Acquistion of data occured over a period of 3.5 years. Two contractors supplied the data. Lasermap Image Plus captured data from the floodway inlet to St.Agathe in June of 1999. Aeroscan International captured the remaining data in the fall of 2001, spring and fall of 2002.
The DEM was split into 2 separate files, North Section and South Section.
Executive Summary
During the month of June 1999, LaserMap Image Plus Inc. (LMIP) performed an airborne laser profiling survey surrounding the Red River, south of Winnipeg Manitoba using an Optech AL71M 1020 laser system. 7he survey was performed under contract to the International Joint Commission and had two primary objectives. The first objective was to establish an accurate DEM for internal analytical purposes, while the second objective was to make the completed DEM freely available to interested end-users. Due to the extreme flatness of the area, the end use associated with the survey had to have an accuracy that could identify subtle elevation differences. The terrain type of the area consisted of, agricultural areas, urban areas and the river valley.
Despite the fact that record rainfalls hampered the start of the operation the project proceeded quickly got on track and was a success. Four missions were aborted due to adverse flying conditions (high winds and rain) but were subsequently re-flown when the weather was conducive to producing the product with the highest possible quality. In total, nineteen (19) flights were flown including re-flights.
All data were processed in the field to assure data quality and any re-flights were known within 24 hours and subsequently re-flown. Prior to demobilizing all data was processed, verified for quality control and ready for the mapping stage of the project. Preliminary grid data was delivered to IJC approximately eleven weeks after returning from the data collection. The delivered data consisted of several sets of ASCII files containing X,Y,Z point data for the ground at a 5 metre interval of the surveyed area. Final maps (hard and soft copy) were delivered October 19,1999.
This project demonstrated that airborne laser data could be acquired and processed in a relatively short time frame, especially when compared to conventional techniques. The project also demonstrated that the airborne laser system is capable of obtaining a high degree of accuracy unlike other sensors and offers a turnkey solution to producing DEM of relatively flat terrain.
This report covers the chronology of events that occurred over the course of the project, as well as a detailed account of the survey of the site, and the results obtained. A summary of problems and the solutions to these problems is presented, and the report offers a set of conclusions.
Survey Chronology
Sunday, May 24
Survey engineers perform initial reconnaissance of existing control and the potential sites for new control for project area.
Wednesday, May 27 New control is established with monuments and subsequent descriptions are established.
Sunday, May 30 Static GPS network is surveyed to establish new control points for project area.
Monday, May 31- Sunday, June 13 Laser survey was conducted including re-flights.
Tuesday, June 1 GPS kinematic profiles are surveyed along Hwy 75. Kinematic checkpoints are established around new control points (mini DTM - PAD 1).
Friday, June 4 GPS kinematic profiles are surveyed along Hwy 330. Kinematic checkpoints are established around new control points (mini DTM - PAD2).
Friday, June 11 GPS fast static survey to establish control points (also check points) for kinematic pads. Kinematic checkpoints are established (mini DTM's PAD3 & PAD4)
Saturday, June 12 GPS fast static survey to establish control points (also check points) for kinematic pads.
Sunday, June 13 GPS fast static survey to establish control points (also check points) for kinematic pads.
Tuesday June 15 ALTM laser system is demobilized.
Monday, June 21 Kinematic profiles are surveyed for Hwy 100, Hwy 59, Road 311 and Road 305. Kinematic checkpoints are established encompassing all quadrants of project area (Mini DTM's PAD5 - PADIO).
Control Survey
Prior to performing the airborne laser survey, first order control was sought to establish base stations for the subsequent survey. Due to the limited amount of first order control it was deemed necessary to establish our own base station control to fully optimize the ALTM's systems accuracy. The criteria for establishing base stations for satellite observations are:,
1. Free from any obstructions such as trees, signs, buildings etc. 2. The distance does not exceed 20 kilometres from the aircraft in the project area 3. Easily accessible by vehicle for field crews
Two base stations were established along Highway 75 which represents the north/south centre of the project, set approximately 10 km. respectively from the north and south extents (station John and Alex see map 1). Each base station was monumented using re-bar approximately one metre in length and buried 10 cm. below ground level.
