Identification of Feature Details in Land Extents Using Handheld GPS

Handheld GPS equipment are becoming affordable and vendors are presently loading them with additional features such as easy data capture, data transfer, easy tracking and analysing capabilities. In order to ascertain the mapping capability using handheld GPS, a case study application was carried out mapping a known land extent and its features using several handheld GPS units of the same make and having same parameters. The mapping exercise was carried out by trained GPS operator groups who surveyed the same area approximately at the same time. GPS survey data were plotted on the served map of the project area which is a land extent of 6.6km2 close to Homagama. Surveys were carried out measurements with varying spatial resolutions. Survey results indicate the issues faced by the GPS operators when maps were prepared with measured data. Results and discussion attempts to enlighten GPS users of the limitations pertaining to mapping and also the accuracies of the present paper associated. The survey used Garmin GPS 12 and Magellan 600 handheld units. Results and accuracies of two or more units when operated parallely are discussed in this paper. Common problems such as signal coverage in the open and under canopy, plotting capability of smaller features to an acceptable level are also discussed. This work was limited to the measurements made in the horizontal direction and therefore no attempts were made to measure elevations.


Introduction
In the modern world, identification of land feature details has become more important than in the past. This has reached a high level of importance due to the advances in faster travel options; breakthroughs in telecommunication developments in data transmission over wired and wireless means etc. Feature data such as buildings, water bodies, forests, locations of tourist significance and sites which are environmentally sensitive, business centers, office locations, roads etc; are some of the land features that require identification. Requirements of feature identifications vary based on a variety of importance such as planning, design, detailed computation, management, knowledge etc. Therefore, the need to know a land feature is always linked to the identification of location and its feature attributes. In a rapidly advancing: world, the location requirements would be to identify with reference to a Global Referencing System. [1] Fulfilment of above requirements of identifying a location can be done using many methods. Conventional surveying, satellite images, aerial photography or paper map approximations, navigational satellite systems are some of them. Global Positioning Systems (GPS) incorporate a cluster of satellites to identify the position of given feature with reference to a global referencing system. In order to identify a location using GPS, a user requires to position the GPS equipment at the target feature. GPS systems ranging from precise RTK (Real-Time Kinematic) to handheld, provide location coordinates with varying degrees of accuracy. The RTK-GPS provide highly accurate data comparable to Total Stations.
The differential GPS accuracies are reported as centimetre level [3]. The handheld GPS receivers have been reported as quite accurate with horizontal accuracies ranging from sub meters, a few meters, to many tens of meters but not as accurate as RTK.
[4] Using four low cost handheld GPS indicated that there were significant differences in accuracy and precision in measuring coordinates. Manufacturer reporting accuracies and the rapidly decreasing prices of handheld GPS units have made them very attractive for most people to use them for position identifications. Today's handheld GPS are filled with many features such as easy data capture, data transfer, increased storage, easy tracking, capturing locational attributes in text or photographic image etc; [2]. The accuracy of GPS measurements depend on many factors such as availability of satellites, canopy, waiting time, atmospheric conditions etc, [5,6]. Data logging GPS units weighing as little as 24g are possible because of spare processing power and memory within today's GPS engine modules. GPS speed calculations have been indicated as about 10 times faster than a car odometer [8].
In this backdrop it is necessary for GPS users to understand the accuracies that are to be expected when carrying out field data collection using GPS. With the Handheld GPS becoming affordable, there is a greater likelihood that the users would be with minimum knowledge and may not be well conversant even to correctly carry out GPS parameter settings. Therefore, it is important to identify the results and issues when collecting GPS feature data under such circumstances.
Present work carried out a Handheld GPS field survey with 5 teams of GPS users who had a brief exposure to what is a GPS, how it works, how to incorporate appropriate settings and how to carry out recording of field data. Each group was used to collect field data pertaining to a common spatial extent. The work was carried out with the use of two types of receivers namely Garmin 12 and Magellan 600. A critical comparison of data was performed to assess the methodologies to be considered during similar field work.

Objective
Objective of the study is to identify the accuracy of handheld GPS field measurements through a repeated set of measurements made by a set of average GPS users.

Study Area
The study area located in the Western Province of Sri Lanka and approximately 21 Km South East of Colombo, between 60 50' 38" and 60 50'

Methodology
Two types of GPS receivers were used for field work; four Garmin GPS12 and one Magellan 600 units. Nominal position accuracy is 15m (RMS) for GPS Garmin 12 and Magellan 600 has a nominal accuracy of 7 m. The 1:10,000 scale Survey Department georeferenced map was the data layer that was used to establish the reference data layer.
The study area premises plan surveyed and plotted to the scale of one chain to an inch was also available to supplement base data [7]. 1:10,000 and 1 chain to an inch maps were georeferenced and overlayed to verify features in each map and then to digitize a base map. Survey map details were physically checked with field measurements of length and breath of buildings and it was identified that there was a maximum difference of 2.3 m and average difference was approximately 0.1 meters. The maps used for the study are in Table 1.
GPS survey of the temple premises, its boundary and buildings were assigned to five groups. Fieldwork groups consisted of a set of personnel who had very limited exposure to formal training on GPS. Groups 1-4 carried out the feature survey using Garmin 12 units and Group 5 carried out the same survey by using both Garmin 12 and Magellan 600 units at the same time.
Details of 1 chain to an inch map were verified by carrying out tape measurements of the buildings. Overall flow of the study is shown in Figure 2. GPS Field survey was conducted over a few weeks during daytime. Field data collectors were given a general instruction to pickup details of buildings, boundaries, roads and other important features pertaining to the temple premises. GPS settings used for the study are in Table 2. Data collected by the groups were entered into a GIS Database.
The present work consisted of carrying out measurements of Northings and Eastings. Therefore, the accuracy reported in this work consists of the accuracy in Northings, Easting and hence in the horizontal direction only. Accuracy value is the deviation of a measurement with reference to the base map.

