Remote Sensing, GPS and GIS Technique to Produce a Bathymetric Map Mark Schnur EES 5053 Remote Sensing Fall 2007 University of Texas at San Antonio, Department of Earth and Environmental Science, San Antonio, Texas Abstract A technique to produce a bathymetric map of a reservoir using sonar, GPS, and GIS was developed. An interpolated map was produced that closely follows the original stream channel. Higher accuracy sonar and GPS will produce a more accurate map. Introduction Sonar is a form of active remote sensing that uses sound propagation under water to navigate, communicate or to detect other vessels. Sonar operates at low frequencies and may be used to measure water depth by measuring the time taken for a sound wave to travel from the transmitter, reflect from the hard bottom layer, and travel to the receiver. Variations of this technique have been used to determine area and volume of a reservoir (USGS 2006), reservoir sediment and watershed erosion (Chang, et al, 2003), and depositional patterns in a regularly flushed reservoir (Jansson and Erlingsson, 2000). Sonar data that includes GPS derived location data can be entered into a GIS and used to produce an interpolated raster surface estimating reservoir depth. The goal of this study is to develop a technique to establish a baseline reservoir depth and bottom contour using readily available hardware and software, and to gain the lessons learned to expand the technique using higher accuracy hardware to determine the sediment depth and volume of a sludge disposal lake. Study Area Boerne City Lake in Boerne, Texas has a surface area of slightly over 100 acres. The lake is a man-made reservoir constructed by the damming of Cibolo Creek to aid in flood control, and subsequently to serve as a public water supply for the city of Boerne (Figure 1). The lake is also used for recreation. The lake depth is generally known to be about 50 feet (15 meters). Establishing a baseline water depth and bottom contour will provide information valuable in managing the reservoir and enable change detection through repeated water depth measurements. Data Aerial imagery rasters and hypsographic contour shapefiles of the Boerne City Lake area were obtained from the Texas Natural Resources Information System. The aerial imagery is 2006 onemeter pixel resolution National Agriculture Imagery Program (NAIP). The hypsographic contours are 20 foot interval digital line graph data derived by digitizing map features as line graph elements from cartographic source materials at 1:24,000 scale. The NAIP image was used to digitize a lake boundary, which was used in processing the SONAR data. 1
Methods Transect points were selected and stored in a point feature class in a geodatabase. X,Y coordinates were added to the feature class, and the points were exported to Microsoft Excel for future upload to a GPS for navigation purposes. Transect lines were drawn between the points for representation on graphics. The lines were stored as a line feature class in the geodatabase. A line was drawn around the lake boundary and stored in the geodatabase as a polygon feature class. The geodatabase and feature classes were set to the same datum and coordinate system as the imagery. Twelve transects were conducted on two different dates across the reservoir in a small boat equipped with a commercially available depth finder and a hand-held recreational GPS. The transect routes were selected to maximize coverage of the lake, and be achievable in one day of data collection. Transects conducted on the first date resulted in inadequate coverage of the reservoir, and additional transects were conducted to better cover the reservoir (Figure 2). The depth finder used was a commercially available Garmin 160C Depth Finder. The unit operates using SONAR at a frequency of 200kHz, with a user selectable cone angle of 14 or 45 degrees (Table 1). A 45 degree cone angle was used for this study, in accordance with the manufacturer s recommendation for better performance in shallow water. The depth data is displayed on the screen. The transducer was mounted inside the hull of the boat, and covered in a layer of petroleum jelly to improve performance. The manufacturers specifications do not list accuracy values. The GPS used was a Garmin hand-held GPS equipped with Wide Area Augmentation System (WAAS) for improved horizontal accuracy. The GPS has an accuracy using WAAS of less than 3 meters (Table 2). Points were recorded approximately every ten meters in the GPS, with a sequential numeric point ID recorded. The water depth displayed by the depth finder was manually recorded in a notebook opposite the point ID. A total of 197 points were collected. The point data was downloaded from the GPS and the water depth entered into a database record corresponding to the point ID. The database file was imported as X,Y data into ArcGIS version 9.2, and exported as a shape file. ArcGIS Spatial Analyst was used to produce an interpolated surface using Inverse Distance Weighted (IDW), spline, and kriging procedures. The Z value field was set to the depth field in the database, and the default settings of each procedure were used, with the lake boundary used as an analysis mask to limit the results to the lake surface. Results and Discussion The transect depth data from the first data collection was used to produce an interpolated surface using the Inverse Distance Weighted procedure in ArcGIS Spatial Analyst. Visual observation showed that the data left large gaps in the interpolated surface (Figure 3), and a second set of transects were conducted to fill in the gaps in the data. The second transects were conducted seven days after the first set to minimize any change in the surface elevation of the lake level. The second transects filled in the gaps in the data (Figure 4). Interpolated surface were also produced using the spline, and kriging procedures. Visual comparison of the surfaces shows that 2
