EUROPEAN COMMISSION JOINT RESEARCH CENTRE Institute for the Protection and Security of the Citizen Agriculture Unit JRC IPSC/G03/P/ASI/asi D(2008)(9481) / Report Results of the field day of 8th GNSS Workshop, Dublin, 10th April 2008 Author: Aleksandra Sima Status: v1.3 Co-author: Circulation: Public Approved: Wim Devos Date: 15/05/2008 Int. ref: file://s:\fmparchive\p\9481.doc Contents 1. Introduction...2 2. Description of the exercise and results...2 2.1. General comments... 2 2.2. GNSS receivers used... 3 2.3. Description of the parcels and results... 3 2.3.1. Parcel 1... 4 2.3.2. Parcel 2... 6 2.3.3. Parcel 3... 8 2.3.4. Parcel 4... 10 2.3.5. Parcel 5... 11 2.4. Measurements of the length of a linear feature... 13 3. Results...15 4. Conclusions...15 Post: Joint Research Centre, IPSC- AGRICULTURE, TP 266, I-21027 Ispra (VA), Italy Telephone: direct line +39 0332 78 6622 Facsimile: +39 0332 78 9029 E-mail: aleksandra.sima@jrc.it URL: http://mars.jrc.it/
1. Introduction This year s field day of the GNSS workshop was a challenging event for many reasons and from many points of view. For the participants: with the very full agenda of the day, the limited time to ask questions and get the answers and the physical efforts to move around the fields for 6 hours, enduring a very variable weather quite a challenging task! For the equipment providers: to be able to present their devices in right way, answer all the questions, give an opportunity to the participants to put their hands on equipment and collect the data at the same time, sometimes having problems with the satellite signal not an easy task! For the organizers: to comfort and manage more than 90 people in the field is never an effortless duty. Despite these challenges, the field day went smoothly and the goal of the exercise was achieved. The priority during the whole field day was to answer all the questions of participants and present the features of the devices rather than collect the data, causing too few measurements to be taken for a parcel to perform a statistical processing of the results. The exercise was meant to be a hands on exercise, not an accuracy test, so the resulting measurements are presented as illustrations only. Also note that the results presented in this document are not straightforward comparable, as the individual measurements were taken with different methods, receiver settings (e.g. with or without Glonass), at different times of the day thus with a different satellite constellation. 2. Description of the exercise and results 2.1. General comments In the field exercises six companies presented their achievements: FieldFact, John Deere, Magellan, Satcon System, Topcon and Trimble. In this document only results of the 3 following equipment providers will be commented on: Magellan, Topcon and Trimble. Unfortunately, we didn t receive any data files from Satcon System. Presentations of John Deere and FieldFact were demonstration sets, oriented on the possibilities of agricultural use of the GNSS system; they did not aim to measure the area of the parcel. Internal ref: JRC IPSC/G03/P/ASI/asi D(2008)(9481) 2
The participating equipment providers were given a detailed description of the field activities before the field exercise. The description included the following points: - description of the challenges on each parcel, e.g. parcel in the forest, one border of the parcel not accessible, approximate size of the parcel, - timing and rotation of the groups on the parcels. Moreover, the orthophotomap (GSD = 0,50m) of the field day area was provided. The equipment providers were left with a free choice of the methods and settings used during the exercise. In the exercise participants were asked to measure the area of reference parcels after receiving a short introduction with instructions and considerations from the manufacturers. All the results of measurements were immediately written down on a data form, the digital data were stored as well and delivered later. 2.2. GNSS receivers used Company Receiver Link Magellan Magellan Magellan Magellan Satcon System Topcon Topcon MobileMapper CX (MMCX) hardware + DigiTerra Explorer (DE) software MobileMapper CX hardware + Fast Survey (FS) software MobileMapper 6 + DigiTerra Explorer MobileMapper 6 + Mobile Mapping SW F@rmphone GMS-2 + ArcPad GR-3 + TopSURV software http://pro.magellangps.com/en/products/ product.asp?prodid=1292 http://pro.magellangps.com/en/products/ product.asp?prodid=1292 http://pro.magellangps.com/en/products/ product.asp?prodid=1293 http://pro.magellangps.com/en/products/ product.asp?prodid=1293 http://www.satconsystem.com/en/hardware/31/ http://www.topconeurope.com/se/index.asp?pageid =99fad9ab9b8a4edcaf9384c66f196026 http://www.topconpositioning.com/products/gps/ geodetic-receivers/integrated/gr-3.html Trimble Trimble 2005 GeoXH http://www.trimble.com/geoxh.shtml Trimble Trimble ProXRT http://www.trimble.com/pathfinderproxrt.shtml 2.3. Description of the parcels and results The parcels design was significantly different from the standard pattern with open horizon, football pitch like test fields. The main scope of this year s exercises was to deal with obstructions of the horizon, causing weaker GNSS signal. Internal ref: JRC IPSC/G03/P/ASI/asi D(2008)(9481) 3
