (Weak)Design B (balanced–moderate density)MediumHybrid0.630.610.640.660.680.722 (Moderate)Design C (distributed–high density)HighDistributed0.780.740.790.810.760.771 (Best) Comparative evaluation and ranking of IoT disaster-response architectures. The other measures offer a complementary diagnostic information and are evaluated based on the components or layer level, but are not used to change the relative hierarchy of the system designs. Checks on sensitivity indicated that adding more metrics do not change trends in comparative trends, which implies that the subset used to test the framework is adequate to support the validity of the comparative and diagnostic feature. Disaster detection complexity metrics evaluation The disaster detection complexity metric is measured using multiple scenarios from the Smart Flood Response case study. Scenario (i) is presented in Fig. 4, where Emergency Action Planning Agent retrieves the Dam level status from the water level sensor ‘WLS’ deployed in region 1 Dam. Moreover, the detection complexity for sending data requests and receiving data responses from the sensor is calculated using (1) as a total of four units. Data flow across multiple regions in the proposed IoT-enabled smart disaster response framework under simulated flood conditions. Sensor response load is also four units. In total, there are 8 units of complexity in the response-request transaction. Scenario (ii) is illustrated in Fig. 5 in blue edges where the EAPA requests data from Region-1’s Dam-1. The complexity of all IoT requests made through Region-1 gateway-1 is 7 units; see the edge count in (18). Different regions response activities Scenario. The sensor response load equals the request load (7 units). Therefore, disaster detection requires a total of 14 units of response-request complexity. Here another Scenario (iii) is derived from Fig. 5. A sudden rise in water level was observed at Dam-1, indicating the onset of a potential flood condition. In this case, the decision-maker controller sends alert signals to EAPA. As a result, EAPA sends a command for opening the Smart Spillway Gate actuator to respond to the disastrous situation (outflow the extra water into the rivers and maintain the level in Dam). Consequently, the disaster detection complexity in Region-1 gateway-1 is 7 units, and responding to floods using an SSG actuator results in a 14-unit load (the same as 18). The average complexity in Eq. (3) for this case study system is calculated to be (4 + 7 + 7)/3 = 18/3 = 6. units, which means it is within the acceptable threshold (indicates average case complexity). Design level disaster detection complexity The disaster detection complexity at the design level metric determines the overall complexity of an IoT system from its design model. Figures 4 and 5 show a system design with different scenarios for disaster responding. The EAPA checks the update of region 1 regarding the current dam-1 (water level) situation. The agents to connect to sensors to access current dam conditions in all regions. The design detection complexity of the entire Internet of Things system is assessed using the DDC metric at the design level, as defined in Eq. 2, which takes into account the unit cost of each edge. The average disaster detection complexity at the design level, as calculated in Eq. 5 for this case study, is found to be (24/25), i.e., less than 1, which is an acceptable threshold and considered the best-case complexity. whereas, less than 0.5 is the worst-case complexity. Therefore, the proposed design model is a straightforward system. Using Table 9, the network load measurements for request and response complexity are illustrated in Fig. 5. However, Table 9 provides detailed measurements for data requests by EAPA, including data packet size, time interval, supported protocol, and specific data format for each sensor and actuator device. However, data packet size is based on the protocol header and travel field size (in bytes). With the help of Table 9; Fig. 5 shows the edge weight (data loads) for scenario (ii) and scenario (ii). Table 9Data formats and packet sizes for flood and earthquake devices in the base layer.Sensor/actuatorsData generation time intervalProtocol supportedProtocol header sizeData formatPacket size (in bytes)WRS1 minNetCDF4 BYTESBINARY8RLS1 minMQTT2 BYTESJSON14ARG1 minWiFi2 BYTESBINARY18AWS1 minRS-4858 BYTESBINARY16HS1 minWiFi2 BYTESTEXT8WLS1 minMQTT2 BYTESJSON4ULS1 minWiFi2 BYTESJSON26LTS1 minFSK2 BYTESXML6VS1 minREST8 BYTESJSON12SS1 minRS-2328 BYTESBINARY10AS1 minPVM5 BYTESBINARY6DHMCONDITIONALREST8 BYTESJSON16SDMACONDITIONALSOAP4 BYTESXML18DASACONDITIONALREST8 BYTESJSON10SSGCONDITIONALMQTT2 BYTESBINARY3JEACONDITIONALMQTT2 BYTESBINARY3 Data formats and packet sizes for flood and earthquake devices in the base layer. Disaster severity identification metric evaluation This metric (6) is evaluated by measuring disaster severity through the definition of individual sensor readings in a single region. In scenario (iv), we take a flood severity checking scenario in region 2 (see Fig. 6). Eight flood sensors are deployed in two gateways (Dam and River) in a region that