Keywords

Introduction

Forest fires have become an escalating concern due to the intensifying frequency and severity of extreme weather conditions, exacerbated by the effects of climate change. This global issue not only results in the loss of human lives and habitats but also contributes to the release of millions of tons of CO2 and other pollutants [1]. The development of effective forest fire emergency management systems becomes therefore crucial in minimizing the impacts of future fire events.

The SAFERS project (Structured Approaches for Forest fire Emergencies in Resilient Societies) addresses this pressing challenge by proposing a comprehensive Emergency Management System (EMS) designed to manage forest fires across all critical phases of the emergency management cycle. Leveraging a service-oriented approach, SAFERS integrates a set of Intelligent Services (ISs), independent modules exploiting Artificial Intelligence (AI), and other advanced capabilities to provide more refined and informative outputs. These services utilize and produce diverse data sources, including Earth Observation from the EU Copernicus program, meteorological forecasts, hazards and risk forecasts, propagation models, data from social media and chatbot applications, as well as real-time data from in situ camera sensors. The outputs generated by the SAFERS IS are stored within a geospatial data lake, harmonized, and made accessible to end users through an interactive web-based dashboard. This integration enables informed decision-making throughout the entire emergency management cycle, enhancing preparedness, response, and recovery efforts.

Compared to other solutions, the SAFERS platform provides two key strengths. First, it encompasses all key phases of the emergency management cycle, ensuring a comprehensive management of the emergencies. Second, the system embraces a modular and service-oriented architecture, allowing each IS to operate independently with minimal dependencies. This modularity is further enhanced by asynchronous communication through a message broker, allowing for more flexibility. To further maximize interoperability, the platform leverages several software standards, including REST API communication for web-based components, INSPIRE metadata, OGC (Open Geospatial Consortium)-compliant services for GIS data management. This standardized approach allows for easier integration with existing systems and applications when necessary, fostering collaboration among various stakeholders involved in forest fire emergency management.

This work aims to provide an overview of the SAFERS ISs and platform, focusing on its architecture and technical aspects. Through this comprehensive analysis, we aim to highlight the innovative features and capabilities of the overall platform, with particular attention to the innovations brought to the domain of forest fire emergency management.

The remainder of this document is organized as follows. Section “The SAFERS Architecture” will delve into the SAFERS architecture, followed by an examination of the ISs and their operational and on-demand functionalities in section “The SAFERS Intelligent Services”. Additionally, the data layer, which plays a critical role in integrating diverse data sources, will be explored in section “The Data Layer”. Section “The SAFERS Dashboard” provides an overview of the SAFERS dashboard, which is a user-friendly web-based interface designed to enhance decision-making and situational awareness during forest fire emergencies. Finally, section “Conclusions” draws the final conclusions, highlighting future directions for further enhancements in this field.

The SAFERS Architecture

In the SAFERS platform, the architectural design shown in Fig. 4.1 is crucial in facilitating efficient communication, data management, and visualization of heterogeneous information.

Fig. 4.1
A model diagram of the SAFERS architecture. The central backend is connected to the rest-A P I, asynchronous message bus, geodata repository, importer, map server, and web dashboard. The A P I has the crowdsourcing solution.

Overview of the SAFERS architecture. The dashboard represents the main entry point, visualizing the data produced by the ISs (top pane). These ISs provide information either via REST or async API. Geospatial data is handled by the GeoData repository (GDR), then processed by the importer and map server for visualization

Full size image

The main functionalities of the SAFERS platform are provided by a set of services independent of one another. The ISs are grouped into three main categories, namely operational services, on-demand services, and crowdsourcing solutions.

First, the operational services represent modules with a simpler communication paradigm: they autonomously provide periodical outputs for specific data. These include Subseasonal Weather Forecasts, Operational Early Warnings, In situ Smoke and Fire Detection Systems, and the Decision Support System.

Second, on-demand services, namely the EO-based Fire Delineation, Postfire Monitoring, and On-demand Wildfire Forecast services, use instead a more complex communication mechanism: they are triggered by a specific message type named map request. These services act on demand, retrieving the necessary data based on user-defined parameters, and producing one or more outputs, depending on the request content. Once the results become available, the other services are notified through a message bus, allowing for the required follow-up actions to take place.

