It uses multi-source remote sensing data (Optical, Radar, Topography, and Climate) representing the year 2022. The model is designed to handle diverse landscapes, including forests, agricultural land, water bodies, and urban areas.

Region: Northeast India (89.4°E - 97.6°E).
Time Period: 2022 (Annual Median).
Input Data: Sentinel-2, Sentinel-1, SRTM Terrain, CHIRPS Rainfall, ERA5 Temperature.
Training Samples: ~10,000 points (Random stratified sampling).

Intended Use

Regional Soil Health Monitoring: rapid assessment of SOC without lab tests.
Carbon Baseline Studies: Establishing 2022 baselines for carbon projects.
Land Use Analysis: Distinguishing between high-carbon forests and low-carbon degraded lands.

Feature Inputs (16 Columns)

To use this model, your input dataframe must contain these 16 columns in this exact order:

Category	Features
Sentinel-2 (Optical)	`S2_B2`, `S2_B3`, `S2_B4`, `S2_B8`, `S2_B11`, `S2_NDVI`
Sentinel-1 (Radar)	`S1_VV`, `S1_VH`
Terrain	`elevation`, `slope`, `aspect`
Climate	`precip_annual`, `temp_mean`
Soil/Location	`soil_texture`, `latitude`, `longitude`

How to Use

python

import xgboost as xgb
from huggingface_hub import hf_hub_download
import pandas as pd

# 1. Download Model
model_path = hf_hub_download(
    repo_id="mona0125/soc-estimation-ne-india", 
    filename="soc_estimation_model_ne_india.json"
)

# 2. Load Model
model = xgb.XGBRegressor()
model.load_model(model_path)

# 3. Predict (Example Data)
# Ensure columns match the Feature Inputs list above!
data = pd.DataFrame({
    'S2_B2': [0.03], 'S2_B3': [0.05], 'S2_B4': [0.04], 'S2_B8': [0.25], 'S2_B11': [0.15], 'S2_NDVI': [0.72],
    'S1_VV': [-8.5], 'S1_VH': [-14.2],
    'elevation': [150], 'slope': [5.5], 'aspect': [120],
    'precip_annual': [1800], 'temp_mean': [24.5],
    'soil_texture': [2], 'latitude': [26.1], 'longitude': [91.7]
})

prediction = model.predict(data)
print(f"Predicted SOC: {prediction[0]:.2f} g/kg")