EIVA NaviModel3. Total Vertical Uncertainty Analysis Tool. Determining the TVU of Multi- and Single-beam Surveys. Lars Dall, EIVA A/S

Transcription

1 EIVA NaviModel3 Total Vertical Uncertainty Analysis Tool Determining the TVU of Multi- and Single-beam Surveys Lars Dall, EIVA A/S

2 Contents: EIVA NaviModel3 - TVU Total Vertical Uncertainty in IHO SP-44 The Implementation of the TVU analysis tool in NM3 TVU analysis, a Case Story Summary

3 Total Vertical Uncertainty in SP-44 In Special Publication number 44 (SP-44) from IHO, the concept of Total Vertical Uncertainty (TVU) is defined as follows: The component of total propagated uncertainty (TPU) calculated in the vertical dimension. TVU is a one-dimensional quantity With TPU (Total Propagated Uncertainty) defined as: The result of uncertainty propagation, when all contributing measurement uncertainties, both random and systematic, have been included in the propagation

4 Total Vertical Uncertainty in SP-44 II SP-44 further states that: Vertical uncertainty is to be understood as the uncertainty of the reduced depths. In determining the vertical uncertainty the sources of individual uncertainties need to be quantified. All uncertainties should be combined statistically to obtain the TVU

5 SP-44 further states that: Recognizing that there are both depth independent and depth dependent errors that affect the uncertainty of the depths, the formula below is to be used to compute, at the 95% confidence level (1.96 * σ), the maximum allowable TVU. The parameters a and b together with the depth d have to be introduced into the formula in order to calculate the maximum allowable TVU for a specific depth: Where: Total Vertical Uncertainty in SP-44 III a represents that portion of the uncertainty that does not vary with depth b is a coefficient which represents that portion of the uncertainty that varies with depth d is the depth 2 a (b* (b * d) represents the portion of the uncertainty that varies with depth d) 2

6 Total Vertical Uncertainty in SP-44 IV Order Special 1a 1b 2 Description of areas Areas where under-keel clearance is critical Areas shallower than 100 m where under-keel clearance is less critical but features of concern to surface shipping may exist Areas shallower than 100 m where under-keel clearance is not considered to be an issue for the type of surface shipping expected to transit the area Areas generally deeper than 100 m where a general description of the sea floor is considered adequate Maximum allowable THU, 95% Confidence level 2 m 5 m + 5% of depth 5 m + 5% of depth 20 m + 10% of depth Maximum allowable TVU, 95% Confidence level a = 0.25 m b = a = 0.5 m b = a = 0.5 m b = a = 1.0 m b = Full Sea floor Search Required Required Not required Not required Feature Detection Cubic features > 1 m Cubic features > 2 m, in depths up to 40 m; 10% of depth beyond 40 m Not Applicable Not Applicable Recommended maximum Line Spacing Not defined as full sea floor search is required Not defined as full sea floor search is required 3 * average depth or 25 m, whichever is greater For bathymetric lidar a spot spacing of 5 * 5 m 4 * average depth Positioning of fixed aids to navigation and topography significant to navigation. (95% Confidence level) Positioning of the Coastline and topography less significant to navigation, (95% Confidence level) Mean position of floating aids to navigation, (95% Confidence level) 2 m 2 m 2 m 5 m 10 m 20 m 20 m 20 m 10 m 10 m 10 m 20 m

7 Implementation of TVU Analysis in NM3 The Total Vertical Uncertainty analysis tool in NaviModel3 has been designed to facilitate a determination of the quality of a hydrographic survey. The part of the quality that is investigated is the TVU. The hydrographic data to be tested could either be acquired by means of single-beam or by multi-beam techniques The basis of the analysis is a base model that must be superior, in terms of TVU, to the survey-spread that is to be tested. In the present context, the base model is termed the Reference Model. The survey-spread to be tested is termed the Test Survey

8 Implementation of TVU Analysis in NM3 II Bearing in mind that the Total Vertical Uncertainty analysis tool is based on comparing a surveyed line against a reference model, it is of the outmost importance to attain a good, superior reference model. This can be achieved in different ways: By utilizing a superior survey configuration (calibration, acquisition method, instrumentation etc.) By surveying the area with multiple lines and in a variety of directions By thoroughly cleaning and editing the data acquired The ideal method would be to carry out a combination of the three

