Bachelor thesis, 15 ECTS
Degree project in Bachelor’s program of Biology and earth science, 180 ECTS Spring term 2021
Evaluating the use of photography for monitoring feeding habits of common
murre (Uria aalge)
Elin Rydevik
Abstract
Seabirds are often used as indicator species for changes in marine ecosystems due to the species visibility and sensitivity to changing conditions, such as changes in prey abundance. They often reside in habitats affected by anthropogenic impacts such as large-scale fisheries and pollution.
Understanding the connection between seabirds and their surrounding environment can give us important insight about the ecology of the ocean and how anthropogenic pressures affects it.
Studying feeding habits, and foraging behavior especially, is useful for understanding seabird´s responses to changing environments. Feeding studies are commonly used in seabird monitoring and requires a lot of time and resources. Monitoring of seabirds are also logistically challenging, and the risk of disturbing bird colonies must be considered. It can be especially complicated when studying cliff nesting seabirds such as the common murre, Uria aalge, the study species for this thesis. Photography as a method for monitoring seabirds may limit the need of people on site, hence minimize disturbance and save time and resources. This study provides insight in whether it is a viable option to use photography instead of on-site field studies when monitoring sea birds. This was accomplished by installing cameras and monitor a feeding study at Stora Karlsö, Sweden, parallel with performing the usual monitoring in the field. This thesis makes it clear that a camera study very well could replace the field study without any larger concerns, although, improvements need to be considered if the study is to maintain a high quality and for results to be reliable.
Key words: Seabird monitoring, time lapse photography, feeding study, prey abundance, indicator species, foraging behavior, Common murre, Stora Karlsö
Bachelor thesis, 15 ECTS
Degree project in Bachelor’s program of Biology and earth science, 180 ECTS Spring term 2021
Innehåll
1 Introduction ... 1
1.1 Background ... 1
1.2 Aim ... 2
2 Material and Methods ... 2
2.1 The study area ... 2
2.2 The Common murre and the feeding study at Stora Karlsö ... 3
2.2.1 The feeding study – collected data and study protocol ... 4
2.2.2 The feeding study - practical implementation ... 4
2.3 Filming the feeding study ... 5
2.4 Video analysis ... 6
2.5 Study protocols and statistical analysis ... 7
2.5.1 Study protocols ... 7
2.5.2 Statistical analysis of fish size error ...8
3 Results ... 9
3.1 Time and resources ... 9
3.2 Technical performance of the cameras ... 9
3.2.1 Storage and construction ... 9
3.2.2 Picture quality ... 10
3.3 Study protocols ... 11
4 Discussion ... 14
4.1 Conclusions ... 15
5 Acknowledgements ... 16
6 References ... 16
1
1 Introduction
1.1 Background
Observations of animal populations are important in data collection for research and monitoring purposes (Lynch et al. 2015), and it may even be considered being the very cornerstone of effective species conservation and management (McMahon et al. 2009). Birds specifically can be, and are in many countries, used as indicators of environmental changes such as habitat change and occurrence of contaminants in their surroundings (Koskimies 1989). Seabirds especially are often used as indicator species for changes in marine ecosystems due to the species visibility and sensitivity to changing conditions in the surrounding ecosystems (Carney and Sydeman 1999; Cury et al. 2011; Southwell and Emmerson 2015). Norway for example, has SEAPOP (Seabird populations), a long-term monitoring program for seabirds. The program aims to gain knowledge about seabirds and surrounding areas to contribute to a better management of the marine environments. Focus lies especially to model effects of human interactions that impact these environments (SEAPOP), an important area of research since seabirds often reside in habitats affected by anthropogenic impacts such as large-scale fisheries and pollution (Parson et al 2008).
