Northern Ontario Development Agreement, Northern Forestry Program,
NODANOTE #18
Accuracy and Precision Tests using Differential GPS for Natural Resource
Applications
by Delio Tortosa and Paul Beach
Table of Contents
1.0 INTRODUCTION
2.0 METHODS AND MATERIALS
2.1 Digital Basemaps and Desktop Mappping Software
2.2 Non-Differential GPS (Standard GPS)
2.3 Post-Processed Differential GPS (PP-DGPS)
2.4 Real-Time Differential GPS (RT-DGPS)2.4.1 United States Coast Cuard (USCG) Real-Time DGPS
2.4.2 Real-Time DGPS using UHF Radio Modems2.5 GPS Data Gathering
3.0 RESULTS
3.1 Accuracy and Precision
3.1.1 Non-DGPS (Standard GPS)
3.1.2 Post-Processed DGPS
3.1.3 Real-Time DGPS using UHF Radio Modems
3.1.4 Real-Time DGPS using the United States Coast Guard
3.1.5 Elevation3.2 Field Trials
3.2.1 Ground Application
3.2.2 Airborne Application
4.0 SUMMARY AND DISCUSSION
4.1 Non-Differential GPS
4.2 Post-Processed and Real-Time DGPS
4.3 United States Coast Guard Real-Time DGPS
4.4 Forest Access Road Mapping
4.5 Airborne Tracking and Mapping
5.0 CONCLUSIONS AND RECOMMENDATIONS
5.1 GPS and DGPS Accuracy
5.2 Ground and Airborne Mapping and Tracking Applications
5.3 Application of Real-Time DGPS for Natural Resource Management
5.4 Potential Forestry and Natural Resource Applicactions5.4.1 Timber Management/Environmental Sensitivity
5.4.2 Multiple Land-Use Conflicts
Acknowledgements
Accuracy and Precision Tests using Differential GPS for Natural Resource Applications
Category: Technology Development & Transfer
Keywords: GPS (Global Positioning System), Differential GPS (DGPS), Real-Time DGPS, Telecommunications, GIS (Geographic Information Systems).
The use of Global Positioning System (GPS) receivers is gaining widespread use in forestry and many other natural resource applications. Although there have been a number of studies to determine the effectiveness of GPS in natural resource applications, there have been few studies to determine the actual level of accuracy of code-correlating GPS receivers when compared to accurately surveyed control points.
Code-correlating receivers are those which have been typically used in forestry applications and have manufacturer-specified accuracies typically ranging between 15-25 metres (standard GPS positioning); in differential mode (differential GPS positioning or DGPS) the manufacture-specified accuracies range from 1-5 metres (GPS receiver survey, GPS World, 1994).
As part of NODA Project 4226, 'Application of Real-Time Differential GPS and Real-Time Tracking for Fire and Resource Management (Tortosa, in press), baseline studies were conducted to determine the level of accuracy of the code-corellating receivers (Garmin SRVY II) to be used for the project. This technical report is a summary of the accuracy tests undertaken and some preliminary results from typical field applications.
2.1 Digital Basemaps and Desktop Mapping Software
One of the principal reasons for using GPS in natural resource applications is to gather position information on geographic features to be represented by points, lines, and polygons. To this end GPS positional data was stored in a digital format and later represented on appropriate digital maps at the various field locations. National Topographic Series (NTS) digital basemaps at 1:250,000 scale were used to represent areas of regional extent, and digital Ontario Base Maps at 1:20,000 were used at a local level. Desktop mapping software (QuikMAP, DOS/Windows) and a GPS linkage software (QuikELINX, DOS/Windows) were used to automatically convert and plot the GPS point, line, and polygon information onto the appropriate basemap.
2.2 Non-Differential GPS (Standard GPS)
Standard GPS positions were collected in order to compare the result with Real-Time DGPS (RT-DGPS) and Post-Processed DGPS (PP- DGPS). For this test, a Garmin SRVY II GPS receiver was set up on a surveyed position (ie. a control point) and both averaged and dynamic position data were collected.
2.3 Post-Processed Differential GPS (PP-DGPS)
Post-Processed DGPS accuracy tests were conducted in order to compare these results with those derived from RT-DGPS. Two Garmin SRVY II GPS receivers were used; one for the Base Station (Control Point #1, figure 1) and one at the remote location (Control Point #2, Figure 1).
