1. University of Southern Mississippi Canadian Hydrographic Service
Carrier Phase GPS Navigation for Hydrographic Surveys, and Seamless Vertical Datums
University of Southern Mississippi
GPS Tide Detection : Implementation of a full integrated solution for hydrographic surveys on the St-Lawrence from data collection to data processing.
Louis Maltais
Canadian Hydrographic Service
Quebec Region
Thanks
GPS Tide Detection : Implementation of a full integrated solution for hydrographic surveys from data collection to data processing.
The project I will present you today is now in operationnal mode. Last survey season was our first year were the GPS tide was the primary system for sounding reduction.

2. Introduction Geographical situation Chart Datum definition
Technique used for Hydrographic surveys on the River
New positionning capability
Seamless Datum definition and establishment
Accuracy
Data collection and processing
Opportunities
Summary
The plan for this paper is the following :
First, I will describe where this project is taking place. After, I will give you a brief definition of what is a chart datum and how we establish a chart datum. On the navigation channel we are using a special technique to know what is the tide at the vessel position. Depending were we work on the river we can be fooled by the water level very rapidly.
Next we will see the new positionning capabilities that we have now. I will present the work we have done on the river.
We will discuss accuracies, data collection and processing.
Finally show some opportunities .

3. Geographical situation
In the Laurentian region we ’ve got two types of surveys.
Channel maintenance surveys and standard hydrographic surveys.
In the channel accuracy is critical. We are talking about IHO special order surveys. Each system used in this type of survey have to be accurate and reliable. We are calibrating 3 to 4 times a year the 3 platforms that work in the channel from April to November.
The first and most important part of the project is in the channel. That part is completed.
Second part is the St-Lawrence estuary and gulf. Some measurement are done but nothing is validated.

4. Traditionnal Chart Datum
Chart datum should be so low that the water level will but seldom fall below it.
Not so low as to cause the charted depths to be unrealistically shallow.
Should vary only gradually from area to area and from chart to adjoining chart, to avoid significant discontinuities.
Represented on the shore by benchmarks.
Let ’s go back to traditionnal hydrography.
The following three criteria are what define a chart datum. A chart datum should be so low that the water level will but seldom fall below it, not so low as to cause the charted depths to be unrealistically shallow, and it should vary only gradually from area to area and from chart to adjoining chart, to avoid significant discontinuities.
In short terms, when we want to define the chart datum at a point, we place a tide gauge at this point, study the tide over a month, simulate tide over years and finally looking at the time serie we can define were the zero is. Chart datum can be modify at any time, nothing is fixed

5. Technique used now for Hydrographic Surveys in the Navigation Channel
Interpolation in space
Extrapolation in time
Survey lines
Tide Staff or Stations
Tide readers talking with the survey vessel giving tide at each 5 minutes.
On the river ,like I was saying before we use a special technique for tide detection at the vessel position. Here ’s a chart in S-57 format showing a section of the river. Let ’s suppose you want to survey in that area %%%
In that section of the river we have those tide stations%%% For that survey we could use those points%%% On shore tide readers are transmitting the water level via radio. On board we ’ve got software were the hydrographer puts in the water level. This software has the position of the tide stations and the position of the vessel updated every second and interpolates a water level at the position of the vessel. Each time a new tide value is put in the system, the trend or slope of the tide is calculated so we can extrapolate in time. Each day the survey team has to deploy on shore to read the tide. Depending of the phase of the tide it can be problematic to send people on shore. For example, try to pick up a tide reader at low tide on a beach in rough seas conditions. It ’s a logistic nightmare !
Hard to read a tide staff at the centimeter level
Need on human ressources on the field are high.

6. New positionning capabilities
Centimeter accuracy was possible in the past using post-processing methods.
Now with Real-Time Kinematic and On-the-Fly algorythms we can get accurate positionning in Real-Time.
Technology improves in GPS positionning. The satellites are still the same but the receivers are now capable of incredible accuracy. Centimeter accuracy in XYZ was possible in the past using post-processing techniques, now in real-time using high-end GPS receivers we can achieve that kind of accuracy. In this project we are working with Thales Navigation equipment. The way a real-time kinematic system works is fairly simple. The base station sits on a kn own geodetic point, broadcast via radio link measurements made at the position of the base station. The rover is collecting measurements, data broadcasted from the base station and computes in real-time the position of the rover. Once you ’ve got the data link working your main concern are now range and ionosphere. The maximum range for the equipement on the market now is about 40 kilometers between base and rover. Once you pass that limit the solutions becomes less reliable. Why bother with the ionosphere ? Because that part of the atmosphere is affecting the GPS signal coming from the satellites. OTF algorythms are still not able to cope with a disturbed ionosphere. So we have to be aware of high solar activity because it influence the ionosphere.
Drawbacks
Limited range
Ionospheric effects