Seabirds are at the top of the food chain and react to changes in lower trophic levels such as overexploitation of fish (e.g. changing prey dynamics) or pollution of the ocean, (e.g. harmful levels of chemicals in prey fish) (Parson et al 2008, Kadin et al. 2016). Understanding the connection between seabirds and their surrounding environment can, and has given us, important insights about ecology of the ocean and anthropogenic impacts, foraging
behaviors, demography and physiology amongst other things (Hentati-Sundberg et al. 2012).
Studying feeding habits and foraging behavior is useful for understanding a species response to changing environments. For example, birds might spend more time foraging when prey density is low (Harding et al. 2007; Kadin et al. 2016). Furthermore, seabirds are considered charismatic and easily gain support from the public due to their beauty and high visibility (Carney and Sydeman 1999). They thereby contribute to ecosystem services such as
ecotourism and birdwatching. Public support, along with being useful research objects, give sea birds a high conservational value (Hentati-Sundberg et al. 2012; Carney and Sydeman 1999).
Monitoring of seabirds have some challenges logistically and the risk of disturbing bird colonies must be considered while conducting research in the field. It may be especially complicated when considering cliff nesting seabirds such as the common murre, Uria aalge, that is our main character in this study (the species is further described in section 2.2.). The cliff ledges are often inaccessible, and doing research requires knowledge about ethology, sensitivity to disturbance and consideration of suitable methods in relation to these issues (Hentati-Sundberg et al. 2012; Carney and Sydeman 1999). It is of uttermost importance to minimize disturbance when conducting studies in sensitive areas, which once again implies careful consideration of which methods that are appropriate to use (Carney and Sydeman 1999; Hentati-Sundberg et al. 2012). If not taken into consideration this could lead to stress, exposure to predation and birds abandoning their breeding area, short or long term (Carney and Sydeman 1999; Rizkalla et al. 2017).
Photography can be one way of monitoring seabirds that helps minimizing interference with the colonies, hence minimizing disturbance. It is commonly used by field biologists for observation purposes. New technologies have been developed over time to increase precision, time efficiency, minimize disturbance to animals and reduce costs (Lynch, Alderman and Hobday 2015). It has been suggested that time-lapse photography can be used as a valuable tool for recording breeding productivity, foraging trip parameters, breeding success and at- nest behaviors (De Pascalis et al. 2018; Southwell and Emmerson 2015).
2
When deciding which technology that is suitable for a specific study you need to consider multiple parameters such as weather proofing, robustness, cost, camera resolution, what type of data you are collecting etc. Feeding studies, such as the one conducted for this thesis, are commonly used in seabird monitoring (Wanless et al. 2005; Kadin et al. 2016; Kadin et al.
2012). Studying feeding habits in the field requires a lot of time and personnel since there must be someone on site and monitor during the entire time of the study.
This study provides insight in whether it is a viable option to use photography instead of on- site field studies when monitoring sea birds at Stora Karlsö in Sweden. The different methods (field and camera study) are compared by performing an annual feeding study in the field as usual while simultaneously filming the birds that are subject of the study. The feeding study is then performed again based on the footage collected. The overall aim of the study is to discuss and decide if the use of camera-monitoring is a viable option to replace the field studies. There are two different cameras used that each have individual benefits and challenges which are assessed in this thesis.
1.2 Aim
The aim of this thesis is to find out whether cameras can be a viable option to study feeding behavior. The expectation is to make the monitoring of Common murre more time and resource efficient and precise.
This study addresses the following questions:
(i) Would camera-based feeding studies be more time and/or resource-efficient than field studies?
(ii) Would a filmed study provide footage with good enough resolution and visibility of the feeding events to gain reliable results?
(iii) Do the potential benefits of filming outweigh the practical disadvantages, such as carrying equipment, installation and potential technical problems?
(iiii) Is it possible to replace the feeding field study at Stora Karlsö, Sweden, and film the feeding study instead each year? If so, what type of camera is preferred?