2.4 Real-Time Differential GPS (RT-DGPS)
Real-time DGPS is the method by which differential GPS accuracies are obtained on a real-time basis. Two methods were employed to test the real-time differential accuracy of the GPS receivers: 1) using a United States Coast Guard radio beacon transmitter and 2) using Ultra High Frequency (UHF) radio modems.
2.4.1 United States Coast Guard (USCG) Real-Time DGPS
The United States and Canadian Coast Guards are developing a RT- DGPS service throughout the Great Lakes, Eastern and Western coasts of North America, and the Gulf of Mexico, which provides the required differential correction to DGPS-capable receivers which are within range of the transmitting radio beacons. This is primarily for use in navigation, but can also be used on land within range of the radio beacon transmitter.
In order to use this service, the GPS receivers have to be designed to accept the appropriate digital format of the data. This format is referred to as the RTCM-104 format (Radio Technical Commission for Maritime Services Special Committee No. 104). In addition, a GPS receiver has to be connected to a radio beacon receiver (or radio modem) which has been tuned to the appropriate USCG frequency. For the accuracy tests, the GPS antenna was placed on the surveyed control point (Figure 1). Similarly a radio beacon antenna was placed on the roof in a nearby location. The radio beacon receiver was tuned to the Whitefish Point transmitter frequency (318 KHz).
2.4.2 Real-Time DGPS using UHF Radio Modems
For locations which are out-of-range of the USCG radio beacon transmitters, it is possible to duplicate the above process using a base station GPS receiver and radio modem to transmit the required DGPS correction at a pre-determined frequency to a remote GPS receiver connected to a radio modem.
The GPS receiver at the base station must have the capability of producing the DGPS corrections in the RTCM-104 digital format. The base station GPS antenna was situated at a surveyed position (control point #1, figure 1), and the remote GPS antenna was located at a second surveyed position (Control Point #2, Figure 1). The base station GPS receiver should be capable of tracking up to 8 GPS satellites in order to provide full satellite coverage.
For each of the previous DGPS and GPS methods described, two types of data were gathered: averaged positions and dynamic positions.
Averaged positions represent a series of averages calculated at 15, 30, and 60 minute time intervals (Table 1, Figure 2) at varying times of the day over a period of two months. This sampling approach was designed to introduce randomness to the geometry of the GPS constellation so that the results would have general application. For all methods, the deviation of the average (in metres) from a known control point were noted for the Northing, Easting, and Elevation.
For the RT-DGPS using radio modems and the Post-Processed DGPS we found that 30 and 60 minute averages were unnecessary and did not improve the results of the 15 minute averages; therefore, no detailed sampling was undertaken for 30 and 60 minute durations.
Dynamic positions represent single DGPS and GPS positions plotted relative to a control point. The GPS data was transferred using the GPS linkage to the mapping software and the number of positions falling within a specified radii of the control point were tabulated and plotted as a cumulative frequency distribution (see Figure 3).
The cummulative frequency distribution of dynamic GPS positions indicate that 90% of the positions fall within 30 metres of the control point (Figure 3).
Averaged Positions show improvement in precision and accuracy from 15 to 30 and 60 minute sampling intervals (Table 1). The data progressively approaches a normal distribution for the 30 and 60 minute sampling intervals (Figure 2).
The cumulative frequency distribution of the dynamic DGPS positions indicate that 90% of the DGPS positions fall within 10 metres of the control point (Figure 3). A frequency distribution of the departure from the control point for the Northing, Easting, and Elevation (Figure 4) indicates a normal distribution for a 15 minute sample (780 positions).
The averaged positions display a symetrical distribution with an accuracy of 0 +/- 1 to 2 metres for the 10 minute sampling interval. This level of accuracy does not improve with 30 and 60 minute sampling intervals since the accuracy represents the hardware limits of the GPS receiver (not a submetre instrument).
3.1.3 Real-Time DGPS using UHF Radio Modems
The cumulative frequency distribution of the dynamic RT-DGPS positions indicate that 90% of the positions fall within a 10 metre radius of the control point (Figure 3). The cumulative frequency distribution is comparable to that obtained from Post- Processed DGPS.