7. Why do we need a Seamless Datum ?
Ellipsoid WGS84
GPS Height relative to WGS84
All thoses great new are nice, but the question now is why do we need a Seamless Datum. Here ’s the explanation. The new positionning possibilities in XYZ are done relative to the WGS-84 ellipsoid. In clear terms I mean latitude ,longitude and height relative to a model of the earth. We are already coping with latitude and longitude using map projections. But for the first time we want to utilize the GPS height value. Let ’s look at the scheme.%%% We are somewhere in Quebec( Deschaillon I think) Looking at the picture we see a tide staff, an automatic tide guage and a boat equipped with RTK GPS receiver. At that location we know the chart datum. The GPS equipment is giving us GPS height relative to WGS-84 and what we want is the tide or in others terms height above chart datum. The missing piece od the puzzle is the separation beetween wgs84 and chart datum. Once we ’ve got that value we can do that simple calculation.
Tide = GPS height - GPS antenna height over water-separation value
The scheme is not perfect but to get the clear picture just imagine the boat in line with the tide staff and the tide guage. So now what we need is a separation value. We need to get the complete coverage in order to be able to work with RTK everywhere.
Separation beetween WGS84 and Chart Datum
Antenna Height
Tide
Chart Datum
Tide = GPS Height - GPS Antenna Height over water - Separation Value

8. Work that have been done on the St-Lawrence
40 Primary Control points (Compensation by NRCAN)
20 Secondary Control points (Validated with the RTK system)
Measurements have been made relative to WGS-84 Ellipsoid
Chart datum value at each control point.
Solution to get the seamless coverage : Krigging software to interpolate between stations. (Like doing a DTM but more rigourous, the surface has to pass throught each control point).
Here ’s the work that have been done on the St-Lawrence to get the seamless coverage with separation values. We called that the undulation or the separation table. 40 primary points where selected from Montreal to Sault-au-cochon(about 100km east of Qubec city). Chart datum is already known at thoses points. GPS measurements were taken using Geodetic GPS receivers. Complete 24 hours observation at each point. Everything was sent to NRCAN for processing and compensation. Compensation is needed because we are working on a large geographic extent. 20 secondary points were added using RTK technique. About 1 hour on each point. The results we got from NRCAN are precise positions solutions for all the 40 control points in XYZ relative to the ellipsoid. For each point we were able to compute a separation value using the known chart datum and the height relative to the ellipsoid. Finally to get the seamless coverage we used krigging software. The geostatistical approch of the krigging is giving us good results. Usually in hydrography we use gridding to generate digital terrain model.That ’s good when you ’ve got a good density of points. In our case with the density of control points we had we found that krigging was the best solution.

9. Accuracy of the Thales LRK GPS equipment
Here ’s a graph showing the accuracy that we can achieve. The distance between Base and rover was about 25 kilometers. The rover was static on a known geodetic point. So the variation that you see is the noise or the limit of the system. Standard deviation is 1,5 centimeter. Depending on the ionospheric conditions the results can vary a bit. I said before tha accuracies are critical when surveying in the channel. A lot of work have been put in comparing RTK to traditionnal tide staff. The hydrographers had some problem admitting that the tide staff method was not perfect. Looking rapidly at the graph with the noise of the GPS solution a lot of people were rejecting the idea of using GPS method to detect a water level. If we reverse everything and try to estimate the accuracy of tide reader observation when you ’ve got waves of half a meter on the tide staff. I thing you see the idea. We are not taking about only few centimeters there. The integration of the RTK value was not obvious. Depending on the sensors on board the integration is different. One thing for sure for use in the navigation channel the integration must be rigourous.

10. LRK Network – 2004 Dredge Channel LRK Base stations
St-François – ID #1
431.1MHz
Neuville – ID #2
430.1MHz
Québec
Grondines – ID #3
431.1MHz
Montmagny
Batiscan
Ste-Croix
Sainte-Marthe – ID #4
430.6MHz
Pointe du Lac
Sorel – ID #5
430.1MHz
Lanoraie
Silo du port Sorel
Verchères – ID #6
431.1MHz
Dredge Channel
LRK Base stations
Longueuil
Montréal
LRK base station repeators

