SPAR native-support validation and boundary tests v2.3.0
Data and reproducibility package for Support-Preserving Forecasting of EGMS Level-3 Displacement Time Series. Current-manuscript reproduction starts with the GitHub <a href="https://github.com/WendongYao/Wendong_2025_Spatiotemporal_InSAR_dataset_prediction/blob/main/experiments_ex
Data and reproducibility package for Support-Preserving Forecasting of EGMS Level-3 Displacement Time Series.
Current-manuscript reproduction starts with the GitHub experiments_ext/README.md guide. The archived experiments/gridding-first and CNN-LSTM workflows are retained for provenance only and are not the reproduction path for the current manuscript.
Version v2.3.0 adds the frozen uniform all-cell SPAR configuration; pre-specified E32N34 partitions 47-50; persistence, DLinear, LASSO, and causal-TCN comparisons; three same-origin within-tile replications; four shortened-history temporal counter-tests; ten analytic matched-interpolation controls; sampler and anchor ablations; direct predictions; source, configuration, and output hashes; and regenerated paper figures.
The primary endpoint is a fixed single-horizon, 12-day-ahead forecast at a valid delivered product cell, with the intervening product value withheld. The release does not claim independent geodetic truth, cross-region transfer, temporal universality, or Europe-wide superiority.
Frozen GitHub commit: 9865ac6c107804abd127f32c4ad76140b497a52d. Deterministic archive SHA256: 6b0138c4c00cb79a1259dadfe01b8d2e34399bcf36abff7ace8ce1027cbc79fc.
Transformer, STGCN, graph-model, and RSASE experiments are outside this release.
Redistributed EGMS content should be attributed as European Union, Copernicus Land Monitoring Service information; https://egms.land.copernicus.eu/.
📤 Share this page
Found this useful? Share it with your network.
Files are hosted on the source repository. Click download to access the full dataset.