Three first order stations comprising two 2D) coordinates and one 3D) coordinate were utilized to establish the airborne base stations. All existing control was located in the northern half of the project. Although this was not ideal, the configuration for a good network design was not compromised due to the geographic and angular separation from each control point (see Map 2). As suggested by the initial RFP, Geodetic Survey of Canada active control station Pinawa was examined and it was concluded that its location was too far away to be effective. However it was utilized in a survey prior to this project and was found to fit other control in the surrounding area to the magnitude of 6 PPM. Although this value is within first order standards the distance from this station to the project area could easily introduce errors in the decimetre level.
In order to assess the quality of the control a least squares adjustment was performed using Geolab TM on the GPS network. To validate the accuracy, a minimally constrained adjustment was performed holding one station fixed. By comparing the published values of the points to that of the "measured values" errors, can easily be identified. The GPS network is of obvious high quality as shown in the following table. By comparing the PPM value to the distance from the fixed station the vector error is easily in the 0-3 PPM range.
Station Fixed Value
A final stage of the adjustment was to produce a constrained adjustment thereby holding all control points fixed to their published first order values. This in essence allows the new points to "fit" all the control in a homogenous network. The final coordinates were used for all GPS processing with the ALTM system.
Airborne ALTM Laser Survey
Overview
LaserMap owns and operates an Opteck AUM 1020 system. The AUM system is comprised of a high frequency optical laser coupled with GPS and an Inertial Navigation System (INS). The 3-dimensional GPS solution (X, Y, Z) is used to position the laser scanner each second, while the INS data are used to determine the systems' orientation. The GPS solution is computed from differential kinematic processing, using data collected simultaneously at the aircraft, and at base stations on the project site. Both base stations were used simultaneously during all flights and GPS trajectories were compared for assurances that the GPS was of high quality.
From the airborne platform, the laser emits pulses at frequencies of up to 5000 Hertz. These pulses are reflected off the ground, vegetation or man-made structures at different time intervals, so the varied distances between the emission and reception can be calculated. With such high pulse emission rates, the laser can obtain as many as 300,000 3 -D points per minute. For this project the nominal flying height was 600 metres above ground. With an aircraft speed approximately 100 knots, and the frequency and width of the laser scan, the ground point density was collected as close as 1.5 metres in open areas, although where overlaps between flight lines occurred, the data are closer than this.
Flight Planning
The optimal flight plan was to fly the project in the north/south direction to minimize the required amount of lines and minimize the turning time in the aircraft. Due to the proximity of the Winnipeg airport and its main approach for the runways, the original flight plan was modified by request of air traffic control and flight lines were changed from a north/south direction to a north-east/south-west direction. Also the flying height restriction for the airspace surrounding the airport necessitated modifying our flying height by lowering it by approximately 100 metres. The by-product of lowering the height is that it allows the size of the actual laser foot print to be smaller, thereby allowing better penetration of the thick forest canopy along the edge of the Red River. The effect of the changes involved 30% more flight lines and improves the quality of the data. New flight plans were filed May 29th. Each day a new flight plan was filed with Air Traffic Controllers to assure that no air traffic would hamper the data collection. Due to an air show that was taking place for three days the southern sections were flown so as not to conflict with the show. Weather also dictated the days flight plans. High winds over 20 knots, were. Assessed for the turbulence and decisions were made if they were conducive to fly. Optimal flying times were in the early evening when the winds had calmed and the air traffic was quiet. Coupled with ideal, flying conditions the GPS constraint had a significant impact. Only ideal "GPS windows" comprised of sufficient number of satellites and good geometry, (i.e. PDOP) were utilized. If a particular flight encompassed a high PDOP then no data were collected until the geometry was of sufficient quality.
