Ontology for a Georeferencing Mobile System for Real

Time Detection and Monitoring of Wildfires

Dulce Pacheco

1,2,3 a

, David Aveiro

1,3,4 b

, Valentim Caires

1,3 c

and Duarte Pinto

1,3 d

ARDITI - Regional Agency for the Development of Research, Technology, and Innovation, 9020-105, Funchal, Portugal

School of Technology and Management, University of Madeira, Caminho da Penteada, 9020-105, Funchal, Portugal

NOVA-LINCS, Universidade NOVA de Lisboa, Campus da Caparica, 2829-516, Caparica, Portugal

Faculty of Exact Sciences and Engineering, University of Madeira, Caminho da Penteada, 9020-105, Funchal, Portugal

Keywords: Georeferencing Mobile System, Ontology, Wildfire.

Abstract: This paper presents the Georeferencing Mobile Wildfire Detection System Ontology (GeMoWilDSOn). This

ontology served as a base for implementing software for a mobile and georeferencing system for real-time

detection and monitoring of wildfires in steep mountainous territories. On average, about 65,000 fires occur

in Europe annually, burning approximately half a million hectares of wild land and forest areas. This growing

tragedy directly reduces the forest biomass and biodiversity, causing severe damage to the ecosystems.

Ontologies help developers speed up the requirements' analysis in the design of a new system. Our work

results in a streamlined ontology focused on fire prevention and fighting with mobile sensors, automatically

georeferenced polygon data, and visible and thermal image captures, specially designed for steep mountainous

terrain, where firefighting can be complex. Our research fills gaps found in related state-of-the-art and

provides innovative contributions such as the concepts of manually drawn areas of fire and shadow, which

are of utmost importance regarding this particularity of steep terrain. Our ontology was validated in three real-

world tests where experts were delighted with the features, captured information, and its representation in the

GUI of the developed system.

1 INTRODUCTION

This paper presents an ontology developed in the

context of a research project financed by European

Union funds. This ontology was a base for

implementing mobile and real-time georeferencing

software for forest fire prevention and fighting in

steep mountainous territories.

Burnt areas in Europe have increased in the last

couple of years, and up to mid-August 2022, more

area has been burnt than in the years before (see

Figure 1). On average, about 65,000 fires occur in

Europe annually, burning approximately half a

million hectares of wild land and forest areas (San-

Miguel-Ayanz, 2012). This disaster also directly

reduces the forest biomass and biodiversity, causing

https://orcid.org/0000-0002-3983-434X

https://orcid.org/0000-0001-6453-3648

https://orcid.org/0000-0002-0871-7212

https://orcid.org/0000-0002-8451-5727

severe damage to the Earth’s Forest ecosystem

(Perez-Mato et al., 2016).

Being able to promptly detect the occurrence of a

wildfire and having the capability to perform an

accurate, real-time tracking of its evolution is vital to

rapidly and efficiently organize the available

resources to control and extinguish it (Arana-Pulido

et al., 2018; Perez-Mato et al., 2016). This task can be

severely compromised in areas of steep terrain, which

not only makes the visual detection and surveillance

of the wildfire fronts or hot spots difficult but also

present localized winds and meteorological

conditions that influence the prediction of wildfire

evolution (Perez-Mato et al., 2016). The archipelagos

of Macaronesia (Freitas et al., 2019) are steep, rocky,

and with profoundly eroded lava gorges running

260

Pacheco, D., Aveiro, D., Caires, V. and Pinto, D.

Ontology for a Georeferencing Mobile System for Real Time Detection and Monitoring of Wildﬁres.

DOI: 10.5220/0011592100003335

In Proceedings of the 14th International Joint Conference on Knowledge Discovery, Knowledge Engineering and Knowledge Management (IC3K 2022) - Volume 2: KEOD, pages 260-268

ISBN: 978-989-758-614-9; ISSN: 2184-3228

down to the sea, which makes the fires' fronts

challenging to spot and monitor.