The distribution of RT-DGPS positions about the control point approximates the shape of an ellipse with the long axis oriented in a north-south direction (Figure 1). This shape is assumed to be caused by the geometry of the GPS constellation at this location in the northern hemisphere.
Averaged positions display a symmetrical distribution for the 15 minute sampling interval with an accuracy of 0 +/- 1 to 2 metres. This accuracy does not improve with 30 and 60 minute sampling intervals for reasons previously mentioned.
3.1.4 Real-Time DGPS using United States Coast Guard
The cumulative frequency distribution of dynamic RT-DGPS positions indicates that 80% of the positions fall within a 10 metre radius of the control point (Figure 3).
The distribution of USCG RT-DGPS positions about the control point approximates the shape of an ellipse with the long axis oriented in a north-south direction (Figure 1). The ellipsoidal distribution is assumed to be caused by the geometry of the GPS constellation at this location in the northern hemisphere.
The averaged positions for the 15 minute sampling interval have an accuracy of -4 (Easting) and -2 (Northing) with a range from 0 to -6 metres (Table 1, Figure 2). This accuracy does not improve with 30 and 60 minute sampling intervals for reasons previously mentioned.
There is an apparent shift of the position of the data cluster and the position averages relative to the control point (Figure 1). This negative shift is not present in the data when other DGPS methods are used. The shift to negative values is attributed to error introduced by the NAD83 (USCG Control Point) to NAD27 datum conversion used by the GPS receiver. The NAD27 1976 adjustment used for the surveyed control points is not included in the NAD27 Canada conversion used by the Garmin SRVY II (or other similar GPS receivers). For the real-time DGPS test using radio modems, both control points are based on the NAD27, 1976 adjusted datum and therefore no apparent shift is observed.
Although determination of elevation was not required for the field applications, a number of test were made to determine the accuracy of the code-correlating receiver.
Averaged elevations, for non-DGPS, over 15 minute sampling intervals display a crude normal distribution with a mean of +3 and a range between -40 and +60 metres (Table 1). For the 30 and 60 minute sampling intervals (i.e. 30 minutes: +5, +/-25 m.; 60 minutes: +8.5, +/-25 m.), the precision improves, but a longer sample interval is required to attain greater confidence (120 minute sample interval). The determination of elevation using standard GPS is subject to higher variability than horizontal positions.
Averaged elevations for Post-Processed DGPS using 5-10 minute sampling intervals display an accuracy of 0 +/- 2 metres (Table 1). These results reflect similar accuracies as acquired for averaged horizontal positions.
Average elevations for the RT-DGPS using UHF radio modems and a sampling interval of 15 minutes, range from 0 to -10 metres below the control point. The negative skew cannot be explained.
Averaged elevations for the U.S. Coast Guard RT-DGPS using 15 minute sampling intervals range from 0 to 10 metres above the control point. This may be due to the NAD83 to NAD27 conversion error using the GPS receiver NAD27 Canada datum previously described.
A real-time DGPS test using UHF radio modems was undertaken on the Abitibi-Price Camp 34 site north of Iroquois Falls during August, 1994 (figure 5). The objective of the test trial was to determine the range of operation and accuracy of the RT-DGPS method using a temporary GPS base station and remote GPS receiver mounted on a vehicle.
The temporary base station was set-up in one of the Abitibi-Price field offices. The GPS antenna was placed on the roof of the office building with an unobstructed view, and a radio modem base station UHF antenna was placed at a height of 10 metres above ground. A temporary control point was established by taking two one-hour averages and averaging the two to arrive at a final position.
Several of the accessible logging roads which traversed the property were driven during a one day period. The RT-DGPS data was then transferred to the mapping system as polylines and overlaid on the OBM digital basemaps provided by Abitibi-Price.
The results indicate that in low relief areas such as those at Camp 34 it is possible to transmit RT-DGPS corrections over distance of up to 50 kilometres with an antenna height of 10 metres using an radio frequency (RF) amplifier of 35 Watts. One can expect a further range of operation with higher antenna heights.
The resulting RT-DGPS positions and polyline fall within and parallel to the boundaries of the roads on the digital OBM (Figure 5). This reflects a level of accuracy for RT-DGPS which falls within the specifications for the digital Ontario Base Maps (+/- 10 metres). This is consistent with the results of the accuracy tests.
