- Email: [email protected]
Online kinematic GNSS data processing for small hydrographic surveys
Ocean Engineering 112 (2016) 335–339
Contents lists available at ScienceDirect
Ocean Engineering journal homepage: www.elsevier.com/locate/oceaneng
Online kinematic GNSS data processing for small hydrographic surveys M. Berber a,n, W. Wright b a b
Department of Civil and Geomatics Engineering, California State University, Fresno, CA, USA Department of Geosciences, Florida Atlantic University, Boca Raton, FL, USA
art ic l e i nf o
a b s t r a c t
Article history: Received 4 July 2015 Accepted 2 October 2015 Available online 4 January 2016
Free access online GNSS (Global Navigation Satellite System) data processing services are becoming popular since only a single GNSS receiver can do the job. These services are user friendly and easy to use. Thus, no training and a GNSS software package purchase are needed. This means less cost to users. Currently, three online GNSS data processing services provide kinematic data processing option. In this study, GNSS data collected for hydrographic surveying is processed using these online services and positioning precision of these services are compared. The results indicate that with 1 s data, decimeter to meter precision can be achieved for both horizontal and vertical coordinates. & 2015 Elsevier Ltd. All rights reserved.
Keywords: Global Navigation Satellite System Online GNSS data processing Real-time kinematic method Hydrographic surveys Precise point positioning
1. Introduction To this date, there are six online GNSS (Global Navigation Satellite System) data processing services (OPUS, APPS, SCOUT, CSRS-PPP, GAPS and AUSPOS) freely available to users to our knowledge. Detailed information about these services may be obtained by accessing their respective website URLs given in Table 1. Online processing services deliver solutions to users without any cost, and with unlimited access. Only access to the Internet and an email address are needed to make use of these services (Ghoddousi-Fard and Dare, 2006). Although some of these services accept most proprietary receiver binary format files, some still require RINEX (Receiver Independent Exchange Format) file or compressed RINEX files. With just one GNSS receiver, the observations are collected and then postprocessed based on differential methods using either reference stations or precise point positioning using globally available precise satellite orbit and clock data, and processing results are returned to the user via e-mail (Ebner and Featherstone, 2008). Among these online services only three of them (APPS, CSRS-PPP and GAPS) provide kinematic data processing option. Hydrographic surveying deals with measurement of depth at known locations, typically for the purpose of creating depth maps of water bodies. In the past, locations of depth soundings were determined using shore-based surveying or dead-reckoning navigation techniques. However, since the development of precise satellite based survey methods (i.e., GNSS), locations of depth measurements are more commonly determined using RTK (Real n
Corresponding author. E-mail address: [email protected] (M. Berber).
http://dx.doi.org/10.1016/j.oceaneng.2015.10.001 0029-8018/& 2015 Elsevier Ltd. All rights reserved.
Time Kinematic) GPS (Global Positioning System). Using RTK GPS for hydrographic surveying is similar to other applications, in that a base station is set up over a known point on land, and a rover receiver is used to measure the positions of unknown points using corrections transmitted from the base station via radio transmission. The rover on the vessel combines the data from the base with its own observations and determines its corrected 3D position in real time (Ghilani and Wolf, 2007). Echosounders measure depth of water by transmitting a sonar pulse from a transducer into the water column and measuring the amount of time it takes for that pulse to return to the transducer. Travel distance through the water column is then calculated by applying the velocity of sound through water, which varies with salinity, pressure, and temperature. In hydrographic surveying, it is crucial to have an accurate velocity measurement at the survey area, which can be measured using either bar check or velocity profiling methods. The bar check method is essentially a determination of velocity by taking a series of measurements through water to a known depth. Velocity profiling uses a cable mounted sensor that directly measures the velocity of sound waves through water passing between a transmitter and receiver in the sensor housing as the unit is lowered through the water column. In this study, bar checks were used to calculate the appropriate velocity, and were performed at approximately hourly intervals. By combining RTK GNSS horizontal and vertical coordinates with echosounding techniques and practices, a user can expect to obtain XYZ coordinates of the bottom of a small body of water with accuracies approaching one decimeter. Achieving the same level of precision is investigated by using online kinematic GNSS data processing services; namely, CSRS PPP, GAPS and APPS.