Data Processing
During each laser profiler flight, the raw laser data are recorded on 8min data cartridge tape. These tapes are capable of storing vast quantities of data at a high capture rate. At the same time the GPS and inertial navigation data are also recorded. These two sources provide high accuracy positional information for the system.
The GPS and inertial data are processed in tandem to achieve the best positional result. Once the position and attitude of the aircraft are known at each epoch (1 -second intervals), then these data are integrated with the laser ranges to provide a position for each data point on the ground. Up to 5000 laser ranges are acquired each second, thereby creating files with millions of data points. The data are processed using the proprietary ALTM laser suite of software to produce an ASCII file of (x,y,z) coordinates. The data can then be transformed into formats compatible with numerous CAD software packages.
Quality control for the project was done at several stages in the processing cycle. The GPS was analyzed using FLYKIN TM kinematic GPS OTF (On the Fly) software as well as ALTM kinematic GPS software to meet pre-determined statistical criteria. Since the GPS data are processed immediately after each flight, any re-flights required due to poor satellite constellation or insufficient returns were integrated into a subsequent day's survey mission. Each flight comprised of setting up both base stations to collect data. By using two base stations this satisfied two criteria, if one base station failed there would be a second backup and secondly both GPS trajectories were compared on each flight as an independent check to verify that the positional data were correct. For all flights the GPS data were of high quality and only the base station closest to the aircraft was utilized. This minimized the absolute error for the aircraft position.
The inertial data were analyzed at the output stage of laser processing . Any peculiarities with respect to velocities or drift were noted and a determination made as to whether a re-flight is necessary. Two re-flights were necessary for this project - an average for this type of work. To ensure that the data collected along the Red River were penetrating the forest canopy, an extra flight with different parameters was flown in a north/south direction. Due to fast winds the data were only collected into the head wind allowing a slow flying rate. By utilizing a slow flying rate more laser points are reflected off the features. The flying height was lowered approximately 100 metres again to see if better forest canopy penetration could occur. The results indicated that due to the thick canopy, flying slower and lower did not significantly improve the results.
The primary quality control tool for the laser ranges is the percentage of returns that are received back at the laser after it has emitted a signal. The acceptable range for returns, typically between 90% and 95% were met for this project. Lower percentages are normal over water and other poor reflectivity surfaces. Due to the significant rainfall at the beginning of the project the heavy wet soil had a weaker return from the reflected IR light but was still in the 85-90% range. To compensate for "less" data after the wet days the flight line overlap was increased to 50% side-lap thereby ensuring that there was 100% total overlap in the data. This in effect doubled the amount of laser returns. When drier conditions prevailed, the overlap was reduced slightly.
Classification
Classification of the LIDAR data were done at the end of the processing cycle. This task processes the data points through an intensive filtering process, and various classes of points are separated. For this project, the classification essentially was limited to vegetation data removal, separating these data from the ground layer data. Where applicable, building data were also removed. After all filtering had been completed a rigorous analysis was performed and a finer refinement of vegetation/building removal was undertaken.
Quality Control
Laser Verification
After each day the laser data were analyzed for any anomalies residing in the data using a CAD package along with statistical analysis tools written by GE0surv Inc. Specifically values located in the overlap sections between adjacent lines were checked for each particular flight. Also data between subsequent days which contained overlap data were verified. This allowed two independent answers to be compared.
Ground checkpoints
Even though all of the laser data are positioned using 3-dimensional GPS and inertial data, LaserMap surveyed a number of checkpoints within the project area. In addition, LaserMap crew's surveyed profiles along some existing roads, as well as DTM "pads" surveyed by ground kinematic GPS techniques. For example, continuous kinematic GPS data were collected by driving along the Highway 330, Highway 75 (through the complete project area) Highway 59, Highway 305 Highway 101 and secondary roads. Seven stationary GPS ground control points were established and tied to local first order control (horizontal and vertical). These secondary control points are used for two purposes, first they are definitive checkpoints against the laser data and secondly they were used for the base stations for the DTM pads. A total of 10 mini DTM pads were surveyed in strategic locations encompassing the project area. The purpose of the DTM pads is to have an independent check against the laser data. The procedure consisted of using GPS kinematic techniques to establish approximately 50 - 100 points spaced 1 - 3 metres apart. This in essence emulates the laser data capture. By comparing two "surfaces" rather than a single point, the overall relative accuracy can be compared.
