A-Train Cloud Segmentation Dataset for Multi-Angle Satellite Analysis

The A-Train Cloud Segmentation Dataset is designed to train deep learning models for volumetric cloud segmentation. It provides spatiotemporally aligned patches combining multi-angle polarimetry from the PARASOL mission's POLDER sensor with vertical cloud profiles from CloudSat's radar. The dataset is produced by the National Aeronautics and Space Administration.

Use Cases

• Training deep learning models for 3D cloud segmentation from satellite imagery.
• Analyzing cloud vertical structure using aligned radar profiles from CloudSat.
• Studying cloud properties via multi-angle polarimetry data from the POLDER sensor.
• Developing algorithms for spatiotemporal analysis of atmospheric data.

Strengths

• Data is spatiotemporally aligned from two distinct satellite instruments (POLDER and CloudSat CPR).
• Specifically designed for training volumetric segmentation models, indicating a curated machine learning focus.
• Produced by the authoritative National Aeronautics and Space Administration.

Limitations

• Critical metadata is missing: no column names, row counts, or dataset size are provided on any platform.
• License information is inconsistent, listed as 'other-license-specified' on one platform and null on another.

Provenance

Source

National Aeronautics and Space Administration

Collection Method

Spatiotemporally aligned patches from POLDER sensor on PARASOL mission and 2B-CLDCLASS product from CloudSat CPR.

Time Range

null

Freshness

2026-03-12 21:01:59.292787

Geography

null