GeoNDC: An Orders-of-Magnitude Compressed Neural Data Cube for Planetary-scale Earth Observation
Authors: Jianbo Qi, Mengyao Li, Baogui Jiang, Yidan Chen, Qiao Wang
Beijing Normal University - Advanced Interdisciplinary Institute of Satellite Applications
We present GeoNDC, a neural earth data cube that compresses 168 GB of global MODIS imagery (2003-2023) into just 0.44 GB while preserving R² > 0.85 reconstruction accuracy. By learning implicit neural representations of planetary-scale vegetation dynamics, GeoNDC enables instant queries of any location on Earth without decompression. Our approach introduces phenological inpainting for cloud removal and computes ecological derivatives through automatic differentiation, achieving what we call "planetary calculus." Extensive experiments on global MODIS and HiGLASS LAI/FPAR datasets demonstrate that GeoNDC achieves compression ratios up to 380:1 while maintaining R² = 0.997 for vegetation indices.
@misc{qi2025geondc,
title={GeoNDC: An Orders-of-Magnitude Compressed Neural Data Cube for Planetary-scale Earth Observation},
author={Qi, Jianbo and Li, Mengyao and Jiang, Baogui and Chen, Yidan and Wang, Qiao},
year={2025},
institution={Beijing Normal University}
}If you use GeoNDC in your research, please cite our paper.
Compression ratio comparison between original MODIS data and GeoNDC compressed model
Cloud removal results showing implicit inpainting of observation gaps
Global NDVI velocity field computed through automatic differentiation
LAI and FPAR accuracy scatter plots against ground measurements