2.1. Site Descriptions
The two camera installations are at the North Head lighthouse, Washington State, and the coast guard watch station at Coos Bay, Oregon, both in the region of large wave energy in the Pacific Northwest (
Figure 1a). The North Head Argus station is located 3.3 km north of the north jetty of the Columbia River and overlooks Benson Beach, a low sloping, dissipative, west-facing beach with open wave exposure to the Pacific Ocean (
Figure 1b). The Coos Bay Argus station is 325 km to the south and includes an estuary opening to the WNW that connects to the long estuary of Coos Bay that winds north and then turns south (
Figure 1c).
Wave heights in the PNW vary seasonally from average heights of 1.5 to 2.0 m in summer months and 2.5 to 4.0 m in winter months (e.g., [
4]). However, storms can be extreme, with significant wave heights commonly ranging from 5 to 10 m and peaking at 14 to 15 m, and these magnitudes have been shown to be increasing [
5]. Due to the long fetch of the Pacific Ocean, the largest storms often have periods of 15 to 17 s, but wave periods can at times exceed 20 s periods during winter months [
4].
For North Head, wave height data for this study were extracted from CSIRO WaveWatch III model data for node 46.0° N, 124.8° W while tide data came from NOAA gauge 9440581 for Cape Disappointment, Washington. For the Coos Bay component of this study, tide data were obtained from NOAA tide gauge 9432780, located just inside the Charleston, Oregon, harbor mouth. Wave data were obtained from CAWCR Global Wave Hindcast model for model node point 44.0° N, 125.2° W.
The north Columbia River jetty’s construction began in 1885. Due to the large viewing distances, only the north jetty is visible in the Argus data. Dredge material disposal for Columbia River sediments has been done at the “Shallow Water Site” (SWS) since 1973 in an attempt to keep sediments within the nearshore littoral system and available to the local beaches. The SWS site is located just to the north side of the navigation channel (location shown in Figure 4).
The Coos Bay jetties’ constructions began in 1890. Both jetties are visible in the Argus data, allowing measurement of wave breaking on both sides of the inlet. Coos Bay has two disposal sites: site E to the south of the offshore channel and site F to the north. Both locations are shown in Figures 6 and 12.
2.2. Argus Camera Sampling
Measurements were made from imagery collected by Argus stations at both sites. An Argus station is a set of cameras, computers, and software algorithms that have been developed and deployed since 1986 to take image data from which relevant nearshore measurements can be made [
6].
For North Head, there was a requirement to sample with sufficient resolution out to a range of 4.5 km, the distance of the SWS disposal site from the North Head lighthouse. Due to the extreme range, five cameras were required to span the 40° field of view (
Figure 1b) to allow sampling from Benson Beach on the left of the camera view out to the SWS location on the right. Each camera had a field of view of only 9.3° (50 mm lens) with a 5% overlap between adjacent images.
Figure 2 shows example snapshot images from each camera.
To make geolocated measurements from imagery requires an understanding of photogrammetry, the quantitative transformation between image, and world locations of imaged features [
7,
8]. In general, this transformation between coordinates is accomplished through multiplication by the projective matrix, P. P contains information about the camera sensor, called the intrinsic calibration, which is performed once prior to camera installation, and information about the camera location and viewing angles, called the extrinsic calibration, that must be performed at installation and any time the camera moves. While camera location can be accurately measured at installation and will not change significantly, the viewing angles must be resolved to around 0.01° to maintain sufficient accuracy at the 3–4 km ranges of interest. For multi-year camera deployments like this, typical camera mounts are only stable to a factor of 10–100 times of this requirement, so quantification of the North Head image data required development of image co-registration algorithms. For the left-most camera, this process was based on the co-registration of land-based features, but for the right-most four cameras, registration was based on water-based features like details of breaker foam that were observable in the overlap regions between cameras, allowing the passing of geometry information from a camera on the left to adjacent cameras on the right. This development was challenging and important, but the details are beyond the scope of this manuscript (details will be featured in a future publication).
With photogrammetry accurately corrected, images from the five cameras could be merged into composite images in which world locations are accurately known.
