Social Media to predict fires

Recent updates in disaster management frameworks and regulations have emphasized the significance of people-centered data. WMO's call for Impact-based forecasting encourages the exploration and analysis of additional signals before and after disasters.  This project aims to utilize the most people-centered data of all: Social Media messages. 

The objectives are: 

1. Enhancing Prediction: While factors like rising temperature and humidity serve as reliable indicators for the onset of the fire season, various events such as large camping or musical festivals, political unrest in fire-prone areas, or frequent electricity issues in buildings can all signify an increased likelihood of anthropogenic fires.  The research questions revolve around identifying previous fire causes, discerning patterns from social media data, and assessing the enhancement in disaster modeling accuracy. 

2. Analyzing Post-fire Effects: Beyond the immediate loss of life and property, wildfires have indirect or "less severe" consequences that still need to be addressed. Some examples here could include post-fire health issues (e.g., respiratory problems), requests for shelter or food assistance, water quality degradation due to runoff, and intangible impacts on local businesses. Evaluating these non-direct losses could serve as the foundation for impact-based forecasting and appropriate insurance policies. 

GeoTagging AI will support students with data collection and topic detection/classification aspects of the project. They are keen on collaborating to identify community pain points, relevant conversations, the social media platforms where these discussions are most active, and running experiments on the effectiveness of taking those new signals into consideration in disaster modeling.