2 Material and Methods
2.1 The study area
Stora Karlsö (57°17´N, 17°58´E) is an island located in the Baltic Sea, outside the west coast of Gotland, Sweden. The area is roughly 2.5 km2 and the island is shaped as a large plateau (figure 1) with many steep rock walls that goes straight down to the beach or into the water.
The island hosts approximately 15000 breeding pairs of common murres, which is the largest colony of Common murre in the Baltic sea. The steep rock walls with ledges are suitable breeding grounds for the species (figure 2) (Olsson and Hentati-Sundberg 2017; Ransgart 2009).
3
Figure 1 (to the left). Stora Karlsö
Figure 2 (to the right). Cliff walls inhabited by common murres.
2.2 The Common murre and the feeding study at Stora Karlsö
The common murre is an alcid bird that can be found in the Northern Hemisphere. The birds breed in large colonies and on steep ledges high up from the ground. The common murre form pairs for their entire life span and returns to the same breeding area each year and even to the same ledge and spot if possible (BirdLife International 2021; SLU Artdatabanken;
Ransgart 2009). The common murre lay one egg per year (Ransgart 2009) and both parents take part in incubation, chick-rearing and foraging. Egg-laying at Stora Karlsö starts in end of April or early May and it then takes around 30 days of incubation before the egg hatches (Kadin et al. 2012; Kadin et al. 2016). The birds live for around 30 years, but birds as old as 43 years has been recorded (Baltic seabird Project 2021). This gives the observer opportunity to follow individuals over time and study their breeding patterns thoroughly. The common murre is a top predator and a highly specialized feeder and almost exclusively feeds of sprat, Sprattus sprattus, and herring, Clupea harengus, in the Baltic sea (Kadin et al. 2012; Kadin et al. 2016). Sometimes, however, it preys on other fish such as sand lances (Ammodytes sp.) and sticklebacks (Gasterosteidae). The Common murre only catches one fish at the time, as it is a so-called single prey loader. This makes the quality of the prey, such as size and energy content, very important (Kadin et al. 2012).
Data collected about eggs, breeding success and weight of chicks of common murres can be used as indicators of the health status of the Baltic sea, supplying us with information about pollutants in the water which may affect thickness and strength of the eggshells (Pirie-Hay and Bond 2014). Changes in abundance of the main prey sprat may reflect in the birds choice of main prey, fledging chicks body mass etc. (Österblom et al. 2006). At Stora Karlsö, a monitoring program has been conducted since 1997. Monitoring includes multiple parameters such as feeding preferences, breeding success, chick survival and body mass, behavior, age and ID (based on ringing) of the birds. Ringing of chicks is also conducted every year. The monitoring program is part of “The Baltic Seabird Project”, a research project that aim to gain a larger understanding about the ecosystems of the Baltic Sea and its environmental health status. The common murre is the most thoroughly studied species in the project but there is also focus on razorbills (Alca torda) and other sea birds.
As a part of the monitoring program at Stora Karlsö there is an annual feeding study in order to monitor feeding preferences, fish size and species, feeding frequencies of chicks and how the pairs divide the work of fishing and brooding between them. This study has been used for several scientific publications such as Österblom et al (2006), Kadin et al (2012) and Kadin et al (2016). The feeding study is performed in the month of June and goes on from dusk to dawn (03:00-23:00), during 2-3 days. This way of direct monitoring and observations in the
4
field demands large amounts of resources in time, personnel and preparations. The feeding study requires at least six people on site due to the nature of the study.
2.2.1 The feeding study – collected data and study protocol
The data collected in the study is noted in a protocol and the following parameters are included: pair number, time of feeding event, fish species, fish size, failed foraging and failed feeding (table 1). There are four categories of fish size included; 1-4. What size category a fish belongs to is measured by using parent bill length as a reference; 1= less than 2 bill lengths, 2=2 1/3 bill lengths, 3= 2 2/3 bill length, 4= 3 or more bill lengths (mean bill length is set at 49mm) (Kadin et al. 2012). The different parameters in the protocol serves different purposes and are explained in table 2.
