Synthetic Multi-Fidelity Climate Modeling Dataset for Physics-Informed Deep Learning (SMCD-PIDL)

This dataset was generated for the study entitled “A Scalable Hybrid Physics-Guided Multi-Fidelity Learning Framework for High-Accuracy and Uncertainty-Aware Climate Anomaly Prediction under Sparse and Heterogeneous Data Conditions.” The dataset represents synthetic climate and environmental observations designed to simulate sparse, heterogeneous, and multi-fidelity data conditions for climate anomaly prediction. It includes atmospheric, environmental, spatial, temporal, and uncertainty-related variables used for training, validation, and performance evaluation of the proposed physics-guided learning framework. The dataset was created to support experimental analysis where real-world climate observations may be incomplete, unevenly distributed, or affected by uncertainty. It can be used to evaluate climate anomaly classification, uncertainty-aware prediction, and hybrid physics-guided machine learning models.