Development of Eider ducks electronic monitoring and mitigation technology to avoid predation on mussel lines

Aquaculture Association of Nova Scotia
AIMAP-2009-M01

Table of contents

• Executive Summary
• Problem Exploration
• Concept Evaluation
• #1 Ultrasonic Acoustic Reflection
• #2 Passive Infrared
• #3 Radar
• #4 Vision
• #5 Underwater Acoustics
• Results
• Site Trials
• Conclusions and Recommendations

Executive Summary

The purpose of this project was to design and develop a system based on unified communications (UC) technology to be employed as a non destructive monitoring method for predatory ducks (mainly Eider ducks). UC technology uses instant messaging, email, voice mail, web conferencing, fax, audio, video, cell phones and/or other applications blended into a single system linked to sensors to immediately inform the employees of any changes that may compromise the production at the farm. This project investigated the requirements, evaluated current concepts and created functional prototypes to determine the effectiveness of duck detection on a mussel farm with UC technology. Five prototypes were evaluated, three were subjected to limited testing on the mussel farm and one was chosen as the most promising method to detect ducks.

The underwater acoustic sensor was deemed the best method to move forward to a potential second phase of development. This low-power cost-effective device has the ability to capture duck noises in a small area and relay qualified positive detections wirelessly to a handheld device (cell phone, computer, etc) or a predetermined deterrent mechanism. With the sensor just under the water, the system is less susceptible to surface wave action, adverse weather conditions or ice-flow damage and can remain covered from ducks or vandals.

There are more tests and refinements that can be performed with the underwater acoustic sensor system. About 12 algorithms for the sensor were created ranging in sensitivity. With further testing the selection of algorithm could be refined to detect a flock of ducks with certainty with low number of false detections or detect a single duck with higher probability of positive detections. Implementation of the sensor platform into a smaller covered self-powering buoy system would also be an asset to the program.

Problem Exploration

Marine birds regularly visit aquaculture sites to feed. They have a good memory of where farms exist and can dive down to depths of up to 60 feet to feed on mussels. The Eider duck is one of the most damaging birds on Nova Scotia’s mussel farms. Each 1.8 Kg Eider duck can eat about 20 % of its weight in mussels per day. The impact of these feeding activities translates into economic loses that can reach values of up to 40 % of total production.

Traditional and current mitigation measures for raft mussel culture involve using protective nets. In the case of long line production systems, this approach is impractical. Similarly, the use of protective socking has been considered, but this interferes with the growth and survival of the mussel. Other approaches are the use of inflatable devices that act as scarecrows and surface or submerged noise generators, but it was found that due to the continuity of these methods the ducks get used to it, thus defeating the purpose. The most common method used is scaring and chasing the ducks on boats or shooting at them. Chasing ducks is costly, disruptive to operations and has negative environmental impacts. It is therefore imperative to develop cost effective, spontaneous and non destructive techniques. This is the reason that led us to conduct the present project.

The design of the system took place by considering the main challenges common to most mussel farms. For example, mussel farm sites are mostly located in remote sheltered locations with low cellular phone coverage and no accessible power source. The detection system would have to be self-powered in the field and yet be able to use a communication system (UC) to exchange information with the handheld device or deterrence system. The size of the mussel farm leases can vary significantly. Thus, the resulting technology would have to be able to cover any size of lease effectively.

Concept Evaluation

Based on the above considerations, many concepts were evaluated:

1. Ultrasonic (acoustic reflection)
2. Passive infrared (sensing of energy radiated by an object relative to its background)
3. Radar (reflected or obstructed microwave energy)
4. Vision (image processing or video signal processing)
5. Sonar (underwater acoustics)

All the concepts above were prototyped for testing with the exception of #4 (vision). It was determined early that detection of a duck in the view of the camera would be spoiled in successive frames due to inevitable wave action, power requirements, and land mass in the horizon.

#1 Ultrasonic Acoustic Reflection

Ultrasonic refection was evaluated to detect duck presence. This method is effective in a limited range (10-12 ft) with a small emitter. The power usage, however, is very large and the emitters would be easily be fouled on the surface of the buoy by turbulent water, and/or ice, thus reducing effectiveness.

#2 Passive Infrared

Infrared units may be able to be used to detect the ducks. However, infrared units are small and could detect the ducks only within a small range. Multiple units have to be used on buoys to survey the field. These units would not be affected by the weather but would be exposed to damage on top of modified buoys. The most negative aspect of passive infrared is that the thermal aperture (gradient) of a duck is significantly reduced due to the insulating nature of the feather covering of the ducks.

#3 Radar

Initial investigations showed that small low-power radar units would not be able to detect the presence of single ducks. Thus, multiple ducks in the area are required in the view to create a reliable detection.  Also, small radar units are much directed in view, limited in range, require high power, are expensive and cannot scan large areas.