Figure 3 shows a merged panorama (pan for short) computed from the snapshot images in
Figure 2. The pan axes are vertical and horizontal angles (tilt and azimuth) and world locations such as the SWS disposal site boundary can be accurately mapped in the composite image.
While snapshots provide an instantaneous feel for the wave conditions of the day, the goal of this paper is to estimate wave breaking statistics (defined below) on the jetty and at the SWS locations. Since wave heights modulate on several minute time scales, any particular snapshot can give a misleading representation of average breaking conditions. To improve our statistical representation of breaking, we use the “brightest” images [
6]. For a “brightest” image, a series of snaps are taken at 2.0 Hz sampling frequency for ten minutes. The brightest intensity from those images for every pixel in the image is found and saved as a “brightest” image.
Figure 4 shows an example of “brightest” panorama for 16 July 2017, a day with relatively low wave conditions. White areas indicate locations where wave breaking occurred at least once during the ten-minute collection, for example, along the shoreline and over a nearshore sand bar, as well as along the sides of the jetty. There are also some streak features due to boat wakes, for example, the long trail of a small boat entering the estuary just beyond the north jetty. The image shows azimuth and tilt angles for illustration. Images are shown with a vertical exaggeration of 4:1 to make features more visible. This has the effect of making breaking on the jetty look more extreme than it is. The boundaries of the SWS disposal site are included as a black line. The “brightest” images will be used to detect wave breaking on the north jetty and over the SWS disposal site.
The North Head Argus station ran from July 2017 to October 2019, for a total of 847 days. On each day, Argus images are collected every half hour for daylight hours, yielding a total of 13,009 image sets. Of these, many were unusable due to fog, rain, darkness, or sun glitter. These were removed by manually viewing every merged pan and flagging those that would be difficult or impossible to use. After manual culling, 5923 images remained, 45.5% of the total count.
Wave breaking patterns depend on environmental conditions, mostly wave height and tide level. After viewing each of the almost 6000 pan images, it was clear that developing a robust quantification of the amount of breaking over the SWS disposal site and the north jetty was going to be challenging. Instead, it was decided to average all of the pans within each set of wave height and tide level bins. For North Head, the tide was partitioned into four elevation bins, −1 to 0, 0 to 1, 1 to 2, and 2 to 3 m. Significant wave heights, H
s, were partitioned into nine 1 m bins from 0 to 9 m. Thus, 36 averaged “brightest” pan images were computed.
Figure 5 shows a useful example of tide between 0 and 1 m elevation and H
s between 3 and 4 m. The number of individual pan images that contributed to the mean was also recorded, in this case 165. The upper limits of each image look like a double image. This is a consequence of the slow dipping of the camera tilt as the camera mounts aged over time. The fact that the jetty has a fixed location demonstrates the usefulness of the geometry correction process described above.
Most of the sampling for Coos Bay is similar to that for North Head. As seen in
Figure 1c, six cameras were used to span almost 180° of viewing, including views back into the estuary. For this paper, we will only consider seaward looking views, so cameras 1 to 3.
Figure 6 shows an example of “brightest” panorama from 20 May 2019, showing some wave breaking on both the north and south jetties as well as on the ocean beaches to the north and south. The boundaries of the two dredge disposal areas are shown offshore with dotted lines (site E is to the south or left, site F is to the north or right).
The Coos Bay site operated for a shorter time than North Head, running from 8 May to 27 November 2019, for a total of 202 days. In total, 6046 image sets were collected. Darkness, rain, and sun glitter made many of them unusable, but there were still 3122 good “brightest” pan images after culling.
A similar bin-averaging was done for Coos Bay, although tidal flow was also felt to have an important influence on wave breaking patterns and was therefore included in the bin choices. For Coos Bay, significant wave heights, Hs, were partitioned into four bins from 0 to 6 m with bin size of 1.5 m. Tide elevation was split into four bins, from −2 to +2 m with 1 m bin sizes. There were no direct estimates of tidal flow, so the finite difference of tidal elevation (dzt/dt) was used as a tidal flow proxy (t measured in days). Tidal flow was partitioned into 6 bins, from −18 (spring ebb) to +18 (spring flood) with bin sizes of 6 m/day (this is not a horizontal flow, but instead is a rate of tidal elevation change).