plays a fundamental role in measuring the severity of a particular disaster. Consequently, the decision-making controller device processes regional sensor data to estimate flood severity using all sensors count in particular areas of the region. As a result, the alert spreading agent is activated to disseminate alerts across high-severity levels, enabling a timely response to disasters. Flood severity level in region 2. According to Fig. 6, there are two WSR-88D Radar Sensors, two 26 g Radar LevSel Sensors, two Automated Rain Gauges, and two WL705 Ultrasonic Level Sensors (total eight) deployed in two areas (dam & river) of region-. All these sensors show different values of different units The decision-maker controller processes regional flood data to estimate flood severity in the Dam & River located in Region 2 and sends severity alerts to the alert Spreading agent for responding to the disaster. For this metric calculation, it is necessary to convert all deployed region 2 sensor values to a common scale. For this purpose, we introduce the “Max Scaling” process. Consider the WSR-88D Radar Sensor (WRS), which provides rainfall readings at 60 nm. Whereas the max sensor value is 80 nm (see Table 10). According to Eq. 19, the Max Scaling will be: = (60 / 80) = 0.75, equivalent to 1. All sensor values range from 0 to 1, where 0 and 1 represent the activated and deactivated sensor states. For Scaled flood sensor readings in a single region, see Table 10, which displays the output values of different sensors. Using Table 5 and Fig. 6, DSI metric (6) is calculated as: Table 10Sensor scaling for DSI evaluation.#SensorsReading outputMax valueScaled value1WRS60 nmmax 80 nm0.75 = 12WRS70 nmmax 80 nm0.875 = 13RLS55 mmax 70 m0.785 = 14RLS69 mmax 70 m0.985 = 15ARG393 mm/hrmax 500 mm/hr0.786 = 16ARG410 mm/hrmax 500 mm/hr0.82 = 17ULS16 mAmax 20 mA0.8 = 18ULS12 mAmax 20 mA0.6 = 1 Sensor scaling for DSI evaluation. Here, DSI > 0.8 units indicates a highly severe disaster level, necessitating urgent alerts for disaster response and preparedness. Disaster timeliness index metric evaluation This metric is evaluated by timely identification and response to disaster events in an IoT network based on the status of its deployed active sensors in a particular region. Figure 7 illustrates a smart IoT-based disaster response network, where multiple regions are connected to a controller that responds to detected disasters through an automatic system. Three regions with multiple pathways and two disasters (floods and earthquakes) are represented in Fig. 7. Purple edges represent region-1. Red and green edges show region-2 flow, whereas orange edges depict region-3 disaster. The first region contains a dam gateway, the second region has dam and river gateways, and the third region is based on hill gateways. The first two regions have flood sensors, whereas earthquake sensors are deployed in the third region. Disaster response. First, we examine a flood case from regions 1 and 2 for Disaster Timeliness Index metric evaluation. The controller detected an increase in water level in Region 1. In the event of an increased water level, the controller sends a command to Emergency Action Planning Agent-1 to address this disastrous situation (purple edges). Region-1 has three sensors (Using Figure (7), ULS, WLS, ARG with 1, 1, 0 sensor status respectively); then Eq. 7 will be: Then, EAPA-1 responds to disasters by actuator signals (opening smart spillway gates) to maintain the water level in Dam-1 and transfer water to other water reservoirs (the excess water is transferred to Dam-1 of the region-2 gateway). As a result, it decreases the flood event was detected in region-1 (purple edges). The floodwater is coming from Region-1 stores in Region-2 Dam-1. As a result, the water level indicated by the controller is increased. Then, EAPA-2 sends a command to open the SSG actuators to manage the disaster by flooding the water into region-2’s river-1. All the deployed flood sensors in River-1 have already exceeded threshold levels (detected by the controller) due to heavy rainfall and dam water outflows in the River-2 region. In this regard, Disaster Alert Spread Agent-1 alerts the victims to vacate the premises (green edges). Region 2 has a total of ten sensors. Suppose, using Figure (7), sensors are deployed at dam RLS = 1, ARG = 1, AWS = 0, WLS = 1, and ULS = 1. Whereas, sensors deployed at river RLS = 0, ARG = 1, AWS = 1, WLS = 1, and ULS = 1 (giving active = 1 and non-active = 0 status discussed in Table 10, then Eq. 7 for region-2 will be: On the other hand, the region-3 gateway (hilly area) has earthquake sensors. If an earthquake occurs in Region 3, then, with the help of Eq. 7, the controller detects the disaster type. Furthermore, it sends a command to DASA-2 to disseminate disaster alerts to citizens and the EOC through earthquake alarm devices (marked with orange edges). Consider sensors in region-3 are (VS = 0, AS = 1, SS = 0, using Figure (7), giving some sensor status to help the system in Disaster Timeliness Index, then Eq. 7 for region-3 will be: The threshold value for the Disaster Timeliness Index metric is between 