1st Brazilian Fire Fighters

https://youtu.be/9CYttnD_lwE

https://imgur.com/a/IvSDJ3l

How We Addressed This Challenge

Due to the relevance of the chosen challenge, we focused on the effects of urban and forest fires on the most vulnerable communities on the planet. Thus, we seek ways to streamline care for these groups, preferably, since they are victims of environmental injustice in this serious danger (IBGE, 2020). This vulnerability may stem from their limited resources to fight fires and protect their lives and property, as in the case of indigenous tribes, quilombolas and low-income communities (e.g., slums, suburbs and neighborhoods with precarious structures and urbanization). The absence of fire prevention and firefighting infrastructure and the location of such communities, close to vegetation areas or on the outskirts of cities, make these groups more susceptible to the emergence of new outbreaks (Carneiro Filho, 2019; IBGE, 2020). Additionally, this vulnerability may be the result of the intrinsic characteristics of some groups, such as the elderly and children, as in both cases they are groups more susceptible to the effects of the smoke emission, in addition to presenting greater difficulties in cases of evacuations on the occurrence of fires.

Moreover, our project addresses one of the main environmental problems in the world nowadays: the high incidence of fires in dry climate regions. This danger is the result of climate change and endangers the lives of millions of people every year (HRW, 2020a), in addition to destroying properties, the native fauna and flora, threatening the planet's biodiversity. Particularly in Brazil, fires have been recurrent in urban and forest regions of different Biomes, mainly in the Brazilian Savannah (Cerrado) and the Amazon (INPE, 2021). Other countries have also been hit hard by the accentuation of the dry and hot climate, such as Greece, which had entire islands threatened in 2021 (European Forest Fire Information System (EFFIS)), as well as the state of California in the USA, which every year sees forest fires reaching a growing and significant number of people (National Interagency Fire Center (NIFC)). In all cases, it is also possible to see an increase in health problems related to smoke emission (HRW, 2020b). Thus, our group focused on designing technological and scientific solutions that could reduce the impact of these environmental incidents, prioritizing marginalized communities.

Detailed Project Description

Our team proposes the development of an integrated platform of solutions for preventing and fighting urban and wildfires called ONFIRE. These solutions make use of science and technology to mitigate the consequences of fires, while seeking to prioritize vulnerable communities. By offering preferential care to these groups through the solutions of the ONFIRE platform, we seek to mitigate inequalities and achieve greater environmental justice. 

For that, one of the main technological solutions applied in ONFIRE are drones (UAVs), which are the equipment that has become popular in recent years, with an affordable cost to think about developing an autonomous fire brigade, capable of supporting fire combat teams in land to protect a particular region or city of a country. 

In addition, their fast displacement skill allows us to use them to speed up care for the vulnerable. UAVs are used in ONFIRE in three different situations: (i) to help in the rapid identification of new fire spots, not yet detectable in satellite images; (ii) to monitor the spread of flames in a unleashed fire and (iii) to assist in the rapid evacuation of communities identified in the "flame path", acting as an audible alarm, while helping people and vehicles indicated as the best escape routes, faster and safer.

Another technological tool used is the databases of space satellite images, which are now widely available by space agencies such as NASA, INPE, among others, which cover different regions of the planet. It would be unfeasible to think of drones covering all monitored areas. On the Other hand, satellite images allow the platform for a more complete scan of fires that have already started. Once a burning area is identified in these images, through databases accesses (Pathfinder (2019), INPE (2021) and SEDAC (2018)), the ONFIRE platform can take care new regions with fires in real time , triggering to monitoring actions.

ONFIRE also employs different scientific approaches. First, the images captured by cameras used to equip the drones are sent to the platform, to be processed and classified, aiming to identify fire spots, people, obstacles, among others. For that, there are several research techniques in the field of digital image processing and pattern recognition that can be used in ONFIRE, such as, Deep Learning (LeCun, 2015).

Another important issue concerns the choice of priority regions to send the surveillance drones aiming to identify new fire spots. It would be impracticable to have an entire city or forest area covered by drones. In such case, scientific bases from space agencies and research institutes (such as an association between the Pathfinder (2019) and IBGE (2020) bases), will feed tools based on Machine Learning (ML) and Data Science (CD), to create a predictive model capable of anticipating the emergence of fire outbreaks, identifying them as areas of greatest risk. In this prediction task, we can mention the use of regression algorithms (linear, logistic, among others) (Sigaud, 2011) and clustering algorithms (k-means, Kohonen) (Xu, 2005) for the development of a recommendation system.