Table 1. Protocol used in the feeding study (clup is short for clupeid).
Year Month Day Ledge Pair Time Fish spec Fish size Failed foraging
Failed
feeding Comments
2018 6 14 A1 39 3:19 clup 3
2018 6 14 A1 3:22 clup x
2018 6 14 A1 22 3:32 clup 1
Table 2. Explanations of the different parameters in the feeding study protocol.
Parameter Research purpose
Time of event Shows frequency of feeding events for each
pair.
Fish species Shows prey choice. Can indicate changes in
abundance of main prey.
Fish size Paired with data on prey energy content
(Røjbek et al, 2014) indicates the energy obtained by the chick.
Failed foraging Can indicate abundance of prey. Can also
show individual behavioral differences between parent birds.
2.2.2 The feeding study - practical implementation
The cliff ledge chosen for the Baltic Seabird project’s feeding study is labelled A1 and is inhabited by around 40 breeding pairs of Common murre. The ledge is located 50 meters above a rocky beach and 3,5 meters below the plateau above it (figure 3).
Figure 3. The A1 ledge circled in red.
5
It takes at least six people taking part in the observation for the shifts to be manageable. Each shift is three hours long and there are two people present by the ledge at each shift (figure 4 and 5).
Figure 4. Looking over the cliff down on the A1 ledge. Figure 5. One is watching, the other writes protocol.
Only pairs of Common murre with hatched chicks are monitored in the feeding study. Every time a parent bird lands on the ledge the time for this is noted in the protocol displayed in table 1, and what type of event that has occurred (e.g. bird arrives with a fish). Binoculars are used to identify ringed birds when spotted on the ledge, however, this is not part of the feeding protocol. It is done spontaneously when possible and noted in another protocol. The observation goes on continuously during the time period of 03:00 to 23:00 with no gaps in the observations, i.e., the ledge is always manned during the study.
2.3 Filming the feeding study
For this thesis I chose to use data from two years, 2018 and 2019. For each year one observation day was chosen for comparison. The data from the two days, and the two
consecutive years, include both filmed and direct observations in the field. I chose to include both years since there was a new camera installed in 2019. This gave me the opportunity to compare the performance of the two different cameras and see how the picture quality (among other things) impact the study results.
For the 2018 study I used a 1.3 MP time-lapse camera with a wide-angle lens; Brinno TLC 200 PRO, and a GoPro camera. The GoPro however, had to be eliminated from the study due to the size of the memory card. The storing capacity was not enough to enable recording 20 hours of time lapse photography.
The Brinno TLC 200 PRO camera was used together with a weatherproof housing (Brinno ATH120) to protect it from weather impact and bird faeces. I used a laser rangefinder to measure the distance between the position of the camera and the ledge (A1) below it. This was done so I could set the focus manually before putting the camera in place. The camera was mounted on a metal rod (figure 6). The Brinno TLC 200 PRO is battery driven, so no cords were needed (cords can be seen in some photos but these are for the GoPro which, as mentioned above, was eliminated from the study). The metal rod was carefully put in place by very slowly lowering it from the cliff above the ledge, in order not to disturb the birds. The metal rod was firmly secured so it would not risk falling or move. When in place the camera was positioned 2,3 meters above the A1 ledge, (figure 7).
6
Figure 6. Both GoPro and Brinno TLC 200 PRO mounted to the metal rod.
Figure 7. Metal rod put in place with cameras mounted.
For the 2018 study I used time lapse photography with the interval of 1 frame per second.
This made it possible to detect and observe the same events as when watching in the field, i.e.
when a bird flies in with a fish, if the feeding is successful, if partners were shifting places etc.
The camera was taking pictures the entire study period (03:00 to 23:00).
