\citep{Foster_Martinez_2018}\citep*{Houser_2010}

2. Wave Measurement Deployments

2.1 Terminal 4 - Wake Waves 

2.2 Giant Marsh - Wind Waves / Wave Attenuation Across an Oyster Reef

The time-lapse camera system consisted of GoPro Hero 5 black mounted in a plastic case with an integrated solar panel, 12v lithium battery, and intervalometer. The system was purchased from CamDo Solutions in summer 2019, and included their BlinkX intervalometer unit. The camera was set to take pictures at a rate of 60 frames per second (fps) for bursts of 10 minutes at the beginning of every hour, during daylight hours. The camera was mounted to a 2-inch steel conduit pipe on the mudflat, at an elevation of roughly 10 feet NAVD88. This is about 3 feet above the highest annual tide event, and about 8 feet above the local mudflat elevation.
The camera was positioned to face offshore, with wave sensor #x in the center of its field of view. An marked orange and white survey rod (markings every 1 foot) was attached to the pole with sensor #x so that camera pixels could be converted to feet during post-processing.
Images were post-processed using the Image Processing Toolbox in Matlab (c) v2016b. Figure X gives an overview of the process for identifying the instantaneous visible rod length during each photograph:
1. raw images were clipped to a subset around the pole and converted to grayscale (to reduce file size). 
2. The grayscale images were converted to a raster of gradient intensity weights, an array of scalars indicating the local intensity of the gradient in grayscale intensity for each pixel. 
3. The gradient intensity image was segmented into a binary image using a Fast-Marching-Method algorithm in Matlab (Soulthier 1996). This process uses a user-specified seed area of the image (e.g. a section of the image containing a piece of the rod) and a user-specified sensitivity threshold, to segment the image into foreground objects (i.e. the rod) and background images (i.e. the water). This produces a binary raster where the background is populated by zeros and the foreground object is populated by ones.
4. Smaller objects mistakenly identified as 'foreground' were removed by setting a cutoff threshold for the number of pixels that amount to a 'foreground' object. Usually these small objects were wave crests whose color was similar to the pole, or the reflection of the pole itself in the water.
The largest challenge with this procedure was setting a user threshold for the FMM method that cut differentiate the pole from its reflection during periods of small waves. 

3. Results and Discussions

3.1 Terminal 4 - Wake Waves    

3.2 Giant Marsh - Wind Waves / Wave Attenuation Across an Oyster Reef