336
M. Berber, W. Wright / Ocean Engineering 112 (2016) 335–339
2. Theory of precise point positioning The position of a standalone receiver tracking dual frequency GNSS signals can be determined using undifferenced code and phase observations as C 1;2 ¼ ρsr þcðdt r dt s Þ þ dI 1;2 þ dT þeðC 1;2 Þ
P 1;2 ¼ ρsr þ cðdt r dt s Þ dI 1;2 þ dT þ λ1;2 N 1;2 þ eðP 1;2 Þ
ð1Þ
where C 1;2 code observations of L1 and L2 pseudoranges, respectively P 1;2 phase observations of L1 and L2 carrier phases, respectively ρsr true geometric distance between receiver and satellite c speed of light in vacuum (299,792,458 m/s) dt r receiver clock error from GNSS time dt s satellite clock error from GNSS time dI 1;2 ionospheric terms on L1 and L2 signals, respectively dT tropospheric effect λ1;2 wavelengths of L1 and L2 signals, respectively N 1;2 integer phase ambiguities relevant L1 and L2 signals eðC 1;2 Þ measurement errors of C 1;2 code observations eðP 1;2 Þ measurement errors of P 1;2 phase observations. If a linear combination model is used, LIF ¼
2 2 f1 f L 2 2 2 L2 2 2 1 f 1 f 2 f 1 f 2
ð2Þ
Table 1 URLs of online GNSS data processing services. Service
URL
OPUS APPS SCOUT CSRS-PPP GAPS AUSPOS
http://www.ngs.noaa.gov/OPUS/index.jsp http://apps.gdgps.net/ http://sopac.ucsd.edu/scout.shtml http://webapp.geod.nrcan.gc.ca/geod/tools-outils/ppp.php http://gaps.gge.unb.ca/ http://www.ga.gov.au/bin/gps.pl
in which L represents both code (C) and phase (P) observations, the frequency-dependent ionospheric terms can be eliminated from Eq. (1). On the other hand, precise orbit and clock information that come from rapid and final orbit solutions of satellites can be considered as known parameters in the observation equations. In this case, the ionosphere-free (IF) observation equations become C IF ¼ ρsr þ cdt r þ dT þeðC IF Þ P IF ¼ ρsr þcdt r þ dT þ λIF N IF þeðP IF Þ
ð3Þ
where λIF is the wavelength of the combined phase and N IF is the non-integer ambiguity of ionosphere-free combination. The linearized adjustment model of Eq. (3) and its solution procedures can be found in e.g. Kouba and Heroux (2001). Normally, before the use of IF observation equations in the adjustment model, they should be corrected for additional systematic effects such as satellite antenna, phase wind-up, relativistic effect and site specific corrections due to ocean loading, solid tides, etc.
3. Applications and results Triple frequency GNSS receivers (Leica GX1230) and a hydrolite single beam echosounder are used to collect the measurements. The hydrolite, manufactured by Seafloor Systems, Inc., is a single beam echo sounder with the transducer attached to the bottom of a GNSS rover rod, which takes point soundings at a rate of 6 Hz and sends the data through a Bluetooth connection to the GNSS data collector. Since RTK corrections supply elevation values for the transducer (after applying an offset for the height of the rod), this provides a relatively user friendly configuration for small hydrographic surveys. Measurements for this study are taken in Turkey Creek in Brevard County, Florida (Fig. 1). The survey area for this study was centered on the Florida East Coast (FEC) railway bridge crossing the creek, and extended about 100 m to the east, and approximately 150 m west of the bridge. This is considered an ideal location for the RTK and Hydrolite setup due to shallow (o5 m depth) and calm inland waters, where conditions and depth favor a smaller boat and survey setup.
Fig. 1. Aerial photo of project site. Colored lines show locations of depth measurements, with color scale representing elevation of points in feet (aerial image from Labins. org). (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)
M. Berber, W. Wright / Ocean Engineering 112 (2016) 335–339
0.8 0.7
337
CSRS
0.6 [m]
0.5 0.4 0.3 0.2 0.1 0 Fig. 3. Differences between LGO produced horizontal coordinates and CSRS produced horizontal coordinates. In the figure, x axis shows number of points (1372) and y axis shows differences in meters.
0 Fig. 2. Schematic of hydrographic survey setup showing RTK GPS base station on land, and rover and hydrolite on survey boat.
