Spring 01 lidar survey results in Prince William Sound

Transcription

1 Spring 01 lidar survey results in Prince William Sound James H. Churnside NOAA Environmental Technology Laboratory 325 Broadway Boulder, CO Tim Veenstra Airborne Technologies P.O. Box Wasilla, AK Abstract Several flights were made over Prince William Sound in March of 2001 to test the effectiveness of lidar for aerial surveys of herring and of infrared imaging for surveys of sea mammals. This first test clearly showed that the lidar was able to detect herring and the infrared camera to detect sea lions. A number of areas for improvement were identified, including a better ability to operate in confined bays and a better spatial resolution with both the lidar and the infrared camera. Each of these can be obtained, and methods are discussed in this report.

2 Introduction In March of 2001, a series of flights was made over Prince William Sound to test the effectiveness of lidar for aerial surveys of herring and of infrared imaging for surveys of sea mammals. We were specifically looking at three questions related to the performance of these instruments from aircraft in the sound. The first question concerns aircraft operations in the areas of interest. The prespawning concentrations of herring are located at the head of several bays. They spend the daytime on the bottom, coming to the near-surface layer at night. Thus, aerial surveys must be conducted at night when the fish are near the surface in areas that are tight to the mountains. If these areas can be covered with fixed-wing aircraft, the cost will be much lower than if a helicopter is required. The helicopter would make it easier to get into tight bays, especially at night. We will investigate the coverage possible in a fixed-wing aircraft in several of the most important bays for pre-spawning herring. In some cases, this may involve adapting the survey altitude and transect lines to adapt to the topography around each bay. The next question involves the depth coverage of the lidar. Generally, the depth penetration is greatest far from shore where the water is cleaner. The fish, however, tend to be high up in bays and very close to the shore. Here, the water is less clear, but the fish are close to the surface. Comparing the depth penetration obtained from the lidar with the depth distribution of fish observed by the acoustic will tell us whether or not all of the fish are being observed by the lidar. The next question involves marine mammals and other predators. We invested the use of thermal infrared cameras and intensified cameras to observe and identify marine mammals in the vicinity of the herring schools at night. The Prince William Sound Science Center has demonstrated that sea lions and whales can be observed from a surface ship (Thomas and Thorne 2001); we were looking at extending this capability to aerial surveys. This report describes the instrumentation that was used in this study, the data collection procedures, the results, and the conclusions that were reached in each of the three questions above. It also includes several recommendations for future work. Instrumentation The NOAA lidar and Airborne Technologies thermal camera were mounted in a small, twin-engine aircraft. It is an Aero Commander with turbo-charged piston engines. It has two camera ports in the bottom, enough electrical power to operate both the lidar and the camera, and a Global Positioning System (GPS) receiver. A photograph of the airplane is presented in Figure 1. Figure 1. Survey aircraft 2

3 Figure 2 Schematic diagram of lidar Figure 3 Lidar in aircraft The lidar system is a non-scanning, radiometric lidar. A block diagram is presented in Figure 2. The major components are 1) the laser and beam-control optics, 2) the receiver optics and detector, and 3) the data collection and display computer. Figure 3 is a photograph of the installation in the airplane. In the background of the figure is the back of the electronics rack with the laser power supply, the computer, and timing control electronics. The optics package is in front of the rack in the photograph. The black box near the bottom is the laser and a cover for the steering mirrors and diverging lens. The silver box near the top of the optics package is the power supply for the various components. The telescope is on the other side of the gold-colored plate. Both the telescope and the laser are directed downward through a hole in the bottom of the aircraft. Churnside, et al. (2001) present more details about the lidar. The laser is a frequency-doubled, Q-switched Nd:YAG laser that produces about 100 mj of green (532 nm) light in a 12-nsec pulse at a rate of 30 pulses per second. The laser is linearly polarized and the beam is diverged, using a lens in front of the laser, so that it is safe at the surface (ANSI, 1993). This irradiance level is also safe for marine mammals (Zorn, et al., 2000). The diverged beam is directed by a pair of mirrors to be parallel to the axis of the telescope. The figure shows a coaxial configuration of the transmitter and receiver. For these flights, a side-by-side configuration was used instead. While the coaxial configuration makes alignment easier at the short ranges available inside an aircraft hanger, there is no difference in the performance of the two configurations in flight. The receiver optics use a 17-cm-diameter refracting telescope. A polarizer is placed on the front of the telescope to select either the component of the return that is co- 3