For 2019 the Baltic Seabird Project installed a permanent camera for filming the birds on the A1 ledge. The camera is a 2x zoom, 2 MP, OnVif CCTV camera; Avtech AVM543 (figure 8).
The camera is connected to a recording unit by ethernet cable. It has a weatherproof body and is mounted on a metal rod 7 meters above the A1 ledge. The camera has infrared lighting which make the picture bright and sharp even when it is dark. The CCTV camera is filming continuously (not a time lapse camera) and are permanently installed during the breeding season. Photography from the same time interval as for the field study was used (03:00 to 23:00).
Figure 8. OnVif CCTV camera; Avtech AVM543 (not the camera used in this particular study).
2.4 Video analysis
The A1 ledge usually hosts 40 breeding pairs each year. All pairs with hatched chicks are included in the feeding study. For this thesis however, I chose to include 12 (2018) and 15 (2019) of the pairs located on the right side of the ledge to make the study more manageable,
7
(figure 9). The reason the number of pairs differs between years is simply because the number of hatched chicks were higher in 2019.
Figure 9. The “map” of breeding pairs from the 2019 feeding study. Pairs with hatched chicks on the right side of
the red line was included for both 2018 and 2019.
When analyzing the photography from both cameras I used the video software “VLC media player”. It allowed me to look at the time lapse photography from the time lapse camera as a film and it is also possible to zoom interactively while watching. I used the same paper protocols as during the field observations and mostly followed the same procedure with a few exceptions. The exceptions were that:
• I both watched and filled out protocols by myself (instead of being in a team of two),
• I had the possibility to stop the film when I had to fill in the protocol or needed to confirm what I just saw,
• I could rewind if I thought I missed something,
• I could decide when to look at the material and do small parts at a time.
Since the goal was to compare the two methods (field studies and camera studies), I avoided to look at the field observation protocols during the time I went through the recorded material. This was to make sure that the human factor was not eliminated. In the field it is possible to miss a feeding event since there is much going on at the ledge at the same time.
The camera studies also needed to include this possibility for it to be comparable.
2.5 Study protocols and statistical analysis
2.5.1 Study protocols
When comparing the field protocols with the camera protocols initially and comparing results between the two different cameras and years, I used MS Excel to create a spreadsheet with the parameters of interest (table 3).
The columns of table 3 are explained below:
• “Feeding number” tells us which number that particular feeding event has, starting with the one earliest in the morning.
• “Camera type” defines if it was the TLC 200 PRO (in the table noted as Brinno) or the Avtech AVM543 (in the document noted as CCTV).
• “Date” is the same for field and camera protocols, the time of a feeding event is noted separately in the column “time field protocol” or “time video protocol”.
• If the field and camera protocol was matching completely this is marked with a Y in the “complete match camera/field protocol” column. “Complete match camera/field protocol, time not included” is marked with a Y if the protocols are matching except
8
from the time. This column was added because of a difference of 2-3 minutes between camera and field protocols for almost every feeding event. This delay was consistent over time and not due to human error (i.e. noting time incorrectly). For this study I only use the result without time included. Time difference is noted in the “time inaccuracy” column.
• “Feeding opportunity missing in video” and “Feeding opportunity missing in field protocol” is marked with Y for yes, and N for no, stating if one or the other has better coverage of all the events during the study.
• Every bird pair on the ledge is marked with a number, the column “pair confused” is marked with a Y if the field and camera protocols showed different pair number. By looking at the footage afterwards it was possible to see which number that was right and wrong.
• “Species of fish match” is marked with a Y if fish species match between protocols,
• “Size of fish error” shows if there is a difference in noted fish size. For example, -1 means that the size noted in the camera protocol was one size smaller than the one noted in the field protocol, 0 means they were the same. For this thesis I choose to consider the results from the field protocol as the correct answers I compare the camera results to. Since the field study is used actively for research it is considered to deliver reliable results.
• “Failed feeding match” and “failed foraging match” is marked with a Y for Yes if field and camera protocols match.