Results
Field measurements of five groups were plotted on base map to view and compare the features in the study area. Field measurements were compared for the positional accuracy of capturing point information both as individual groups and also as average of the sets of data from five groups. Figure 3 shows the maps prepared with the six sets of field data collected by the five groups, indicating the deviations from one another. In this study though each group had collected over 111 data points, only 40 points with reference to clearly identifiable base map points were taken for accuracy comparison shown in Figure 4. Figure 5 shows the clustering in the plot of Northings and Easting accuracies pertaining to each observation.
The entire set of data were plotted to identify the horizontal accuracy variation with percentage of the number of times the measurements were taken (Figure 6). It was noted that the horizontal accuracy was less than 14 meters for 80% of the time whereas for 90 % of the time the error was less than 20 meters. The Horizontal accuracy was less than 34 meters for the entire set of readings. The accuracy range in the Northings and Eastings when compared with the base map indicated that the Northings accuracy was between -23 and +34 meters, whereas the Eastings accuracy was between -16 and +34 meters. The points measured by Garmin 12 and Magellan 600 were plotted (Figure 7) to assess the and (Table 3) The points that were representing building outline and perimeter of the premises were plotted and the centroids of each polygon drawn by individual groups were computed. Centroid computation done for the base map polygons was compared with group measurements. Figure 8 shows the horizontal accuracy of the centroid when compared with the polygon area. The centroid accuracy for majority of data were m though there were several instances the value reached +20 m and -50m respectively, including a significant change in the field measurement when compared with the base map Polygon areas were computed for the existing buildings and also for several moderate type polygons which were created by identifying the measurements and the survey plan coordinates. The polygon area comparison is shown in Figure 9. Percentage Accuracy is the percentage accuracy of either the measured area   Figure 13 indicate that as expected a greater accuracy could be noted in the case of no canopy, scattered canopy locations also showed a reasonable clustering of readings. The cluster group measurements were observed as shown in Table 4.
In order to identify a map developed with the GPS field measurements, average readings from all groups were taken after removing points which indicated very clear differences when compared with the other group measurements. The field data, plot prepared with the averaged data set is shown in Figure 14. The field plot indicates the differences observed when measuring buildings, roads and boundaries.  had to be reduced to 40 for comparison. This leads to difficulties in the averaging of the observations. Therefore, there is a clear need of an introductory training on the identification of points to capture field information. GPS readings from all groups which were measured on random days and random daytimes showed capability of identifying the approximate order of positioning. Results indicated to the possibility of improving the plotting of collected positional data through additional field notes, sketches and pictures. There was a significant clarity of identification with respect to the measurement of regular features than with irregular boundary details such as of perimeter of ground or premises.
6.2 Positional accuracies from the field measurements showed that the values of Northings, Eastings and the resultant position consisted of errors that were on the high side when compared with the nominal position accuracy indicated by the manufacturers. This may be due to the number of points used for this comparison or due to the field performance of each equipment. The accuracy values of measurements indicated a clustering around the origin though a tendency to cluster in the all negative and all positive quadrants could be observed.
Centroid measurements were computed in order to identify the shift in position of a particular feature. Values computed showed the general shift only. Azimuth readings may have to be captured to find directional changes or rotations.
6.3 Area, Perimeter and length measurements indicated the accuracy level that could be achieved in case of capturing polygon or line features. Therefore according to study results, the accuracy of polygon area measurements become lower than 50 % on for polygons greater than 6000 m2 in area. A similar observation can be made in case of the perimeter. Therefore, these could be used to assess the accuracy of features captured by the handheld GPS and make decisions on the field data collection.
6.4 Comparison of Garmin 12 and Magellan 600 did not show a significant difference in horizontal data capturing. This may require more field information for these results to be conclusive. However these data can be used as a guideline for future activities. As the study comparatively studied only the horizontal positioning, it may be necessary to study other features such as vertical accuracies, tracking, data storage, transfer, etc; prior to making final decisions on the type of an equipment.
6.5 Comparison of GPS performance recorded in relation to canopy cover and the attempts to identify notable behaviour lead to the observation that during this study, the no canopy had shown greater precision and accuracy. Scattered canopy measurements indicated a clustered data set with m accuracy through many outliers were also present. As it is important to identify the effects of canopy on the field measurements, a further study with more parameters such as weather, satellite coverage could be helpful for conclusions.
6.6 The field map produced with the use of averaged observations and screened for outliers showed a map with shape and locational disparities having accuracy values as indicated in the results. This provides a manager an opportunity to compare an averaged map with base data to assess the output with respect to the resource management requirements. Also the study through a set of group observations shows the deviations that could be expected during different measurement efforts. The comparison of 1:10,000 maps and 1:50,000 maps show that handheld GPS may be opted by resource managers who would find the scale of 1:50,000 as suitable for their work. However, managers with good careful assessment capabilities can capture better information from the plots to a scale of 1:10,000 along with possible ranges of sensitivity noted in the field measurements.

Conclusions
7.1 A comparison of field data capturing efforts by a set of average users indicated the need to provide adequate introductory training on the identification of feature details, canopy cover, regular shapes etc.
7.2 The accuracy of horizontal positioning that was identified during the study was 20m for 90%of the time and 14 m for 80%of the time.
A 5 meter or better accuracy would be observed only for 35% of time.
7.3 The selection of features for detailed conclusions such as area or perimeter should be carefully carried out by understanding the accuracies that would be achieved from such measurements. This study results could be used as a guideline.