#4 Vision

It is not possible to use a computer vision system to look out from the farm to detect the presence of ducks in the area. This method can detect the ducks as they come close to the farm and can be configured to view in all directions. The method would however be affected by poor visibility, moving buoys, platforms and weather. The range of viewing and processing could be very large (line of sight) but would require considerable power and processing power. The camera system would require few units per farm but would depend on location.

#5 Underwater Acoustics

It is possible to use underwater microphones (hydrophones) to detect the presence of ducks in the area. This method flags the detection late after the ducks have already landed near the farm. It is however robust as it is slightly or not affected by weather, visibility, and environment. This method also could detect ducks over a relatively large area (100-200 ft) since low frequency sounds can travel considerable distances in salt-water.

Results

Based on the concept evaluation it was decided that underwater acoustics was the best method. Further development and testing of a system based in this technology was performed for the remaining of the project.

The hydrophone picked up the noise of the feet movement and the vocalizations of ducks. In analysis of the sample, a Fast-Fourier Transform (FFT) was taken on the sounds with a Kaiser window (filter). The FFT looked at the sound in the frequency domain over time. The FFT window in the lower-right of the graphic showed more energy at 3 KHz when the ducks quacked. With further work in the spectrograph (in the upper-right window) the vocal sounds from the various ducks could be isolated and determined.

After close examination of the sounds, there were at least two distinct quacking calls. One group at 3 KHz and another group between 2000-2500 Hz. This difference provided classification criteria of duck type from the pitch of the calls (which is interesting but not required). As for the feet movement, it will take more experimenting and fine tuning the data, but the paddling feet noise could be seen in the 100 Hz range. This sound however, may be muted by the sea conditions and low frequency wave noises emitted by the buoy. Therefore, for practical reasons it was decided to focus on the quacking sounds.

The prototype was created to process data from a hydrophone. The unit consisted of an analog front-end, microprocessor, and wireless radio. When in use the signals from the hydrophone are amplified and processed using FFTs in the microprocessor. If the energy in the frequency of interest is present for the duration and frequency levels described above, the noise floor changes and the detection is flagged. This detection is then signaled wirelessly to a handheld device or deterrence system. The data of the frequencies typical of duck sounds was used to create up to 12 algorithms with varying range of sensitivity to be included in the microprocessors.

To test the UC capabilities of the system, a wireless video and data capture system was created to log the events. A video camera was mounted inside a waterproof enclosure along with wireless communication, Global positioning systems, and digital storage.

When the sensor (hydrophone) triggered a detection, the unit would send a wireless signal to the video camera which in turn is commanded to wake and fire a five-second still frame of its current view. Due to the range capabilities of the wireless unit, the camera could be located over one kilometer away.

The camera system would use the GPS to log the time of the event and which sensor tripped the detection to digital storage. At every five -minute interval between 5AM and 7PM the unit would fire a true negative five -second frame to check if ducks are visually near the units without being detected (true negatives).

Site Trials

The prototypes built were tested to evaluate the concepts described above. A platform containing a radar, passive infrared and underwater acoustics was assembled and positioned in a small boat anchored inside a mussel lease. Two field tests were performed. During the first trial at Country Harbour, NS technical problems interfered with the proper functioning of the system and unfortunately no useful data was generated. Once fixed, the system was deployed again, this time at a mussel lease in Marie Joseph, NS.

During the field trials in Marie Joseph, triggering data was successfully collected but there were only two ducks observed in the area. It was observed that the system fired several false positives. It was concluded that the sensitivity of the algorithms used was not appropriate for this location and the detections were not necessarily attributed to the actual presence of ducks. With the most sensitive algorithm the acoustic underwater sensor would false detect about three  times a minute from sea conditions natural noises. The least sensitive algorithm would not flag detections with only two ducks in the area.

Conclusions and Recommendations

The information generated in this project was very useful to determine if an electronic approach is the correct way of dealing with the monitoring of predatory ducks in mussel farms. It was found that underwater acoustic technology is promising but more practical work is needed. It is recommended that a second phase to this project should be undertaken. The second phase would further refine the required sensitivity and algorithms required along with more extensive field trials and perhaps even develop more algorithms if none of them are useful.  Not all mollusk-eating ducks make the same noises and this should be characterized. After fine tuning the functionality, the design could be refined to fit inside a modified power-assisted buoy so that the system can be self sustaining, covered and protected. A successful second phase could be what is needed to determine if an electronic monitoring system will be useful in a commercial basis. It will also indicate if it would be feasible to couple the monitoring system with an automated deterrence device(s) (Possible phase 3). The information generated so far will be communicated through the present report to interested parties and a copy will be available through the AANS website.