-0.2
CSRS
-0.6 -0.8 -1 -1.2
Fig. 4. Differences between LGO produced vertical coordinates and CSRS produced vertical coordinates. In the figure, x axis shows number of points (1369) and y axis shows differences in meters.
3.5 3
GAPS
[m]
2.5 2 1.5 1 0.5 0 Fig. 5. Differences between LGO produced horizontal coordinates and GAPS produced horizontal coordinates. In the figure, x axis shows number of points (1364) and y axis shows differences in meters.
3 2
GAPS
1 [m]
During the survey, the base station (shown in Fig. 2 and as the coordinate with highest elevation in Fig. 1) transmitted RTK signals to the rover attached to the hydrolite system mounted on the survey boat. GNSS data provided the location and elevation of the rover antenna (and therefore transducer), and depth measurements collected by hydrolite system are subtracted from the GNSS elevations to provide elevation of river bottom (see Fig. 2), with the resulting elevation map shown in Fig. 1. When using online processing services, an initial 30 min of non-moving RTK data is needed to allow for proper convergence of carrier phase ambiguities (see Section 2). We performed this initialization in the field at a point near the base station, and then continued with the kinematic hydrographic portion of the survey, which was collected at 1 s intervals. For this study, the GNSS rover receiver was set to collect data using the RTK corrections from the base station. In Leica Geo Office (LGO), after checking survey settings (e.g., base station coordinates, antenna height, etc.), the corrections were removed in order to reveal the raw, uncorrected coordinates, which were then exported to RINEX format files. This RINEX file is submitted to three online services (CSRS, GAPS and APPS). While submitting the file, kinematic option and epoch of GNSS data is chosen for epoch information under processing options. Although some services provide more processing options such as elevation angle cutoff and solution output rate, the basic processing options are selected in order to be consistent between processing services. Then, a comparison between the corrections from the RTK base station and each online processing service is made. In this study, coordinates produced using typical RTK corrections from the base station in LGO are considered the “truth” and the differences between LGO results and the results obtained from online GNSS data processing services are computed. These results are shown in Figs. 3–8. In CSRS results (Figs. 3 and 4), although there is a jump with CSRS horizontal determinations, most data is consistent within a few decimeters. CSRS vertical determinations portray almost the same pattern as horizontal results even though the results are close to meter range. As can be seen in Figs. 5 and 6, GAPS results are more erratic compared to CSRS results. Horizontal determinations are between 0.5 and 2 m range, with a maximum of 2.85 m, and vertical coordinates are determined in the order of plus or minus 2–3 m. APPS results (Figs. 7 and 8) show the highest amount of error among these three online GNSS data processing services in terms
[m]
-0.4
0 -1 -2 -3 -4
Fig. 6. Differences between LGO produced vertical coordinates and GAPS produced vertical coordinates. In the figure, x axis shows number of points (1333) and y axis shows differences in meters.
of both horizontal and vertical coordinates. In general, horizontal coordinates are in agreement within 1.5 m, with some errors approaching 5 m, and vertical coordinates are only good to about plus or minus 3 m.
338
M. Berber, W. Wright / Ocean Engineering 112 (2016) 335–339
5
[m]
4
APPS
3
Table 2 RMS differences (m).
CSRS GAPS APPS
2
Horizontal
Vertical
0.21 1.18 1.76
0.87 1.24 1.51
1 0
[m]
Fig. 7. Differences between LGO produced horizontal coordinates and APPS produced horizontal coordinates. In the figure, x axis shows number of points (1199) and y axis shows differences in meters.
4 3 2 1 0 -1 -2 -3
APPS
Fig. 8. Differences between LGO produced vertical coordinates and APPS produced vertical coordinates. In the figure, x axis shows number of points (1084) and y axis shows differences in meters.
