File size: 1,834 Bytes
1cd194f
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
---
license: cc-by-4.0
pretty_name: ASR CYGNSS-SMAP H-GMM reproducibility data
task_categories:
- tabular-regression
tags:
- cygnss
- smap
- gnss-r
- ocean-wind
- remote-sensing
- gaussian-mixture-model
---

# ASR CYGNSS-SMAP H-GMM reproducibility repository

This public dataset repository stores **derived CYGNSS-SMAP collocations, model
checkpoints, evaluation statistics, and manuscript figures** for the manuscript
on year-adaptive CYGNSS sea-surface wind retrieval.

## Upstream public data

- CYGNSS Level-2 Science Data Record v3.2 (`CYGNSS_L2_V3.2`), NASA PO.DAAC,
  DOI: 10.5067/CYGNS-L2X32.
- JPL SMAP Level-2B CAP Sea Surface Salinity and extreme-wind product v5.0
  (`SMAP_JPL_L2B_SSS_CAP_V5`), NASA PO.DAAC, DOI: 10.5067/SMP50-2TOCS.

The upstream raw HDF5/netCDF granules are **not mirrored** here. They are
retrieved temporarily in Google Colab, transformed into quality-controlled
collocation shards, and deleted after the derived shard has been uploaded.

## Repository structure

- `data/collocations/`: monthly Parquet shards for 2021-2025.
- `training/`: balanced historical training sample.
- `models/`: hard-gated H-GMM and chronological calibrators.
- `predictions/`: 2025 raw and strict out-of-time test predictions.
- `metrics/`: JSON/CSV statistical outputs and bootstrap intervals.
- `figures/`: vector PDF and 600-dpi PNG manuscript figures.
- `manifests/`: provenance, checksums, split definitions, and processing logs.
- `code/`: exact Colab notebook and support package.

## Important limitations

SMAP is treated as an external reference product rather than error-free ground
truth. The strict primary protocol uses the earliest 5% of 2025 observations as
chronological anchors and evaluates all later observations without anchor/test
overlap.

Repository: `wuff-mann/ASR-CYGNSS-HGMM-Reproducibility`