4 polarized with the laser or the cross-polarized component. We used the cross-polarized component, because our experience suggests that this component produces the best contrast between fish and the scattering from small particles in the water. The telescope collects the light onto an interference filter to reject background light. An aperture at the focus of the primary lens also limits background light by limiting the field of view of the telescope to match the divergence of the transmitted laser beam. The resulting light is incident on a photomultiplier tube (pmt), which converts the light into an electrical current. A 50-Ω load resistor converts the current in a voltage, which can be digitized in the computer. High-speed digitizers exist that plug directly into the bus of personal computers, but these are limited to 8 bits of resolution. This produces 255 possible levels, which is not as much dynamic range as we would like for fish lidar applications. Therefore, we fed the detector output into a logarithmic amplifier. The output of the logarithmic amplifier is fed into the digitizer. The particular amplifier we used has a response of V log = log 10 ( V linear ) ( 1) It has an input voltage range of -0.2 mv to -2 V, which corresponds to an output voltage range of about V to V. Since the output voltage range is well within the range of an 8-bit digitizer, the logarithmic amplifier increases the maximum possible dynamic range from 255 to about The computer records the raw lidar data. It also records aircraft position from the aircraft Global Positioning System (GPS), GPS time, the voltage applied to the photomultiplier tube, and the attitude of the aircraft as measured by tilt meters and laser gyroscopes on the optical package. The applied voltage on the photomultiplier tube is used to find the gain of the tube, which is necessary for calibration. The computer is also used to display the data during the flight. The thermal imager is a Raytheon Palm IR250, which has an un-cooled ferroelectric detector (320x240) sensitive from 7 to 14 microns. A thermoelectric cooler provides thermal stabilization. The imager outputs an SMPTE-170M video signal that is recorded onto digital video tape. The video is updated at a 30 Hz rate. A computer controls the imager via an RS-232 serial communication port. A 75-mm lens on the imager provides a 12x9 degree field of view. The imager is mounted to the floor in a nadir point of view. Figure 4 shows the installation of the thermal imager and control rack. Figure 4 Thermal video system in aircraft 4

5 At the same time the video signal is recorded onto digital video tape, GPS information of the flight track is being recorded onto one of the two available sound tracks. A 9-inch black and white monitor enables the operator to view the thermal imager signal in real time. Any item of interest can be marked on the GPS track by an external trigger connected to the computer. This enables fast post-process review of suspected targets. The second sound track of the DV recorder is used to record all aircraft intercom and radio communications. This communications recording aids in post process correlation between the lidar and imaging data with what the aircraft and vessel personal are viewing. A 3-chip digital video camera was mounted in the aircraft to aid in target identification during day flights. The signal was displayed in real time on a 9-inch color monitor and was also recorded on digital video tape in the same way as the thermal video. The intensified, or night vision, camera that was planned for this survey was not delivered until well after the end of the survey. Operations Data were collected during five flights in March of Figure 5 presents the tracks for all flights except one made over the Copper River Delta. The points represent the centers of the data segments used in the processing. Those segments are 500 lidar shots long, which corresponds to a length on the water of just over 1 km. The first flight, in violet in the figure, was on March 7. Data were collected in the bays on the north end of Montague Island. The next day, a flight was made over the Copper River Delta to get a measurement of the penetration of the lidar in the Copper River. Some holes in the ice over the river were seen, but these were not thought to be representative of the conditions during the salmon run. The next flight, in blue in the figure, was the next day. This was in the same general area as the first flight. After this flight, we experienced almost 2 weeks of weather that was unsuitable for surveying. This included fog, show, and icing conditions. We finally returned to Anchorage. As might be expected, as soon as we gave up and returned to Anchorage, the weather improved. The fourth flight, in green in the figure, was part of the return flight from Anchorage on March 21. On the way to Cordova, we covered the north side of Port Gravina, where herring had been seen from the ship. The final flight, in red in the figure, was made that same night, and covered a much larger area. One of the main operational questions about the lidar concerns the depth to which fish can be detected. The strength of the lidar return from the fish, the configuration of the lidar, and the clarity of the water all affect this depth. However, we can make an estimate of the maximum penetration depth for each of our two lidar configurations (day and night). We calculate the median lidar return from each depth over a horizontal distance of about 1000 m. The depth at which this median return first goes below some 5

