US Wave by Month

publication date

MarineCadastre.gov Data Registry

Source data by URL

The hindcast data processing consisted of the following main steps:

1. Read, filter, and calculate individual years annual/monthly statistics from annual data files and save

a. Using the NREL Developer API the hindcast data output was read directly from the AWS repository.

b. Each regional annual file was read individually, filtering the data from a 3-hour time step to a daily time step (every 8th time step). This was required for efficient processing speed and reasonable storage needs.

c. Of the ten wave data variables available only 5 were used in this data processing:

i. Mean Wave Direction - Direction Normal to the wave crests (compass direction, where wave is coming from)

ii. Significant Wave Height - Wave height based on the zeroth spectral moment (i.e., H_m0)

iii. Mean Absolute Period - Resolved Spectral Moment (m_0/m_1)

iv. Peak Period - The period associated with the maximum value of the wave energy spectrum

v. Maximum Energy Direction - The direction from which the most wave energy is travelling (compass direction, where wave is coming from)

d. While reading the data bad (missing) values were removed. For the West Coast and Hawaii regions these values were represented with either -9 or -999, in the Atlantic region these were represented by nan.

e. All the daily time step values were collected and then averaged for the whole year and for each month of the year. This averaged data was then written to a local CSV file with the hindcast node ID, decimal degree coordinates, and 5 wave variable statistics.

i. The two wave direction variables were handled slightly differently (see below).

f. For the two direction variables:

i. Averaging the direction (in degrees) without vector data would not prove accurate averages. Instead, the directions were binned into 16 direction bins. Each bin covered 22.5 degrees and was represented by the degree value in the middle of that range, for example, the bin covering 78.75-101.25 was written as 90, and the range of +/-11.25 around 0 (north) was written as 0. This binning was done to support identifying the mode later when annual/monthly statistics were calculated.

ii. Instead of calculating and saving the annual and monthly average, the number of occurrences for each directional bin throughout that year was tallied for the annual and monthly timescales and was saved as a CSV.

2. Calculate annual and monthly statistics for the full 32-year period

a. Once all annual and monthly files were prepared, the statistics for each wave variable were prepared to cover the full 32-year period at an annual and monthly level.

b. For the period and height variables:

i. The mean value for each of the individual years was averaged together and saved for each hindcast data point in a CSV for the annual and monthly means.

c. For the direction variables:

i. Using the frequency data by directional bin, the annual and monthly values for all 32 years were totaled for each bin. The bin with the largest frequency over the 32 years (mode) was selected and the direction for that bin was saved in a CSV for the annual and monthly means.

ii. At the same time, another set of CSV files were saved that included the totaled frequencies for each bin both on an annual and monthly level covering the full 32-year period. These statistics were used later to help generate the gridded summary products.

1. Hind Cast Points with annual statistics

a. From one of the annual CSV files, a point feature class was created using the latitude and longitude of each hindcast data point. The hindcast record id was also incorporated as a unique point id.

b. The annual statistic for each of the five variables was then incorporated as an attribute for each point.

c. Any points where all of the five statistics where NULL (mainly just points near the shoreline, that were also NULL in the hindcast data) were removed.

2. Tables of Monthly statistics for each variable

a. Each of the monthly statistic CSV files were converted to a geodatabase table.

b. The table consisted of the Point ID (that matches the point in the annual points feature class) and then twelve fields with the statistics for each month.

c. Any table records where the statistics were all NULL, were removed from the table (these were the same records that were removed from the annual point dataset)

3. Summary Hexagon Grid Creation

a. A hexagon grid was created using 3 resolutions (1 km2, 10 km2, and 100 km2) using the ArcGIS Pro - Generate Tessellation geoprocessing tool, with the following settings:

i. Spatial Reference - EPSG:3857 WGS84 Web Mercator (Auxiliary Sphere) - using this ensures that the hexagons will look correct in a map service and works for a projected coordinate system for all regions.

ii. Shape Type - Hexagon

iii. Extent - same as the hindcast point extent for that region

iv. Cell size - 1 km2, 10 km2, or 100 km2

b. An attribute field was added to each of the hexagon grids to store annual wave direction mode, mean absolute period, and mean significant wave height summary statistics.

c. Wave statistics were then summarized based on all the hindcast data points that fell within each hindcast hexagon cell (for each resolution).

i. For the mean absolute period and mean significant wave height:

1. Used the ArcGIS Pro Summarize Within geoprocessing tool to calculate the mean for each hexagon cell and each variable:

a. Input - hexagon grid feature class

b. Summary Features - annual hind cast point feature class

c. Kept all input polygon features

d. Statistics - mean for both period and height variables

2. Joined this output to the original hexagon layer and populated both variable attribute fields with these results

ii. For the wave direction mode:

1. Created a temporary point feature class from the annual wave direction bin frequency data (count of occurrence in each bin over the whole 32-years).

2. Used the ArcGIS Pro Summarize Within geoprocessing tool to calculate the sum for each cell:

a. Input - hexagon grid feature class

b. Summary Features - temporary points of direction bin frequency

c. Kept all input polygon features

d. Statistics - sum for each of the 16 direction bin frequency fields

3. For each hexagon grid cell, the direction bin that had the largest frequency (sum of all point frequencies within that cell) was identified as the mode, and the direction mode was set to be that bin's mid-point direction in degrees.

4. Joined this output to the main hexagon layer and populated Wave Direction Annual Mode attribute field with these results.

Added data for the Alaska region