Historical databases will support the identification of areas classified with a higher risk of new fires. ONFIRE will prioritize marginalized people that live or stay in the area under surveillance by drones. The risk area data will be cross-checked with geopolitical information that point to the existence of vulnerable communities in the regions identified as being at high risk (IBGE, 2020). As a result, a prioritization of areas targeted for surveillance will be generated, that is, those with the greatest risk and greatest vulnerability. Based on this information, teams of drones for medium altitude flights (80 to 120 m), equipped with thermal cameras, will be directed to these priority target areas to carry out surveillance in search of new fire spots.

Once new fire spots are identified, an alarm will be sent for the firefighting agents of the endangered region. Thus, it is expected that vulnerable regions get faster treatment, lowering the chances of increasing the fire and all the serious consequences of those.

While mitigating the risks, this measure is not enough to grant that big fires will not occur in the cities and forest areas under supervision. We also propose the use of science and technology in those cases where the fire is already established. Once the fire is identified by satellite images, high altitude drones (200m or higher), with thermal cameras, would be sent to monitor and capture real time data about the fire spreading. This data will help the firefighters teams on land, as well as feed a fire spreading prediction model. The outcomes of such model will also be used to guide the land teams, improving their strategies and decisions.

The fire propagation model will be based on computational techniques and AI, such as cellular automata modeling (Chopard, B., 1998) and differential equations (ØKSENDAL, 2003), to achieve good prediction accuracy. This kind of modeling has a good maturation on basic research and it can be easily applied to real fires by using the appropriate data to have a good accuracy. The figure below shows a simulation made using a cellular automata model for fire spreading to illustrate the predictive model to be applied in ONFIRE. 

The space agencies databases and other databases (Pathfinder (2019), INPE, (2021) e SEDAC (2018)) with local information about the monitored areas will provide important real time information like wind speed and direction, humidity and vegetation type. Furthermore, the real time data provided by the drones will be processed and used to feed and to iteratively adapt this model periodically (for example, at each ten minutes).

Once the model identifies a city or forest area that will probably be reached by the flames in minutes or hours, the system will send a new alert to the civil defence team. In this way, the evacuation of people located in those high risk areas can be realized as soon as possible. In the meantime, the information can be sent to satellite navigation softwares (like Google Maps and Waze) for vehicles to avoid the endangered region, by the flames or the smoke.  

Besides, the system can cross this information again with the geopolitics databases (IBGE, 2020) to identify the existence of vulnerable communities in the flame and smoke path. In the case of realizing that these communities are endangered, rescue drones would be sent to assist in evacuation with priority.

Arriving on those spots, the drone squad will emit sound and visual signals to alert the people present in the area, as well as direct pedestrians and vehicles to safer routes, acting as an independent fire brigade. To construct these safe routes, ONFIRE needs to be feeded by other database information, related to the local map of streets, in the case of cities, or paths and roads, in the case of forest areas.

Survey

Knowing how important it is to validate our project and ask for public opinion, our team interviewed 60 outsiders (from 10 to 70 years old) through “Instagram”. All of the interviewees agreed about how fires offer greater risks for those who live in marginalized communities. Furthermore, 95% of the participants think that investing in surveillance and monitoring systems would help minimize the environmental injustice and approximately 99% of them believe that drones would be the best option to do that.

When a fire occurs, it is necessary to apply an assertive and fast response to make sure that the problem does not escalate and reach people, especially vulnerable groups because they are more susceptible to the consequences of this environmental hazard. Using data based on past fires enables planning assertive firefighting strategies. In this way, the proposed solution (ONFIRE) is programmed to extract important information from the available databases to the decision making process. In addition to this process, the databases are essential to the training of the prediction models, which will be able to establish what areas are in greater danger and how the fire will evolve. In this way, many important databases were identified that contain relevant data to feed different steps of the proposed solution. Succinctly, the solution ONFIRE is divided in three steps: Surveillance, Tracking and Evacuation. The databases employed in the ONFIRE platform are described bellow with the corresponding steps where they are applied.