Although differences between standard RTK methods and online processing services may seem large, we would like to point out that in this project, o1 s of data is collected at each point from a moving rover during the hydrographic portion of the survey. If each point was occupied for a longer interval, the results would most likely improve for both horizontal and vertical coordinates. In a terrestrial RTK survey, a higher number of data epochs may be collected, which allows a better chance of getting integer ambiguities determined properly. When using GNSS methods, height result errors are typically 2–3 times greater than horizontal coordinate error (Berber et al., 2012). While processing the data using online data processing services, we experienced that if a data set is not consistent throughout, i.e., if there are changes in observation number, or if observations do not fit the expected order defined in the header, online data processing services return error messages or do not process the data at all. It means that these services may be unforgiving when it comes to RINEX file format. Additionally, because short period of data (1 s data) is processed in this study, in some cases these services could not solve integer ambiguities (see Section 2) and as a consequence erroneous results are generated; for instance, 10 m or more errors with elevation determinations. These erroneous results are manually removed from the data set. As a result, the number of observations used is slightly different with each service. Quality of data is the determining factor for the output results; yet, as mentioned already in this paper the precision of online GNSS data processing services is investigated. Hence, regardless of data quality, these services should return the same results. Meaning that if data quality is good, then all services should return well processed data. Also, if data quality is not good, it is expected that these services might yield some erroneous, but similar results. This means that here relative precision is compared among these services. Root Mean Square (RMS) differences are tabulated for each service in Table 2. As can be seen, the values are in the order of decimeters to meters. The RMS differences shown in Table 2 appear large because positions computed using online kinematic GNSS data processing
Table 3 Biases exhibited by online processors (m).
CSRS GAPS APPS
Horizontal
Vertical
0.08 0.84 2.77
0.14 1.22 1.24
will not be optimal until the carrier phase ambiguities have converged. For points with shorter occupation times, positions will effectively be calculated using only the pseudorange observations (see Section 2). Allowing the GNSS antenna to occupy a point for a longer period of time may make it possible to resolve the ambiguities required to recover positions using the more precise carrier phase observations. As mentioned above, in this project, only 1 s data were collected. It means that for the solutions of coordinates of these points pseudorange observations are used. As well known, use of pseudorange observations yields meter level precision, which is in agreement with Table 2 values. It appears that the differences in the above graphics exhibit systematic positive and negative shifts which mean that some of the processing are biased. In order to get additional insight about the performance of the online processors in terms of their biases and precision, we calculated the average differences and removed these average differences from the data sets and then calculated the MSE again which is equivalent to calculating the variance of a sample. As a check, we recomputed the RMSE values by squaring the bias and adding the variance, and compared these results against the RMSE values listed in Table 2 and these differences gave the biases which are listed in Table 3. As can be seen with the results in Table 3, the outputs produced by these online processors are biased. In terms of size of biases, values in Table 3 are aligned with the results of Table 2 which is another indicative of power of the processors used by these services.
4. Conclusions Precision of three online GNSS data processing services (CSRS PPP, GAPS and APPS) are compared to RTK data solutions from a base station for hydrographic surveying purposes. In this study, kinematic data collected at a rate common for hydrographic surveying (1 s) is processed using these three online processing services and compared to RTK data collected using a base station. The results of RMS error are in the order of decimeters to meters for both horizontal and vertical coordinates. Among the three services, CSRS produced the best results and APPS produced the worst results. Further analyses on biases portrayed the same picture in terms of power of the processors used by these services. However, the source of these large differences is likely the high rate of data collection, meaning less time for the antenna to occupy a single point. Less occupation time likely does not allow time for carrier phase ambiguities to converge, and as a consequence pseudorange observations are used for coordinate determinations. Use of pseudorange observations typically results
M. Berber, W. Wright / Ocean Engineering 112 (2016) 335–339
in meter level precision, which is seen in the results obtained from the three services. Depending on specific project requirements, the precision provided by these online services would not be good enough for most hydrographic survey purposes. If better precision is needed, it is our recommendation that a surveyor use either a traditional RTK base and rover configuration, or allow for a longer period of data collection.
Acknowledgments Authors would like to thank CSRS PPP team especially Pierre Tetreault for their help with processing the RINEX data. Our thanks also go to Marco Mendonca for his help to process the RINEX file using GAPS software, and Seafloor Systems for equipment support.
339
References Berber, M., Ustun, A., Yetkin, M., 2012. Comparison of accuracy of GPS techniques. Measurement 45 (2012), 1742–1746. Ebner, R., Featherstone, W.E., 2008. How well can online GPS PPP post-processing services be used to establish geodetic survey control networks? J. Appl. Geod. 2, 149–157. Ghilani, C.D., Wolf, P.R., 2007. Elementary Surveying—An Introduction to Geomatics, 12th ed. Pearson Prentice Hall, New Jersey. Ghoddousi-Fard, R., Dare, P., 2006. Online GPS processing services: an initial study. GPS Solut. 10, 12–20. Kouba, J., Heroux, P., 2001. Precise point positioning using IGS orbit and clock products. GPS Solut. 5 (2), 12–28.