6 Figure 5 Survey flight tracks color coded by flight: violet March 7, blue March 8, green March 21 day, red March 21, night threshold is considered to be the maximum penetration depth. The threshold is selected manually for each configuration. Above this depth, there is a good likelihood of seeing a strong fish return above the noise. Below this depth, only a very strong fish return would be detectable above the background light and noise. Figure 6 is a map of the maximum penetration depth calculated in this fashion for the day flight on March 21. We see that the maximum depth was just over 30 m in Sheep Bay. The penetration in Port Gravina was much less, especially along the northern edge of the bay. Figure 7 is a map for the night flight. We see that the penetration depths are generally larger; the lidar can penetrate deeper at night without the interference from sunlight. As before, the minimum penetration is in Port Gravina. The penetration around Montague Island is better. We were somewhat surprised to see that the penetration in the center of the channel between Knowles Head and Montague Island was less than that on either side. Generally, however, the lidar should be seeing deep enough to detect schools of herring at night. There were some limitations because of the fixed-wing aircraft in the small bays where we expected to find the herring. For the region around Montague Island, the flight altitude was adjusted to be 1200 feet instead of our normal 1000-foot altitude. In St. 6

7 Figure 6 Lidar penetration depth for daytime flight on March 21. Figure 7 Lidar penetration depth for nighttime flight on March 21 7

8 Mathews Bay, this is still not high enough to be safe, and we have made some modifications to the lidar so that it can be operated from 3000 feet. These modifications will allow us to safely cover a much broader range of bays with no loss of performance at night. In the northern Sound, there is still higher elevation, such as the 3720 ft peak at the head of St. Matthews Bay, that would limit access into certain areas. To get complete coverage of the bays that are associated with herring, we would need to go to an altitude of 6000 feet, which would reduce the signal level to 25%. Because of the exponential dependence of penetration depth on signal level, the loss of penetration would only be a few meters. The other issue was the speed of the aircraft. The results of the next section show that the scales of interest are very small compared with the normal bin length of the lidar data. To a certain extent, the lidar bin length can be shortened. The speed of the aircraft still means that few lidar pulses are returned from each school. The redundancy of multiple pulses greatly reduces the uncertainty in the lidar measurements, so high aircraft speed can reduce the sensitivity of the lidar to small schools unless they are very dense. For this application, it appears that a helicopter would be preferable to a fixedwing aircraft if the cost were acceptable. We are currently investigating modifications to the lidar to increase its performance in restricted areas and at fixed-wing speeds. Results Figure 8 is a chart showing the locations of lidar detections on the first flight in red. In this case, a lidar detection is noted in any data segment where the peak return is greater than the baseline water return by more than 80 %. Most of the data within Zaikof Bay contain returns, while almost none of the areas outside of the bay do. A typical echogram of the lidar return within the bay is Figure 8 Lidar detections (red) and non-detections (blue) for flight of March 7 8

9 presented in Figure 9. It shows a scattering layer at a depth of about 5 m that varies in intensity. Based on previous experience in more open waters, we would tend to identify the areas of greater return as fish. This is because the plankton layers we have seen tend to have much less internal structure. The ship did not find fish in this area, however, and we think that turbulent mixing of a plankton layer into a more Figure 9 Lidar echogram from Zaikov Bay patchy structure might be causing the observed lidar returns. Sampling of the plankton will be necessary to verify this conjecture. Figure 10 is a map showing the locations of detections on the third flight in red. There were still no schools of fish observed on the west side of the island. The single detection in Stockdale Harbor is a single large animal at a depth of about 7 m, and another at a depth of about 10 m. An echogram of these returns is presented in Figure 11. The deeper return is smaller than the other; this is because it is a smaller animal or because it was not in the center of the beam. The two returns off of Middle Point are a single scattering layer in the last 400 m of one segment and the first 200 m of the other. The longer piece is shown in Figure 12. Figure 10 Fish detections (red) and non-detections (blue) for March 8 9

10 Figure 11 Lidar echogram from Stockdale Harbor Figure 12 Lidar echogram from near Middle Point Figure 13 is a map of the detections for the fourth flight. There were three areas noticed. The first is a single data segment off of Red Head. This is a very weak layer without much internal structure, and is almost certainly plankton. The next is a larger region on the north side of Port Gravina. This region contains layers about 5 m deep with harder returns buried within them. Figure 14 is an example from near the mouth of St. Mathews Bay. The last area is a section about 5 km long near the mouth of Sheep Bay. This is also a region of patchy layers containing more solid returns. Figure 15 is an example from this region. Figure 13 Fish detections (red) and non-detections (blue) for March 21 day flight The last flight was the most complete. Figure 5 shows that this flight covered a region around Montague Island as well as Port Gravina and Sheep Bay. No lidar detections were seen around Montague Island, and Figure 16 shows the eastern portion of the flight track with lidar detections noted. First, we note two detections off of Goose Island. These are very weak patchy layers, and are almost certainly plankton. The 10

