Data Formats¶

Processing Levels¶

Data progresses through multiple processing levels from raw model output to atlas-ready summaries:

Data Level	Description	Format	Public Access
`00_raw`	Original model outputs (FVCOM / WW3+SWAN)	NetCDF	No
`a1_std`	Standardized with consistent naming and metadata	NetCDF	No
`b1_vap`	Value-added products with derived variables	HDF5	Yes
`b4_vap_summary_parquet`	Summary statistics for analytics	Parquet	Yes
`b5_vap_atlas_summary_parquet`	Marine Energy Atlas visualization data	Parquet	Yes

HDF5 Structure¶

Public datasets are distributed as HDF5 files (.h5). Each file contains:

time_index - Timestamps as ISO 8601 strings
meta - Structured array with latitude, longitude, and spatial metadata per grid point
Variable datasets - 2D arrays with shape (time, location), stored as scaled integers with a scale_factor attribute

Example file structure:

Cook_Inlet_2005.h5
├── time_index          (8760,)    # hourly timestamps for 1 year
├── meta                (392002,)  # grid point metadata
├── sea_water_speed     (8760, 392002)  # current speed at each timestep
├── sea_water_power_density  (8760, 392002)
└── ...

To retrieve physical values, divide by the scale factor:

physical_value = dataset[...] / dataset.attrs['scale_factor']

Parquet Format¶

Summary statistics are available as Parquet files for efficient analytical queries. Parquet files are partitioned by location and contain one row per grid point with columns for each summary variable.

File Naming Conventions¶

Tidal Files¶

{Location}_{Year}.h5

Examples: Cook_Inlet_2005.h5, Puget_Sound_2015.h5

Wave Files¶

{Region}_wave_{Year}.h5

Examples: West_Coast_wave_2010.h5, Atlantic_wave_1995.h5

Data Dimensions¶

Tidal¶

Dimension	Description
time	Hourly or half-hourly timestamps over 1 year
element	Unstructured triangular grid cell centers
depth	10 sigma layers (terrain-following vertical coordinate)

Wave¶

Dimension	Description
time	3-hourly timestamps over 32 years
location	Unstructured grid points (200m–10km spacing)