Improving Accuracy: Acquisition of Roads Data in a Rural County | ||||||||||||||||||||||||||
|
Author Paul Watkins American River College, Geography 350: Data Acquisition in GIS; Fall 2008 Contact Information: 285 Ursula Drive, Sutter Creek, CA 95685, (209) 256-0742, email: watkinsp@comcast.net | ||||||||||||||||||||||||||
|
Abstract This paper evaluates the collection of GPS data in the El Dorado National Forest in Amador County California for the purpose of creating a GIS roads centerline layer. GPS data was captured using a Trimble GeoExplorer 2008 Series GeoXT collecting streaming vertices along about 60 miles of the centerline of primary Forest Service and county maintained roads. GPS data was uploaded into a GPS-enabled personal geodatabase and post processed for differential correction, using ArcPad, GPS Correct, GPS Analyst extension and ArcMap 9.3. A resulting roads layer was compared to its background aerial imagery, secondary gis data from the USDA Forest Service and Amador County Division heritage data. Statistical analysis of GPS positions was also evaluated for accuracy. Data acquired was used as a source in interpreting road location for desktop digitizing a roads layer displayable at 1:600 scale. | ||||||||||||||||||||||||||
|
Introduction The focus of this paper is to describe my efforts to modify and append the Amador County GIS road centerline layer with accurate and complete road geometry and attribution for that part of the road network contained within the boundaries of the El Dorado National Forest in Amador County. Study Area  | ||||||||||||||||||||||||||
|
Background Concurrent with my enrollment in the Geog 350, Data Acquisition class, I have been working as an intern for the Amador County GIS Division. In late June of 2008 I was given the assignment of improving the accuracy and completeness of the centerline roads layer for the county. This assignment involved a number of specific tasks which included comparison of the existing GIS roads layer with other county information including parcel data and databases from the Survey Department and Public Works Agency. Additionally, I was in communication with staff from the local cities, utility districts and the U. S Forest Service. During this time I acquired a new roads layer from GIS staff of the U.S Forest Service, El Dorado National Forest. Upon acquisition of the data I overlayed the Forest Service roads shapefile with the Amador County roads feature class on a background 2006 high resolution (1 foot per pixel) color orthophoto of the County. Although I was very happy to acquire a very recent data set from the agency of jurisdiction for these roads, I was faced with a certain dilemma. The GIS Coordinator for Amador County was attempting to achieve a county GIS layer that was displayable at a scale of 1:600. The U.S Forest Service data was developed to be displayed at a 1:24,000 map scale with a plus or minus 40 feet degree of accuracy. Motorized trails: Data Collection Standards with Global Positioning Systems (GPS) and Geographic Information System (GIS) Integration (USDA Forest Service, 2004) refers to this objective and outlines work flow and tasks, and specifies the Data Dictionary to be used for National Forests in California. County Public Works Agency "heritage" GPS based data was also available for county maintained roads within the Forest. Pre-existing Study Area Data.  | ||||||||||||||||||||||||||
|
Methods The methodology adopted for this data acquisition task was determined after experimentation with various GPS receivers and techniques. On October 11th and 13th I began familiarization exercises using a GeoExplorer 2008 Series GeoXT (Trimble) and a Venture HC (Garmin) GPS receiver. I did this by acquiring the datasheet for a National Geodetic Survey (NGS) marker at Westover Airport in Amador County and taking a series of waypoints to compare with the known coordinates which have a 2-3 cm accuracy. I returned to the site on December 12 and 13 for additional measurements. The table below shows the results of this effort.
On October 19th I travelled to the El Dorado National Forest and tested collection of data along Ellis Road. I used the Garmin "recreational grade" (WAAS enabled) GPS receiver in order to collect static waypoints at sites identifiable in a high resolution aerial photograph. I used the Trimble receiver with an attached Hurricaine antenna to collect streaming vertices while I drove the road in a vehicle.  The methods tested during these preparatory exercises determined those used for the actual data collection. On October 29th, I returned to the El Dorado National Forest equipped with both receivers and a strategy for efficient and accurate collection of roads data. The Trimble was uploaded with GPS Correct and ArcPad version 7.1 software and a GPS-enabled roads linear feature class. It was connected to an external antenna mounted to my Jeep Wrangler on a 2 meter range pole. For the next three and a half hours I collected about 1800 GPS positions as I drove forest roads with various levels of tree canopy and topographical characteristics. With little traffic I was able to vary vehicle speed between stopped and 15 miles per hour depending on the strength of radio signal reception. I was able to gauge the quality of GPS signal received by both visual displays and audible tones from the Trimble. Whenever possible I tried to collect (at a minimum) positions at changes of direction in the road. Early in the day, I stopped at the spillway of Bear River Reservoir and took static location waypoints using both receivers at the exact northeast corner of the bridge crossing the spillway. The location is clearly visible on the reference aerial photograph and is in the center of the study area. I also photographed the site where the points were taken. I collected the waypoint using the Garmin averaging for 30 seconds showing an estimated position error of 10 feet. 