Several key technologies and methods are

commonly used for wildfire detection, surveillance,

and prediction (Perez-Mato et al., 2016). These

methods include fixed and mobile ground-based

solutions, aerial platforms, and satellite imaging or

sensing (Perez-Mato et al., 2016). However, most of

them have specific limitations that might affect their

performance and the usefulness of the generated

information (Perez-Mato et al., 2016). Furthermore,

algorithms used to predict wildfire spread relies on

accurate near real-time input data to maximize their

reliability (Perez-Mato et al., 2016). Unfortunately,

most previously described technologies cannot

provide data quickly or accurately enough (Perez-

Mato et al., 2016).

Figure 1: EFFIS Weekly Cumulative Burnt Areas in

Europe, in 2022.

The georeferencing of the wildfire could also see

some improvements. Most ground and airborne

solutions only provide images of the fire, captured by

thermographic or visual cameras, and leave the fire

georeferencing task to the staff supervising those

images (Perez-Mato et al., 2016). This adds a manual

step to the process, which can be very slow and

inaccurate in most cases, severely affecting the rapid

decision-making scenarios required during a wildfire

extinction process (Perez-Mato et al., 2016).

Perez-Mato and colleagues introduced a rapidly

deployable mobile unit (RDMU) prototype. It uses a

thermographic camera to autonomously detect, track,

and georeference wildfires within its detection range

(Perez-Mato et al., 2016). The data collected by these

RDMUs is complex, needs to be safely kept, and be

rapidly available for the firefighting staff members,

both in real-time, to help in the fire extinguishing

efforts, and a posteriori, to analyze the wildfire

progression and behavior.

The list of requirements for the functionality of

fire safety systems has been increasing in the last

couple of years, while the time of implementation of

software projects has been reduced (Nikulina et al.,

2019). Ontologies may help developers speed up the

requirements analysis in the design of a new system.

Furthermore, ontologies may contribute to solving

the problem of integrating knowledge from various

sources and presenting it by subject area in an explicit

form (Nikulina et al., 2019). In turn, it facilitates

knowledge development, understanding, and

maintenance, reducing duplications and

inconsistencies (Nikulina et al., 2019).

This paper presents the ontology designed to

formalize information supplied by a set of RDMUs,

as well as decisions and commands issued by the

intervenients. All information and commands are

made available according to the needs of staff in an

intuitive GUI.

The following sections present: the context of our

work; a review of related work regarding ontologies

for fires and sensors-based systems; our ontology

proposal, which addresses the concrete needs of our

practical project and its validation; a comparison

between our approach and related work; and

conclusions and future work.

2 SENSOR-BASED WILDFIRES

DETECTION SYSTEMS

This section overviews the most widely used sensor-

based wildfire detection systems. Most of these

systems may also serve as: (1) tracking tools once the

fire has started; or (2), in a post-fire moment, for

firefighting staff and investigators to go back in the

data and analyze the fire progression. The main

features and limitations of these systems are

presented.

2.1 Ground Detection Systems

Fire detection systems based on the ground are

usually composed of cameras attached to

watchtowers or similar infrastructures, which can

provide the necessary power and communication

interfaces (Perez-Mato et al., 2016). To be able to

surveil large extensions of a forest, several of these

observation points need to be installed, ideally

accounting for the field of view (FOV) of each

Ontology for a Georeferencing Mobile System for Real Time Detection and Monitoring of Wildﬁres

261

camera and considering the individual pan and tilt

ranges, as well as the presence of any surrounding

obstacles (Perez-Mato et al., 2016).

According to the literature (Perez-Mato et al.,

2016), ground-based wildfire detection systems

present the following main limitations: (1) cameras

are permanently exposed to weather conditions and

are prone to vandalism, which demands regular and

pricey maintenance to keep the cameras operational;

(2) the limited FOV demands numerous observation

points to cover a large area; (3) the georeferencing of

the detected fire is usually performed manually by

firefighting staff, which is prone to error.

An alternative to ground-based wildfire detection

systems is vehicles with thermographic cameras

installed. This configuration increases the mobility

and reconfigurability of the observation points and

reduces the possibility of damage caused by weather

and vandalism (Perez-Mato et al., 2016). However,

they still suffer from the same subjectivity regarding

wildfire georeferencing and pose a high risk for

vehicle operators (Perez-Mato et al., 2016).

2.2 Aerial Detection Systems

Human-crewed helicopters and aircraft are often used

during wildfires for firefighting and monitoring tasks,

as they can spray water over the fire and serve as

high-altitude observation points (Perez-Mato et al.,

2016). Aircraft flying over a wildfire area provides a

much larger FOV and can be dynamically moved

from one place to another as the wildfire evolves.