Last, SAFERS also incorporates crowdsourcing solutions, including a chatbot and a social media module, to gather information directly or indirectly from citizens. The chatbot enables structured geolocated and multimedia data collection, while the social module gathers and classifies real-time Twitter posts, extracting relevant emergency events and estimating their impacts.

To support efficient data management, the platform includes a data layer, which serves as a central storage system for heterogeneous data. This module adheres to INSPIRE metadata standards,Footnote 1 ensuring proper organization and accessibility of diverse datasets. When applicable, the raw data stored in the GeoData Repository (GDR) is processed by the Importer and Mapper modules. These components allow the creation and management of OGC-compliantFootnote 2 layers, enabling the integration and presentation of spatial data within the system.

The main communication point between the user dashboard and the services is the central backend, which facilitates the exchange of information, allowing users to interact with various platform components seamlessly. This backend ensures the smooth flow of data and requests throughout the system, while a web-based dashboard is provided to users for intuitive data visualization. The dashboard supports the display of geospatial information through Web Map Service (WMS) and Web Map Tile Service (WMTS) layers. Additionally, it incorporates data from crowdsourcing solutions and real-time information from in situ cameras, enabling users to gain valuable insights and make informed decisions.

The SAFERS Intelligent Services

The ISs have the purpose of providing heterogeneous data in a timely manner to the SAFERS Dashboard. These services are carefully designed to address specific aspects of the emergency management cycle, enhancing preparedness, response, and mitigation efforts.

Operational Services

These play a critical role in providing high-quality information relevant to forest fires. Data is essential for preventing, preparing, and responding to potentially dangerous wildfire hazard weather situations. To achieve this, an automated processing chain has been established, allowing for the accessibility of necessary resources within the SAFERS platform, provisioned by four services. First, the Subseasonal Weather Forecasts service provides forecast data collected from the ECMWF,Footnote 3 focusing on relevant variables for forest fire hazard and risk prediction. The data includes high-resolution deterministic forecasts with hourly updates up to 72 h, as well as medium-range probabilistic predictions for lead times up to 15 days (360 h). Additionally, extended range/subseasonal probabilistic predictions with lead times up to 46 days (1104 h) are provided twice a week. All weather data is postprocessed, automatically uploaded to the SAFERS GeoData Repository, and made available for visualization and sharing with other partner services.

Second, the Operational Early Warnings service augments operational wildfire hazard and risk mapping to improve the dissemination, uptake, and updating of early warnings. By integrating open datasets from complementing sources and utilizing novel risk models, this service delivers valuable information to stakeholders, enabling daily assessment and preparation in case of a wildfire. The datasets used include the European Forest Fire Information System (EFFIS),Footnote 4 Fire Weather Index (FWI), European weather forecasts from FMI, and real-time information about active fires. The wildfire risk modeling module integrates current and forecast weather information, and it is expected to incorporate lightning forecasts.

Third, the Smoke and Fire Detection System (smoCAM) utilizes standard surveillance cameras to monitor large areas for wildfire and smoke, automatically detecting fire and smoke plumes in various outdoor conditions. This AI-based service is capable of accurately detecting smoke and fire, even in challenging environmental conditions such as clouds and fog. SmoCAM is deployed in six test sites, each equipped with a PTZ (Pan-Tilt-Zoom) camera and an LTE router to enable remote connections. The camera patrols different points of view to widen the monitored area, and relevant information describing the detected smoke and fire is transmitted as JSON messages to the SAFERS bus. The messages are then integrated and displayed on the dashboard, providing real-time updates on detected wildfire events.

Last, the Decision Support System (DSS) analyses data from various sources and provides valuable recommendations to stakeholders. It utilizes semantic reasoning on top of the SAFERS domain ontology to derive new insights from preexisting knowledge. The DSS has been designed to meet user requirements and address specific scenarios, such as FWI forecasts, Fine Fuel Moisture, duff moisture, and drought anomalies. The system’s functionalities have been implemented to accommodate the SAFERS ontology, derived from previous works [2]. The DSS plays a crucial role in aiding decision-makers during forest fire emergencies, contributing to the resilience and effectiveness of the overall system.