The Kirkwood Forest, situated 10 kilometres north of Thessalon, was selected as the site to complete aerial trials using the Forest Pest Management Institute (FPMI) aircraft. Digital Ontario Base Maps were available for the area, and there was a grid of forest access roads which could serve as a guide for the aircraft. The intent of the field trial was to test the accuracy of the Garmin SRVY II GPS system and compare the results to the standard accuracy of an OBM.
A differential base station for the Garmin SRVY II GPS receiver was located near a landing strip, and the GPS and UHF antennae were located on the roof of a trailer. A reference control point was established using the method described previously.
The aircraft flew within a radius of 10 kilometres of the base stations without any loss of the UHF signals or differential GPS corrections throughout the survey. Four forest access roads oriented in a north-south direction were flown twice in a north and south direction.
Using the GPS/desktop mapping system, the RT-DGPS data was transferred to the desktop mapping software, converted into points and polylines and overlaid on the OBM road network. The results indicate that the nearly coincident flight lines display minor differences with the position of the OBM secondary.
The results of our field testing indicate that 90% of the uncorrected GPS positions fall within 30 metres of the surveyed control point. Selective Availability (S/A), although generally quoted as ranging between +/- 50 metres, is cyclical over time so that there are periods of short duration when S/A may approach zero.
With sampling intervals of 30 and 60 minutes the GPS positions approach a normal distribution with an average close to 0 (Figure 2). This is useful for remote forestry operations where no survey control is available, and can be used to establish survey control which meets the 1:20,000 OBM horizontal position accuracy requirements for the representation of geographic information (i.e., roads). A minimum of 60 minute sampling interval with a PDOP ranging from 1.0-1.5 and tracking 6-8 satellites, is recommended in order to achieve this accuracy. For elevation, a minimum 2 hour sampling time with similar PDOP and satellites, is recommended in order to meet the 1:20,000 OBM vertical position accuracy.
4.2 Post-Processed and Real-Time DGPS
The accuracy of both these methods is similar since both techniques are alike: for post-processed differential GPS the data correction is completed in the post-processing software on the computer, whereas for real-time DGPS the GPS data correction is done immediately in the firmware of the GPS receiver. A 15 minute sampling interval is suitable in order to achieve a 0 +/-1 metre level of accuracy, and there is no improvement with a 30 and 60 minute sampling interval since this is the accuracy limits imposed by the GPS receiver hardware (i.e. not capable of submetre accuracy).
4.3 United States Coast Guard Real-Time DGPS
Tests demonstrate that accuracies in the order of +/- 5 metres can be achieved over 80 kilometre baselines, however, averaged RT-DGPS positions will be skewed upto several metres due to the NAD83 to NAD27 datum conversion error previously discussed. A 15 minute sampling interval is suitable to achieve this level of accuracy, and there is no clear improvement with 30 and 60 minute sampling intervals. The radio beacon system can only be used in- land for about 50 Km and is dependent on the local topography and relief. Only a few radio beacons in the United States and Canada are transmitting the DGPS corrections, so that this method is currently not widely available to the public.
4.4 Forest Access Road Mapping
Forest access road mapping at the Abitibi-Price Camp 34 demonstrates the accuracy of the real-time DGPS method, and by analogy, the post-processed DGPS method, when applied to a 1:20,000 OBM with both original and SAP-derived road locations (Figure 5).
Figure 5 shows a portion of two OBM map sheets with the OBM road on the southern sheet continuing on the northern sheet as a road derived from information provided by supplementary air photos (SAP). The DGPS-derived road, represented as a dashed line, coincides with the centre of the OBM road. The SAP-derived road approximately coincides with the DGPS-derived road. In other parts of the area where the DGPS-derived road was mapped in both directions, the SAP-derived road is displaced from the DGPS- mapped road by up to 50 metres in places.
These results demonstrate that DGPS methods are just as accurate as SAP-derived forest road mapping. DGPS road updates can be completed very quickly and accurately providing forest products companies with a rapid, low cost alternative to the use of supplementary air photos or other mapping techniques.
