Uncertainty Quantification for Deep Learning-Based Soil Moisture Forecasting Using Conformal Prediction

This repository contains the code used in the paper:
"Uncertainty Quantification for Deep Learning-Based Soil Moisture Forecasting Using Conformal Prediction."

The code implements conformal prediction with spatial, temporal, and spatio-temporal localization to quantify uncertainty in soil moisture forecasting under exchangeability violations.

Environment Setup (Conda)

Create and activate the environment:

conda env create -f environment.yml
conda activate my-env

Data Source

The soil moisture data were obtained from the International Soil Moisture Network (ISMN). ISMN integrates ground-based observations from multiple monitoring networks worldwide, ensuring high-quality and standardized soil moisture data.

Reproducibility Note

Results may vary slightly when running the notebook in different environments or software setups (e.g., due to differences in library versions or hardware). However, the overall findings remain consistent: STCQR achieves coverage closer to the nominal level while maintaining sharper (narrower) prediction intervals.