0.5 units. Thus, this smart IoT framework is efficient enough to detect disaster types based on sensor status, enabling it to respond to victims and assist relevant authorities accordingly. Sensing complexity metrics evaluation The sensing complexity metric plays a crucial role in measuring greater predictability and lower sensing risks for disaster response. We take scenario (v) to evaluate the regional and disaster-level sensing complexity. Figure 7 depicts disaster-sensing scenarios in multiple regions. At every stage of an emergency, sensed data from flood and earthquake sensors can be a significant source of information. It also illustrates how quickly a region recovers after a disaster. Let’s calculate the sensing complexity for all disaster-affected regions. Figure 7 shows three regions with the total number of activated sensors. So, S1 = 2, S2 =8, S3 =1, and Rn = 3. Then, regional complexity in Eq. We took the worst case (3.67 units) for regional sensing complexity in Eq. 24. The regional sensing complexity has the worst (< 5.0), average (< 5.0 to > 8.0), and best (> 8.0) case. Let’s calculate the sensing complexity for a disaster where a flood has occurred. Figures (7) shows two regions (region-1, region-2) with activated sensors. So, S1 = 2, S2 =8, and T = 2. Then, regional complexity in Eq. (10) We took the average case (5.0 units) for disaster sensing complexity in Eq. 25. The disaster sensing complexity has the worst (< 5.0), average (< 5.0 to > 8.0), and best (> 8.0) cases. These threshold values are set by the SDMA. Decision data integrity metric evaluation The DDI metric is evaluated using scenario (iv) in Fig. 6, where EAPA demands hydrological, geological, and metropolitan data from sensors deployed at region-2 gateways to forecast floods. Data coming from both gateways are accessed by the decision-making controller for pre-disaster processing. Data transmitted between gateways to the controller involved the data integrity (authenticity) factor, which Smart Decision Making Agent handles. Consider a situation where EAPA sent 7 times data requests there 3 data authentic failures occurred. The proposed IoT network can face throughput errors, such as packets sent from sensor devices may be incomplete or corrupted during the entire transaction period. As a consequence, to verify the integrity (authenticity) of data at decision layer Eq. (11) will be calculated as: In conclusion, 0.7 (approximately 1) units of integrity indicate that the data is approximately the same as stored and processed. Our calculated value represents the maximum integrity level (the ideal value of integrity is 1), as the value of integrity ranges from 0 to 1. Disaster sensor health monitoring metric evaluation To evaluate this metric, Scenario (vi) in Fig. 8 shows that the Smart Decision-Making Agent keeps tracking the sensor device health. Sometimes sensor readings reach or exceed threshold values or suddenly stop providing any value (which may indicate damage), indicating that a disaster has occurred in that region (which can be minor or major). As a result, EAPA sends signals to the device health manager to repair the failed devices or instruct regional authorities to do so immediately (green nodes path). From Fig. 8, one out of five sensors stops working, then DHM check the sensor health at base layer Eq. (12) which is calculated as: Smart flood management framework: Region 1 dam. The calculated 0.8 reading represents the best-case scenario, where no severe-level disaster occurred or just minor device failure occurred in the particular region. Whenever the threshold is reached or exceeded (i.e., the value is greater than 0.5), the metric is classified as indicating that sensor devices are in good health (the best case scenario). Alert dissemination efficiency metric evaluation The Alert Dissemination Efficiency metric is measured using Scenario (vii) in Fig. 8, where a a disaster has occurred, and EAPA sends commands to the flood alert spreading agent for alert dissemination (non-black edges). Where this metric (13) shows how different framework entities E (i.e., public E1, government E2, and NGOs E3) are. In the given Scenario, the Smart Decision-Making Agent continuously tracks the sensor device’s performance and sends signals to the device health manager to repair the failed devices (green nodes on the path). Scenario (vii): Over the disaster occurrence, EAPA sends commands to the flood alert spreading agent for alert dissemination (non-black edges) efficient in receiving alerting messages through communication channels C (i.e., broadcast media C1, internet media. C2, print media C3, and social media C4). According to Fig. 9, the ADE metric in Eq. (13) will be evaluated as: Hierarchical structure of the smart disaster response framework. The higher efficiency of a smart IoT-based disaster response network reduces the operating cost of the system. Our calculations (0.833 = 1) show the best-case alert dissemination efficiency. Whereas, if the value goes below. 