Contents lists available at ScienceDirect
Ocean Engineering journal homepage: www.elsevier.com/locate/oceaneng
Online kinematic GNSS data processing for small hydrographic surveys M. Berber a,n, W. Wright b a b
Department of Civil and Geomatics Engineering, California State University, Fresno, CA, USA Department of Geosciences, Florida Atlantic University, Boca Raton, FL, USA
art ic l e i nf o
a b s t r a c t
Article history: Received 4 July 2015 Accepted 2 October 2015 Available online 4 January 2016
Free access online GNSS (Global Navigation Satellite System) data processing services are becoming popular since only a single GNSS receiver can do the job. These services are user friendly and easy to use. Thus, no training and a GNSS software package purchase are needed. This means less cost to users. Currently, three online GNSS data processing services provide kinematic data processing option. In this study, GNSS data collected for hydrographic surveying is processed using these online services and positioning precision of these services are compared. The results indicate that with 1 s data, decimeter to meter precision can be achieved for both horizontal and vertical coordinates. & 2015 Elsevier Ltd. All rights reserved.
Keywords: Global Navigation Satellite System Online GNSS data processing Real-time kinematic method Hydrographic surveys Precise point positioning
1. Introduction To this date, there are six online GNSS (Global Navigation Satellite System) data processing services (OPUS, APPS, SCOUT, CSRS-PPP, GAPS and AUSPOS) freely available to users to our knowledge. Detailed information about these services may be obtained by accessing their respective website URLs given in Table 1. Online processing services deliver solutions to users without any cost, and with unlimited access. Only access to the Internet and an email address are needed to make use of these services (Ghoddousi-Fard and Dare, 2006). Although some of these services accept most proprietary receiver binary format files, some still require RINEX (Receiver Independent Exchange Format) file or compressed RINEX files. With just one GNSS receiver, the observations are collected and then postprocessed based on differential methods using either reference stations or precise point positioning using globally available precise satellite orbit and clock data, and processing results are returned to the user via e-mail (Ebner and Featherstone, 2008). Among these online services only three of them (APPS, CSRS-PPP and GAPS) provide kinematic data processing option. Hydrographic surveying deals with measurement of depth at known locations, typically for the purpose of creating depth maps of water bodies. In the past, locations of depth soundings were determined using shore-based surveying or dead-reckoning navigation techniques. However, since the development of precise satellite based survey methods (i.e., GNSS), locations of depth measurements are more commonly determined using RTK (Real n
Corresponding author. E-mail address: [email protected] (M. Berber).
http://dx.doi.org/10.1016/j.oceaneng.2015.10.001 0029-8018/& 2015 Elsevier Ltd. All rights reserved.
Time Kinematic) GPS (Global Positioning System). Using RTK GPS for hydrographic surveying is similar to other applications, in that a base station is set up over a known point on land, and a rover receiver is used to measure the positions of unknown points using corrections transmitted from the base station via radio transmission. The rover on the vessel combines the data from the base with its own observations and determines its corrected 3D position in real time (Ghilani and Wolf, 2007). Echosounders measure depth of water by transmitting a sonar pulse from a transducer into the water column and measuring the amount of time it takes for that pulse to return to the transducer. Travel distance through the water column is then calculated by applying the velocity of sound through water, which varies with salinity, pressure, and temperature. In hydrographic surveying, it is crucial to have an accurate velocity measurement at the survey area, which can be measured using either bar check or velocity profiling methods. The bar check method is essentially a determination of velocity by taking a series of measurements through water to a known depth. Velocity profiling uses a cable mounted sensor that directly measures the velocity of sound waves through water passing between a transmitter and receiver in the sensor housing as the unit is lowered through the water column. In this study, bar checks were used to calculate the appropriate velocity, and were performed at approximately hourly intervals. By combining RTK GNSS horizontal and vertical coordinates with echosounding techniques and practices, a user can expect to obtain XYZ coordinates of the bottom of a small body of water with accuracies approaching one decimeter. Achieving the same level of precision is investigated by using online kinematic GNSS data processing services; namely, CSRS PPP, GAPS and APPS.
