TAMIDS Wildfire Data Science Challenge Awards

Finalist teams, judges, and organizers at the Wildfire Data Science Challenge final event on April 18, 2023

The month-long 2023 TAMIDS Student Data Science Competition, the Wildfire Data Science Challenge culminated in the final event on April 18, 2023. The competition was held over a five-week period in March and April 2023. From an initial registration of 150 participants, there were 13 team submissions from which 8 were selected for presentation to the judges at the final event where prizes totaling $9,500 were awarded to the winning teams. The competition received generous support from Chevron, the Texas A&M University Departments of Statistics and Electrical & Computer Engineering, and the Texas A&M University School of Performance, Visualization & Fine Arts.

The Wildfire Data Science Challenge

The Texas A&M Forest Service (TFS) was the first state forestry agency to be established as part of a land-grant college. In collaboration with TFS, the Texas A&M Institute of Data Science (TAMIDS) organized the Wildfire Data Science Challenge to seek innovative and sustainable solutions that can help better predict and manage wildfires and ultimately mitigate the effects of fires on both people and the environment. Participants were expected to address one or both of the following challenges and provide actionable advice to decision-makers based on their analysis.

  • Predicting wildfire behavior: Wildfire behavior is influenced by a complex interplay of factors, and it is difficult to predict how a wildfire will behave in a specific location. This makes it challenging to develop accurate models and forecasting tools.
  • Communication with end-users: Researchers need to effectively communicate their findings to end-users such as land managers, policymakers, and the public, which can be challenging to achieve in a timely and effective way.

Rebekah Zehnder, TFS Geospatial Services Manager and competition judge says “It’s exciting to see the enthusiasm students have for applying their knowledge and skills to solve real-world problems.  I hope this is part of the beginning of a long relationship between TAMIDS and Texas A&M Forest Service”. Dr. Nick Duffield, Director of the Texas A&M Institute of Data Science, Organizer and Judge for the competition states: “The TAMIDS Data Science Competition encourages participants to define and solve a problem within the board focus area of the competition setting. I was very impressed this year with how the team proposed distinct and interesting problems, drew on varied methods from data analytics, machine learning, and optimization, and developed software tools to visualize and explore the data and their analysis results.”

The Winning Undergraduate Teams

Left to right: Anjana Mittal, Catherine Duong, Kara Leyendecker, Sophia Zhong, and Jaya Sundarrajan

First Place ($2,000): Data Minds: Catherine Duong, Statistics, College of Arts & Science, Kara Leyendecker, Bioinformatics, College of Arts & Science, Anjana Mittal, Statistics, College of Arts & Science, Jaya Sundarrajan, Computer Science & Engineering, College of Engineering, Sophia Zhong, College of Arts and Science. 

“Wildfires have significant consequences on humans, wildlife, natural resources, and the economy. Therefore, it is imperative to be able to predict and forecast the occurrence of wildfires. Our project addresses this concern by using machine learning and predictive analytics to forecast wildfire behavior using the U.S. Wildfire dataset from Kaggle. Predictive modeling and machine learning can provide valuable insights into developing efficient evacuation strategies, evaluating societal and environmental impacts, and enacting policies prioritizing public health and safety”.

Link to presentation

Left to right: Vardaan Kola, Josiah Kuncheriah, Shane Olson, and Ram Pangaluri.

Second Place ($1,000): Aggieland Data Dunkers: Vardaan Kola, Computer Science & Engineering, College of Engineering, Josian Kuncheriah, Industrial Systems Engineering, College of Engineering, Shane Olson, Nuclear Engineering, College of Engineering, Ram Pangaluri, Computer Science Engineering, College of Engineering.

“The primary goal of our team was to find trends in number of fires, acres burned, and fire location. Another main goal of our project was to see if there were any relationships between demographic features and wildfire risk. We limited our focus to the United States and made special focus on Texas. Most of the time and energy of this project was dedicated to finding good data and changing it into a usable form. This data was then transferred to a data visualization program to be analyzed and presented in a series of interactive dashboards. From these dashboards our team was able to learn new insights and we can now recommend new strategies for fire departments and the general public to better prepare and fight wildfires and even predict future fire hotspots.”

