SBIR Phase I: User-generated real time qualitative data processing for climate impacted model validation, integration, and augmentation

The broader impact of this SBIR Phase I project is the development of an integrated methodology to use resident’s experiences about flood (and other climate change) events to validate modeling in real time, inform policy, and provide design insights for infrastructure development. It provides an integrated solution for capturing the valuable information captured by people’s direct experiences (photos, stories, and data) with climate change that are otherwise underutilized.

The team will develop a community knowledge platform that can process a mix of text and photo data submitted by residents and process it into formats usable for understanding on-the-ground impacts, flood occurrence and severity, and deliver that data to planners and modelers developing ways to better manage floods. Data processing occurs behind the scenes and allows residents to engage with the planning processes impacting their communities in new ways, increase the access of underrepresented communities, and improve equity in decision-making.

The project will stimulate research in data sciences, generate new types of jobs in civic data systems, and improve the efficiency of public infrastructure investments. User’s cell phones will become powerful local data collection tools allowing a direct line of communications and building trust between government decision makers, scientists, and residents.

Advances in data science allows the analysis of heterogeneous qualitative and image data to incorporate user generated posts into large scale infrastructure planning around climate resilience. Currently, descriptive data and photos submitted by users are manually analyzed for content. Through novel use of natural language processing (NLP), spatial data analysis, artificial intelligent (AI) and computer vision of flood event photos, and development of an application programming interface (API) to curate data for hydrological model developers, this project automates the process of extracting the full value of community generated posts of flood events.

When successful, hyperlocal user generated posts will be processed in real time to deliver detailed on-the-ground data on flood events to planners, for model validation, and community members themselves. The product builds innovative technologies to permit processing at scale so that any community experiencing flood events can generate real time flood data and monitor the impact of infrastructure as hydrological baselines continue to shift.

The project develops new machine learning NLP to automate the analysis of qualitative text data, keyword detection for sentiment analysis and impact, AI to extract flood characteristics from photos, and API for protecting model IP while allowing integration with external data for validation purposes.

This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.