9 Implementation of TVU Analysis in NM3 III In NaviModel3 this would mean that data for the reference model that has been acquired with superior accuracy and coverage, is meticulously cleaned, by utilizing the best possible combination of manual and automatic cleaning techniques. Further, in order to arrive at the best possible base for the testing, the cell size of the reference model should be adequately small ( as small as possible (with sufficient data in all cells)). With smaller cell-sizes, undesired influences from a generalization of the observed data into cells with only one attribute value, is minimized

10 Implementation of TVU Analysis in NM3 IV Data for the testing, the test survey, must be acquired with a well calibrated, yet typical survey spread. It is recommended that the test survey is not run parallel to any of the runlines associated with the reference model. In the postprocessing phase, the test data should also be cleaned, in order to remove insignificant influences from outliers and gross errors. To prepare for the testing the test survey must be exported from NaviEdit, in a special format, named Ascii XYZ,Angle,Quality, once the required editing and cleaning has been completed.

11 Implementation of TVU Analysis in NM3 V Once the reference model is loaded into NaviModel3 and the test survey has been created, the Total Vertical Uncertainty analysis can be performed as follows:

12 Implementation of TVU Analysis in NM3 VI The TVU test contains, apart from general information, three subsets: Relative to IHO (SP-44), all three orders (default is on) Relative to USACE standards (default is off) User defined order test, based on requirements similar to IHO (default is off) It is also possible to define the opening angle (of the test survey) to be investigated as well as the number of bins for visualization in the graphical histogram representation

13 Implementation of TVU Analysis in NM3 VII The TVU general data is comprised by: Statistical results from comparison between reference model and test data (mean diff., std. max, min, 95% confidence etc.) The report associated with each subset contains: IHO standards: Calculation of limit and test result for all orders (with predefined values for a and b) User Defined Test: Calculation of limit and test result, relative to user defined values for a and b USACE: statistics associated with the Engineering and Design for Hydrographic Surveying, dated April, 1, 2004 from US Army Corps of Engineers. The standards state some requirements to 95% confidence figures associated with surveys for navigation channels and dredging support (results are in US Survey feet)

14 Implementation of TVU Analysis in NM3 VIII The histogram window is divided in three: The Confidence plot shows the quality of beams as a function of the beam angle. The X- axis is the beam angle (in degrees), whereas the Y-axis depicts the associated error with the grey area being the depth of 95% of all beams. The blue line visualizes the mean error The Error probability histogram shows the error distribution of all beams within the selected angle limit. The X-axis depicts the error in meters, whereas the Y-axis and shows probability The Beam count histogram shows the probability distribution of a given number of beams in the reference model. The X-axis shows number of beams, whereas the Y-axis gives the probability in percent of any given number

15 Implementation of TVU Analysis in NM3 IX Reports associated with the Total Vertical Uncertainty analysis can be generated Also plots associated with each of the histograms can be created

16 TVU Analysis, a Case Story The case story is based on a recent project, where the main requirements (of relevance), were specified in the Scope of Works as follows: Maximum allowable TVU within 95% confidence level: a = 0.05 m, b = (expected depth of 20 m (d) equals a limit of m (95% confidence)) Minimum resolution requirement: 40 soundings/m 2 The processed and gridded bathymetry (the DTM) shall have a grid resolution of 0.2 m At least 95% of all nodes (cells) shall be populated with at least 3 soundings

17 3D positioning: POS MV 320 (primary system) & Javad Delta 3GT (secondary system) Gyro Compass: POS MV 320 Motion sensor: POS MV 320 Multi-beam Echo-sounder: Reson 7125 (with online Reson SVP 70) SoundVelocity probe: Reson SVP15 TVU Analysis, a Case Story II The survey equipment system set-up chosen to meet these extremely strict requirements, comprised the following instrumentation:

18 TVU Analysis, a Case Story III Further, in order to meet requirements: Only instruments and software solutions with the highest performance are used. The methods proposed have been chosen in order to take full advantage of the superiority of the hard- and software parts of the system. The total overall specifications of the proposed system can thus be regarded optimum with respect to commercially available software and instrumentation Survey sensors are physically mounted on a rigid pole. Variations in offsets and mount angles due to deformation of the structure can thus be neglected. Also impact on uncertainties on offsets measurements can be regarded insignificant The MBE swath width is limited to +/- 45 º, to minimize errors from un-modeled ray bending and grazing angle The system performance is based a high accuracy system calibration (calibration will not effect the overall system performance) The system performance is based on valid up to date sound velocity profiles