336
M. Berber, W. Wright / Ocean Engineering 112 (2016) 335–339
2. Theory of precise point positioning The position of a standalone receiver tracking dual frequency GNSS signals can be determined using undifferenced code and phase observations as C 1;2 ¼ ρsr þcðdt r dt s Þ þ dI 1;2 þ dT þeðC 1;2 Þ
P 1;2 ¼ ρsr þ cðdt r dt s Þ dI 1;2 þ dT þ λ1;2 N 1;2 þ eðP 1;2 Þ
ð1Þ
where C 1;2 code observations of L1 and L2 pseudoranges, respectively P 1;2 phase observations of L1 and L2 carrier phases, respectively ρsr true geometric distance between receiver and satellite c speed of light in vacuum (299,792,458 m/s) dt r receiver clock error from GNSS time dt s satellite clock error from GNSS time dI 1;2 ionospheric terms on L1 and L2 signals, respectively dT tropospheric effect λ1;2 wavelengths of L1 and L2 signals, respectively N 1;2 integer phase ambiguities relevant L1 and L2 signals eðC 1;2 Þ measurement errors of C 1;2 code observations eðP 1;2 Þ measurement errors of P 1;2 phase observations. If a linear combination model is used, LIF ¼
2 2 f1 f L 2 2 2 L2 2 2 1 f 1 f 2 f 1 f 2
ð2Þ
Table 1 URLs of online GNSS data processing services. Service
URL
OPUS APPS SCOUT CSRS-PPP GAPS AUSPOS
http://www.ngs.noaa.gov/OPUS/index.jsp http://apps.gdgps.net/ http://sopac.ucsd.edu/scout.shtml http://webapp.geod.nrcan.gc.ca/geod/tools-outils/ppp.php http://gaps.gge.unb.ca/ http://www.ga.gov.au/bin/gps.pl
in which L represents both code (C) and phase (P) observations, the frequency-dependent ionospheric terms can be eliminated from Eq. (1). On the other hand, precise orbit and clock information that come from rapid and final orbit solutions of satellites can be considered as known parameters in the observation equations. In this case, the ionosphere-free (IF) observation equations become C IF ¼ ρsr þ cdt r þ dT þeðC IF Þ P IF ¼ ρsr þcdt r þ dT þ λIF N IF þeðP IF Þ
ð3Þ
where λIF is the wavelength of the combined phase and N IF is the non-integer ambiguity of ionosphere-free combination. The linearized adjustment model of Eq. (3) and its solution procedures can be found in e.g. Kouba and Heroux (2001). Normally, before the use of IF observation equations in the adjustment model, they should be corrected for additional systematic effects such as satellite antenna, phase wind-up, relativistic effect and site specific corrections due to ocean loading, solid tides, etc.
3. Applications and results Triple frequency GNSS receivers (Leica GX1230) and a hydrolite single beam echosounder are used to collect the measurements. The hydrolite, manufactured by Seafloor Systems, Inc., is a single beam echo sounder with the transducer attached to the bottom of a GNSS rover rod, which takes point soundings at a rate of 6 Hz and sends the data through a Bluetooth connection to the GNSS data collector. Since RTK corrections supply elevation values for the transducer (after applying an offset for the height of the rod), this provides a relatively user friendly configuration for small hydrographic surveys. Measurements for this study are taken in Turkey Creek in Brevard County, Florida (Fig. 1). The survey area for this study was centered on the Florida East Coast (FEC) railway bridge crossing the creek, and extended about 100 m to the east, and approximately 150 m west of the bridge. This is considered an ideal location for the RTK and Hydrolite setup due to shallow (o5 m depth) and calm inland waters, where conditions and depth favor a smaller boat and survey setup.
Fig. 1. Aerial photo of project site. Colored lines show locations of depth measurements, with color scale representing elevation of points in feet (aerial image from Labins. org). (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)
M. Berber, W. Wright / Ocean Engineering 112 (2016) 335–339
0.8 0.7
337
CSRS
0.6 [m]
0.5 0.4 0.3 0.2 0.1 0 Fig. 3. Differences between LGO produced horizontal coordinates and CSRS produced horizontal coordinates. In the figure, x axis shows number of points (1372) and y axis shows differences in meters.
0 Fig. 2. Schematic of hydrographic survey setup showing RTK GPS base station on land, and rover and hydrolite on survey boat.
-0.2
CSRS
-0.6 -0.8 -1 -1.2
Fig. 4. Differences between LGO produced vertical coordinates and CSRS produced vertical coordinates. In the figure, x axis shows number of points (1369) and y axis shows differences in meters.