• (IBGE, 2020) Instituto Brasileiro de Geografia e Estatística - IBGE. 2020. Base de Informações Geográficas e Estatísticas sobre os indígenas e quilombolas para enfrentamento à Covid-19. Rio de Janeiro, RJ: Ministério da Economia (ME). https://www.ibge.gov.br/geociencias/organizacao-do-territorio/tipologias-do-territorio/27480-base-de-informacoes-sobre-os-povos-indigenas-e-quilombolas.html?=&t=acesso-ao-produto. Accessed 02/10/2021.
• (Pathfinder, 2019) National Aeronautics and Space Administration - NASA. 2019. Wildfires Data Pathfinder. Washington, D.C.: Earthdata - NASA Data Pathfinders. https://earthdata.nasa.gov/learn/pathfinders/wildfire-data-pathfinder. Accessed 02/10/2021.
• (INPE, 2021) Instituto Nacional de Pesquisas Espaciais - INPE. 2021 (real time). Portal de Dados Abertos do Programa de Queimadas. São José dos Campos, SP: Ministério da Ciência, Tecnologia e Inovações (MCTI). https://queimadas.dgi.inpe.br/queimadas/dados-abertos/#. Accessed 02/10/2021.
• (SEDAC, 2010) Center for International Earth Science Information Network - CIESIN - Columbia University, and Information Technology Outreach Services - ITOS - University of Georgia. 2013. Global Roads Open Access Data Set, Version 1 (gROADSv1). Palisades, NY: NASA Socioeconomic Data and Applications Center (SEDAC). https://doi.org/10.7927/H4VD6WCT. Accessed 02/10/2021.
• (SEDAC, 2018) Center for International Earth Science Information Network - CIESIN - Columbia University. 2018. Global Fire Emissions Indicators, Country-Level Tabular Data: 1997-2015. Palisades, NY: NASA Socioeconomic Data and Applications Center (SEDAC). https://doi.org/10.7927/H4V69GJ5. Accessed 02/10/2021.

The databases related to wildfires SEDAC (2018), Pathfinder (2019) and INPE (2021) will be applied in the Surveillance state, along with the database IBGE (2020) referring to vulnerable communities. In the Monitoring state, the database Pathfinder (2019) will help the system with the physical characteristics of the environment. Finally, the IBGE (2020), SEDAC (2010) and Pathfinder (2019) bases will be applied in the Evacuation state.

Our journey had promoted a change of mind regarding natural disasters and the way to attack them considering the different needs of communities. Inequality between communities goes beyond financial differences. It impacts all life aspects, including environmental disasters. We can use our scientific knowledge and technologies already developed to change the world, specifically by mitigating environmental injustices. Moreover, in the database search performed this weekend, we learned that there is more data available to help to solve this issue than we might think about.

Our participation in NASA SpaceApps started with the initial motivation to use drones and AI for firefighting. We were more inclined to choose another challenge, which proposes to investigate the use of drones for the development of cities. But, from our initial brainstorm and analyzing the challenges that most interested us, we realized the importance of analyzing this environmental problem from the perspective of marginalized populations. Therefore, we changed the challenge chosen initially and we believe that this added a lot of value to the project and to our perceptions about the problem.

After consulting and analyzing some of the available bases, we made a correlation with the motivation of the challenge that deals with the reduction of environmental injustices, and together with our previous knowledge about drones, imaging and, mainly, computational techniques of AI and ML for the construction of predictive models. Thus, we developed an integrated proposal that combines the advantages of these solutions to seek to reduce the inequalities faced by marginalized communities in the face of fires, whether in the city or in the countryside.

We highlight the teamwork, combined with the different skills of the members o four group, which allowed us to achieve success in executing this project and overcome the setbacks that arose. Also important in this success are the stages of brainstorming, planning, organizing and dividing tasks according to each skill.

Finally, we would like to thank Luiz Gustavo Almeida Martins, Kleber Del Claro, Maurício Escarpinati, and Ana Carolina who helped us in different ways, with suggestions or even some scientific clarification. We are also grateful for the mentoring and lectures that were offered by the Brazilian organization.