Link to presentation

Christopher Abib (right), pictured with Arash Abdollahzadeh & Nick Duffield.

Third Place ($500.00): IllFigureItOutLater: Christopher Abib, Industrial Systems Engineering, College of Engineering 

“Wildfires are widely known to be a threat to people, infrastructure, and the environment because of the massive damage they can cause. While society has developed many technologies and techniques to put out fires the real danger comes from the volatility of wildfires. Today, it still poses a challenge to predict their occurrence and efficiently communicate their presence. While prediction is a key facet towards solving this problem, this report will focus on the communication aspect of it. In order to make the most effective use of the complex predictions made, the information must be both processed and properly distributed. Many avenues exist to communicate information in 2023 to the numerous variety of end-users. In this project, a popular social media platform is used to deliver daily updates about local wildfire incidents to end-users.”

Link to presentation

The Winning Graduate Teams

Left to right: Chih-Shen ChengXukai Zhang, and Bahareh Alizadeh.

First Place ($2,000): Infernotix: Bahareh Allizadeh, Construction Science,School of Architecture, Chih-Shen Cheng, Civil Engineering, College of Engineering,  Xukai Zhang, Civil Engineering, College of Engineering.

Wildfires pose significant threats to human life, property, and the environment, with their frequency and severity being exacerbated by climate change. Timely detection and accurate prediction of wildfire location and propagation are crucial for enabling rapid response and re- covery efforts. This study aims to develop a novel approach for the timely prediction of wildfire location and propagation using artificial intelligence (AI) and pixel-based Support Vector Machine (SVM). A convolutional neural network (CNN) is employed to identify fire locations and intensities from remote sensing imagery through segmentation. Additionally, a pixel-based SVM is proposed to estimate fire propagation in a time period based on the current state of the fire and predicted environmental data, such as wind speed and direction, precipitation, and temperature. The proposed approaches hold considerable potential to enhance the efficiency and accuracy of wildfire detection and propagation prediction. Ultimately, the developed method has the potential to facilitate prompt responses to ongoing wildfires, thereby safeguarding human lives and mitigating financial losses.”

Link to presentation

Left to right: Weishan Bai, Jiaxin Du, Chunwu Zhu, and Tianchen Huang.

Second Place ($1,000): UrbanDS: Weishan Bai, Jiaxin Du, Tianchen Huang, Chunwu Zhu, Urban & Regional Science, School of Architecture. 

“The aim of this research is to investigate the non-linear relationship and spatial heterogeneity of the natural and built environment factors to wildfire duration in Texas using interpretable machine learning and geographically weighted random forest. Our results find that grids with longer wildfire duration are significantly concentrated on the certain area in northern, eastern, and southwestern Texas. The random forest model results show that proportion of forest land, proportion of grass land and distance to the seashore are among the highest influential natural environment factors, distance to weather station and distance to dispatch office are two of the most influential built environment factors. The partial dependent plots reveal the non-linear relationship between natural and built environment factors to wildfire duration. Geographically weighted random forest model and local variance importance plot shows the spatial heterogeneity of the effect of natural and built environment factors. For example, the proportion of agricultural land shows larger importance in the wildfire-concentrated grid in northern Texas but is less important in eastern and southwestern Texas.

Link to presentation

Left to right: Prathik Vijaykumar, Shubham Jain, and Sambandh Dhal.

Third Place ($500.00): Samstrikers: Krishna Chaitanua Gadepally, Electrical & Computer Engineering, College of Engineering, Samband Bhusan Dhal, Computer Engineering, College of Engineering,Shubham Jain, Water Management & Hydrological Science, College of Arts and Science,  Prathik Vijaykumar, Computer Engineering, College of Engineering. 