 The purpose of this task was to ascertain the accuracy of the aerial photograph relative to the GPS position data. This exercise was repeated with comparable results at a helicopter landing spot in another part of the study area. Upon return to my residence, I downloaded the GPS positions from both receivers. Using the Trimble GPS Analyst extension, I brought the GPS enabled roads feature class into ArcMap, differentially corrected the positions through post processing and rebuilt the features using the corrected positions. The GPS Analyst determined the static point taken at the Bear River Reservoir spillway to be accurate within 0.48 meters. I also downloaded the averaged waypoint from the Garmin using EXpert GPS and created a shapefile which I brought into ArcMap. The (Garmin)point on the spillway bridge displayed about 16 feet from the GPS position collection point visible in the aerial photograph.
| Results Results from the first outing on
October 19th creating a centerline along Ellis Road were not successfully
post processed and are not included here. The resulting road feature is
included in the appendix. From the second (October 29th) outing about 2.5%
of the 1800 Trimble GPS positions taken were accurate within 50
centimeters; 40% were submeter and 47% were within two meters accuracy.
Display of the road layer on the background aerial photograph showed most
positions clearly near the road centerline. Similar results were obtained
on the third outing where over 3200 positions were collected. The
resulting GPS-enabled roads feature class displayed partially here are
variable accross the study area. Typical
Differential Correction Results
A Validation Report shows statistics for all GPS-enabled points and linear features collected during the data collection. 
| Analysis The chosen method of collecting the
GPS data proved to be efficient once I increased my familiarity with the
equipment and software. Inability to maintain a strong radio signal was
both anticipated and experienced under heavy tree canopy and in certain
topographical positions in the Mokelumne River canyon. Mission planning
probably would have benefited the data collection as signal availability
varied in specific areas throughout the day. External time constraints
negated the ability to benefit from this, however. Adjusting the "slider"
on the Trimble to favor productivity at some expense to precision turned
out to be an acceptable compromise resulting in useful data. Comparison of
roads created from the GPS capture to secondary Forest Service and
pre-existing Amador County data accross the study area generally indicated
improved data accuracy. Reference to the Validation
Report will show 18 of 25 features invalid. The initial test points
taken at Westover Field are included but were not collected with the
external Hurricaine antenna attached. The GPS-enabled linear features were
initally set to be invalid if any positions exceeded 3 meters.
Irregardless, the average position accuracy remained very good for all
linear features. A very small percentage of GPS positions required
correction. Even with relatively good coincidence of the GPS-enabled roads
feature class on the road centerline in the background imagery, display at
a 1:600 scale does not meet the desired appearance standard for the Amador
County GIS Division. A secondary benefit of the effort was the analysis of
the accuracy of the aerial photographed used by the county GIS Division.
It was checked against known NGS coordinates at Westover field and at two
other locations well distributed throughout the county. At all locations,
the aerial photographed appeared to be accurately georeferenced within one
meter. The accuracy of the aerial photograph precipitated the following
outcome. Using all available data, the GIS roads centerline layer was
digitized at the desktop using the sketch and tangent curve tools. 
| Conclusions Although the GPS data collected
was more accurate than either the Forest Service or Public Works Agency
data, the road geometry needed additional editing based on the aerial
photography for the region. The improved GPS road features along with the
2006 aerial photograph are good tools for interpreting and creating a
digitized roads layer. In the Amador County GIS Division the preferred
method for finishing the roads line work is desktop digitizing using the
sketch and tangent curve tool while creating minimal vertices.
| References USDA Forest Service. (2004, April). Motorized Trails: Data Collection Standards. Retrieved October 13, 2008, from National Forests in California: http://fs.fed.us/r5/routedesignation/data-collection-motorized trails
| Appendices
 | |||||||||||||||||||||