• Carneiro Filho, A.; Souza, O. B. “Atlas de pressões e ameaças às terras indígenas na Amazônia brasileira”. Brasília: ISA, 2009.
• Chopard, B., & Droz, M. (1998). Cellular automata (Vol. 1). Berlin, Germany: Springer.
• (HRW, 2020a) Human Rights Watch. “Brazil: Amazon Fires Affect Health of Thousands - Bolsonaro Government Fuels Deforestation, Flouts Climate Commitments”. Nova Iorque, NY: Human Rights Watch - HRW. August, 2020. Retrieved from: <https://www.hrw.org/news/2020/08/26/brazil-amazon-fires-affect-health-thousands> Accessed 03/10/2021.
• (HRW, 2020b) Human Rights Watch. “The Air is Unbearable - Health Impacts of Deforestation-Related Fires in the Brazilian Amazon”. Nova Iorque, NY: Human Rights Watch - HRW. August, 2020. Retrieved from: <https://www.hrw.org/report/2020/08/26/air-unbearable/health-impacts-deforestation-related-fires-brazilian-amazon>. Accessed 03/10/2021.
• (IBGE, 2020) Instituto Brasileiro de Geografia e Estatística - IBGE. 2020. Base de Informações Geográficas e Estatísticas sobre os indígenas e quilombolas para enfrentamento à Covid-19. Rio de Janeiro, RJ: Ministério da Economia (ME). https://www.ibge.gov.br/geociencias/organizacao-do-territorio/tipologias-do-territorio/27480-base-de-informacoes-sobre-os-povos-indigenas-e-quilombolas.html?=&t=acesso-ao-produto. Accessed 02/10/2021.
• (INPE, 2021) Instituto Nacional de Pesquisas Espaciais - INPE. 2021 (real time). Portal de Dados Abertos do Programa de Queimadas. São José dos Campos, SP: Ministério da Ciência, Tecnologia e Inovações (MCTI). https://queimadas.dgi.inpe.br/queimadas/dados-abertos/#. Accessed 02/10/2021.
• LeCun, Y., Bengio, Y., & Hinton, G. (2015). Deep learning. Nature, 521(7553), 436-444.
• ØKSENDAL, Bernt. (2003) Stochastic differential equations. In: Stochastic differential equations. Springer, Berlin, Heidelberg. p. 65-84.
• (Pathfinder, 2019) National Aeronautics and Space Administration - NASA. 2019. Wildfires Data Pathfinder. Washington, D.C.: Earthdata - NASA Data Pathfinders. https://earthdata.nasa.gov/learn/pathfinders/wildfire-data-pathfinder. Accessed 02/10/2021.
• (SEDAC, 2010) Center for International Earth Science Information Network - CIESIN - Columbia University, and Information Technology Outreach Services - ITOS - University of Georgia. 2013. Global Roads Open Access Data Set, Version 1 (gROADSv1). Palisades, NY: NASA Socioeconomic Data and Applications Center (SEDAC). https://doi.org/10.7927/H4VD6WCT. Accessed 02/10/2021.
• (SEDAC, 2018) Center for International Earth Science Information Network - CIESIN - Columbia University. 2018. Global Fire Emissions Indicators, Country-Level Tabular Data: 1997-2015. Palisades, NY: NASA Socioeconomic Data and Applications Center (SEDAC). https://doi.org/10.7927/H4V69GJ5. Accessed 02/10/2021.
• Sigaud, O., Salaün, C., & Padois, V. (2011). On-line regression algorithms for learning mechanical models of robots: a survey. Robotics and Autonomous Systems, 59(12), 1115-1129.
• Xu, R., & Wunsch, D. (2005). Survey of clustering algorithms. IEEE Transactions on neural networks, 16(3), 645-678.
• Zabaniotou, A., Pritsa, A., & Kyriakou, E. A. (2021). “Observational Evidence of the Need for Gender-Sensitive Approaches to Wildfires Locally and Globally: Case Study of 2018 Wildfire in Mati, Greece”. Sustainability, 13(3), 1556.

#EnvironmentalJustice #MarginalizedCommunities #Fires #DataAnalytics #FireSimulation #ArtificialIntelligence #CellularAutomata #PredictiveModels #drones #UnmannedAerialVehicle

This project has been submitted for consideration during the Judging process.