0.5 indicates the worst case. So, here we have 0.8 units, which indicates the high-level efficiency of the proposed alert system. Disaster system recoverability metric evaluation Any component or device in the system can experience sudden failure during the flood response activity. Therefore, there is an instant need to recover that particular device and check the system working status with the help of the disaster system recoverability metric in Eq. (14). Consider three cases (best, average & and worst) for each region to measure the DSR metric (see scenario (viii) in Fig. 9). In region-1, 7 devices are working, where 2 components are identified in error state, and the. Region-1 in purple, Region-2 in blue, and Region-3 in orange edges indicate the system component status as recovered (1)/failed (0). There, EAPA requests all device statuses and checks all components to determine whether they are in a normal or error state. As a result, DHM recovers the failed components or indicates the System Fail = 0 / System Recovered = 1 status. The remaining 5 are still working. then the recoverability metric (14) is calculated as: In region-2, 15 devices are working, where 9 components are identified in error state, and the remaining 6 are still working, then the recoverability metric (14) is measured as In region-3, 8 devices are working, where 1 component is identified in error state, and the remaining 7 are still working, then the recoverability metric (14) is calculated as: As seen here, the measurement of the DSR metric indicates that regions 1 and 3 are in a stable state, with the average & best cases, respectively. Furthermore, calculations indicate that these regions have stable IoT networks and the most recovered components to respond to disasters in a timely manner. However, we found the the worst-case scenario for region 2, indicating network failure and insufficient recovery capability during disaster conditions. While the proposed metrics have been validated in hypothetical scenarios of the smart flood response case study, this validation remains limited. Among the potential areas for future work, we consider enhancing reliability by benchmarking the proposed metrics against existing solutions in disaster response, such as those presented by Sood et al.5 and Tonmoy et al.16. These comparisons would show more exactly the aftereffect tuning on scalability, responsiveness, or compatibility with QoS (Quality-of-service). Furthermore, we can also utilise real-world empirical data from historical disaster events (such as rainfall intensity, river overflow records, and actual sensor logs) to take a further step in demonstrating the usefulness and robustness of the proposed evaluation Framework in execution. Although the case study is centered around flood-response, the evaluation framework is not limited to this area. The comparative study in this paper can be further applied to other disaster scenarios through redefinition of sensor configurations, distribution of gateway and response mechanisms. This shows how the proposed framework can be adapted to a variety of disaster-response applications that can be based on IoT. Validity and generalisability considerations The given case study of flood-response is grounded on a controllable and representative simulation scenario, in which the sensor density, connectivity of the gateway, and component failure are set to typical IoT-based disaster-response architectures. Although these parameters have been chosen on the basis of those parameters which have been widely reported in the literature, they are not based on a particular real-world deployment. The main aim of the given study is to show that the suggested graph-based evaluation framework can be applied and have a comparative ability instead of offering absolute performance standards. The metric values/thresholds (e.g., DDC, DSI, DSR) are then used in a relative manner to compare across architectural configurations under similar assumptions. Even though the findings offer valuable insights into the effect of system design decisions, the overall generalisability of the framework can also be enhanced by means of validation with real-world deployments of IoT or large-scale system simulations. Further research will be aimed at incorporating real sensor data, disaster traces, and simulation environments to test the suggested metrics in the context of realistic operation and various disaster conditions. Summary of evaluation results across metrics In the previous subsections, all design quality received a single measurement at baseline, in-disaster and post-response. Table 11 shows the consolidated results of all metrics. Table 11Summary of design quality metrics across scenarios.Metric No.Design quality metricBaseline scenarioDuring floodPost-response scenarioOverall assessment4.1Disaster detection complexity (DDC)0.80 (Good)0.45 (Poor)0.68 (Fair)Needs improvement4.2Disaster severity identification (DSI)0.84 (Good)0.49 (Poor)0.71 (Fair)Moderately reliable4.3Disaster timeliness index (DTI)0.77 (Good)0.42 (Poor)0.63 (Fair)Needs strengthening4.4Scalability (SC)0.88 (Excellent)0.65 (Fair)0.82 (Good)Strong performance4.5Disaster data integrity (DDI)0.81 (Good)0.52 (Fair)0.75 (Good)Reliable4.6Disaster sensor health monitoring (DSHM)0.90 (Excellent)0.55 (Borderline)0.72 (Good)Acceptable4.7Alert dissemination efficiency (ADE)0.82 (Good)0.50 (Fair)0.73 (Good)Stable4.8Disaster system resilience (DSR)0.79 (Good)0.41 (Poor)0.69 (Fair)Needs reinforcement4.9Disaster component reusability DCR)0.83 (Good)0.44 (Poor)0.66 (Fair)Mixed4.10Fault tolerance0.85 (Good)0.53 (Fair)0.74 (Good)Reliable4.11Data throughput efficiency0.81 (Good)0.48 (Poor)0.70 (Fair)Moderately reliable4.12Energy consumption optimization0.79 (Good)0.54 (Fair)0.72 (Good)Acceptable4.13Security & privacy assurance0.83 (Good)0.51 (Fair)0.76 (Good)Reliable4.14Scalability & adaptability (extended)0.87 (Excellent)0.60 (Fair)0.81 (Good)Strong4.15Interoperability of IoT devices0.82 (Good)0.47 (Poor)0.71 (Fair)Needs Improvement4.16Accuracy of early warning systems (EWS)0.78 (Good)0.43 (Poor)0.68 (Fair)Needs Strengthening4.17Timeliness of emergency alerts0.80 (Good)0.46 (Poor)0.69 (Fair)Mixed4.18Cost efficiency of the proposed design0.84 (Good)0.56 (Fair)0.77 (Good)Economically Viable Summary of design quality metrics across scenarios. The radar plot is shown in Fig. 10. relative values for all 18 design quality metrics for each of three scenarios: baseline, flood stage and response. The figure displays robust quantities, such as Scalability and Sensor Health Monitoring, alongside weaker ones, including Communication Reliability and Early Warning Systems, during the event. This visualisation provides a comprehensive overview of the resilience and recovery of systems. Radar chart showing comparative performance of all design quality metrics. Heatmap in Fig. 11 with results of 18 design quality metrics for the scenarios, scored using a green-to-red color grading. Larger scores (green) indicate better performance, while smaller scores (red) indicate a deficiency in the case of a flood. The heatmap facilitates the easy identification of strengths and weaknesses, complementing the radar chart’s analysis by comparing different characteristics. Heatmap summarizing metric performance. Comparative analysis with existing evaluation approaches The current disaster-response models mostly focus on the architectural design, the sensor deployment, the risks modelling or the dissemination of alerts, yet they lack an assessment of design quality based on a quantifiable metric. As an illustration, IoT-based and coastal-resilience frameworks suggested by Sood et al.5 and Tonmoy et al.16 respectively, both of them do not include structured metrics to quantify complexity, sensing behaviour, component recovery or reusability. In the same fashion, the IDSS framework created by Jung et al.17 provides better decision support without providing a quantifiable evaluation of the system quality. Conversely, the current research proposes a coherent set of nine measures that are projected to QoS attributes and make it possible to measure complexity, integrity, sensing accuracy, alert efficiency and system resilience. Table X (added) is a summary of the fact that the previous frameworks normally consider only 1–2 dimensions of evaluation, and our framework measures all large stages of SDR lifecycle quantitatively. The second distinction aspect is that it uses the abstract graph schema representation, which was not used in the earlier techniques. This graph-based model allows multilayer dependencies sensor-network-decision-response to be studied in fine detail, and interactions, bottlenecks and propagation of failures to be studied. Thus, relative to the preceding study that offers conceptual direction exclusively, our model adds a quantifiable, systematic, and generalisable methodology of the evaluation, enhancing the empirical and analytic foundation of smart disaster-response system development. Table 12 gives a comparative study of the current IoT-based frameworks of disaster-response to the proposed model of graph-driven evaluation. It points to the fact that earlier research is not structured and metric-based, does not use sensor-health monitoring, recoverability analysis, or graph-based dependency modelling. Contrary to it, the proposed framework proposes nine measurable design-quality metrics encompassing the entire SDR lifecycle and provide a more detailed and scalable assessment method. Table 12Comparative analysis.Study / FrameworkMetrics providedGraph-based modellingMeasures sensor healthMeasures data integrityMeasures alert efficiencyMeasures recoverabilityDesign-level evaluationLimitationsSood et al. (2020)NoNoNoPartialNoNoNoFocus on evacuation, no design-level evaluationTonmoy et al. (2020)Limited (resilience factors)NoNoNoNoNoNoSpecific to coastal zonesHong et al. (2019)NoNoNoNoNoNoNoLacks performance metricsJung et al. (2020)NoNoNoNoPartial (decision stage)NoNoDecision support onlyThis workNine metricsYesYesYesYesYesYesComprehensive, quantifiable evaluation Comparative analysis. Graph-based modeling of existing disaster-response frameworks In order to make visible the non-trivial comparative analysis, the chosen existing IoT-enabled disaster-response models were formally modelled through the proposed abstract graph schema. Each framework had system components that were aligned to graph nodes representing