 |
| A high grass LiDAR land cover
classification checkpoint. |
|
In 2007, Penn State
contracted Dewberry, a full-service engineering and surveying firm,
headquartered in Fairfax, Va., to provide independent accuracy testing and
qualitative analysis of the data collected for the PAMAP Program. Dewberry, in
turn, contracted Herbert, Rowland & Grubic Inc. (HRG), a multidisciplinary
engineering and related services firm based in Harrisburg, Pa., to perform the
statewide data collection of checkpoints that were needed to support quality
assurance and quality control (QA/QC) analysis of both the PAMAP LiDAR data and
orthophotography. Due to the multiple data vendors involved in the PAMAP
initiative and the varying dates of capture, this “ground truth” testing was
essential in guaranteeing that the public was getting a reliably accurate
dataset. The HRG project team had just three years and a limited budget to
cover a project area consisting of 46,000 square miles in 67 counties. Meeting
the objectives required the right technology and an innovative approach.
An equal distribution of checkpoints across the data collection area was a key
project requirement. For the orthophotography checkpoints, the HRG team
selected photo-identifiable ground features in the office prior to the survey.
Using the PAMAP orthophotography in Esri ArcMap software, survey technicians
“heads up” digitized each checkpoint over existing surface features visible on
the imagery before field visits. These points were then imported into Panasonic
Toughbooks with Delorme Street Atlas software so that they could be easily
navigated to while in the field. The use of this GPS navigation software was
essential as field personnel moved through unfamiliar terrain and greatly
increased efficiency.
The LiDAR points proved more challenging since they essentially had to be
chosen “on the fly” in the field. The HRG team had to select points that fit
the criteria for the five land cover classes and were also distributed well
spatially while also meeting criteria set by Dewberry.
The criteria were based on the 1998 National Standard for Spatial Data Accuracy
(NSSDA) and specified that horizontal accuracy would be tested by comparing the
planimetric coordinates of well-defined points in the dataset with coordinates
of the same points from an independent source of higher accuracy. Vertical
accuracy would be tested by comparing the elevations in the dataset with
elevations of the same points as determined from an independent source of
higher accuracy. Since the NSSDA was developed before widespread use of LiDAR,
the project team also relied on supplemental guidelines issued by the American
Society for Photogrammetry and Remote Sensing (ASPRS), the National Digital
Elevation Program (NDEP) and FEMA for vertical accuracy testing of LiDAR data.
These guidelines recommend surveying 100 vertical checkpoints for each LiDAR
collection block, which equated to 20 checkpoints each for the five land cover
types specified in the project (bare ground, high grass, brush, woods and
urban).
 |
| A Pennsylvania topo map showing the
distribution of orthophoto checkpoints. |
|
The accuracy of all checkpoint surveying required the
use of dual-frequency GPS equipment. GPS offers substantial benefits over
conventional (angle/distance) survey methods, but at the time the survey began,
only two methods of GPS surveying were commonly used: static and radio-based
RTK. Neither method was satisfactory for the purposes of this survey.
Time constraints and a high volume of checkpoints over an expansive, statewide
project area ruled out the use of static GPS due to the lengthy observation
times required. The other drawback to this method was that there was no way to
verify the quality of the data being collected until it was downloaded and processed
in the office. This was of particular concern in relation to the checkpoints in
the “woods” land cover class. To survey these points, an inter-visible pair of
control points had to be set and observed with GPS. The control points were
then occupied with an electronic total station, and the checkpoint was measured
under tree cover where GPS is ineffective.
RTK GPS was impractical due to the large geographic area of the project; most
broadcast radio signals are limited to a range of two to five miles. Rapidly
and cost-effectively collecting a high volume of checkpoints over an expansive
area that had highly accurate horizontal and vertical positions required a
different approach.
Cellular modems were
being successfully employed by other industries to provide real-time data for
communication and remote monitoring. The sending and receiving of data with
this method was not limited by distance and was possible anywhere wireless data
service could be obtained. The use of this technology for GPS surveying at the
beginning of the project was in its infancy. Still, the HRG team believed that
applying this technology to the RTK survey method could provide substantial
improvements in efficiency while achieving the required accuracy.
HRG personnel determined that Airlink cellular modems via a nine-pin serial
port could be readily interfaced with the Trimble R8 and 5800 GPS receivers
used for field surveys. Once the modems were configured properly, real-time
data could be transmitted by one modem at the base receiver to its “mate”
connected to a rover receiver realizing a modem “pair.” After establishing the
connection through a wireless network, a high-quality “fixed” RTK solution was
achieved, which gave the field surveyors real-time positions on any given
checkpoint or survey control point used to conventionally locate LiDAR
checkpoints in the “wooded” land cover classification.
The next challenge was to find a way to obtain redundant observations on each
of the PAMAP checkpoints to ensure the quality of the survey data. Two accepted
methods to achieve this redundancy are to observe each point twice at different
times and/or from different base locations. However, the large project area and
the distances traveled made observing the checkpoints at different times
impractical. Instead, the team developed a method for observing each point from
two base locations in a single visit. The two field members working on the
collection each carried two Airlink modems along with their GPS rover. Two
Trimble R8 base stations were then set up at different locations with a single
Airlink modem, and each of the base modems was configured to transmit to both
of the field members simultaneously. This enabled each of them to observe each
checkpoint twice, without having to revisit the point.
The combination of cellular modems and the creative approach to redundancy
greatly increased the efficiency of the workflow and accuracy of the final
dataset while allowing the crews to set survey control pairs “on the fly” for
observing wooded LiDAR checkpoints.
Comments (0)