This makes them much more versatile and efficient

than ground-based solutions (Perez-Mato et al.,

2016).

The main limitation posed by these systems is the

danger to the aircraft crew due to the proximity to the

active fire and local turbulence, which may affect the

aircraft’s stability (Perez-Mato et al., 2016). This has

been a direct consequence of many fatal accidents in

the past.

Uncrewed Aerial Vehicles (UAVs) are now

widely used for monitoring tasks under risky

situations (Perez-Mato et al., 2016). However, there

is concern about addressing a loss of control if they

share the same airspace with crewed aircraft (Perez-

Mato et al., 2016). Other typical limitations of

commercially available UAVs are their limited

autonomy, low payload capacity, and limited ability

to withstand strong winds or turbulence (Perez-Mato

et al., 2016).

Wildfire tracking using satellite imagery and

multispectral sensing has often been employed when

large extensions of land or forest are affected by a

severe fire (Perez-Mato et al., 2016). The primary

limitations associated with satellite-based remote

sensing are the time it takes for the image to be

available to the firefighters' staff and the low

periodicity of images captured (Perez-Mato et al.,

2016).

3 RELATED WORK

This section presents the most relevant ontologies

within the scope of our project.

3.1 Ontologies of Fire Prevention

Systems

The literature presents a few ontologies in the field of

fire prevention and safety. A review of this field was

compiled by Nikulina and colleagues (Nikulina et al.,

2019). Some ontologies (Chandra et al., 2022;

García-Castro & Corcho, 2008; Souza, 2014) are

concerned with wildfires, while others focus on fires

in buildings (Bitencourt et al., 2018; Fitkau &

Hartmann, 2021; Nunavath et al., 2016; Tay et al.,

2016).

3.1.1 Wildfires

An ontology of a Semantic Sensor Network for Forest

Fire Management was presented in a study to

semantically enhance fire detection alert inference

methods by integrating meteorological information

and deep knowledge mining from observational data

(Chandra et al., 2022). The authors (Chandra et al.,

2022) discovered that the system's running time rises

while processing large ontologies owing to the

enormous amount of information, which indicates

that the framework's scalability must be improved.

Researchers (Chandra et al., 2022) have considered

adopting similar processing techniques, allowing

several processors to examine various portions of the

ontologies concurrently, improving the execution

time.

The Fire Ontology Network (García-Castro &

Corcho, 2008) and the Fire Ontology (Souza, 2014)

are intended to fight forest fires and address the use

case of wildland fire risk management. However,

some parts of these ontologies may be used for other

purposes, such as fires in buildings.

The Fire Ontology (Souza, 2014) represents the

set of concepts about the fire occurring in natural

vegetation, its characteristics, causes, and effects,

focusing on the Cerrado vegetation domain. There are

53 classes and 19 properties in this ontology. It

KEOD 2022 - 14th International Conference on Knowledge Engineering and Ontology Development

262

focuses on fire characteristics like area burned, fire

frequency, fire intensity, fire severity, flame height,

speed, and spread.

The Fire Ontology Network (García-Castro &

Corcho, 2008) supports the project use case on forest

fire risk management. This approach mainly reuses

SWEET ontology (ESIP Semantic Team, 2022),

which covers the following domains: fire, forest and

vegetation, weather, geography, water body,

infrastructure, location, and time. A SpatialObject

class was added to represent objects that have a

location, classes were identified to be considered as

spatial objects (bodies of water, landforms,

infrastructures, and fire), and the definition of

datasets (to make them cover a region and a temporal

extent) was extended (ESIP Semantic Team, 2022).

3.1.2 Fires in Buildings

The approaches Emergency Fire Ontology

(Bitencourt et al., 2018), Building Fire Emergency

Response (BFER) (Nunavath et al., 2016), and

Building Ontology (Tay et al., 2016) propose

ontologies for emergency fire situations, especially in

buildings. They describe the emergency protocols

aiming to enable end-users to respond quickly to fire

emergencies in facilities. However, they do not focus

on preventive fire safety.