Five parcels (see Figure 1.) with the following characteristic were used for the purpose of the exercise: Figure 1. Parcels used for the purpose of the field exercises. Parcels 6 and 7 were used by John Deere and FieldFact for demonstration of the agricultural use of GNSS system. The overall performance of all receivers was estimated for each parcel on a basis of all the delivered data and treated equally regardless of the provider, method, correction etc. All the equipment providers were using external antennas. 2.3.1. Parcel 1 This is a small, elongated, parcel with a size of 0.2623ha, permanently fenced, with a forest on SE border. Internal ref: JRC IPSC/G03/P/ASI/asi D(2008)(9481) 4
Figure 2. Parcel 1 (yellow on Figure 1.) Outlines of the measurements taken on this parcel are presented on Figure 3. The influence of the trees obstructing the SE border is clearly visible on the measurements collected by Magellan. It looks like Topcon team dropped or lost one of the vertices on the SW end, which is critical when using stop&go method. This measurement was eliminated from the below described calculation. Figure 3. Outlines of the measurements taken on Parcel 1. Internal ref: JRC IPSC/G03/P/ASI/asi D(2008)(9481) 5
A bias of 5% of area of the parcel was noticed all the results were smaller than the reference area established by the Ordnance Survey with RTK measurements, however within the perimeter buffer limit allowed by the Regulation. See Table 1. for details. 2.3.2. Parcel 2 reference AREA [ha] reference Perimeter [m] 0.2623 331 Receiver area measured [ha] error on perimeter [m] MMCX DE (Egnos) 0.2578-0.14 MMCX DE (Egnos) 0.2479-0.44 MMCX FS (Egnos) 0.2502-0.36 MMCX FS (Egnos) 0.2457-0.50 MMCX FS (Egnos) 0.2412-0.64 MMCX FS (Egnos) 0.2578-0.14 0.2501 = mean area 5% = bias = approx. of the buffer 1.15m (2.8xstdev) 1 Table 1.Evaluation of performance of all the receivers, from which the data collected on Parcel 1 were made available. The biggest from all the test fields: 3.4692ha, with one (simulated) inaccessible border and open horizon all round. Figure 4. Parcel 2 (dark green on Figure 1.) a view from the side of the inaccessible border. 1 Buffer is defined as a reproducibility limit at 95% confidence level, however there were too few observations to compute it. In this situation standard deviation of the dataset multiplied by a factor of 2.8 can be used as a good approximation of the buffer. Internal ref: JRC IPSC/G03/P/ASI/asi D(2008)(9481) 6
Due to some misunderstanding the problematic border was approached in a different way than it was intended. The situation is presented on Figure 4 below: the red line represents the designed and surveyed border of the parcel, the rest of the lines are the outlines of the measurements taken. Figure 5. Parcel 2 - misunderstanding in the definition of the border: original reference border in red, the correction of the border in pink. Measurements taken are presented in blue and orange. In order to correctly evaluate the performance of receivers on this parcel, a new reference area was established by subtracting the small part of the parcel (see Figure 5.) which was omitted during the field exercises, as equal to: 3.4367ha. No bias was observed on this parcel, the approximation of the buffer is below the value of 1.5m allowed by the Regulation. Internal ref: JRC IPSC/G03/P/ASI/asi D(2008)(9481) 7
reference AREA [ha] reference Perimeter [m] 3.4367 758 Receiver area measured [ha] error on perimeter [m] geoxh (Egnos) 3.4050-0.42 geoxtr (post-processed) 3.3950-0.55 GMS-2 (+Glonass) 3.4631 0.35 MMCX FS (Egnos) 3.4679 0.41 MMCX FS (Egnos) 3.4875 0.67 MMCX DE (Egnos) 3.4009-0.47 MMCX DE (Egnos) 3.4243-0.16 3.4348 = mean area 0% = bias 1.28m = approx. of the buffer (2.8xstdev) Table 2. Evaluation of performance of all the receivers on Parcel 2. 2.3.3. Parcel 3 Parcel 3 is an elongated, 12 meters wide parcel with area of 0.2734ha and an open horizon. Figure 6. Parcel 3 (red on Figure 1.) one border marked with pegs, rest of the borders permanent. Only two files resulting from the measurements on this parcel were delivered by the equipment providers. The approximation of the buffer was computed as 0.07m! See details in Table 3. Internal ref: JRC IPSC/G03/P/ASI/asi D(2008)(9481) 8
Figure 7. Outlines of the measurements taken on Parcel 3. reference AREA [ha] reference Perimeter [m] 0.2735 481 Receiver area measured [ha] error on perimeter [m] MMCX DE (Egnos) 0.2722-0.03 MMCX DE (Egnos) 0.2747 0.02 0.2734 = mean area 0% = bias 0.07m = approx. of the buffer (2.8xstdev) Table 3. Evaluation of performance of all the receivers, from which the data collected on Parcel 3 were made available. Internal ref: JRC IPSC/G03/P/ASI/asi D(2008)(9481) 9