The locations of these control features are on the attached map. All of the control features were used to check the LIDAR data, which agreed within a maximum of 10-15 Cms. Typical with airborne laser data the height bias associated with airborne processing was non-existent. Therefore no correction was made to adjust the absolute position of the laser data.
Problems and Solutions
Weather played a significant role at the start of the project due to high amounts of rain. This had a two fold affect thereby delaying the start of the project and secondly enhancing the foliage on the rapidly budding trees. Also, due to the rain the soil was saturated which was rectified by flying with more overlap thereby increasing the number of points, The integrated inertial system (IMU) also had a few minor problems. For unexplained reasons, the IMU had two system crashes during the project. This problem has occurred on occasion in the past with the same system, most recently while at Optech during servicing. In most cases powering down the system and powering it up again, while in flight, rectifies the problem. The inertial system then re-initializes and can realign itself in the air. It must be noted that no data were collected during IMU outages.
The ALTM system, unlike any other sensor, is unique in the fact that it can penetrate forest canopy in most areas. However extremely dense vegetation (where the sun cannot penetrate to the ground) makes it impossible to obtain a result from even the ALTM system. The areas surrounding the Red River were extremely dense but results indicate that penetration did occur. To increase penetration for this type of heavy vegetation, we prefer the LIDAR system be flown during leaf free conditions.
System Cheek and Calibration
It is LMIP's policy to perform calibrations twice a year unless the system performance indicates additional calibration is necessary. Prior to the Red River project, LMIP had undertaken two calibration surveys within the last three months of the project time frame. The system check was performed in Winnipeg before the project commenced and strategic and systematic procedures were followed to find any bias that may exist in the systems parameters. With all calibrations and system checks, the system performed to specifications.
Mapping
During June we had several discussions with colleagues at Natural Resources Canada concerning the datums. In Canada the computations for the application of the new vertical datum used in the United States (NAVD88) are apparently not completed. Natural Resources was reluctant to provide elevations based on NAVD88 for bench marks in the Winnipeg area. As a result we agreed with Dr. Halliday that we would keep all of the coordinate values on the 1983 North American Datum (NAD83)and retain the elevations based on the old vertical datum for North America (CGVD28).
Processing of the data commenced immediately and continued for a couple of weeks before it became obvious the automated vegetation removal software was interpreting the berms and dykes as "thick hedges" or "rows of trees" and was removing these data. We decided that this information was critical to the project and reviewed all of the sheets processed to date. A number had very few berms or dykes and were quickly corrected. Several however, required re-processing completely. Thence forward we processed all of the vegetation removal using the tracking video record to ensure this did not occur again. This was more time consuming in respect it required the interactive participation of the proc6ssing technicians, however, they were also able to conduct some of the quality control processes at the same time. So the additional time spent was not overly onerous. However, it did set back our delivery schedule. Both the IJC and R. Halliday & Associates were appraised of the situation immediately.
Once a sufficient area had been processed, we were able to commence gridding the data to the required 5 metre grid. Processing and gridding continued through July and August. Note, that although the data are delivered in a five metre grid format, the original data were collected more densely and are archived in raw data format.
As there were no specifications regarding the plots, LaserMap compiled a sheet format and suggested a plot scale of 1:5,000, with each sheet covering 4knis by 4 kms. The scale was chosen as being sufficiently small to keep the number of map sheets produced to a reasonable number (43), while being sufficiently large to allow the 1 foot contour interval to be reasonably plotted (without contours being so close together they run into each other). A sample sheet was plotted at 1:5,000 scale and submitted for comments. Comments were limited to a request for more contour numbers. LaserMap modified the parameters for adding contour numbers, as requested. Sheet numbering was based on the southwest coordinate value for each sheet. An index box of surrounding sheet numbers was included in the sheet surround of each sheet.
