Valley Harpy Eagle| Automated Detection of Hazards
Automated Detection of Hazards
SENT.in (Satellite Environmental Threats Interface)
Summary
It has been known for many years that environmental preservation and care for nature are extremely important for humanity, both for the present and for the future and, recently, large-scale forest fires, such as in Australia, Brazil, and the United States, have been highlighted and caught the world's attention. As a consequence, we have developed SENT.in, a tool that will make it possible to predict in real-time the chance of a fire outbreak occurring in a given location. Eventually, the tool will have a wide range of ancillary data, which will make it easier for civil society and governments to make decisions.
How We Addressed This Challenge
We developed a Machine Learning system capable of pointing down in a heat map the risk of fire in a given location. This process occurs due to the correlation made between the information obtained in satellite data, such as precipitation, days without rain, or season. This technology is highly important since it points out the susceptibility of that specific region having a wildfire, something that could support decision-makers and scientists to take prevention measures.
Furthermore, we plan to add more complex data to have a higher confidence level on the Machine, and include socioeconomic and geographical data, to deepen the tool utility. Besides, we not only want to detect or predict where such forest fires could occur but also issue alerts so that the authorities and the population are aware of the risks that these disasters can impose.
How We Developed This Project
Recently, the world has been undergoing a wave of wildfires of gigantic magnitudes, such as the Bushfires in Australia, which led to the death of over 3 billion animals and at least 33 people, or the Amazon Rainforest Wildfires, with an estimated 906 thousand hectares of forest burnt. With that in mind, we recognize the extreme significance of developing a system capable of predicting and detecting wildfires, aligning such information with data that can assist decision-makers in determining the best approach to the problem.
During our developing process, we tried to convey as most databases as possible, to guarantee a detailed and efficient detection mechanism. Among our tools, we used satellite data from INPE and NASA to evaluate the fire risk in the selected regions. Furthermore, we developed our tool by using a wide array of softwares, like Power BI, Github, Jupyter, and others. The coding language that we used to build our project was Python due to its simplicity and plethora of formulas that facilitate the coding process.
However, our group encountered some barriers during the development of the program. We created a separate document to describe what these problems were but in summary, due to the short time for formulation and a small team of developers, we decided to leave out of our initial prototype some variables as human factors, since the monitoring and categorization of every anthropogenic action is a difficult task, and does not guarantee reliable records. And another example would be the automation of creating heat maps. Currently, we obtain the final data set, import it into the program, configure it, and produce the graph; but we hope to automate this process in the future.
Finally, despite such difficulties, our team could have several achievements. Among them are being able to relate the variables linked to fires not only empirically but also through correspondence graphs. Also, we have been able to interpret these data and place them on a heat map, a process that we want to automate in the future.
How We Used Space Agency Data in This Project
Our developers utilized NASA FIRMS Active Fire Data to have a base for our machine learning model. However, our primary data source was INPE BDQueimadas, which uses satellite imagery from NASA's satellites and some other sensory data.
Also, we consulted data sources from the space agencies JAXA and CSA ASC, but we plan to utilize this information in a future iteration due to the short timeslot.
Project Demo
Project Slides: https://drive.google.com/file/d/1NXqK4i0EHGm7HxbREH7Q7yUbYBEMD8E4/view?usp=sharing
Demo Vídeo (30 seconds): https://www.youtube.com/watch?v=bb7i3ltq2vI
Complementary Materials
Regional Project Video (São José dos Campos): https://www.youtube.com/watch?v=g-xXVGIBCWQ&feature=youtu.be
Project Development Implications: https://drive.google.com/file/d/1sWUqcURla2NWTkR8UM3t7xTlD3z9iAui/view?usp=sharing
Project Code
Data & Resources
1. Banco de Dados de Queimadas (INPE) - http://queimadas.dgi.inpe.br/queimadas/bdqueimadas#
2. Global Fire Emissions Database (GFED) - https://globalfiredata.org/pages/amazon-dashboard/#faq_methods
3. NASA FIRMS - https://firms.modaps.eosdis.nasa.gov/active_fire/#firms-shapefile
4. Fire spot identification based on hotspot sequential pattern and burned area classification - https://www.researchgate.net/publication/328917720_Fire_spot_identification_based_on_hotspot_sequential_pattern_and_burned_area_classification
5. Potential influence of meteorological variables on forest fire risk in Serbia during the period 2000-2017 - https://www.degruyter.com/view/journals/geo/11/1/article-p414.xml?language=en
6. Correlations between the meteorological elements and the occurrences of forest fires in the urban area of Juiz de Fora, MG - https://www.scielo.br/scielo.php?script=sci_arttext&pid=S0100-67622011000100017
7. Understanding Fire Danger - https://www.nps.gov/articles/understanding-fire-danger.htm
8. CSA ASC Open Data MOPPIT - ftp://data.asc-csa.gc.ca/users/OpenData_DonneesOuvertes/pub/MOPITT/
9. Climate at a Glance (NOAA) - https://www.ncdc.noaa.gov/cag/national/time-series
10. NOAA North American Drought Monitor - https://www.ncdc.noaa.gov/temp-and-precip/drought/nadm/indices