RCAC receives NASA grant to elucidate the effects of wildfires on water systems
Purdue University’s Rosen Center for Advanced Computing (RCAC) is part of a major NASA grant awarded to develop a new Cyberinfrastructure (CI) tool for post-fire water management and decision-making. This tool, named HydroFlame, will allow researchers to predict the amount of damage wildfires will cause to local watersheds, giving government officials the data they need to protect freshwater supplies.
Lan Zhao, Senior Research 
Scientist, and I Luk Kim, Senior Computational Scientist, both of RCAC, are co-investigators on the HydroFlame project. The Principal Investigator (PI) is Adnan Rajib, an assistant professor in the Department of Civil Engineering at the University of Texas at Arlington (UTA). Other team members include individuals from the U.S. Geological Survey and several government and non-government agencies across the western United States. NASA awarded the group $824,020 over three years to develop and refine HydroFlame. Utilizing NASA satellites, HydroFlame will integrate emerging remote sensing data on droughts and fires with watershed models, providing a one-stop platform that enables end-users to analyze, visualize, and predict the effects of wildfires on water systems. Zhao and Kim will lead the development of the new CI platform.
“Satellite data on fire and water are complex and challenging to interpret,” says Zhao. “Students, researchers, and practitioners in fire-affected communities often do not have the expertise to overcome this barrier. We aim to develop a powerful tool that simplifies this process, making it easier to utilize satellite data in managing and mitigating fire hazards.”
This project comes in response to the ever-increasing threat of wildfires on our ecosystem, which includes the sullying of freshwater sources and drinking water. Currently, wildfires cause the country to lose between $400 and $900 billion each year in economic costs. This number is expected to rise as the severity and frequency of wildfires increases.
“This is a project that will make an impact nationwide,” says Rajib. “Wildfires and other climate-charged disasters are becoming more destructive and more frequent. No community should be thinking that it’s not going to happen to them.”
Rajib, who is also the director of UTA’s H2I Lab, continues, “Wildfires dramatically alter the natural flow of water and contaminate freshwater sources with debris from burned areas. In severely affected regions, it can take years for water sources to recover. There is currently no easy way to predict such cascading hazards of wildfires, meaning communities are in the dark when protecting freshwater resources.”
This issue is precisely what HydroFlame aims to address. The platform will use a process-based hydrologic model to simulate daily streamflow and water quality variations across large river networks. A built-in data-discovery tool will continuously track potential wildfires from satellite remote sensing. When a fire is detected, the model will map burn areas within watersheds, calculating changes in vegetation, evapotranspiration, and soil moisture—sourced from multiple satellite datasets and adjusted for varying burn severity. These fire-induced alterations will then be dynamically integrated into model simulations. Finally, the model will be paired with a Machine Learning tool to predict post-fire streamflow and water quality in numerous unmonitored streams and rivers, potentially in near real-time and up to 14 days in advance.
"As we embark on developing the HydroFlame platform,” says Kim, “we are leveraging cutting-edge technologies to ensure both performance and scalability. For the frontend, Nuxt.js will provide a fast and flexible development process, enabling a smooth user experience. FastAPI will serve as our backend REST API framework, seamlessly connecting the frontend to robust Python-based data processing and machine learning simulation code. To ensure the platform can handle large-scale computation, HydroFlame will be deployed on Anvil, a national advanced computing resource at Purdue University, ensuring maximum efficiency and speed for scientific workflows."
Significant features of the HydroFlame platform include:
- Open Science Geospatial Capabilities— HydroFlame will provide open access to its data, models, and tools, enabling researchers and practitioners to collaborate, innovate, and advance the science of fire and water systems.
- Convergence of Diverse Data Sources and Types— HydroFlame will integrate diverse data types, sources, formats, and resolutions—pulled from satellite remote sensors and geospatial databases—with models of varying complexity to provide a holistic understanding of fire-hydrology interactions.
- Powerful Data Visualization and Analytics— HydroFlame will offer a suite of data visualization and analysis tools to help researchers and practitioners explore, analyze, and interpret complex fire and water data, fostering data-driven decision-making.
Once completed, HydroFlame will be available via a web application. Users will be able to access and utilize data surrounding three different scenarios: historical fire scenarios, potential fire scenarios, and near real-time fire predictions. After selecting which region on the map they are interested in, users will then be able to view any of the three scenarios. For prediction analyses, users can submit simulation jobs, which will be sent (on the back end) to high-performance computing (HPC) resources provided by Purdue. The simulations will use machine learning algorithms, a sub-sect of artificial intelligence, to supply post-fire streamflow and water quality predictions. The benefit of developing the HydroFlame platform as an easy-to-use web application is that users will not need any prior knowledge of HPC, coding, or simulation model development. This puts the relevant research and data at the fingertips of the decision-makers who need it.
On top of being easy to use, HydroFlame will also adhere to the ‘FAIR Guiding Principles for scientific data management and stewardship.’ FAIR stands for Findability, Accessibility, Interoperability, and Reusability, and these data guidelines ensure that scientists can find and utilize the most relevant data for their research.
The HydroFlame platform will be deployed entirely on Purdue’s HPC resources, namely the Geddes composable system and Anvil supercomputer. RCAC manages and operates all centrally maintained research computing resources at Purdue. To learn more about the benefits of HPC, please visit our “Why HPC” page.
The Anvil supercomputer is funded by the NSF. For more information regarding Anvil, please visit Purdue’s Anvil website.
Written by: Jonathan Poole, poole43@purdue.edu