Data Quality Assurance (DQA) For Multibeam and Hydrographic Equipment

What do we mean by DQA? It is the planned process used to monitor and verify that data acquisition meets expectations established by measurement and performance standards. DQA techniques are NOT calibration -- rather they are tools to help determine if the survey data being acquired are likely to be valid representations of the physical environment.

It can happen that a hydrographic instrument appears to be running properly, but, due to a variety of reasons, is not. DQA offers a standardized method of checking the operation of survey equipment. These techniques are very important in minimizing, in a timely fashion, the amount of spurious data obtained, thereby saving the hydrographer the time and expense of having to repeat field work in order to meet the survey requirements.

HSTP has developed DQA techniques for: (1) direct measuring, sound velocity instrumentation, (2) absolute pressure instrumentation, and (3) shallow water multibeam sonars. The first two methods rely on comparison with readily available and low cost external references; the third relies on generally valid internal consistency.

The Digibar Sound Velocity Probe, manufactured by Odom Hydrographic Systems, Inc., is a manually operated instrument that measures sound velocity directly. It is a relatively low-cost device that is well-suited for operating in shallow water.

The DQA technique consists of measuring sound speed in a sample of fresh (potable) water with the Digibar and measuring the temperature of the fresh water with a calibrated thermometer with one-degree Celsius graduations. The DQA test is passed if the sound speed value measured by the Digibar is within acceptance limits. An excerpt of these limits is shown in Figure 1.

Figure 1. Sound speed limits for several fresh water temperatures.

These limits were computed using the National Institute of Standards and Technology (NIST) table of fresh water sound speed vs temperature, 5-28 degrees C.

For example, if the thermometer reads 15 degrees C, then the test is passed if the Digibar reading is between 1461 and 1471 m/s. The upper and lower bounds at a each temperature were obtained by evaluating the NIST sound speed at temperatures 1.5 degrees C above and below the specific temperature. Based on the graduations of the thermometer and its stated accuracy, 1.5 degrees represents three standard deviations of temperature measurement errors.

In the last seven years, this technique has detected three instances where the Digibar was functioning but was giving incorrect measurements. Figure 2 shows the DQA results for several Digibars.

Figure 2. DQA results for several Digibars.

The Digibar sound speed is plotted vs the temperature recorded from the thermometer. The two bands shown are the DQA acceptance bounds. Note that two of the measurements are well outside the bounds.

The Office of Coast Survey employs divers as one means to investigate probable hazards to surface navigation and to document the position and least depth of such items. The divers are equipped with a portable, precision, absolute pressure gauge.

The DQA procedure for these gauges involves two on-deck observations:

(a) The on-deck pressure reading of the gauge in psia.

(b) The ship's micro barometer reading in millibars.

These two values are input to a computer program. The program uses the calibration data for the particular gauge to convert the gauge reading to a more accurate value and then records the corrected diver gauge and the barometric pressures in a cumulative log file on the computer. The ships are requested to perform the DQA procedure on a daily basis to keep adding to the log file. Also, the two pressure readings are taken before and after each dive and entered into the program.

The program displays all the pressure pairs to-date on the computer screen (Figure 3).

Shown are the barometer readings vs the on-deck gauge readings. The central line shows the exact linear relationship between pressure in psia vs. millibars (1 bar = 14.504 psia). The two other (dashed) lines represent the tolerance bounds for passing the DQA test. They are obtained by shifting the central line 0.3 psia horizontally to the right and left. The 0.3 psia quantity represents 3 standard deviations of pressure measurement error based on the range and stated accuracy of the pressure gauge. The most recent measurement is denoted by the large circle. This graphic output for a particular gauge not only indicates an out-of-bound point, but also, any tendency to "drift."

The multibeam sonar data used for this analysis were obtained from a NOAA survey performed by Science Applications International Corp. (SAIC) in the East River and Long Island Sound, N.Y. Two RESON SeaBat-9001 transducers (each with 60 beams) were configured into a 9002 model with each sonar head rotated 30 degrees from the vertical to create a total included survey angle of 150 degrees.