“In this study, the burn area of large wildfires across the contiguous United States from 2011 to 2020 was analyzed using a predictive Deep Neural Network model with climatological and geological features as model covariates. We found that land cover and the location of wildfire occurrence were the most influential factors in determining the severity and extent of wildfires in the United States. Using the predictive model, the extent of future wildfires can be predicted based on climate and land cover changes. Overall, our data science project provides important insights into wildfires and the need for a multidisciplinary approach to tackle this issue. Our findings highlight the importance of data collection, data modeling, and on-ground action as key areas to work on in order to develop an effective strategy to mitigate the impact of forest fires. We believe that to scale this project on a practical level, it is essential to improve the quantity and quality of data by collecting data on climate, drought index, land cover, forest cover, and topography using both in-situ and remote methods at higher resolution.”

Link to presentation

Many thanks to the other finalist teams for their hard work and participation: 

  • Blazeabaters: Allison Moore, Statistics, College of Arts & Sciences, Luis Loo, Electrical & Computer Engineering, College of Engineering, Manuel Sojan, Statistics, College of Arts & Sciences, Manuel Sojan, Statistics, College of Arts & Sciences, Sreekar Annaluru, Industrial & Systems Engineering, College of Engineering
  • Datanators: Arpan Pal, Mathematics, College of Arts & Sciences, Benjamin Warren, Mathematics, College of Arts & Sciences, Mansi Bezbaruah, Mathematics, College of Arts & Sciences, Rohit Dube, Industrial & Systems Engineering, College of Engineering, Tushar Pandey, Mathematics, College of Arts & Sciences.

Special Team Prizes and Midpoint Prizes

  • Best Presentation Design ($500): Data Minds 
  • Best Use of Additional Data ($500 each team): Infernotix & UrbanDS
  • Best Supplementary Materials ($500 each team): Aggieland Data Dunkers & Samstrikers
  • Midpoint Event Winning Teams ($250 each team): Infernotix, Team Nirvana & Samstrikers

Judges and Organization

Left to right: Marina Romanyuk, Toryn Schafer, Rebekah Zehnder & Jian Tao

The panel of judges comprised Nick Duffield, Director, Texas A&M Institute of Data Science and Royce E Wisenbaker Professor I, Department of Electrical and Computer Engineering, Marina Romanyuk ‘19, Data Scientist, Chevron, Toryn Schafer, Assistant Professor, Department of Statistics, Venkatesh (Venky) Shankar, Professor of Marketing and Director of Research at the Center for Retailing Studies, Mays Business School. Jian Tao, Assistant Professor,  School of Performance, Visualization and Fine Arts, and Assistant Director for Project Development at the Texas A&M Institute of Data Science, Tianbao Yang, Associate Professor, Department of Computer Science and Engineering, Rebekah Zehnder, Geospatial Services Manager, Texas Forest Service, Zhe Zhang, Assistant Professor, Department of Geography. Ann McNamara, Associate Dean for Research & Creative works and Professor, School of Performance, Visualization & Fine Arts at Texas A&M University, was the judge for the one-page graphic mid-point competition. Detailed biographies of the organizers, judges and speakers are available here.

Industry Sponsor Presentation from Chevron

Arash Abdollahzadeh ’20, Machine Learning Engineering at Chevron and graduate of the Texas A&M BS in Electrical and Computer Engineering program gave an informative presentation on Data Science careers at Chevron, which generated many questions from the student audience. The presentation is available here.

Marina Romanyk ‘20, Data Scientist, Chevron and competition judge said that: “As a TAMU alum from the Department of Statistics, it is always a pleasure to return to Aggieland and see what phenomenal things students are up to. It is a particular treat to come and witness such a high caliber of budding data scientists showcase their hard work. I am looking forward to seeing what these bright students accomplish in their careers and beyond.”


TAMIDS gratefully acknowledges support for the 2023 Data Science Competition: Chevron, The Texas A&M School of Performance, Visualization & Fine Arts, and the Texas A&M Departments of Electrical and Computer Engineering and Statistics.