CircuitNotion/agri-ghg-yield-gb

Agricultural Yield & GHG Emissions Model

Multi-output Gradient Boosting regressor that predicts:

Target	Unit	Test R²
Crop Yield	kg/ha	0.946
N₂O Emission	kg N₂O-N/ha	0.693
CO₂ Emission	kg CO₂-C/ha	0.922
CH₄ Emission	kg CH₄/ha	0.604

Average R²: 0.791

Features (16)

Soil: pH, N, P, K, Ca, OM, CEC, SM, SS
Weather: T, Rainfall, RH, Rad, WS, WD
Management: Fert

Files

• trained_model.pkl — MultiOutputRegressor(GradientBoostingRegressor)
• feature_scaler.pkl — RobustScaler fitted on training features

Quick start

import pickle
import pandas as pd

with open("trained_model.pkl", "rb") as f:
    model = pickle.load(f)
with open("feature_scaler.pkl", "rb") as f:
    scaler = pickle.load(f)

features = ["pH","N","P","K","Ca","OM","CEC","SM","SS",
            "T","Rainfall","RH","Rad","WS","WD","Fert"]

# example row
X = pd.DataFrame([[6.5,40,20,150,800,3.0,15,25,2.0,
                   22,100,60,18,2.0,180,100]], columns=features)
pred = model.predict(scaler.transform(X))
# pred columns: Crop_Yield, N2O_Emission, CO2_Emission, CH4_Emission
print(pred)

Or download from the Hub:

from huggingface_hub import hf_hub_download
import pickle

model_path = hf_hub_download("CircuitNotion/agri-ghg-yield-gb", "trained_model.pkl")
scaler_path = hf_hub_download("CircuitNotion/agri-ghg-yield-gb", "feature_scaler.pkl")

Training data

Trained on CircuitNotion/agri-ghg-yield-synthetic (6,000 synthetic agronomic samples).

Code

Full pipeline: ntirushwajeanmarc/agriculture-ghg-yield-ml