Hydrographic Survey Management Guidelines

4 Error management and equipment calibration

The management of uncertainty consists of several steps and elements. The main uncertainty management steps are:

• Pre-analysis (survey design), where an estimate of the error contributions from random error sources is made in order to determine whether the equipment proposed for the survey will allow the resultant survey to meet or exceed specifications (refer to “CUBE Bathymetric data Processing and Analysis (CHS February 2012)”),
• control of errors through calibration and good survey practices,
• evaluation, where the survey results are analysed to determine whether surveys specifications have been met and
• documentation, where individual objects, such as soundings, navigational aids, etc., are given uncertainty attributes, and/or the metadata elements of uncertainty for the whole survey project are documented so that potential users may determine if the data set is fit for their particular purpose. For example, TPU contributors such as the SBET RMS error data file, the tidal uncertainty, Vessel configuration file, etc.

The main uncertainty elements are:

• Systematic errors or biases, which are due to system offsets that have not been properly accounted for by proper survey procedure such as system calibration,
• random errors, which are due to noise in the measurement process used,
• blunders or accidental errors, which might be due to system failures, environmental conditions and human errors that are beyond our ability to control.

4.1 Systematic Errors

Systematic errors can be reduced to negligible levels through careful calibration procedures prior to each survey. In addition, a verification that systematic errors or biases are mitigated should be conducted regularly (daily in some cases) and whenever changes are made to hardware, software or firmware as the survey progresses.

All equipment to be used during the field survey must be field-tested and, if need be, calibrated prior to leaving for the field survey site. No matter how well an instrument may have been adjusted when it left the manufacturer or the maintenance shop, there is always the possibility of damage to equipment during shipping due to vibration, shock or temperature change. Some equipment may also require field testing on site prior to survey operations. Users should refer to the various owners’ manuals and to the QMS hydrographic survey procedures to ensure that the instruments are properly tested and functioning according to established specifications.

As a minimum, the following should be done:

• GPS receivers are to be checked over known control monuments
• All distance-measuring equipment shall be checked against a calibration baseline. The baseline should be of a higher accuracy than the manufacturer’s specified accuracy. Additionally a zero constant value for the instrument should be established.
• Barometers and psychrometers must be calibrated.
• Range finders should be checked against known distances.
• Theodolites shall be checked for errors due to poor horizontal collimation, parallax, vertical index error and for other instrumental imperfections. All moving parts shall be checked to ensure proper operation.
• Levels should be peg-tested and rods be calibrated by comparison with a standard.
• The water level gauges and staffs to be used during the survey should be checked.
• Echo sounders should be checked under field conditions, preferably on the work platforms that they are installed on.
• Sound speed profilers must be compared on an annual basis and problematic probes must be calibrated (consult manufacturer for recommended re-calibration interval).
• Bar checks, poles and lead lines should be measured to ensure the distance markers are correctly fastened.

All computers, computer programs, and networks should be checked to ensure their working order in a remote field environment.

The various software packages used to acquire, process and analyze the various bathymetric data collected should be checked to ensure that they work properly and that the approved updates (service packs, hotfixes) and drivers are being used. The latest release versions should be used at the discretion of the Project Manager. Backups for these various software packages should be available and stored in a secure place in case of need.

All instruments will have to be re-verified and if necessary re-calibrated after being repaired or after changes have been made (electronic card replacements, computer adjustments, etc.). A record of repairs and changes shall be maintained.

4.1.1 Calibration of sounding systems

The echo sounding systems will require an initial calibration before the sounding platform is to be operational in the field.

4.1.1.1 Multi-transducer and Multibeam systems

Multi-transducer and multibeam systems are complex echo sounding systems and need careful adjustment and calibration prior to leaving for the field.

Alignment and offset parameters (distances) between the various sensors of the multi-transducer or multibeam system and the reference point of the vessel must be physically measured and input into the system’s computer. These measurements are usually done after the system has been installed or when some of the equipment has been replaced or relocated.

The following should be done at the beginning of the field season and before leaving for the field or whenever a piece of equipment is replaced or repaired:

• attitude sensor testing and verification,
• bar check to measure transducer draft,
• transducer or boom alignment,
• gyro alignment determined using accurate land-survey techniques or, other methods such as a patch test,
• squat and settlement tests, and
• check and calibrate the time delays between the positioning system, the recorded soundings and the attitude sensor readings.
• Initiate a patch calibration procedure

A patch test should be done to quantify any residual biases in the initial alignment and offsets. This test consists of running reciprocal lines at various speeds, depths and bottom terrain and comparing the results obtained. Lines run perpendicular to smooth sloping bottom (preferably capturing the apex) will be used to check data latency and pitch. Reciprocal lines run along a smooth flat bottom will verify roll. Lines run on both sides of a defined feature such as a wreck, can determine gyro offset.

The data obtained should be carefully analyzed and adjustments made accordingly. The results of this test should be documented, recorded and made available upon request. Note that this test can also be done in the work area.

More information on patch test calibration can be found in “The Calibration of Shallow Water Multibeam Echo-Sounding Systems” by André Godin, 1997.

Once all adjustments and calibrations have been performed, the system should be used to run a series of parallel and perpendicular sounding lines over a reference bottom surface where the depths have been previously determined and verified with an independent system. The results obtained should compare favourably and be within the accuracy requirements of the survey order as specified in Table 1 of the CHS Standards for Hydrographic surveys.

