Water Surface Detection for CZMIL

Various calculations and settings depend on the water surface elevation in a dataset, such as determining the water surface window. When processing data, an algorithm is used to detect and classify the water shots in a dataset, identifying them as the water surface.

• Flight lines are read from the CWF file and divided into segments consisting of 50,000 shots each.

• Each segment goes through waveform processing (peak detection) where all shots are georeferenced and classified.

• The segment is then used to generate a min/max coverage of the shots, based on northing/easting positions of all shots in the segment.

• All shots in the segment are assigned to the appropriate grid cell based on northing/easting positions.

• Each segment is processed by deep learning in order to determine the land/water classification of all 50,000 shots.

The water surface detection algorithm builds a localized mean water level elevation in each grid cell based on the water shots which are assigned to each cell. In order for this to be done, the mean water level elevation must be present on at least 1 channel in the cell, (IR/Deep or Shallows).

A separate mean water level elevation (MWLE) is generated based on the IR channel, the deep channel or possibly the shallow channels, if the Use Shallow Channels for Water Surface Detection parameter is enabled, in which case, the IR/Deep channels are ignored. The application of the algorithm for generating the MWLE for each channel is explained below.

IR Channel Mean Water Level Elevation

To calculate the mean water level elevation (MWLE) for the IR channel, the application first completes the following calculations using the water shots in the IR channel:

• For all valid water shots in the grid cell, store the first return elevation from the IR channel

• Compute the percentage of valid IR shots to the total number of shots in the grid cell

• Compute the percentage of valid IR water shot elevations to the total number of valid IR shots in the grid cell

• Compute the percentage of valid IR water shots that have multiple returns to the total number of valid IR water shot elevations in the grid cell

Next, the above calculations will be used to determine if all of the criteria listed below is met:

• The total number of valid IR shot elevations is greater than 2

• The percentage of valid IR water shots to the total number of valid IR shots in the grid cell is greater than or equal to the percentage defined for the Valid Percentage of Shots in the Water Grid Cell parameter. The default value is 75%.

• All valid IR shots are water shots

• Less than 5% of valid IR water shots have multiple returns

For Nova data, the standard deviation cannot be too high, otherwise there is too much variance in the valid IR shot elevations for the mean to be considered reliable.

If all of the conditions are met, the MWLE will be computed from the valid IR water shot elevations and set for the current cell, making it a qualified grid cell for the IR channel.

Deep Channel Mean Water Level Elevation

To calculate the mean water level elevation (MWLE) for the Deep channel, the application first completes the following calculations using the water shots in the Deep channel:

• For all valid water shots in the grid cell, store the first return elevation from the Deep channel if the number of returns is 2 or less

• Compute the percentage of valid Deep shots to the total number of shots in the grid cell

• Compute the percentage of valid Deep water shot elevations to the total number of valid Deep shots in the grid cell

Next, the above calculations will be used to determine if all of the criteria listed below is met:

• The total number of valid Deep shot elevations is greater than 2

• The percentage of valid Deep water shot elevations to the total number of valid Deep shots in the grid cell is greater than or equal to the percentage defined for the Valid Percentage of Shots in the Water Grid Cell parameter. The default value is 75%.

• All valid Deep shots are water shots

For Nova data, the standard deviation cannot be too high, otherwise there is too much variance in the valid Deep shot elevations for the mean to be considered reliable.

If all of the conditions are met, the MWLE will be computed from the valid Deep water shot elevations and set for the current cell, making it a qualified grid cell for the Deep channel.

Shallow Channels Mean Water Level Elevation

The mean water level elevation (MWLE) for the Shallow channels is mutually exclusive of the elevation values computed from the IR or Deep channels. To calculate the MWLE for the Shallow channels, the application first completes the following calculations using the water shots in the Deep channel:

• For all valid water shots in the grid cell, store the first return elevations from all shallow channels if the number of returns is 2 or less

• Compute the percentage of valid shallow returns to the total number of shallow returns in the grid cell.

• Compute the percentage of valid shallow water shot elevations to the total number of valid shallow returns in the grid cell

Next, the above calculations will be used to determine if all of the criteria listed below is met:

• The total number of valid shallow elevations is greater than 2

• The percentage of valid Shallow return elevations to the total number of valid Shallow returns in the grid cell is greater than or equal to the percentage defined for the Valid Percentage of Shots in the Water Grid Cell parameter. The default value is 75%.

• All valid shallow return elevations are water shots

• The standard deviation is low enough to be considered reliable. If the deviation is too high, there is too much variance in the valid Shallow elevations for the mean to be considered reliable.

If all of the conditions are met, the MWLE will be computed from the valid Shallow water return elevations and set for the current cell, making it a qualified grid cell for the Shallow channels.

The settings in the algorithm are biased towards open water and assume that there will be strong IR/Deep coverage throughout. It also assumes that the quality of the data in the IR/Deep channels is high.

In cases where a grid cell cannot be qualified on the IR or Deep channels, the Water Grid Cell Search Radius lidar parameter can be used to look for neighbouring qualified grid cells to use instead. In open water that has very little variation in elevations, this is sufficient, however, in an environment with variable elevations in close proximity, using a neighbouring qualified grid cell may result in a wrong mean water level elevation being used in the local grid cell.

In situations along a coast, it is highly probable that a grid cell will contain both land and water shots. As such, the above mean water surface detection logic may not be able to generate a mean water level elevation in the local grid cell and will always have to rely on a neighbouring qualified grid cell. In these cases, a water surface control file can be used to explicitly define the water surface along the coastline.

A water surface control file may also be required for areas of shallow water. In these cases, the algorithm may be too restrictive in determining the water surface because there may not be enough areas that the data is 100% water or it may have unreliable IR/Deep coverage. Using neighbouring grid cells may not work if one cannot be found in range, and if the neighbouring grid cell is not at the correct elevation.

The following list identifies the priority order that will be used by the algorithm when selecting data to calculate the water surface:

• Water surface control file (if supplied and has water surface elevation for a particular shot)