On-Demand Services

SAFERS also offers services that can be activated on demand to provide rapid mapping, severity assessment, and wildfire forecast functionalities, as shown in Fig. 4.2. First, the EO-based delineation service leverages Sentinel-2 satellite imagery to generate thematic maps for burned area delineation [3], severity estimation, fire front detection, and land cover maps [4] through machine learning models. By employing state-of-the-art segmentation networks, this service generates output maps for a given area and time interval accurately, and in a short amount of time. The service is deployed using a multimodule containerized architecture, ensuring fault-tolerance and distribution of components, and asynchronous communication through the SAFERS message bus.

Fig. 4.2
4 heat maps of the on-demand services. 1. The forecast of wildfire with a closed area. 2. Severity estimation with a selected bright region. 3. Post-fire analysis area with dark regions. 4. Fule map delineation with a small area.

Example outputs derived from the on-demand services: (1) wildfire forecast simulation, (2) severity estimation, (3) postfire dNBR analysis, and (4) fuel map delineation

Full size image

Similarly, the Postfire Monitoring service generates long-term assessments during the recovery period, providing dNBR and other layers to better observe and understand the regrowth in the affected areas.

The Wildfire Forecast service is instead achieved through the implementation of PROPAGATOR [5], an operational tool based on cellular automata. It rapidly simulates the potential spread of wildfires, considering various factors such as ignition point, topography, fuel cover, and meteorological data. The resulting probabilistic fire fronts stem from averaging several realizations of the stochastic core, and the service provides useful accessory data, including maximum and mean Rate of Spread (RoS) and Fireline Intensity (FI) of the affected area. The simulation also allows for the incorporation of firefighting actions, such as Canadair and helicopter drops, and waterlines. The IS is deployed with async communication using the message broker, enabling efficient orchestration of simulation requests. The results are uploaded to the data repository, and the dashboard receives timely updates from both services.

Crowdsourcing Solutions

Currently, operational procedures primarily involve one-way communication from authorities and monitoring agencies to citizens, lacking active citizen engagement. This deficiency leads to inadequate risk awareness and preparedness among citizens, hindering disaster impact reduction in the case of natural hazards. Valuable information shared by citizens on social media during emergencies often gets lost in the vast sea of data. To address these challenges, SAFERS proposes implementing two essential crowdsourcing solutions: the Social Module and the Chatbot.spiepr Par26

The Social Module is designed to fetch and analyze real-time content posted on Twitter,Footnote 5 with a particular focus on wildfires and other natural disasters. By combining text and image analysis with location information, the module utilizes natural language and image processing to extract meaningful data related to emergency situations, early warning signals, and damage assessment. First rule-based filtering excludes erroneous or misleading data, retaining only informative content relevant to emergency scenarios. The module then classifies social media posts using text mining and deep learning techniques, mapping them into relevant categories [6], and groups them by meaningful events [7, 8]. Implemented and deployed in a cloud platform, the Social Module seamlessly integrates with the SAFERS message bus through secured web-based (HTTPS) REST APIs, ensuring efficient data sharing and enhancing the overall emergency response capabilities.

The Chatbot [8] facilitates instead structured geolocated and multimedia data collection from first responders and volunteers, serving as an innovative and user-centric tool by enhancing communication and collaboration during disaster management efforts. It enables two-way communication between citizens, field forces, and control centers, crucial for effective disaster risk reduction. Built on a robust technology stack, the Chatbot ensures scalability and seamless integration with the Telegram API, enabling efficient real-time communication and personalized functionalities for different user roles. Two primary user roles are supported: professionals with specialized training and citizens. For professionals, the chatbot offers advanced features like real-time location sharing, reporting activities, detailed measurements, damage assessments, and instructions exchange with the control room, facilitating efficient emergency coordination. Citizens, with limited emergency response training, can actively contribute by reporting events, sharing real-time information, receiving geolocated broadcast messages, and accessing safety guidelines through the Chatbot.

Both the Social Module and the Chatbot integrate with the project-level API Gateway and OAuth2Footnote 6 authentication module to ensure secure data management and user authentication, safeguarding user accounts and associated data.

The Data Layer

Every service introduced so far generates heterogeneous outputs that need to be stored and processed to be visualized and exploited by expert users. This is the task of the data layer, which consists of two main components: the Geodata Repository (GDR), and the Importer-Mapper coupling, displayed in Fig. 4.3.