The Preventive Fire Safety Ontology (PrevFis)

contains general descriptions which present the

topology of a building, as well as part of preventive

fire safety, which is crucial for structural fire safety

(Fitkau & Hartmann, 2021). Fitkau and Hartmann

(2021) describe a general ontology based on a

detailed rule-based data source, using the ontology

development METHONTOLOGY (Fernández-

López et al., 1997). This work (Fitkau & Hartmann,

2021) reports on real-world use cases successfully

presented and concluded in PrevFis, collected in close

cooperation with fire safety specialists.

3.2 Ontology of Sensor-based Systems

The Semantic Sensor Network (SSN) is an ontology

developed by a W3C group for describing sensors and

their observations, the involved procedures, the

studied features of interest, the samples used to do so,

the observed properties, and actuators (Haller et al.,

2017). SSN follows a horizontal and vertical

modularization architecture by including a

lightweight but self-contained core ontology for its

elementary classes and properties, called SOSA

(Sensor, Observation, Sample, and Actuator) (Haller

et al., 2017). With their different scope and different

degrees of axiomatization, SSN and SOSA can

support a wide range of applications and use cases,

including satellite imagery, large-scale scientific

monitoring, industrial and household infrastructures,

social sensing, citizen science, observation-driven

ontology engineering, and the Web of Things (Haller

et al., 2017).

4 PROPOSED ONTOLOGY

4.1 Research Approach

Our proposed ontology of a georeferencing mobile

wildfire detection and monitoring system was

designed with the help of experts in the field of

wildfire detection, prevention, and fighting, who

were involved in the evaluation of each iteration. The

initial core of our ontology was based on previous

work on wildfire ontologies (Chandra et al., 2022;

García-Castro & Corcho, 2008; Souza, 2014) and the

SSN ontology (Haller et al., 2017). Thus, the

alignment of the asserted knowledge was an

incremental process built over several iterations.

The development of this ontology was based on

METHONTOLOGY, a well-known methodology

used to build ontologies from scratch (Fernández-

López et al., 1997). It identifies a set of activities

during the ontology development process: planify,

specify, acquire knowledge, conceptualize,

formalize, integrate, implement, evaluate, document,

and maintain. METHONTOLOGY proposes the

following steps: specification, conceptualization,

formalization, integration, implementation, and

maintenance (Fernández-López et al., 1997).

In communication with the domain experts, an

initial conceptual model was developed , using the

traditional Entity Relationship notation (Chen, 1976)

and methods of DEMO and Enterprise Engineering,

namely the Organizational Essence Revealing (OER)

method (Dietz & Mulder, 2020). The knowledge

represented in the initial conceptualization was then

formalized in a more detailed way in an ontology

represented in the Concepts and Relationships

Diagram (CDR) (Gouveia et al., 2021; Pacheco et al.,

2022), an adaptation of the diagram of the Generic

Ontology Specification Language presented in Dietz

and Mulder (2020).

The CDR is a generic, global, and synthetic view

of an entire domain’s concepts while abstracting from

their attributes (Pacheco et al., 2022). In the CDR, a

concept is represented by a collapsible box whose

expansion discloses its attributes, one per line. The

value type of an attribute is specified to the left of the

Ontology for a Georeferencing Mobile System for Real Time Detection and Monitoring of Wildﬁres

263

line, while to the right is the attribute's name. The

value type can be any of the following options:

category, reference, document, text, doc & text,

number, date, or boolean. Arrows express

relationships, which will always consist of an

attribute in one concept whose instances will

reference instances of the other concept. Cardinalities

are represented with arrows pointing to relationships’

“one side”. A dark-filled circle attached to a concept

in one connector means that an instance of this

concept, in order to exist, depends on an instance of

the concept at the other end of the connector (Gouveia

et al., 2021). The specialization/ generalization

relationship is depicted using a connector with a

pointed line (Gouveia et al., 2021; Pacheco et al.,

2022).

In Figure 2, we can find our ontology's concepts,

relationships, and attributes, together with their value

types, essential for implementing software systems.

4.2 Ontology Description

Our proposal constitutes a Georeferencing Mobile

Wildfire Detection System Ontology

(GeMoWilDSOn). This ontology covers a mobile

system's classes, concepts, and attributes to detect,

monitor, and prevent wildfires. The GeMoWilDSOn

includes information on the sensors, their

configurations, deployments, geographic regions,

campaigns, notifications, logs of executed

commands, locations, and captures (georeferenced

polygons, drawn polygons, and images).