3.5 3
GAPS
[m]
2.5 2 1.5 1 0.5 0 Fig. 5. Differences between LGO produced horizontal coordinates and GAPS produced horizontal coordinates. In the figure, x axis shows number of points (1364) and y axis shows differences in meters.
3 2
GAPS
1 [m]
During the survey, the base station (shown in Fig. 2 and as the coordinate with highest elevation in Fig. 1) transmitted RTK signals to the rover attached to the hydrolite system mounted on the survey boat. GNSS data provided the location and elevation of the rover antenna (and therefore transducer), and depth measurements collected by hydrolite system are subtracted from the GNSS elevations to provide elevation of river bottom (see Fig. 2), with the resulting elevation map shown in Fig. 1. When using online processing services, an initial 30 min of non-moving RTK data is needed to allow for proper convergence of carrier phase ambiguities (see Section 2). We performed this initialization in the field at a point near the base station, and then continued with the kinematic hydrographic portion of the survey, which was collected at 1 s intervals. For this study, the GNSS rover receiver was set to collect data using the RTK corrections from the base station. In Leica Geo Office (LGO), after checking survey settings (e.g., base station coordinates, antenna height, etc.), the corrections were removed in order to reveal the raw, uncorrected coordinates, which were then exported to RINEX format files. This RINEX file is submitted to three online services (CSRS, GAPS and APPS). While submitting the file, kinematic option and epoch of GNSS data is chosen for epoch information under processing options. Although some services provide more processing options such as elevation angle cutoff and solution output rate, the basic processing options are selected in order to be consistent between processing services. Then, a comparison between the corrections from the RTK base station and each online processing service is made. In this study, coordinates produced using typical RTK corrections from the base station in LGO are considered the “truth” and the differences between LGO results and the results obtained from online GNSS data processing services are computed. These results are shown in Figs. 3–8. In CSRS results (Figs. 3 and 4), although there is a jump with CSRS horizontal determinations, most data is consistent within a few decimeters. CSRS vertical determinations portray almost the same pattern as horizontal results even though the results are close to meter range. As can be seen in Figs. 5 and 6, GAPS results are more erratic compared to CSRS results. Horizontal determinations are between 0.5 and 2 m range, with a maximum of 2.85 m, and vertical coordinates are determined in the order of plus or minus 2–3 m. APPS results (Figs. 7 and 8) show the highest amount of error among these three online GNSS data processing services in terms
[m]
-0.4
0 -1 -2 -3 -4
Fig. 6. Differences between LGO produced vertical coordinates and GAPS produced vertical coordinates. In the figure, x axis shows number of points (1333) and y axis shows differences in meters.
of both horizontal and vertical coordinates. In general, horizontal coordinates are in agreement within 1.5 m, with some errors approaching 5 m, and vertical coordinates are only good to about plus or minus 3 m.
338
M. Berber, W. Wright / Ocean Engineering 112 (2016) 335–339
5
[m]
4
APPS
3
Table 2 RMS differences (m).
CSRS GAPS APPS
2
Horizontal
Vertical
0.21 1.18 1.76
0.87 1.24 1.51
1 0
[m]
Fig. 7. Differences between LGO produced horizontal coordinates and APPS produced horizontal coordinates. In the figure, x axis shows number of points (1199) and y axis shows differences in meters.
4 3 2 1 0 -1 -2 -3
APPS
Fig. 8. Differences between LGO produced vertical coordinates and APPS produced vertical coordinates. In the figure, x axis shows number of points (1084) and y axis shows differences in meters.
