Experimenting with predicting street flooding during the 2017 Subsurface Hackathon in Houston
Switch branches/tags
Nothing to show
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Permalink
Failed to load latest commit information.
data
README.md
create_features.py
create_training_data.py
extract_vertices.py
features_and_class.csv
merge_street_flood_data.py
prediction.geojson
prediction_features.geojson
train_model.py

README.md

Houston_street_flooding

Experimenting with predicting street flooding during the 2017 Subsurface Hackathon in Houston

The basic idea

  • Hydrologic models and historical discharge predict flooding in floodplains very well
  • Hydrologic models do not predict local street flooding and ponding
    • Clogged storm drains, etc
  • Detailed topography can help predict localized flooding
    • e.g. Is the area very flat?
    • Are we in a local low point (even one that drains)?

So, let's train a classifier on features extracted from LIDAR-based DEMs

  • Local slope, relief, etc in a moving window.

What do we train it on?

Train it on data from all of Houston, predict on a small area (time constraints)

Presentation

https://docs.google.com/presentation/d/1djExE4mQNk-_jCY1pz77aR7MoX_Mmvv3-FiGXXmeTbU/edit?usp=sharing

Prediction for small area

http://bl.ocks.org/d/d0ca5528514c43265517f36440d990f5