sensing, gateway, decision, and response layers with data flows, control interactions and alert pathways depicted as directed edges. This common graph-based abstraction ensures that every framework is tested on the same structural assumptions, which makes the comparison of them using the presented quality metrics of design objective. Metric-based differentiation and alignment with known system characteristics The suggested metrics were subsequently used on the graphical representation of the chosen frameworks so as to achieve quantitative scores in important design dimensions such as; complexity of detection, efficiency of sensing, efficiency of alert dissemination, system recoverability, and fault tolerance. The scores obtained are distinguishable between different frameworks and can be compared to their established architectural features. Frameworks based on centralized coordination have less sensing complexity but more detection complexity and less fault tolerance, distributed and redundancy-conscious designs have better recoverability and alert timeliness at the expense of increased complexity in their architecture. The varying metric performance and output validate that the suggested framework can diagnose the strengths and weaknesses of the current disaster-response systems, but not deliver generic or homogenous appraisals. Table 13 shows quantitative comparison of some disaster-response frameworks with IoT-enabled approaches to disaster response is expressed with the suggested abstract graph schema. The metric values reported indicate the behaviors of differentiated systems and represent the known architectural limitations and strengths, as it shows the diagnostic power of the proposed evaluation framework. Table 13Metric-based comparison of existing disaster-response frameworks.FrameworkDDC ↓SC ↑ADE ↑DSR ↑FT ↑Diagnosed strengthDiagnosed limitationFramework A (centralized)HighModerateModerateLowLowSimplicity, low overheadSingle-point failureFramework B (hybrid)ModerateHighHighModerateModerateBalanced designPartial redundancyFramework C (distributed)LowHighHighHighHighResilience, fast alertsHigher complexity Metric-based comparison of existing disaster-response frameworks. Discussions The primary aim of this research was to define a quantification system for a smart disaster response framework to determine its design quality. To achieve our design quality goals, we developed a set of metrics for evaluating the framework. We utilised the disaster response framework, with an IoT network as its foundation, for our study. We presented the entire framework as a graph. We defined nine metrics, of which the first is disaster detection complexity metrics to measure software complexity. Second, the disaster severity identification metric to measure the disaster’s nature and level of severity to respond accordingly. Third, the disaster type identifica- tion metric determines the disaster type. Fourth, sensing complexity metrics defines for measuring disasters and regional complexities. Fifth, the decision data integrity metric is used to measure data accuracy, completeness, and consistency. Sixth, the disaster sensor health monitoring metric is designed to track the health of the deployed sensor. Seventh, the alert dissemination efficiency metric is based on alert spreading performance. Eighth, the disaster component reusability metric indicates the level to which system components can be reused repeatedly. The ninth and final metric is disaster system recoverability, which assesses the stability of all system components and measures the overall system status based on the recovered or failed components. The defined quality metrics are assessed using a smart flood response IoT system case study. Using different scenarios from the case study, we evaluated the proposed metrics. The design quality metrics presented in this study aimed to evaluate the effectiveness and efficiency of the framework in addressing various aspects of disaster response. The experimental results demonstrate that the proposed model offers a comprehensive and adaptable disaster evaluation paradigm. Previous works have focused on sensor-driven monitoring36, risk-specific assessment37, and federated concepts for decentralised intelligence38, but rarely in a comprehensive model that can also drive resilience. Our work fills this gap by merging 18 design quality measures that assess exposure, communication, resilience and cost-effectiveness. The results presented synthesise the overall strengths and weaknesses observed in the system, summarised as the scalability of sensors, sensor health monitoring capability, and data integrity provided by the IMS technology. This balanced perspective highlights the novelty of our contribution: a multi-metric, holistic evaluation framework that not only diagnoses where smart disaster management systems perform well, but also identifies major weaknesses in cases of stress. Although validation is anchored in simulated disaster scenarios, the metrics designs enable replicable benchmarking in various smart city contexts. Though the proposed metrics are proven by a representative smart-flood IoT