Figure 2: Concepts and Relationships Diagram.

KEOD 2022 - 14th International Conference on Knowledge Engineering and Ontology Development

264

The concept Region refers to a specific

geographic area where the Wildfire Detection System

will be used, the relevant attributes are the name of

said region and the relief, that is, the type of terrain in

said region.

A Campaign is an organized fire prevention

activity in a set period (with a start and end date), that

applies to a specific Region.

The Wildfire Detection System has as one of its

primary tools Sensors (which include the ability to get

visible and thermal images of the terrain). Each of

these Sensors has a set of intrinsic properties of the

equipment, namely serial number, a software and

hardware version, a specific camera lens length

specification, the expected field of view of the lens,

and the battery capacity.

The concept of Sensor Deploy regards the use of

the before-mentioned Sensors in specific Campaigns,

deployed in a given Location within the region. These

deployments take place within a specific timeframe

(that can be the same as the Campaign itself) and

therefore need a start and end timestamp. Other

relevant attributes include: the latitude and longitude

where it has been deployed (as well as a less exact

approximation of said latitude and longitude); the

specific height of the deployment; the type of the

communication (for example, medium band); the

alarm identification code; the team responsible for the

deployment of that sensor; the current battery levels;

and the state of the deployment (if it is ongoing, has

ended or has any issues).

Each Sensor Deploy also needs a Sensor

Configuration where the parameterization for that

specific Sensor is set, although, during the

deployment time, this configuration may be changed

multiple times. The Sensor Configuration includes a

timestamp of when it took place, a yaw and pitch

range, a set roll (movement and position of the sensor

within its axis), a capture frequency for the polygons

and images, and the sweep status (if it is in sweep or

fixed mode).

Despite the default Sensor Configuration, it is

also possible to send specific Commands to take

effect in the Sensor Deployment. These Commands

can be of four different types; 1) Change the Capture

Frequency, where the time between captures can be

altered; 2) Change Yaw, where the range of the Yaw

rotation can be modified; 3) Change Pitch, where the

range of rotation of the Pitch can be changed; and 4)

Change Sweep Mode where the type of Sweep mode

can be changed.

When a Sensor Deploy is in use and depending on

the set Capture Frequency, the Sensor will make

multiple image captures. These Images contain

multiple attributes, namely, the visible image, the

thermal image, the yaw, pitch, and roll of when it took

place, the temperate, pressure, and humidity of the

area, as well as wind direction and speed. These

captures are processed by georeferencing algorithms

that take information from the thermal image,

position of the sensor, and terrain information,

namely contour lines, to produce polygons that

represent areas that are for sure covered by fire and

shadow areas that, due to the steepness of the

mountains, one cannot be sure that fire is there or not.

In the georef polygon concept, the attribute type

indicates if the area is of type fire or shadow.

Operators may complement the capture with

manually drawn polygons, both of fire and shadow

types. Experienced fire prevention and fight

coordinators will know, considering fire progression,

local meteorological conditions, knowledge of the

terrain, and other direct observations (by aircraft

pilots, for example), how to complement the

automatically identified fire and shadow areas. Such

complementary information is essential for a more

informed evolution of the fire and necessary decisions

on how to allocate firefighting resources and possible

changes of the sensors' positions. These

modifications, might allow a complete view of the

situation and more effective automatic generation of

fire and shadow areas by the algorithms.

The concept of Location, before mentioned in the

Sennsor Deploy, also has its own set of attributes with

the latitude, longitude, height, and accuracy that are

optimal for sensor deployment in said location. This

optimal location may or may not be used in a specific

deployment due to logistic reasons.

Each Location can also have a Range Polygon

associated, which is a polygon that delimits the total

visible area that the sensor can capture, considering

the surrounding topography, like mountains that

could block the view. This information is mainly used

to pick the best location for deploying each sensor,

maximizing the covered area.

The final concept is the Notification, which

reports the association of a Sensor Deployment and a

specific predetermined Location (or not).