11 Figure 14 Lidar echogram from St. Mathews Bay Figure 15 Lidar echogram from Sheep Bay Figure 16 Fish detections (red) and non-detections (blue) for March 21 night flight 11

12 second region is off Red Head, near where a plankton layer was noticed during the day flight. The nighttime detections contain are very localized and strong signals within a plankton layer, however, and these are very probably schools of fish. An example is shown in Figure 17. Most of the lidar detections were located in a rather large region south of Knowles Head. Figure 18 is an example of the signals in this region. These show fairly strong and localized regions of scattering. Further examination reveals that these are generally at the ends of scattering layers. There were large scattering layers over much of this region during the night flight; much more than during the day. We also note that these detections are clustered along the boundary between the shallow water just south of Knowles Head and the deeper water just to the west. Finally, we recall that the ship did not find fish in this area. These clues lead us to the conclusion that these lidar returns are probably strong Figure 18 Lidar echogram from near Knowles Head Figure 17 Lidar echogram from near Red Head patchiness in the plankton caused by turbulent mixing at the edge of the shallow water. The ship located herring near the mouth of St. Mathews Bay. A distribution of echosounder returns is presented in Figure 19. This is close to the area off Red Head where fish returns were seen in the lidar. The scale of the acoustic returns and the lidar returns is very different, however. The entire survey region in the figure is about 1000 m 12

13 by 3000 m. The aerial survey does not cover an area this small easily, especially this close to hills that are above the flight altitude. The primary purpose of the thermal imager was to identify sea lions in the water. Of particular interest were rafts of sea lions at night that would not be included in scat samples at haul outs. No such rafts were seen. Sea lions were observed on the shore. Figure 20 is an image of several sea lions on the south side of Zaikov Bay. These show up as light spots in the image. These were identified by visual observation. It is clear that greater spatial resolution is desired to better identify sea lions, especially at night. Figure 19 Distribution of acoustic returns near mouth of St. Mathews Bay Figure 20 Infrared image of sea lions Figure 21 Infrared image of near-shore mixing The infrared imager may also be used for other purposes. Figure 21 shows mixing of cold (dark) water draining off the shore and mixing with warmer sea water. This type of information may be of use in understanding the complex relationships between fish, sea lions, and the physical environment. Note also that there are several warm spots in the image that could be mammals, but identification is difficult with this resolution. Further studies will require better spatial resolution, even if it is at the expense of area coverage. Seabird activity is also apparent on the thermal camera. Birds can be observed sitting on the water surface as well as flying. The presence of birds around forage fish and feeding sea-mammals has been readily observed and documented. A better thermal camera has been acquired, and will be available for this program. An example of an image from this camera is presented as Figure 22. This image is the wakes of two humpback whales swimming side by side in the Gulf of Alaska. Each rise 13

14 Figure 22 Infrared image of whale wakes of the tail pushes cool water to the surface. The size of each cool spot on the surface is about 10 m; the cool spots are turbulent and expand as they rise to the surface. This wake is a distinctive pattern that was observed quite often, and can persist for several minutes after the whale has passed. If a similar pattern can be identified for sea lions, they can be detected from an aircraft even when they are not on the surface. Conclusions Aircraft Operations - The fixed wing would have done well if the herring had been in large layers in Zaikov and Rocky Bays. For small schools in St. Mathews Bay, a helicopter would have been a better platform because of its lower speed and ability to operate in confined areas. The helicopter is more expensive, however, and we are looking into ways to operate the lidar from greater altitudes and with better spatial resolution. Depth Coverage - Maximum depth penetration at night was generally between 20 and 40 m. This is probably sufficient to detect most of the pre-spawning herring stock. Predator Detection - The thermal imager was able to see sea lions on the shore, but none were observed in the water. Better resolution is recommended to improve the detection probability. The "night vision" camera was not tested because of delays in delivery. 14

15 References ANSI (1993), Safe Use of Lasers, Standard Z (American National Standards Institute, New York). Churnside, J. H., J. J. Wilson, and V. V. Tatarskii (2001), Airborne lidar for fisheries applications, Opt. Eng. 40, Shifrin, K. S. (1988), Physical Optics of Ocean Water (American Institute of Physics, New York) p21. Thomas, G.L., and R.E. Thorne (2001), Night-time predation by Steller sea lions, Nature 411, Zorn, H. M., J. H. Churnside, and C. W. Oliver (2000), "Laser safety thresholds for cetateans and pinnipeds," Marine Mammal Sci. 16,