Research map / Probabilistic and scientific modeling / Molecules and proteins

ml4xcube: Machine Learning Toolkits for Earth System Data Cubes

AAAI 2025 · Probabilistic and scientific modeling › Molecules and proteins · approach: Statistical analysis of ML practice · 1 citations (OpenAlex)

Read the paperCheck your idea against 108,967 papers

AI-extracted summary. Written by AI from the paper's full text. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Introduces ml4xcube, an open-source modular Python toolkit that brings the full ML pipeline (exploration, preparation, processing, modeling, evaluation) to Earth System Data Cubes, demonstrated via three land surface temperature use cases.

Problem

Machine learning is underused on Earth System Data Cubes (ESDCs) because existing tools lack end-to-end support, handle ESS-specific issues poorly (data gaps, spatio-temporal autocorrelation), and have high entry barriers for domain experts.

Approach

A modular, xarray/xcube-based framework covering the whole ESDC ML workflow, with domain-specific tools such as SVR-based gap filling, block train/test splitting to counter spatio-temporal autocorrelation, and distributed multi-GPU training.

Key results

• Demonstrates the toolkit through three use cases: linear regression for LST prediction, a CNN time-series model for LST, and distributed multi-GPU training of that CNN; no quantitative performance numbers are reported.

Limitations

• Does not guarantee performance, fairness, correctness or safety of models built with it
• Inherits ESDC limitations such as equal treatment of spatial, temporal and nominal dimensions and remapping to a common grid
• Still requires considerable ML and data science expertise from users

Experimental setup

Datasets: DeepESDL Earth System Data Cube (land surface temperature, air temperature 2m, soil moisture, radiation)

Metrics: metrics from sklearn/PyTorch/TensorFlow (unspecified)

Baselines: XCast nd framework scores

Builds on: xcube xarray dask scikit-learn PyTorch

Closest papers by meaning

• SolarCube: An Integrative Benchmark Dataset Harnessing Satellite and In-situ Observations for Large-scale Solar Energy Forecasting (NeurIPS 2024 · similarity 0.84)
• Terra: A Multimodal Spatio-Temporal Dataset Spanning the Earth (NeurIPS 2024 · similarity 0.83)
• DroughtSet: Understanding Drought Through Spatial-Temporal Learning (AAAI 2025 · similarity 0.83)
• Common Task Framework For a Critical Evaluation of Scientific Machine Learning Algorithms (NeurIPS 2025 · similarity 0.83)
• No Location Left Behind: Measuring and Improving the Fairness of Implicit Representations for Earth Data (ICLR 2025 · similarity 0.82)
• Identifying Spatio-Temporal Drivers of Extreme Events (NeurIPS 2024 · similarity 0.81)