4.5 Airborne Tracking and Mapping
Aerial trials in the Kirkwood Forest north of Thessalon resulted in continuous reception of the RT-DGPS corrections throughout the survey. An overlay of the DGPS-derived flightlines on the 1:20,000 OBM for the area indicate a small displacement between the GPS track and the OBM secondary roads (figure 6).
Although navigation was not possible due to the location and setup of the GPS receiver (not accessible to pilot), the main limitation for accurate real-time DGPS navigation is the low resolution of the course deviation index (CDI) scale bar on the instrument (+/- 50 metres). In order to be fully effective for RT-DGPS navigation, a course deviation index lightbar with a resolution of +/- 1 metre is required.
For aerial spraying applications in boreal forest, a DGPS system with an accuracy of +/- 10 metres is sufficient to improve the current navigation methods (ie. use of a navigator aircraft), provide a map showing the spray block, flight lines and spray swath, and provide an audit check which conforms to the accuracy of a 1:20,000 Ontario Base Map.
5.0 Conclusions and Recommendations
Both GPS and DGPS static position data approach a normal distribution about a control point. By using averaged positions it is possible to achieve increased accuracies. Averaged GPS positions fall within +/- 5 metres of a control point; single position DGPS accuracies fall within +/- 10 metres of the control point. This provides a benchmark test for GPS receivers which complies with the specified horizontal accuracies required for 1:10,000 and 1:20,000 OBM.
Given the limited variation in the manufacturers' specified accuracies for GPS receivers (GPS receiver survey, GPS World 1994), we would expect that other code corellating GPS receivers similar to the Garmin SRVY II should exhibit comparable GPS and DGPS accuracies since each GPS receiver uses essentially the same mathematical functions to calculate a position. However, we recommend that prior to undertaking any field work with a GPS receiver, where the data is to be used on Ontario Base Maps, that benchmark tests similar to those done for this project be completed in order to assess the accuracy of the instrument.
5.2 Ground and Airborne Mapping and Tracking Applications
From the airborne and ground trials completed during the project, the real-time DGPS accuracies obtained using the GPS/desktop mapping system compare favourably with the accuracy of the Ontario Base Maps at 1:20,000 scale.
The trial tests also demonstrated that the level of accuracy of the GPS/desktop mapping system compares favourably with higher accuracy GPS receivers used for aerial spray applications for map scales of 1:20,000 and 1:10,000 (Tortosa, in press).
For companies and governments involved in aerial spraying and other natural resource applications, both the accuracy achieved with post-processing and real-time DGPS methods, and the map output would serve to easily depict the location of the roads, flightlines, spray blocks, and spray swaths on a map base appropriate for the survey (figure 7). The ability to provide accurate maps of the surveys while in the field would allow resource professionals to achieve a high level of confidence in the survey and offer the opportunity to correct any errors or omissions. Finally, there is a cost benefit to government agencies since the audit aircraft for aerial spraying can be eliminated.
5.3 Application of Real-Time DGPS for Natural Resource Management
Although technically feasible, the necessity of using real-time DGPS is based on a number of factors, including:
A) The availability of, and access to, USA/Canada Coast Guard differential corrections. The Coast Guard DGPS program is still under development and the range is restricted to the Great Lakes and near shore areas.
B) The extent to which navigation is a critical component to the application. Many natural resource professionals have a requirement for DGPS accuracy to access and locate forest values which require real-time DGPS navigation.
C) The number of applications which require Differential GPS. If a large number of resource professionals require differential GPS accuracy, then it may be more efficient to use real-time DGPS for all applications, eliminating the need for continuous collection of GPS data for post-processing.
D) The topographic relief and nature of the terrain. The use of Ultra high frequency radio spectrum for real-time DGPS results in a short range in many areas of the Canadian Shield.
E) The accuracy requirements of a particular application. Many natural resource applications can achieve the required accuracy using standard GPS positioning (ie. mapping fire perimeters), so that there is no need for DGPS.
We believe that a flexible approach to the implementation of DGPS technology is the most prudent method of incorporating this new technology into an organization. An easily transported real-time (and post-processing) DGPS system can meet specific application requirements under many terrain conditions. These may be either permanent or temporary operations. The GPS/desktop mapping system can be relocated quickly and easily into remote field operations, and is capable of establishing a reference position (control point) at the remote location which meets the accuracy requirements for Ontario Base Maps.