Finally, once all of the processing was complete, an independent quality control review was made, by an Arc/Info expert who had not been involved with the data processing, to ensure there were no blunders. Then CD-ROMs were written of the data. As the data were gridded at 5 metre intervals, each 4 km x 4 km mapsheet resulted in a file size of approximately 3 mbs (Arc/InfoGrid files). All of the 43 map sheets, therefore, were easily accommodated on a single CD-ROM. Five copies were disbursed as per instructions of Mr. Halliday. Plot files were made of each map sheet and five copies of each produced to be shipped according to instructions also received from Mr. Halliday. Two CD-ROMS were delivered with each sheet containing the contours as ARC/INFO coverages and DWG files.
SPECIFIC NOTES
As mentioned in the narrative above, all data are delivered in NAD83 horizontal datum on a standard UTM grid in zone 14. The elevation data are in CGVD28.
Plots were made, as requested, in the same, NAD83 and CGVD28. However, the coordinate system of the plots is in metres. The elevations of the contours have been processed and are plotted as elevations in international feet. (1 metre equal to 3.2808398954 feet.)
Conclusions
This project was a text-book example of the use of LIDAR for a flood plain mapping project. Lasermap Image Plus achieved accuracies that were in actual fact superior to the requested requirements. The project was delivered on budget, and while deliveries were approximately one month late due to circumstances outlined above, this was mitigated through an advance delivery of most of the digital files so the consultants could commence using the data in priority areas.
LaserMap Image Plus was pleased to work on this very successful project and appreciates the professionalism of all the parties involved in assisting us to provide the International Joint Commission with an impressive data set.
METADATA REPORT FROM AEROSCAN INTERNATIONAL
A Digital Terrain Models (DTM) is being produced over the Red River valley, south of Winnipeg, Manitoba, Canada. The extent of the area to be surveyed is from Sainte-Agathe to the American Border. East-west, the width varies from 19 to 42 kilometres over the river valley. Lidar was used to acquire the data, which is an airborne laser survey system mounted on a helicopter. Aeroscan International Inc. started collecting lidar data during the fall of 2001. A snowstorm occurred before the completion of the acquisition, as a result, the decision was made to post-pone the rest of the acquisition, so the ground accuracies wouldn't be compromised. Approximately 46% of the project has been sucessfully surveyed, the rest is still pending. The remaining portion of the project is expected to be surveyed in October 2002. After the snow melted in the spring, water levels were high in the Morris area. During that period, Aeroscan surveyed another section of the Red River, near Selkirk, Manitoba.
This document describes what has been achieved up to date, and shall be updated after project completion.
Aeroscan's proprietary Lidar system combines laser, GPS and Inertial technology to measure elevations on the earth's surface. The system is mounted on a helicopter and flies at about 250 metres above ground. Typical flying speed is 35 metres per second and the laser measures 10000 useable points per second. The typical spacing between measured points on the ground is 1 metre across track and 1.5 metre along track. Three-dimensional points are collected, from which a bald earth ground model can be derived. The final product consists of a 5 metre horizontal grid in an ASCII XYZ format that forms the Digital Terrain Model (DTM). The "ground thin" data points were also produced, as well as the point cloud data.
The helicopter positions are calculated by post-processing the raw GPS data at 2 Hz interval on the WGS-84 ellipsoid. The GSD95 undulation model is used to assign a Mean Sea Level elevation to each helicopter position. These positions are combined with the inertial data in a Kalman filter to generate a time tagged post-processed position and orientation solution for the laser receiver. These values are then used with the laser ranges and mirror angles to compute all the individual X,Y,Z laser returns, which form the point cloud. At this stage, the projection used is Universal Transverse Mercator, zone 14, metres. The point cloud data set is then run through a series of algorithms. First the erroneous points that are definitely below the ground surface are removed. Second, the laser returns are separated between the ones that hit the ground and those that hit vegetation or buildings. Third, the ground hits are thinned, removing points that have an elevation difference of less than 5 cm compared to the other points within 2 meters horizontally. This thinning method allows us to keep the maximum number of points in steeper terrain and remove redundant points in flat areas. This data set forms the bald earth ("ground thin") model.