4.3 Validation

The ontology was validated in real-world tests that

took place in three archipelagos in Macaronesia:

Cabo Verde, Madeira, and Canaries. These tests

relied on the involvement and contribution of teams

of experts from different local wildfire prevention and

fighting services, in particular, Instituto das Florestas

e Conservação da Natureza (Madeira), Cabildo de

Ontology for a Georeferencing Mobile System for Real Time Detection and Monitoring of Wildﬁres

265

Gran Canaria (Canaries), and Serviço Nacional de

Proteção Civil e Bombeiros (Cabo Verde).

A complete hardware and software system was

developed, allowing real-world tests and ontology

validation. Figure 3 shows a diagram of the system's

architecture.

Figure 3: Architecture of the system.

The system’s hardware is constituted by a set of

mobile sensors (cameras that capture both visual and

thermal images) and two web servers. One of the

servers, implemented by our project partner, handles

sensor information processing and command

reception. The other server, developed by us,

aggregates the database, sensor information reception

interface as well as a GUI component.

The GUI is composed of three visors. The

MainVisor shows a satellite or contour line view of

the terrain, as well as real sensor positions (camera

symbols), their range of observation, and fire and

shadow polygons, both the automatically generated

and the drawn ones. In a layer selection feature, one

can choose to see all types of polygons or just a few.

This visor also has a timeline feature where one can

navigate through a set time interval, to view, analyze,

and eventually draw/edit/delete the polygons. The

NodeVisor displays more detailed information about

the selected sensor and also allows the user/operator

to send commands to change some of its settings. The

ImageVisor shows the images, both visible and

thermal, that were captured by the selected sensor in

the time and date that is currently selected in the

timeline navigation bar. Figure 4 shows a screenshot

of the main GUI.

Figure 4: System’s GUI with thermal image.

The real-world tests in the three Macaronesian

regions started in Santiago Island (Cabo Verde), in

June 2022, focusing on the mobile device itself.

Around 40 participants were involved in these tests.

In August 2022, the tests took place in Madeira Island

where the integrated testing of the entire system was

carried out for the first time, including the sensors and

the two web servers. The tests in Madeira included

around 25 participants. The entire system was

improved based on the experience acquired, and the

final test occurred in September 2022, in Gran

Canaria (Canary Islands). In this last field test,

participated six fire prevention and fighting experts,

representing the three regions involved in the project.

The last two tests included the detection of

wildfire ignitions by the system and the analysis, by

the experts in fire prevention and fighting, of the:

comprehensiveness and completeness of the data

collected, and the functionalities of the GUI.

The feedback given both on the system's ontology

and functioning was very positive, and experts did not

identify any missing information.

4.4 Discussion

Our work proposes the GeMoWilDSOn, an ontology

for georeferencing mobile wildfire detection and

monitoring systems. We foresee that the number of

different mobile sensor-based systems to detect and

monitor fires will increase shortly. In our search for

existing ontologies in the scope of our work, we

verified that a few ontologies existed regarding fires,

but none foreseeing mobile georeferencing of fires

based on thermal images and terrain information. Fire

Ontology (Souza, 2014) and Fire Ontology Network

(García-Castro & Corcho, 2008) focus mainly, or

only on fire, natural resources, and infrastructure

concepts. Semantic Sensor Network Ontology (Haller

et al., 2017) and Semantic Sensor Network for Forest

Fire Management (Chandra et al., 2022) are primarily

focused on the sensors’ domain. These ontologies are

somewhat generic, complex, and more of a normative

kind, giving freedom for implementation. They do not

easily translate to a streamlined data model that can

be efficiently implemented in a real-time system

where fast performance is essential and of utmost

value.

Although a sensor-based approach like the one

proposed by Chandra and colleagues (2022), with

bases on SSN and SOSA, can be highly versatile and

used in practically any domain where sensors are

involved, the fact of being so generic also means it is

not optimized for any particular scenario.

KEOD 2022 - 14th International Conference on Knowledge Engineering and Ontology Development

266

GeMoWilDSOn was tailored for forest fires' fight

and prevention with sensors that capture images and

georeferentiation. It can be easily adapted to expand

the range of sensors/capabilities for other contexts, if

so is necessary. Nevertheless, as it is,

GeMoWilDSOn was based and is totally aligned with

the classes prescribed in the sensor-based approach

regarding our implementation scope, thus making our

solution grounded and validated according to the

state-of-the-art.