2.3.4. Parcel 4 Parcel located in a forest, borders marked with pegs, size of 0.1978ha Figure 8. Parcel 4 (violet on Figure 1.) the vegetation season hasn t fully started at the time of the exercise. The outlines of the measurements (shown on Figure 9.) are significantly different from the shape of the reference parcel, but surprisingly the approximation of the buffer is as low as 0.8m. Figure 9. Outlines of the measurements taken on Parcel 4. Internal ref: JRC IPSC/G03/P/ASI/asi D(2008)(9481) 10
reference AREA [ha] reference Perimeter [m] 0.1978 190 Receiver area measured [ha] error on perimeter [m] GMS-2 (+Glonass) 0.1996 0.09 MMCX FS (Egnos) 0.1886-0.48 MMCX FS (Egnos) 0.1938-0.21 MMCX FS (Egnos) 0.2015 0.19 0.1959 = mean area 1% = bias 0.80m = approx. of the buffer (2.8xstdev) Table 4. Evaluation of performance of all the receivers on Parcel 4 2.3.5. Parcel 5 This is a parcel with a forest on the southern border, size of 0.2998, one border marked with pegs, rest of the borders were permanent. Figure 10. Parcel 5 (light green on Figure 1.) forest on southern border of the field. The mean result from all the measurements is 6% smaller (biased) than the reference. The approximation of the buffer is: 1.46m still below the limit allowed by the Regulation, although very close to that limit. See details in Table 5. Internal ref: JRC IPSC/G03/P/ASI/asi D(2008)(9481) 11
Figure 11. Outlines of the measurements taken on Parcel 5. reference AREA [ha] reference Perimeter [m] 0.2998 341 Receiver area measured [ha] error on perimeter [m] geoxh (KOREC s VRS) 0.2790-0.61 geoxtr (Egnos) 0.2910-0.26 GMS-2 (+Glonass) 0.2760-0.70 MMCX DE (Egnos) 0.2864-0.39 0.2831 = mean area 6% = bias 1.46m = approx. of the buffer (2.8xstdev) Table 5. Evaluation of performance of all the receivers on Parcel 5 Internal ref: JRC IPSC/G03/P/ASI/asi D(2008)(9481) 12
2.4. Measurements of the length of a linear feature This exercise was improvised on the last moment as one team of equipment providers had cancelled its participation. A linear feature (horizontal length of 175.2m) was marked with pegs every ~20m. The length of this line was measured by the participants in 3 different ways: - with a tape, - with a GNSS receiver with stop and go method: beginning and end of the feature, logged only - with a GNSS receiver with a kinematic method: 1s logging interval. Figure 12. Pegs marking the segments of the linear feature measured during the exercise. The results confirmed that the most accurate tool for that distance was a tape (with 0.04% of the length of the feature). The measurements taken with the GNSS receiver (Trimble, GeoXT with TerraSync) with stop&go method gave 3 times larger errors (for this distance), but still on acceptable level of 0.13% of the length of the feature. As expected, the kinematic method gave significantly larger errors, on level of 1.15%. Surprisingly the distance measured by the GNSS instruments was always smaller than the reference. However it should be underline, that measuring with GNSS receiver is more time-efficient and needs less effort than using a tape. Moreover, the result can be seen immediately and a digital record of the measurement is available and possible to trace back. Internal ref: JRC IPSC/G03/P/ASI/asi D(2008)(9481) 13
Tape [m] GNSS GNSSstop&go [m] kinematic [m] 175.50 168.00 174.80 175.45 174.76 175.19 174.40 175.06 174.89 175.20 174.93 175.02 175.25 174.37 175.45 175.35 Mean [m] 175.12 173.19 174.98 absolute error [m] -0.08-2.01-0.22 relative error [%] 0.04 1.15 0.13 Table 6. Results obtained with a tape in comparison to the ones obtained with a GNSS receiver. Internal ref: JRC IPSC/G03/P/ASI/asi D(2008)(9481) 14
3. Results The poorest measurements were observed on the parcels, where the conditions of measurements (visibility of satellites) are unstable for the whole parcel (all the borders). Trees on part of the borders of Parcel 1 and Parcel 5 had a stronger impact on the result of the measurements than the forest covering the complete Parcel 4. Both results from Parcel 1 and Parcel 4 were significantly biased (more than 5%). A surprisingly good result was achieved on Parcel 3 (buffer of 0.07m on perimeter) please note, that the result for that parcel is only based on 2 measurements. All the results were below the maximum threshold buffer of 1,5 meter allowed by the Regulation. It is worth repeating that these results should be treated as illustration to the exercise only. Before using any tool for a purpose of on the spot checks a full scale test, designed along with the JRC recommendations (http://mars.jrc.it/documents/pac/8307.pdf) should be performed. 4. Conclusions We believe that the goal of the exercise has been fully achieved: - Participants of the workshop had an opportunity to get familiar with a variety of the latest GNSS equipment, - Participants had an opportunity to discuss some burning issues with the equipment providers and among each other and exchange their experiences. - Awareness was raised on the various field conditions that might influence the measurement results The overall performance of the equipment was always within the limit allowed by the Regulation, even under sometimes quite difficult meteorological conditions and dealing with very demanding parcels (small and with obstructed horizon). (end document) Internal ref: JRC IPSC/G03/P/ASI/asi D(2008)(9481) 15