19 TVU Analysis, a Case Story IV In order to substantiate, on an á priori basis, that the accuracy-requirements can be met, with the proposed equipment, error-budgets have been prepared for the vertical as well as for the horizontal component. In the error-budgets, errors originating from the specifications of the instrument are random as opposed to the systematic errors originating from calibrations. These systematic values, given in the budgets are directly linked to the acceptance criteria of the calibrations A few reservations have been made in the TVU budget, shown below, the majority of these are based upon the fact, that the TPU-values are linked to a relative approach

20 TVU Analysis, a Case Story V It is assumed, that the time-delay is 0 ms (the time-tagging on the multi-beam data and on the position data is configured so that NaviPac/NaviScan and the MBE are using the same, accurate time reference The accuracy by which the on-line system is capable of tagging the sensor-data timewise is optimized. The latency value(s) can hence be set to 0 The heave value is corrected in the post-processing phase (delayed heave from POSMV), whereby contributions originating from absolute drift are minimized The multi-beam accuracy is the accuracy by which a single beam can be determined in a totally controlled environment Sound Velocity determination is related to the accuracy by which the sound velocity profile in the water column can be determined. Influences from Sound Velocity interpolation is ignored, since it is assumed, that the characteristics of the water column is (close to) identical to those determined Influences from offsets, geoid model, reference station coordinates all have a relative contribution of 0

21 TVU Analysis, a Case Story VI The 95% confidence values have been calculated according to the following formulas: Total 1.96* 1.96 Total 2 ributed Normaldist E 2 ributed Normaldist Systematic E Systematic

22 TVU Analysis, a Case Story VII The surveylines were covering the survey area (thick green line) in an East/West and in a North/South direction respectively, with a line-spacing of 15 m. In addition 2 times two extra lines cross the area from corner to corner. All survey lines were run twice, in opposite directions. This yields a 800% coverage as the basis for the reference model(s) of the TVU analysis, in accordance with the overall requirements to such a model. Further on the data and the model: The data (reference model as well as test survey-data) has been thoroughly cleaned and edited in the post-processing phase The reference model was based on approximately 65 million depth observations in an area of a little more than 200 * 200 meters. This is equivalent to more than 1000 observations per square meters. Similarly, the test survey(s) consisted of almost 1.5 million depths, in the full swath

23 TVU Analysis, a Case Story VIII The TVU analysis was performed on identical observation sets, in a variety of ways, in order to investigate the effect of employing different post-processing methods: Different cleaning methods (manual and S-can based) Different approaches to bathymetric (RTK-height) smoothing Different approaches to heave correction Post-processed versus online 3D position Different cell-size in reference model 1. Real time heave, unsmoothed bathy 2. Real time heave, smoothed bathy 3. Delayed (true) heave, unsmoothed bathy 4. Delayed (true) heave, smoothed bathy 5. Pospac merged positions with heave, unsmoothed bathy 6. Pospac merged positions with heave and smoothed bathy 7. Pospac merged position without heave For each method, two models were generated, one where data had been the subject of manual cleaning and one where S-can cleaning had been performed. Investigations regarding the consequence of changing the cell size was furthermore done with one of the model types.

24 TVU Analysis, a Case Story IX The first analysis was done on the basis of models that were generated from real time heave, with the bathy value not being the subject of smoothing. Two test surveys were investigated for both systems, one in a relatively flat area (N017) and one that included the excavation (N01). From the results below, it is clear that for the flat area investigation the results are indicators of the vertical uncertainty only, whereas the result from line N01 includes position error as well, since a large position change will involve a large depth change in the dredged areas. Observe furthermore how the S-can cleaned models have a TVU-value that is approximately 10% better than the manually cleaned models. Also observe how the results that include a 2 * 60 º opening angle are only slightly less accurate than those from a 2 * 45 º coverage.

25 TVU Analysis, a Case Story X The bathy value is arrived at (online in NaviPac/NaviScan as well as offline in NaviEdit) by subtracting the Heave value from the RTK-height, in order not to compensate twice for the high frequency movements (the heave). Normally the bathy value is then smoothed in order to remove undesired high frequency noise, with the noise thought to be originating from the GPS. As can be seen here, the accuracy is actually deteriorating after the smoothing, thus indicating that the RTK height values are OK. In the present situation, with an integrated system, the RTK height and the motion data are originating from the same source, with the same time tagging reference. In such a situation smoothing of the bathy is not improving the vertical accuracy.