Internally, elevation comparisons were made between post-processed kinematic ground points and the Lidar bald earth model. The XY coordinates of the kinematic points were transposed on to the surface model (at the intersection of the triangle plane), and given an elevation. The differences between these two elevations were computed. Experience has shown from past projects, that the surface elevations are constantly approximately 15 centimetres lower than the kinematic elevations. To make our surface elevations tie to the ground, we applied a + 15 cm bulk shift to the whole data set. At this point, statistics shows the internal accuracy to be 15 cm RMS.
External checks were made, comparing elevations from the Lidar and kinematic data sets to survey grade road elevations from the Department of Transport. Although the Lidar and Kinematic were consistent within one another, they mistied by 24 cm when compared to the DOT data. After investigations, it was found that the wrong geoid undulation model was used during Aeroscan's initial geodetic network. The HT97 model with NAD83 (CRSR 96) reference system was used instead of the GSD95 model with the ITRF 92 reference system. The later being consistent with the model used in our post-processing software. Therefore all the base station's ellipsoidal elevations used to position the helicopter were too low by 24 centimetres. A +24 cm was applied to the Lidar data set to make it consistent with the government published elevations.
The whole area was divided into blocks, so the millions of points could be handled easier. 52 blocks were designed with an approximate size of 6 km by 6 km each. At this stage, 24 of the 52 blocks have been finalized.
From the ground thin data set, a 5 metre by 5 metre horizontal grid was created.
The point cloud data set was also provided, with each point cloud file representing a flight line.
Because we did not have the Ground Thin data from the IJC commission we could not evaluate the Manitoba Department of Highways kinematic data with the IJC grid in the same manner. We did however do a comparison and found that the IJC data did not correspond to the Highways data. We created a 5 metre grid of the Highways data and overlayed it with the IJC data. The output was viewed as a histogram using the Z-values in a range legend. We also looked at the results of the overlap area in the same manner. In both cases there was a noticeable shift in the histogram from been evenly distributed. By applying various correction values and comparing the resulting histograms it was determined that the IJC data needed to be adjusted by +0.078m. This value resulted in the best fit for both the overlap and Highways data to have the highest percent of points matching within the -0.15 - 0.15 range. Mosaicked the Aeroscan and Lasermap grids ensuring that the Aeroscan grid took precedence in areas of overlap.
To ensure that the LiDAR met specifications outlined in the Identification Section of the Supplemental Information, compared the ground-thinned data with survey-grade GPS data collected by Acres Manitoba for the floodway and by Manitoba Transportation and Government Services for highways 75 and 201. Used the CREATETIN command with GENERATE to convert ground-thinned ASCII tables to tins. Converted the tins to point coverages with TINARC. Used a custom AML script to select ground-thinned points within 1.5-meters of each GPS point and report the minimum difference between the GPS and LiDAR elevations.
Converted 5-meter grid ASCII tables to point coverages using the GENERATE command. Created 5-meter grids from points with POINTGRID. For the tiled data, clipped each 5-meter grid to its tile boundary. Used the MERGE command to mosaic the tiled and floodway grids. Assigned values to small Nodata slivers along tile boundaries using FOCALMEAN with a 3x3 cell neighborhood.
To reclassify major lakes and rivers within the LiDAR extent to Nodata, extracted major water body polygons from Manitoba Conservation's 1:20 000 base data. Converted polygons to 5-meter grids using the POLYGRID command and assigned all water cells a value of 9999. Created a subset of water bodies larger than 3000 square-meters. Merged the resulting water grid with the LiDAR DEM. Reclassified the 9999 cells to Nodata.
Used the MERGE command to mosaic the resulting grid with the pre-existing LiDAR DEM.