| license: cc-by-4.0 | |
| task_categories: | |
| - time-series-forecasting | |
| - tabular-regression | |
| tags: | |
| - oceanography | |
| - subsurface-temperature | |
| - argo-validation | |
| - glorys | |
| - physics-constrained-ml | |
| size_categories: | |
| - 10GB<n<100GB | |
| # ADRISHTA: 3D Subsurface Ocean AI Dataset (North Indian Ocean 2024–2026) | |
| This repository contains the complete open scientific assets for the ADRISHTA Subsurface Ocean AI system: | |
| - **Master Zarr Store** (`zarr/oceanembed_multiyear_2024_2026.zarr`): Harmonized 7-channel satellite surface inputs + 15 standard INCOIS target depths at 0.25° resolution across 2024-01-07 to 2026-10-03 (148 daily/weekly snapshots). | |
| - **In-Situ ARGO Ground Truth** (`argo/`): 134 real physical CTD float soundings from the Coriolis GDAC / INCOIS array. | |
| - **Model Checkpoints** (`checkpoints/`): Frozen Physics-Constrained and Baseline OceanEmbedNet weights. | |
| - **Raw Multi-Parameter Copernicus Data** (`raw/`): 3D potential temperature, salinity, currents, and sea surface height NetCDFs. | |