26 TVU Analysis, a Case Story XI The next analyses are made on the basis of delayed heave values from the POSMV. These are generated by the POSMV with a delay of 3 minutes. The NaviPac POSMV interface that is distributing the POSMV data received to NaviPac and to NaviScan can be configured to log this information, to be subsequently be merged in NaviEdit. A recalculation of the bathy value must then take place (re. previous slide). Surprisingly, the results appear close to those associated with real-time heave. It must however be expected that in a situation with rougher weather and with accelerations perpendicular to the sailing direction, delayed heave might involve an improvement relative to the real-time heave.

27 TVU Analysis, a Case Story XII The final analyses were made based on gyro, motion and position data originating from the POSPac software. Processing was performed on the basis of raw online data and Rinex data logged at the 5 reference stations in the area. Processed data were subsequently merged into NaviEdit as new position, motion, GPS height and gyro data. Subsequently the bathy value was recalculated.

28 TVU Analysis, a Case Story XIII For the POSPac based models compared to the flat area, no improvements are seen. For data including the excavation a substantial improvement can be observed. This is most likely caused by improvements in the position on the POSPac data. Again, smoothing of the bathy is not improving the vertical accuracy and the S-can cleaning yields better results than the time-consuming manual cleaning.

29 TVU Analysis, a Case Story XIV Pospac without heave is based on modelling the vertical movements on the basis of RTK-height only. As can be seen this processing method gives the best results in all situations.

30 TVU Analysis, a Case Story XV Pospac without heave would essentially mean that the heave components should be removed from the vertical error-budget.

31 TVU Analysis, a Case Story XVI The final investigation shows the consequence of changing the cell size in the reference model. Whereas the previous comparisons were based on 10 cm grid cells, 5 and 20 cm were used for the comparisons here. The 20 cm cell models result in all situations in higher TVU values, because of the undesired influences from the generalization of the observed data that a DTM expresses. The 5 cm cell models on the other hand can be considered closer to the raw data and will thus give lower TVU results. The danger is that the models and thereby the analysis becomes statistically weak when decreasing the cell size to a value that will result in a low observation population in the cells.

32 TVU Analysis, a Case Story XVII The population in different cell-size models are visualized here. As can be seen, the 5 cm cell-size model appears to be on the limit when it comes to statistical significance for the TVU analysis. 5 cm cell-size 10 cm cell-size 20 cm cell-size

33 TVU Analysis, a Case Story XVIII

34 TVU Analysis, a Case Story XIX Beam counts can be performed relative to an entire model or relative to a boundary. The beam count data can be saved to pdf or to a csv-file for further manipulation.

35 TVU Analysis, a Case Story XX The requirement to number of beams per m 2 can be visualized in the DTM window of NM3 by using the Colourmode Density (below right). The beam count statistics can also be applied to this requirement, by using a 1 m cell-size model (below left)

36 TVU Analysis, Summary The Total Vertical Uncertainty analysis tool: Is designed for testing of fulfilment relative to requirements to TVU (IHO SP-44 etc.) The reference model must be superior to the test survey with respect to hard- and software as well as acquisition method, observation density, cleaning method etc. The cell-size of the reference model must be adequately small in order to minimize undesired influences from a generalization of the observations into a single cell value The TVU Case Story: The effect of employing different post-processing method was investigated on similar datasets. All results fulfilled the extremely strict requirements and were better than the ápriori error budgets S-Can cleaning provides significant improvements to TVU compared to manual cleaning Smoothing the RTK-based bathy value does not improve the TVU when RTK height and motion data originate from identical source, with identical time tagging reference Using delayed heave data does not give improvements to TVU relative to online heave when acquiring during calm conditions and when sailing without abrupt turns POSPac data can be used to optimize TVU by using GPS-height only (no heave) Cell size must optimized for statistical significance for the TVU analysis Other: Tools to investigate beam count, data density etc. are integrated into NaviModel3

37 Further Information EIVA Training and Documentation Site: EIVA Knowledgebase: Frequently Asked Questions: Tutorial on the TVU Analysis Tool in NaviModel3: EIVA NaviModel3, Total Vertical Uncertainty Analysis Tool (this document): Press Releases at Hydro International: Release of NaviModel3 DTM Software, February 2010: Release of NaviModel3 DTM Software, version 3.2, September 2011:

38