Tests should also be run to ascertain that the system can in fact detect targets of the minimum size required by the specified order of the survey.

4.1.1.2 Single beam

Single beam echo sounders installed in the sounding platforms that will be used during the actual survey should be given a trial run to ensure that the echo sounders and the rest of the equipment perform according to their specifications. Bar checking the system to measure draft and verify speed of sound is necessary. Where higher order results are required, system checks such as squat and settlement tests, gyro or attitude sensor alignment and sensor offsets are necessary.

4.2 Random Errors

Random errors (equipment noise), unlike systematic errors, cannot be eliminated. However, it is possible to estimate the affects that random errors will have on the outcome of the survey (pre-analysis), to mitigate or reduce their effects (survey design and instrument selection) and to evaluate the quality of the survey, both as the survey progresses (real-time quality assurance – RTQA) and at the conclusion of the survey. A post-survey assessment allows attribution of each object with uncertainty values and a global assessment of the survey quality for the purposes of metadata creation. This in turn provides users of the data set the information they need to determine if it is fit for their purpose.

Pre-analysis can be done, e.g. using an EXCEL spreadsheet. Values for each random error contribution can be assigned at 1 sigma (standard deviation, RMS, CEP, etc.) and propagated as the square root of the sum of squares (the variances). This is known as propagation of random variances. For single beam echo sounders this is quite simple (e.g. see Professional Paper No. 25 from UKHO). For multibeam echo sounders, it’s a bit more complex (e.g. see R. Hare, A. Godin and L. Mayer, “Accuracy Estimation of Canadian Swath (Multi-beam) and Sweep (Multi-Transducer) Sounding Systems”, 1995).

Several navigation and logging packages have built-in real-time assessment tools that look at standard deviation of the bathymetric surface. Combined Uncertainty and Bathymetry Estimator (CUBE) is an example of a real-time assessment tool implemented in acquisition and processing packages (see § 4.3.1). As new data are added, the standard deviation of the bathymetric surface can be reduced because of the increased number of measurements. Consider that the standard deviation of the mean is always smaller than the standard deviation of a single measurement.

4.2.1 Total Propagated Uncertainty (TPU)

As stated in the CHS Standards for Hydrographic Surveys, the accuracy of the reduced depths must be determined. In determining this accuracy all sources of individual errors need to be quantified in order to determine the Total Propagated Uncertainty (TPU). The following are some of the sources of errors that must be considered:

• static draft setting,
• dynamic draft, including settlement and squat,
• speed of sound changes,
• instrumental errors,
• roll, pitch and heave,
• navigation timing,
• water level readings or GPS heights,
• tide or separation models.

Position and depth precision estimates are calculated separately as the square-root of the sum of the variances (root-sum-square (RSS)) from all error sources. This approach assumes therefore that all errors are uncorrelated, unbiased and follow a normal (Gaussian) distribution. In the instances where such an assumption is not valid, an estimate of the error introduced into the estimation process will be given.

As there are some error sources which are not well understood, initial estimates will be based on some elementary testing of a limited number of data sets. In order to validate these estimates, independent ground-truth (validation) data sets will be used as a method of post-calibrating the error estimation process.

In order to estimate the effects of temporal change on bathymetry, the approach of Velberg [1993] is followed. This approach requires that the error due to the dynamic nature of the seafloor is known or can be estimated.

In order to form an error budget, estimates of all independent random errors sources, at the same level of confidence, are required. More information on TPE and the computation process can be found in the document “Estimation of Bathymetric Accuracy Attributes and their Implementation in the Source Data Base, 1996, R. M. Hare”.

4.3 Blunders

The last category of errors is blunders, or accidental errors. Unlike systematic errors, these cannot be eliminated by careful survey procedures. Unlike random errors, they cannot be estimated in advance of the survey and they cannot be reduced by making additional measurements. Blunders can, however, be detected and removed if sufficient redundancy exists in the survey network and there is sufficient precision in the survey measurements. In control networks, the term minimally detectable blunder (MDB) is often used. For bathymetry, very dense data sets (e.g. high-resolution multibeam) make it possible to detect blunders where they do not fit (statistically) with the values of their neighbours. There are numerous statistical approaches to blunder detection in multibeam data.

For single-beam echo sounder surveys, the lack of data density means that detecting small blunders is much less likely. The reliability of multibeam surveys is much higher due to the detection capability. This is the reason such surveys are recommended in harbours, approaches and critical areas.

4.3.1 CUBE (Combined Uncertainty and Bathymetry Estimator)

CUBE uses soundings and their associated uncertainty estimates (see § 4.2) as input and through spatial and uncertainty weighting, while also relying on the very high data density of multibeam data sets, outputs a bathymetry gridded surface and its associated uncertainty (error) surface. In addition, it tracks the statistical hypotheses for each depth point, and where there is more than one estimate, makes an attempt to determine which the most likely value is. This makes it a very powerful tool for identifying and removing blunders in the data. Once these have been removed from the data, CUBE is rerun to generate the final bathymetry and uncertainty surfaces. See “CUBE Bathymetric data Processing and Analysis (CHS February 2012)”. The uncertainty surface is a quantification of the survey quality, which can be compared against specifications and used as input to the metadata for the survey.