Apppublic

obichimav/lake-salt-risk

sourceHugging Facemitupdated 2d agoView on Hugging Face

0likes

App README

🧂 Lake Salt Risk

A Gradio app that maps which lakes are getting saltier from road salt and trains a random forest to predict salinization risk. Predictions update in real time as you move the sliders.

How to use

Tab	What you do	What you get
1 · Data	Upload two CSVs (measurements + lake characteristics)	Data loaded and columns auto-detected
2 · Trends	Click Calculate trends	Map of rising/stable/falling lakes, sorted table, CSV for QGIS
3 · Train model	Pick target & features, click Train random forest	Accuracy scores, feature importance, risk map, downloadable predictions
4 · Live prediction	Move sliders or start from a real lake	Real-time prediction — try "what if impervious surface doubled?"

Dataset

Dugan et al. (2017) — 529 lakes, North America & Europe, 1940–2016. Download from EDI Data Portal (free Google / ORCID login required). Paper: doi:10.1038/sdata.2017.101

Run locally

bash

pip install -r requirements.txt
python app.py

Keep basemap.json in the same folder as app.py — it holds the map outlines so maps draw without internet.

Model library

After training, name and save up to 6 models in the built-in SQLite library. A model can only overwrite an existing entry of the same name if its accuracy is strictly higher. When the library is full, the lowest-scoring model is evicted to make room.

Data citation

Dugan, H., S. Bartlett, S. Burke, J. Doubek, F. Krivak-Tetley, N. Skaff, and J. Summers. 2017. Global Lake Ecological Observatory Network: Long term chloride concentration from 529 lakes and reservoirs around North America and Europe: 1940-2016. Environmental Data Initiative. https://doi.org/10.6073/pasta/c560d99864cd7fd303153a11d2ba0ddd