Fig. 4.3
A model diagram of the data layer components. It represents a geodata repository connected to a data lake and intelligent services. The dashboard is connected to the importer and mapper with the S Q L database. Pop-up notifications for new resources available.

Data layer components and their interaction with the other modules

Full size image

The GDR consists of a management component that facilitates data ingestion and retrieval by serving as a central repository for metadata. It enables efficient data discovery and exploration within the data repository. This component is built on CKAN,Footnote 7 an open-source data management system that has been customized with plugins and extensions to meet the specific requirements of the GDR. CKAN provides a user-friendly web-based GUI and API for users to upload, edit, delete, and search data. The large amount of data in various formats is securely stored in a cloud big data storage system, using a data lake approach and employing the HDFS file system to preserve data in its original form within Hadoop clusters [9]. To ensure effective communication and collaboration between microservices, the GDR exploits the message bus to notify other services about new data availability and changes to existing datasets.

The Importer-Mapper consists instead of two interconnected elements. The Mapper represents a map server that exposes geospatial data through OGC-compliant services, i.e., WMS, WMTS, WFS, or WCS, implemented through GeoServer.Footnote 8 The Importer routine handles changes in the raw data stored in the GeoData Repository and updates the content available through the Mapper. It can handle data in various formats, transforming and publishing it as layers. The Importer component provides additional APIs to enhance its functionality. These APIs allow users to retrieve the list of available layers, delete layers from both GeoServer and the GDR, obtain metadata for specific layers, download raw files associated with layers, and access the temporal series of values for specific coordinates from a list of layers. By supporting multiple data formats and offering standardized services, the Importer-Mapper components ensure seamless integration and retrieval of geospatial data from different sources.

The SAFERS Dashboard

The SAFERS dashboard, illustrated in Fig. 4.4, stands as a pivotal interface for managing and visualizing critical data throughout various phases of forest fire emergencies effectively. Developed on React JS framework and integrating essential libraries, this dashboard ensures a seamless and intuitive user experience. The dashboard handles the complexities of spatiotemporal data representation, with the central map serving as a geographical anchor, and the top-level date and time selectors as global temporal reference for the whole system.

Fig. 4.4
2 screenshots of the SAFERS dashboard. Left. The main screen points out locations. Right. A new map is created with specifications.

Screenshots of the SAFERS dashboard, displaying the main screen (left), and an example form to create a new map request

Full size image

The visualization methods employed within the dashboard offer multiple approaches to better understand the data. Dynamic filters in the top bar enhance data exploration, enabling users to refine their focus and select increasingly specific information. This allows for a more precise analysis, crucial in making informed decisions during emergency situations. The central map not only provides essential geographical context but also allows users to interact directly with the data.

A user-friendly list pane on the left side offers a streamlined approach to navigation, ensuring that users can effortlessly access the necessary information. The dashboard integrates and handles OGC-compliant layers, complete with their respective legends and metadata. This feature enables users to grasp spatial relationships and patterns. Moreover, the use of INSPIRE metadata standards ensures seamless data sharing and interoperability among European stakeholders. To further enhance the temporal analysis, the inclusion of a dedicated time widget grants users the ability to explore data across different timeframes, including the map layers. Additionally, the dashboard allows retrieving data at specific points, offering the tools for the analysis on the time series, displaying the evolution of a given metric in a specific location (e.g., precipitation, temperature).

Lastly, the dashboard also provides a dedicated summary page, offering a comprehensive overview of ongoing emergency situations. This page consolidates critical information, providing stakeholders with a comprehensive snapshot, and allowing users to grasp the current situation at a glance.

Conclusions

The SAFERS project presents a comprehensive and modular forest fire management system, addressing the challenges posed by increasing forest fires and extreme weather events. The SAFERS ISs form a powerful backbone, offering essential functionalities across all emergency management phases. Operational services provide high-quality weather forecasts for early warnings and risk prediction, while on-demand services facilitate rapid mapping and severity assessment using satellite imagery. Crowdsourcing solutions actively engage citizens in data collection and real-time communication. Overall, the SAFERS platform leverages intelligent technologies like AI and EO, empowering informed decision-making and fostering resilience in forest fire emergency management, contributing to more resilient societies and mitigating future fire impacts.