Skip to content

Ho1yShif/cgi_flood_prediction_mitigation

Repository files navigation

CGI Flood Prediction and Mitigation

This repository stores code for a CGI Externship Project completed in the Fall 2022 semester via NJBDA in collaboration with the Rutgers MBS Exchange Program

Overarching Goal: Deliver a strategy to predict and mitigate flood effects in vulnerable NJ areas in order to help business and families suffering from the effects of flood events and storms such as Hurricane Ida

Data Sources: USGS, NJ Department of Environmental Protection, NJGIN, USDA, FEMA, NJ.gov

Tools Used: Python, Jupyter, ArcGIS, Time Series Modeling

3-Pronged Approach

Water Level Prediction

  • Goal: Model Gauge (water level) in order to predict future floods, which are defined by the USGS as events with water levels at or above 8.5 ft
  • The best-performing model was XGBoost with a R-squared of 99% and a mean absolute error of 0.02 ft
  • Other models we explored include LSTM, SARIMA, and Rolling Window Regression
  • Caveat: XGBoost is likely overfitting due to correlated features; this is addressed in the next steps we provided in our slide deck

Flood Mapping

  • Goal: Build flood susceptibility map to view the most vulnerable area in North Trenton
  • Assunpink Creek, which runs through North Trenton, is a main flood source
  • The deep red area on our map is prone to flooding due to the proximity to rivers
  • Every year, water levels tend to increase from December through May

Flood Financial Analysis

  • Goal:
  • Point

Takeaways and Next Steps

Water Level Prediction:

  • Tune XGBoost model and fix correlated features
  • Analyze feature importance to simplify future models
  • Leverage exponential smoothing models
  • Explore other locations with accessible datasets

Flood Susceptibility Map

  • Build county or city-level flood map
  • Explore more FEMA flood datasets
  • Select the city/county with highest flood susceptibility
  • Add interactive map features

Flood Financial Analysis

  • Explore FEMA claims dataset at the county and city level
  • Study flood damage by industry
  • Risk management strategies for buildings in susceptible areas
  • Explore how floods affect financial and qualitative aspects of a business

Team Members

  • Jinglin Zhao (Team Lead)
  • Shifra Isaacs
  • Alvin Radoncic
  • Priscilla Ramos
  • Topu Saha

About

Repository for NJBDA CGI project, Fall 2022

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published