IDW and spline produced similar results (Figure 5). Kriging was markedly different, not only in the interpolated depth but in the calculated area. The kriging result was not confined to the lake area but produced a rectangular area (Figure 6). The differences were not investigated further in this study since the results were internal to the GIS software and not in the remote sensing realm. The greatest depths in the interpolated surfaces appear to match the old Cibolo Creek stream channel when overlaid with the topographic contours (Figures 4 and 5). This is consistent with the expected results since the stream ran along the lowest elevation, which is now the maximum depth of the lake. In order to expand this study for a peer-reviewed publication, several aspects will need to be refined. Horizontal and vertical accuracy was not considered in this study due to equipment and data limitations. The recreational grade GPS does not collect the information necessary for differential correction of horizontal accuracy. Vertical accuracy was not considered due to no ready source of lake surface elevation. Accuracy controls will need to be established. The USGS study is a good resource and includes accuracy controls and equations. Transect intervals need to be defined using an established formula. The USGS study used the formula max length/200, which was established as adequate for their purposes. The sampling interval also needs to be greatly increased, which will be accomplished using a survey grade GPS with a custom software program written for this purpose. A sampling interval of one point per second will be adequate. A survey grade echo sounder which outputs data to a GPS data logger will automate the data collection process. The interpolation tools in ArcGIS will need to be investigated further and a suitable interpolation tool chosen to produce the best result. Conclusions This study established a technique for producing a bathymetric map of a lake using remote sensing, GPS, and GIS. The map shows that the maximum lake depth is consistent with the stream channel present before the lake was created by damming the stream. Higher accuracy hardware will enable accuracy controls and continued improvements to the technique will produce a better result. References ArcGis, ESRI, www.esri.com Chang TJ, Bayes TD, McKeever S, Investigating reservoir sediment and watershed erosion using a geographical information system, Hydrological Processes 17 (5): 979-987 Apr 15 2003 Garmin 160c Fish Finder, www.garmin.com/products/ff160c/ Garmin etrex Vista GPS, www.garmin.com/products/etrexvista/ Jansson MB, Erlingsson U, Measurement and quantification of a sedimentation budget for a reservoir with regular flushing, Regulated Rivers-Research & Management 16 (3): 279-306 May-Jun 2000 3
USGS, Procedural Documentation and Accuracy Assessment of Bathymetric Maps and Area/Capacity Tables for Small Reservoirs, SIR 2006-5208, 2006 4
Figures and Tables Figure 1. Boerne City Lake, Texas. 5
Figure 2. Transects across Boerne City Lake. Blue dots are transects on the first date, red dots are transects on the second date. Table 1. Garmin 160C Specifications Frequency: 200 khz Transmit power: 150 W (RMS), 1200 W (peak to peak) Voltage range: 10-18 Maximum depth: 900 ft Cone angle: 14 or 45 degrees (dual beam) 6
Table 2. Garmin hand-held GPS Specifications Channels 12 parallel channel GPS receiver continuously tracks and uses up to 12 satellites to compute and update your position Integrated SBAS real-time Update rate 1 Hz, continuous Time to first 15 sec. warm, 45 sec. cold fix Protocols RS232 with NMEA 0183, RTCM 104 DGPS data format and proprietary GARMIN Accuracy < 15 meters, 95% typical (GPS) Accuracy < 3 meters, 95% typical (WAAS) Antenna Built-in patch 7
Figure 3. Interpolated water depth produced with the Inverse Distance Weighted procedure using the points from the first transects. Gaps are present in the map. 8
Figure 4. Interpolated water depth produced with the Inverse Distance Weighted procedure using the points from the second transects. Most of the gaps are filled in. 9
Figure 5. Interpolated water depth produced with the spline procedure using the points from the second transects. The result is similar to the IDW procedure. 10
Figure 6. Interpolated water depth produced with the kriging procedure using the points from the second transects. The result was not confined to the lake area but produced a rectangular area, and is markedly different from IDW and spline. 11