Although differences between standard RTK methods and online processing services may seem large, we would like to point out that in this project, o1 s of data is collected at each point from a moving rover during the hydrographic portion of the survey. If each point was occupied for a longer interval, the results would most likely improve for both horizontal and vertical coordinates. In a terrestrial RTK survey, a higher number of data epochs may be collected, which allows a better chance of getting integer ambiguities determined properly. When using GNSS methods, height result errors are typically 2–3 times greater than horizontal coordinate error (Berber et al., 2012). While processing the data using online data processing services, we experienced that if a data set is not consistent throughout, i.e., if there are changes in observation number, or if observations do not fit the expected order defined in the header, online data processing services return error messages or do not process the data at all. It means that these services may be unforgiving when it comes to RINEX file format. Additionally, because short period of data (1 s data) is processed in this study, in some cases these services could not solve integer ambiguities (see Section 2) and as a consequence erroneous results are generated; for instance, 10 m or more errors with elevation determinations. These erroneous results are manually removed from the data set. As a result, the number of observations used is slightly different with each service. Quality of data is the determining factor for the output results; yet, as mentioned already in this paper the precision of online GNSS data processing services is investigated. Hence, regardless of data quality, these services should return the same results. Meaning that if data quality is good, then all services should return well processed data. Also, if data quality is not good, it is expected that these services might yield some erroneous, but similar results. This means that here relative precision is compared among these services. Root Mean Square (RMS) differences are tabulated for each service in Table 2. As can be seen, the values are in the order of decimeters to meters. The RMS differences shown in Table 2 appear large because positions computed using online kinematic GNSS data processing
Table 3 Biases exhibited by online processors (m).
CSRS GAPS APPS
Horizontal
Vertical
0.08 0.84 2.77
0.14 1.22 1.24
will not be optimal until the carrier phase ambiguities have converged. For points with shorter occupation times, positions will effectively be calculated using only the pseudorange observations (see Section 2). Allowing the GNSS antenna to occupy a point for a longer period of time may make it possible to resolve the ambiguities required to recover positions using the more precise carrier phase observations. As mentioned above, in this project, only 1 s data were collected. It means that for the solutions of coordinates of these points pseudorange observations are used. As well known, use of pseudorange observations yields meter level precision, which is in agreement with Table 2 values. It appears that the differences in the above graphics exhibit systematic positive and negative shifts which mean that some of the processing are biased. In order to get additional insight about the performance of the online processors in terms of their biases and precision, we calculated the average differences and removed these average differences from the data sets and then calculated the MSE again which is equivalent to calculating the variance of a sample. As a check, we recomputed the RMSE values by squaring the bias and adding the variance, and compared these results against the RMSE values listed in Table 2 and these differences gave the biases which are listed in Table 3. As can be seen with the results in Table 3, the outputs produced by these online processors are biased. In terms of size of biases, values in Table 3 are aligned with the results of Table 2 which is another indicative of power of the processors used by these services.
4. Conclusions Precision of three online GNSS data processing services (CSRS PPP, GAPS and APPS) are compared to RTK data solutions from a base station for hydrographic surveying purposes. In this study, kinematic data collected at a rate common for hydrographic surveying (1 s) is processed using these three online processing services and compared to RTK data collected using a base station. The results of RMS error are in the order of decimeters to meters for both horizontal and vertical coordinates. Among the three services, CSRS produced the best results and APPS produced the worst results. Further analyses on biases portrayed the same picture in terms of power of the processors used by these services. However, the source of these large differences is likely the high rate of data collection, meaning less time for the antenna to occupy a single point. Less occupation time likely does not allow time for carrier phase ambiguities to converge, and as a consequence pseudorange observations are used for coordinate determinations. Use of pseudorange observations typically results
M. Berber, W. Wright / Ocean Engineering 112 (2016) 335–339
in meter level precision, which is seen in the results obtained from the three services. Depending on specific project requirements, the precision provided by these online services would not be good enough for most hydrographic survey purposes. If better precision is needed, it is our recommendation that a surveyor use either a traditional RTK base and rover configuration, or allow for a longer period of data collection.
Acknowledgments Authors would like to thank CSRS PPP team especially Pierre Tetreault for their help with processing the RINEX data. Our thanks also go to Marco Mendonca for his help to process the RINEX file using GAPS software, and Seafloor Systems for equipment support.
339
References Berber, M., Ustun, A., Yetkin, M., 2012. Comparison of accuracy of GPS techniques. Measurement 45 (2012), 1742–1746. Ebner, R., Featherstone, W.E., 2008. How well can online GPS PPP post-processing services be used to establish geodetic survey control networks? J. Appl. Geod. 2, 149–157. Ghilani, C.D., Wolf, P.R., 2007. Elementary Surveying—An Introduction to Geomatics, 12th ed. Pearson Prentice Hall, New Jersey. Ghoddousi-Fard, R., Dare, P., 2006. Online GPS processing services: an initial study. GPS Solut. 10, 12–20. Kouba, J., Heroux, P., 2001. Precise point positioning using IGS orbit and clock products. GPS Solut. 5 (2), 12–28.