Our proposal of the GeMoWilDSOn fills the

found gaps in state-of-the-art by being an ontology

that encompasses both the fire and sensor domains in

a streamlined way. This ontology was already

validated in real tests in the field with a software

system close to being completed. An important

innovation and research contribution of our ontology

is the possibility of complementing captured

information and georeferenced polygons with

manually drawn polygons to complement

automatically generated information.

5 CONCLUSIONS AND FUTURE

WORK

This paper presents GeMoWilDSOn, an ontology for

georeferencing mobile wildfire detection and

monitoring systems. We analyzed related work in the

context of our practical research project for

developing an innovative mobile sensor-based

system with associated APIs, database, and GUI. We

identified some concepts in previous work, but we

needed to introduce innovative ones and simplify

some views of related work. The result is a

streamlined ontology focused on fire prevention and

fighting with mobile sensors, specially designed for

steep mountainous terrain, where firefighting can be

complicated and complex. The innovative

contribution of manually drawn areas of fire and

shadow is of utmost importance concerning the

particularities of steep terrain.

Our ontology was validated in three real-world

tests where experts were delighted with captured

information and its representation in the GUI of the

developed system.

Our project is currently in its final stages. We are

now implementing the final task consisting of a

sensor location management and advisor component.

With this new component, it will be possible to

manually add to the system possible preferred and

advisable locations to deploy the mobile sensors.

Another important feature will be the possibility of

the system, considering current fire and shadow areas,

as well as terrain contour lines, to automatically

advise changes of sensor locations, so those shadow

areas (and the size of shadows) are reduced. Thus fire

areas will be more clearly identified and represented.

We expect to extend our proposed ontology with new

concepts needed for these practical features.

Another line of future work that we foresee, in the

context of a subsequent research project, is the

inclusion of UAV-based sensors. Very recent

developments in UAV technology in terms of

stabilization and autonomy will allow the integration,

in our architecture, of information provided by these

aerial sensors. Hence, allowing for a completer

information on fire location and evolution and

improving the overall effectiveness of our system,

fire prevention, and fight in general.

ACKNOWLEDGEMENTS

Special thanks to the participants in the study that

contributed with their fruitful insights and feedback.

This work was supported by the Regional

Development European Fund (INTERREG MAC),

project GesFoGO MAC/3.5b/227 (MAC-2014-

2020).

REFERENCES

Arana-Pulido, V., Cabrera-Almeida, F., Perez-Mato, J.,

Dorta-Naranjo, B. P., Hernandez-Rodriguez, S., &

Jimenez-Yguacel, E. (2018). Challenges of an

autonomous wildfire geolocation system based on

synthetic vision technology. Sensors (Switzerland),

18(11). https://doi.org/10.3390/s18113631

Bitencourt, K., Araújo Durão, F., Mendonça, M., & Laique

Bomfim de Souza Santana, L. (2018). An ontological

model for fire emergency situations. IEICE

Transactions on Information and Systems, E101D(1),

108–115.

https://doi.org/10.1587/transinf.2017SWP0003

Chandra, R., Abhishek, K., Agarwal, S., & Singh, N.

(2022). Semantic Sensor Network Ontology based

Decision Support System for Forest Fire Management.

ArXiv, 1–20.

https://doi.org/10.48550/arxiv.2204.03059

Chen, P. P. S. (1976). The Entity-Relationship Model—

toward a Unified View of Data. ACM Transactions on

Database Systems (TODS), 1(1), 9–36.

https://doi.org/10.1145/320434.320440

Dietz, J. L. G., & Mulder, H. B. F. (2020). Enterprise

Ontology: A Human-Centric Approach to

Understanding the Essence of Organisation. Springer

Ontology for a Georeferencing Mobile System for Real Time Detection and Monitoring of Wildﬁres

267

International Publishing. https://doi.org/10.1007/978-

3-030-38854-6

EFFIS Weekly Cumulative Burnt Areas. (n.d.). Retrieved

August 12, 2022, from

https://effis.jrc.ec.europa.eu/apps/effis.statistics/season

altrend/EU

ESIP Semantic Team (2022). Semantic Web for Earth and

Environment Technology Ontology - SWEET.

https://bioportal.bioontology.org/ontologies/SWEET

Fernández-López, M., Gómez-Pérez, A., & Juristo, N.