11. Data Collection All sensors on board logged as usual
Traditionnal DGPS replaced by RTK
Using NMEA quality indicator
GPS Tide values displayed (Not the full rigourous solution)
Only the raw position data is stored
Our 5 bases station are now online giving us RTK capability from Montreal to Sault-au-Cochon. All the vessels working in the channel have similar setup. Rangnig from 6 to 33 tranducers the platforms are equipped with motion sensors, Thales Navigation LRK receivers, speed log and tide interpolating software. For data collection we use hypack from coastal oceanographics. All sensors on board are logged as usual with hypack. We are still logging 2 tides values for validation purposes. I will explain why later. For interfacing we are still using NMEA GGA sentences. NMEA quality flags are used to warn the hydrographer he is not in RTK mode. WE log position exactly like DGPS data but we provide the hydrographer an approximate GPSTide value and a graph showing the values of the last minute. Looking at this info the hydrographer can see an aproximate water level at the vessel position that he can compare with other sources of tide information. I'm talking about an approximate value because the GPS tide value in that case is just derived from GPSHeight, separation value and antenna offset. So we will not log that value we will only keep the raw position data.
Latitude, longitude and height above ellipsoid.

12. What are we getting from RTK ?
XYZ position of the antenna relative to WGS 84
What is included in that Z value ?
What are we getting from RTK ?
The easy answer is XYZ position of the antenna relative to WGS84. we know what to do with X and Y data but waht is included in that Z value ? Or in others words waht to we have to remove from that Z value to get GPS tide ?
Antenna height : really inportant to know your antenna height relative to the water plane.
Heave: If your survey vessel is affected by heave at the time of the fix with the GPS satellites, heave will be included in the z value
Pitch an roll effects: If your vessel is pitching and rolling the antenna height relative to the water plane will be affected.
Static draft: Function of the load on board. Fuel, people.
Dynamic draft : Function of speed over water
Swell : Or waht no traditionnal motion sensor is able to detect because the wavelenght is too long
Some of theses values are already measured by others sensors, we have to make sure that we are not doing double correction.
Antenna height
Heave
Pitch and Roll effects on antenna height
Static Draft of the vessel
Dynamic Draft of the vessel
Swell
Some of those values are already measured by others sensors, we have to make sure that we are not doing double correction.

13. Integration Choices :
1- Single Beam without motion sensor
Reduce the antenna height to the water level using the HIPS VCF entries for the navigation antenna offset.
Result is ‘GPSTide’ water level with heave and dynamic draft still included.
2- Single Beam with pitch and roll sensor
Remove vessel motion from recorded antenna height with the pitch and roll data.
Remove the dynamic draft (squat / lift) from the antenna height.
Result is ‘GPSTide’ water level with tide and heave included.
3- Multibeam and Multitransducer with motion sensor
Remove the heave from the antenna height by applying the recorded heave data.
Result is ‘GPSTide’ water level with only the true tide effects remaining.
For data processing we use HIPS for Caris. I<ve worked hard with Mike Gourley from Caris to find the best way to integrate the additionnal computation needed to have the complete rigourous solution. Here's the approch we took. It's ranging from single beam to multibeam.
1- Single beam without motion sensor
In that case it's a direct reduction. We've got no motion sensor and no speed log so we can't correct pitch roll and dynamic draft. So thoses value will still be included in the final GPS tide time series.
2- Single beam with pitch and roll sensor
In that case we have pitch and roll cvalues so we will be abvle to correct for thoses effects. If a speed log is installed we wuill be able to remove dynamic draft effect also. The result will GPStide with tide and heave included
3- for multibeam and multitransducers with motion sensors that is the ful rigourous solution.
-Correction for pitch,roll,heave and dynamic draft
The result is a GPStide with only the true tide effects remaining. That means that you can compare the time serie with other traditionnal values

14. Channel survey vessels
CCGS F.C.G Smith
HYDROGRAPHIC SURVEY CATAMARAN 34.8m
33 Transducers - 6 MCS Navitronics
Frequency 200kHz; Depth: m
Beamwidth 8 degrees
Heave / Roll / Pitch compensated with TSS, Gyro and Speedlog
St. Lawrence River channel
Montreal
Quebec
Grondines

15. Channel survey vessels
CCGS GC-03
HYDROGRAPHIC SURVEY CATAMARAN 18.5m
12 Transducers - 2 MCS Navitronics
Frequency 200kHz; Depth: m
Beamwidth 8 degrees
Heave / Roll / Pitch compensated with TSS 335B, Gyro and Speedlog
St. Lawrence River channel
Montreal
Quebec
Grondines

16. Channel survey vessels
Morillon
HYDROGRAPHIC SURVEY LAUNCH
6 Transducers - MCS2000/F6 Navitronics
Frequency 200kHz; Depth: m
Beamwidth 4.5 degrees
Honeywell HMR3000 Digital compass / Roll-Pitch
Montreal
Quebec
Grondines
St. Lawrence River channel

