Upload results/alphaearth_coverage_report.txt with huggingface_hub
Browse files
results/alphaearth_coverage_report.txt
ADDED
|
@@ -0,0 +1,45 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
# AlphaEarth Embedding Integration Report
|
| 2 |
+
# Provenance:
|
| 3 |
+
# Script: /media/drn2/External/TARA-Oceans/03_analyses/WorldModelApp/scripts/t05_alphaearth_integration_20260127_120000.py
|
| 4 |
+
# Input AE: /media/drn2/External/TARA-Oceans/PythiaTIfreeLA4SR_TARA/alphaearth_embeddings_gee_pfam_20260124_175558.tsv
|
| 5 |
+
# Input Meta: /media/drn2/External/TARA-Oceans/03_analyses/WorldModelApp/data/consolidated_metadata.tsv
|
| 6 |
+
# Input IDs: /media/drn2/External/TARA-Oceans/03_analyses/WorldModelApp/data/consolidated_sample_ids.npy
|
| 7 |
+
# Date: 2026-01-27 09:17:34
|
| 8 |
+
# Integrity Check: PASSED
|
| 9 |
+
|
| 10 |
+
SUMMARY
|
| 11 |
+
=======
|
| 12 |
+
Total consolidated samples: 1810
|
| 13 |
+
AlphaEarth coverage: 995 / 1810 (55.0%)
|
| 14 |
+
Uncovered (all-NaN): 815 / 1810 (45.0%)
|
| 15 |
+
Partial NaN rows: 0 (binary coverage -- complete or empty)
|
| 16 |
+
Embedding dimensions: 64 (A00-A63)
|
| 17 |
+
|
| 18 |
+
PER-BASIN COVERAGE
|
| 19 |
+
==================
|
| 20 |
+
Arctic : 13 / 38 ( 34.2%)
|
| 21 |
+
Atlantic : 368 / 644 ( 57.1%)
|
| 22 |
+
Indian : 77 / 195 ( 39.5%)
|
| 23 |
+
Mediterranean : 242 / 242 (100.0%)
|
| 24 |
+
Pacific : 277 / 617 ( 44.9%)
|
| 25 |
+
Red_Sea : 4 / 24 ( 16.7%)
|
| 26 |
+
Southern : 14 / 50 ( 28.0%)
|
| 27 |
+
|
| 28 |
+
EMBEDDING STATISTICS (995 covered samples)
|
| 29 |
+
==========================================
|
| 30 |
+
Value range: [-0.4550, 0.4444]
|
| 31 |
+
Per-dim mean range: [-0.1589, 0.1896]
|
| 32 |
+
Per-dim std range: [0.0677, 0.1859]
|
| 33 |
+
Degenerate dims (std < 1e-8): 0
|
| 34 |
+
|
| 35 |
+
IMPLICATIONS FOR MODELING
|
| 36 |
+
========================
|
| 37 |
+
- 55% coverage means AlphaEarth can serve as primary features
|
| 38 |
+
- Majority coverage (995/1810) enables AlphaEarth as a standard feature set
|
| 39 |
+
- Fully covered basins: Mediterranean
|
| 40 |
+
|
| 41 |
+
OUTPUT FILES
|
| 42 |
+
============
|
| 43 |
+
alphaearth_aligned.npy: (1810 x 64) float64, NaN for uncovered samples
|
| 44 |
+
alphaearth_mask.npy: (1810,) boolean, True where AE available
|
| 45 |
+
|