R2Sonic 2026, Operation Manual

Page: 171 / 254

Page 171 of 254
Version 6.3 Rev r012
Date 05-11-2022
11.8 Improving the Patch Test and Patch Test results
Section 11.7 described areas that should be addressed to improve the results of the patch
test when collecting the data. Further improvement will come with the number of data
collections and the manner in which the patch test is computed.
11.8.1 Need to collect sufficient data
Too many times, surveyors will collect just a few lines of data for each test. One of the major
issues, detailed above, is the variability of the position accuracy of DGPS. Another issue,
detailed above, is the steering of the vessel during the data collection and the relationship of
the sonar head to the feature or slope on each data collection.
To overcome the variability of the DGPS positioning and vessel steering, it follows that the
more tests that are performed, the greater will be the reliability of the test results. Below, is
an example of a multibeam calibration, which included five data collections for each test.
ROLL PITCH YAW
0.73 -0.73 1.02
0.73 -0.99 0.90
0.76 -2.16 0.81
0.76 -1.07 2.26
0.74 -0.83 0.94
Pitch mean with erroneous value = -1.16 (SD = 0.58); without erroneous value of -2.16 = -0.91 (SD = 0.13) Yaw
mean with erroneous value = 1.19 (SD = 0.61); without erroneous value of 2.26 = 0.92 (SD = 0.08)
Consider the above patch test and what the result would have been if only two collections
were made and those were the ones that contained the highlighted values, which can
clearly be seen to be outside of the trend. Having more data to work with will create a more
reliable result.
The more data collected, the more evident will be any out of trend values that may reflect a
DGPS wobble, a steering issue, or variability of the positioning of the pole. Enough data
should be collected to provide a reliable statistical result, i.e. mean and standard deviation.
Collecting enough data to compute six of each test, allows the exclusion of any one ‘out of
trend’ result to yield a mean and standard deviation derived from five computations; this
would be a statistically viable sampling.