EcoViz — Emissions Data Visualization & Modelling

Interactive dashboards for exploring greenhouse gas emissions, plus a regression benchmark that predicts emissions from historical data. Combines gridded US methane inventory data (NetCDF) with global per-country CO₂ emissions, and includes a Gemini-backed assistant for asking questions about the data on screen.

Scripts

File	What it does
carbon.py	Streamlit CO₂ dashboard — multi-country selection, year-range slider, Plotly charts
emissionprototype.py	Streamlit methane dashboard over the gridded NetCDF inventory + Gemini Q&A
methaneemission2012-28.py	Dash app comparing methane across 2012–2018
newshit.py	Full pipeline — methane/CO₂ dashboard plus the ML model suite and Gemini Q&A
MLMODELSCO2GRAND.py	Regression benchmark across ~8 model families
idkwidoing.py	Same benchmark without the Streamlit layer (script form)

Models benchmarked

Linear Regression · Random Forest · SVR · K-Nearest Neighbors · XGBoost · LightGBM · CatBoost · a Keras dense network, combined through a StackingRegressor. Scored with RMSE and R² under cross-validation.

Methane sectors

Label	NetCDF variable
Mobile Combustion	emi_ch4_1A_Combustion_Mobile
Stationary Combustion	emi_ch4_1A_Combustion_Stationary
Natural Gas Production	emi_ch4_1B2b_Natural_Gas_Production
Enteric Fermentation	emi_ch4_3A_Enteric_Fermentation
Municipal Landfills	emi_ch4_5A1_Landfills_MSW

Running it

pip install -r requirement.txt
export GEMINI_API_KEY="your-key"
streamlit run carbon.py

Get a key from Google AI Studio; see .env.example. The key is read from the environment — do not hardcode it.

requirement.txt covers the dashboards only. The ML scripts additionally need scikit-learn scipy seaborn xgboost lightgbm catboost tensorflow.

Stack

Python · Streamlit · Dash · Plotly · netCDF4 · pandas · scikit-learn · XGBoost · LightGBM · CatBoost · TensorFlow · Google Gemini API

Before you run

Several scripts contain hardcoded absolute paths from the original development machine — /Users/admin/Downloads/gridded GHGI/, /Users/admin/Desktop/datasetsNASAPROGRAMS/.... Point base_path / file_paths / the read_csv calls at your local copies first. The .nc methane files are not bundled here; the CO₂ CSVs are.