scenario, such a simulated test is mostly a controlled setting of experiment to showcase the metric functionality, interdependencies and interpretability. This kind of validation by simulation is a common initial phase in design-quality testing models, where a real world application may not be possible in the early phases. The empirical testing needs to be done to enable the establishment of generalisability39. With the growing complexity of modern industrial systems, the occurrence of concurrent faults has increased, necessitating intelligent, data-driven diagnostic frameworks beyond traditional sensor-only approaches40,41. In this work, a concrete future path is proposed that includes the systematic benchmarking of the already developed IoT-based disaster-response frameworks (e.g., developed by Sood et al. and Tonmoy et al.), and an empirical validation on the basis of actual sensor logs, historical hydrological data, and multi-region disaster data. By implementing these measures, it will be possible to conduct comparative assessment among the cities, the types of disasters and deployment settings, thus enhancing the external validity levels and proving the ability to scale in the circumstances of operational settings. Conclusion This article presented a comprehensive evaluation methodology for smart disaster response systems. The method is based on using abstract graph schema representations and applying a set of design quality metrics to assess the effectiveness and efficiency of the disaster response system. The metrics presented in this study include complexity, disaster severity identification, Disaster Timeliness Index, disaster sensing complexity, alert dissemination efficiency, sensor health monitoring, disaster recoverability efficiency, and framework component reusability. The proposed evaluation technique offers a valuable tool for designers and policymakers to assess the quality of smart disaster response systems and pinpoint areas for improvement. By incorporating these metrics into the design process, it is possible to ensure that disaster response systems are optimized for effective and efficient eration in the face of disaster events. The design quality metrics presented in this study provide a comprehensive evaluation of smart disaster response systems and their effectiveness in responding to disasters. Using abstract graph schema representations and these metrics can help develop high-quality disaster response systems capable of addressing the complex and unpredictable nature of disaster events. The suggested aggregation framework is conducive to flexible metric weighting, which enables the adaptation of various disaster-response needs and enhances the real-world applicability. One of the limitations of this study is that it uses a simulated case study with pre-determined parameters. The next-generation of research will be devoted to the validation of the proposed framework by the use of real-world deployments and high-fidelity simulations to increase the external validity and the generalisability. The framework augments current methods of reliability assessment by enhancing the component-level failure analysis to system-level interaction and performance analysis in IoT-based disaster-response settings. In this research proposes a systematic graph-based evaluation framework and develops nine quality metrics for SDRs. Although the framework’s ability has been demonstrated through a theoretical validation example, further validation is still needed. In future work, we will also attempt to benchmark the proposed metrics against other IoT-based disaster response models and further evaluate these metrics using empirical data and deployment contexts. It would lead to higher generalizability, broader applicability and practicability when building robust and high-quality smart disaster response systems. The suggested framework is flexible across various disaster fields through adjusting the type of sensors, communication plan and decision parameters in the graph model. Its future use will be in other disaster situations like earthquakes, wildfires, and landslides to further confirm its overall applicability. Author contributions I.Q. and Q.A. developed the core conceptual framework and wrote the main manuscript text. T.A. and T.A. (Turki Alghamdi) contributed to the methodology design and literature review. N.A. and S.S.A. conducted the analysis of IoT design quality metrics and helped in model validation. M.S. assisted with data collection and visualization, including the preparation of Figures. A.M.I. provided critical revisions, technical supervision, and final manuscript editing. All authors reviewed and approved the final version of the manuscript. Data availability The data supporting this study’s findings are available from the corresponding author upon reasonable request. Declarations Competing interests The authors declare no competing interests. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. References Associated Data Data Citations Data Availability Statement The data supporting this study’s findings are available from the corresponding author upon reasonable request.