5.4 Potential Forestry and Natural Resource Applications
5.4.1 Timber Management/Environmental Sensitivity
New regulations and legislation in Ontario will require much stricter requirements in timber harvesting operations. Currently in areas such as the Abitibi-Price Camp 34, the layout for timber harvesting is determined by pace and compass which are used to demarcate the boundaries of selected areas. These tend to be straight line, simple polygons which are relatively easy to layout in areas of low topography. In addition, the layout of these blocks does not require any detailed knowledge of stand- type. New requirements for selective logging practices and the designation of areas for biodiversity, result in complex polygons with multiple stand type. Accurately surveying the boundaries of these polygons becomes a more complex problem which involves rapid changes in direction and accurate positioning. The use of real-time DGPS provides a technological tool to map out the required boundaries and transfer the GPS locations to the OBM/FRI map. At the completion of the harvest the same area can be re- surveyed using DGPS on the ground or in an aircraft to determine compliance with the previously established boundary.
5.4.2 Multiple Land-Use Conflicts
In areas where there are competing uses for the land, and which result in conflicting interests, geographic information systems (GIS) are being used to provide alternative solutions to meet the requirements of various stakeholders (Megasin Lake Environmental Assessment, MNR, 1993). The use of GIS results in options characterized by multiple and complex buffer zones. The practical implementation of the selected method to resolve a land-use conflict will, in turn, determine its feasibility. For example, complex buffer zones determined using digital elevation modelling and visibility analysis techniques pose a significant challenge to the layout, even when using ground based real-time DGPS surveys.
An alternative solution is to blend a GPS real-time tracking capability with real-time DGPS to produce an automated tracking system which can provide positioning information to a harvestor operator and/or to a control centre. By having the harvest block polygon defined on the desktop mapping/GIS software, the DGPS position of the harvester can be compared with the polygon boundaries in order to avoid the edge of a buffer. The buffer zone can be as complex as required, no detailed ground-based surveys need to be completed, and compliance of the harvested block is met once the harvester reaches a buffer boundary.
The author would like to thank Mr. Paul McBay and staff of the Ontario Ministry of Natural Resources' Regional Fire Centre and the Ranger Lake Attack Base for their support and assistance throughout the project; Mr. Art Robinson provided assistance during the aerial trials and information during the analysis and reporting stages of the project. Mr. George Stanclik and Mr. Erik Turk of Abitibi-Price provided digital basemaps and logistical support during the field trials in the Iroquois Falls, Camp 34. Thanks are extended to Dr. Taylor Scarr for providing operational examples of post-processed DGPS for an aerial spraying application in boreal forests.
Funding for this project was provided through the Northern Ontario Development Agreement, Northern Forestry Program.
The views, conclusions and recommendations are those of the Author(s) and should not be construed as either policy or endorsement by Natural Resources Canada, Canadian Forest Service, or the Ontario Ministry of Natural Resources.
Figure 1: Dynamic, Real-Time Differential GPS positions using the U.S. Coast Guard and UHF Radio Modems relative to second order Control Points. Garmin SRVY II GPS receivers used.
Figure 2: Histograms of the averaged DGPS and GPS positions (Easting and Northing) for 15 and 60 minute sampling intervals, respectively.
Figure 3: Cummulative frequency distributions of DGPS and GPS dynamic positions with distance from a control point.
Figure 4: Frequency distribution of the departure from the control point for the Easting, Northing and Elevation for a 15 minute sampling interval of DGPS dynamic positions.
Figure 5: Digital OBM showing both OBM and SAP-derived roads with the real-time differential GPS track overlayed.
Figure 6: Airborne real-time differential GPS positions using ultra high frequency (UHF) radio telemetry and the Garmin SRVY II GPS receiver for a portion of the Kirkwood Forest north of Thessalon.
Figure 7: Track of aircraft over a spray block determined using post-processed differential GPS. Booms on/off are not shown. Diameter of circle represent 200 spray swath (two planes). Courtesy of Taylor Scarr, Ontario Ministry of Natural Resources, Aerial spray program, June 1995.
Table 1: Data summary for 15, 30, and 60 minute sampling intervals using various GPS methods.