The ocean bottom is assumed to be uniform and flat-faceted. For each port and starboard swath, we consider the depth differences of adjacent beams - - i.e., depth differences for beams 1-2, 2-3, 3-4, ..., 59-60. These differences are averaged over six-minute intervals (coincident with tide measurements). For each six-minute window, the standard deviation of the depth difference for each pair of adjacent beams is computed. Figure 4 shows how this mean standard deviation of depth differences increases as the off-nadir angle increases.

The data are for a six-minute interval. Shown are the original data (in red) and a second-order fit (in black). Area C is relatively shoal and uniform.

The present International Hydrographic Organization (IHO) depth standard for hydrographic surveys requires that for measured depths <= 30 meters, 90% of the depth errors must be <= 30 cm. Using as reference a normal distribution, the corresponding IHO standard deviation of the depth measurements is 30/1.645 = 18.2 cm.

Since we are considering depth differences, and the law of propagation of variances states that the variance of the difference of two random variables is the sum of the individual variances, the variance of the depth differences satisfying the IHO requirement is

(18.2)2 + (18.2)2 = (18.2 2)2 ÷ (25.7)2 cm.

We approximate this total IHO variance as the sum of two squared contributions:

(18.2 2)2 = (Random Error)2 + (System Bias)2

where

1. Random error is associated with slant range measurements,

2. System bias is associated with incorrect refraction adjustments in the conversion from measured slant ranges to depths.

In this analysis, an estimate of the system bias error is derived from the difference quantity: Depth at nadir minus Depth at adjacent beam pairs. This quantity is averaged over the six-minute time interval and fitted to a second order curve. As expected, it increases with the off-nadir angle. The system bias value is obtained from the second order term of this second order fit.

If there was no system bias, then the entire IHO variance could be allotted for the random error. However, in reality, the IHO allowed variance must be diminished by the system bias variance. The allowable IHO standard deviation is also shown in Figure 4 (in blue) as a function of the off-nadir beam. Note that the allowable IHO standard deviation at nadir is approximately 26 cm and is diminished as the angle increases due to the contribution of the system bias error.

The two curves intersect at off-nadir beam number 32 which translates to an angle of 48.75 degrees. This is the maximum angle on the port transducer that can be used with assurance. Note that this analysis constitutes a one-way test only. If the data at a given angle fail the test, they definitely fail IHO. However, the data at angles that pass the test are not guaranteed to meet IHO.

Figure 5 shows the same DQA analysis for another region, Area D, which is characterized by greater depths with many wrecks.

Here we have an example of an outlier for standard deviation at one beam pair. This point may not be an error, i.e., it might indicate a real feature, However, it does not fit into our model of a uniform, flat-faceted bottom. Our method of using curve fits allows us to disregard such points in the analysis. Also ignored are beams where there are an insufficient number of samples over the six-minute time window. The curves in figure 5 intersect at off-nadir beam number 35 which is equivalent to a starboard angle of 53.25 degrees.

The entire survey area consists of 4 regions, areas, A, B, C, D. Figure 6 shows the total allowable included survey angle (port + starboard) determined by the DQA analysis as a function of position.

Figure 6. Total allowable survey angle vs position

The mainly blue region is area C (relatively shoal, uniform). Area D (deep, many wrecks) shown on the right, has both red and blue. Areas A and B in the East River, on the left, are characterized by highly variable depths and this accounts for the mixed results.

The analysis of the data set demonstrates that maximum allowable included survey angle varies inversely with depth. In the shoaler, uniform area C, the allowable included angle is larger than in the deeper or highly variable areas. Such information is useful as a survey planning tool to determine the spacing of track lines. It could also be used to justify NOT investing in a more expensive sonar system with a larger survey angle because, in certain non-uniform or deep areas, the data from the outer beams would likely fail the DQA test.

Revised Thursday October 25 2001by OCS Webmaster