(1997). METHONTOLOGY: From Ontological Art

Towards Ontological Engineering. Engineering

Workshop on Ontological Engineering (AAAI97).

Fitkau, I., & Hartmann, T. (2021). Building Ontology for

Preventive Fire Safety. EG-ICE 2021 Workshop on

Intelligent Computing in Engineering, Proceedings, 8,

0–9.

Freitas, R., Romeiras, M., Silva, L., Cordeiro, R., Madeira,

P., González, J. A., Wirtz, P., Falcón, J. M., Brito, A.,

Floeter, S. R., Afonso, P., Porteiro, F., Viera-

Rodríguez, M. A., Neto, A. I., Haroun, R., Farminhão,

J. N. M., Rebelo, A. C., Baptista, L., Melo, C. S., …

Ávila, S. P. (2019). Restructuring of the ‘Macaronesia’

biogeographic unit: A marine multi-taxon

biogeographical approach. Scientific Reports, 9(1), 1–

18. https://doi.org/10.1038/s41598-019-51786-6

García-Castro, R., & Corcho, O. (2008). Fire Ontology

Network.

http://linkeddata4.dia.fi.upm.es/ssg4env/index.php/ont

ologies/12-fire-ontology-network/default.htm

Gouveia, B., Aveiro, D., Pacheco, D., Pinto, D., & Gouveia,

D. (2021). Fact Model in DEMO - Urban Law Case and

Proposal of Representation Improvements. Lecture

Notes in Business Information Processing, 411, 173–

190. https://doi.org/10.1007/978-3-030-74196-9_10

Haller, A., Janowicz, K., Cox, S., Phuoc, D. Le, Taylor, K.,

& Lefrançois, M. (2017). Semantic Sensor Network

Ontology. W3C Recommendation.

https://www.w3.org/TR/vocab-ssn/

Nikulina, Y., Shulga, T., Sytnik, A., Frolova, N., &

Toropova, O. (2019). Ontologies of the Fire Safety

Domain. In O. Dolinina, A. Brovko, V. Pechenkin, A.

Lvov, V. Zhmud, & V. Kreinovich (Eds.), Studies in

Systems, Decision, and Control (Vol. 199, pp. 457–

467). Springer. https://doi.org/10.1007/978-3-030-

12072-6_37

Nunavath, V., Prinz, A., Comes, T., & Radianti, J. (2016).

Representing Fire Emergency Response Knowledge

Through a Domain Modelling Approach. NOKOBIT

2016, 24(1), 28–30.

Pacheco, D., Aveiro, D., & Pinto, D. (2022). Evaluation of

the perceived quality and functionality of Fact Model

diagrams in DEMO. Lecture Notes in Business

Information Processing, 3, 1–19.

Perez-Mato, J., Arana, V., & Cabrera-Almeida, F. (2016).

Real-time autonomous wildfire monitoring and

georeferencing using rapidly deployable mobile units.

IEEE Aerospace and Electronic Systems Magazine,

31(2), 6–15.

https://doi.org/10.1109/TAES.2016.150057

San-Miguel-Ayanz, J., Schulte, E., Schmuck, G., Camia,

A., Strobl, P., Liberta, G., Giovando, C., Boca, R.,

Sedano, F., Kempeneers, P., McInerney, D., Withmore,

C., de Oliveira, S. S., Rodrigues, M., Durrant, T., Corti,

P., Oehler, F., Vilar, L., & Amatulli, G. (2012).

Comprehensive Monitoring of Wildfires in Europe:

The European Forest Fire Information System (EFFIS).

Approaches to Managing Disaster - Assessing Hazards,

Emergencies and Disaster Impacts, May 2014.

https://doi.org/10.5772/28441

Souza, A. (2014). Fire Ontology.

https://bioportal.bioontology.org/ontologies/FIRE

Tay, N. N. W., Kubota, N., & Botzheim, J. (2016). Building

Ontology for Fire Emergency Planning and Support. E-

Journal of Advanced Maintenance, 8(2), 13–22.

KEOD 2022 - 14th International Conference on Knowledge Engineering and Ontology Development

268