17. Time Series Cleaning and Smoothing
Heave
Time Series Cleaning and Smoothing
Basic cleaning
Reject with interpolation
Reject without interpolation
Pitch
Roll
Once the computations are done we can go in the attitude editor to check our values. In that part of the processing the hydrographer will clean and inspect his time series. In that example the survey vessel met a big container ship. On purpose we let the logging go on. The sampling rate of the GPS receiver 2 hertz. With that sampling rate we are not able to measure the attitude of the survey vessel very well. But the motion sensor did a pretty good job because of the high sampling rate and shorth wavelentght of the movement. So in that case will leave the modeling of the container ship wave to teh motion sensor. Simply reject with inetrpolation the green part. On top of that we will smooth the GPS tide time serie. Before the container ship everything was calm and after it's noisier. All thoses little vertical movement are well model by the motion sensor so to remove that component from the GPStide series we will use fast fourier transform or moving average, that is the purple line that you see on the graph. That's a pretty extreme example. From what we saw a 90 second moving average will do a good job in most cases. Using that smothing capability will will ensure that we are not double correcting. With this approch we can compare GPS tide with traditionnal tide zone calculations.
Smoothing capabilities
Fast Fourier Transform
Moving Average
Tide
GPS Tide
With this approch we can do a direct comparison between GPS Tide and traditionnal tide measurements or tide zones calculations.
GPS Height

18. Opportunities and future work
Bin format from NGS is used to store the values of the separation table (NRCAN has adopted the same format)
With that approch we have compatibility with any RTK receiver giving XYZ relative to the WGS-84
Reference to any datum is possible (once you know the relation to WGS-84)
Squat measurement on large vessels
We want to test the integration in more dynamic conditions.
Carry the positionning quality flag the post-processing software.
To store our separation values we are using a format from National Geodetic Survey in the US. THe BIN format is used by NGS to store GEOID99 values. In our case we are just using the format to store our separation values. Natural ressources Canada has also adopted that format to store their geoid values. With the approch I described we have compability with any RTK receiver giving us XYZ relative to WGS84 ellipsoid using NMEA sentences. Using RTK we can reference sounding to any datum possible. If you know the relation to the WGS84 ellipsoid you can compute another separation table. For example separation of mean sea level instead of chart datum. That is possible because the raw GPS height was recorded. The next big project we have is squat measurements of large ships. Having 5 permanent bases stations we can put 6 rovers at stategic points on a container ship for example and monitor is squat all along the restricted channel. The objective of that project is to propose an adapted squat formula for large ship using the St-Lawrence river. This year we have logged a lot of data with a multibeam in the channel and it seems the integration we are doing is working well. Now we have to test with the new True heave solution propose by applanix and see what kind of results we are getting. We want also to carry GPS quality flags in the post processing software to help in data cleaning.

19. Batiscan
Tide Gauge
SPINE
1.50
Finally, ajust the forecast at each node and interpolate in time and space at vessel time and position
1.60
New system to provide a water level via hydrodynamic model and automatic tide gauges. The goal is to provide the hydrographer with another water level to validate the GPS tide value.
1.40
-0.10
Observations are avalaible at 10:00:00 (3mins)
Computation of ajustement at each tide gauge at 10:00:00
Hydrodynamic model nodes
1.23
1.30
Forecast at 10:00:00
1.24
0.95
-0.05
1.33
1.40
Forecast at 10:07:30
Bécancour
Tide Gauge
1.25
1.18
Observations at 10:00:00
1.28
1.35
-0.07
-0.07
Ajustement at 10:00:00
Vessel Data
UTC Time : 10:01:32
Latitude :
Longitude :
Interpolation of ajustement at each node
1.00
1.09
Forecast are available at each node of the hydrodynamic model (7.5 mins)

20. Summary
Centimeter accuracy in positionning in real-time is now possible.
GPS tide detection: Separation model between Ellipsoid an Chart Datum is needed.
Establishment of Seamless Datum is an issue.
Integration of GPS Tide implies additional computations
Limited range
Sun spot
In summary, centimeter accuracy in XYZ in real-time is now possible. Almost every GPS receiver manufacturer has a solution to propose.
If you want to detect tide using GPS you need a separation model
The establishment and validation of the seamless datum is an important issue, a lot of work to get good accuracy.
Range is limited, in our case 40 kilometers.
Integration of GPS tide implies additionnal computations
Solar activity is going down now, so solutions should improve in the future and perhaps rande will be extended.
Thank you for your attention.