• “Pair” shows the number of the bird pair on the ledge.
Table 3. Table for comparison of data between field protocols, camera protocols and between different cameras.
Only a subset of the dataset is shown.
2.5.2 Statistical analysis of fish size error
The performance of the two cameras was analyzed by means of analyzing their error of estimated fish size:
• Originally a categorical data with fish size classes 1-4
• Error = field observation size – camera observation size
• Mean error in camera below or above 0 would indicate that cameras are misconceiving sizes compared to the field protocols.
• Larger variance between one camera than another would indicate that the lower variance camera is preferred.
• Mean value and variance were tested by performing a permutation test (also called randomization test). The test was performed in the R software for statistical computing v. 4.0. A probability distribution was created by randomly mixing data values and categories (cameras), 1000 permutations was done.
9
3 Results
The installing of the time-lapse camera triggered some minor reactions from the birds such as warning calls and cautious looks, but no birds left the ledge. No reactions were seen when installing the CCTV camera. During the feeding study the birds showed no behavioral
changes.
3.1 Time and resources
The time spent for watching the time lapse film material, and filling out protocols from 2018, was 29 hours. Total time, including tending to practical issues with the computer, media player etc., was 34 hours. The time spent for watching the film material and filling out protocols from 2019 was 23 hours. Total time, including tending to practical issues with the computer, media player etc., was 24 hours. See table 4.
Table 4. Time spent watching films and filling out protocols and total time spent, including tending to practical issues with the computer, media player etc.
I estimate that the preparations for the camera study, compared to the field study, takes the same amount of time. The following results were obtained when comparing the resources needed for the two different cameras and field observations. The results below are based on three days observation, which is the maximum number of days for the feeding study:
The field study requires 6 people and a total of 60 hours (if the study is performed over three days) to perform the study. All people must be on the island at the time of the study.
The time lapse film study requires 1-2 people on the island at the time of preparing and installing the camera. It then takes 1 person and a total of 102 hours (if the study is
performed over three days) to go through the films and write protocol. This person does not have to be at the island.
The CCTV camera study requires 1-2 people on the island at the time of preparing and installing the camera. It then takes 1 person and a total of 72 hours (if the study is performed over three days) to go through the films and write protocol, this person does not have to be at the island.
3.2 Technical performance of the cameras
3.2.1 Storage and construction
Brinno TLC 200 PRO: With a 64GB memory card it was possible to record for 20 hours without running out of storage space. The weatherproof housing kept the camera from getting wet or dirty and the mount on the metal rod was sturdy and kept the camera in place.
The batteries had enough power to keep the camera running for 20 hours and it still had power left.
Avtech AVM543: The recording unit can store around 10 Terabyte, there were no problems concerning storage or power supply. The weatherproof housing kept the camera from getting wet or dirty and the mount on the metal rod was sturdy and kept the camera in place.
Time-lapse, 2018 CCTV, 2019 Time watching films, h 29 23
Total time, h 34 24
10 3.2.2 Picture quality
Picture quality differed between cameras. When looking at zoomed out pictures from the study, the difference did not seem too big (figure 10). When watching the films on a big screen and filling out protocols however, it became clear that the Avtech camera has considerably sharper footage than the time-lapse camera.
Figure 10. The A1 ledge. Brinno TLC 200 PRO and Avtech Avm543
For the time-lapse camera, the picture quality was quite poor over-all, even when light conditions were good. It was sometimes hard to distinguish the birds and pairs on the ledge.
The time-lapse camera had a hard time adjusting to changing light conditions and the picture easily got overexposed. The CCTV camera had better picture quality over all and adjusted well to changing light conditions. Identifying fish size and species was challenging for both cameras, but the CCTV camera footage was sharper hence made it easier to identify these parameters (figure 11). It was not possible to read rings by using footage from any of the cameras.
Figure 11. Common murres carrying fish. Brinno TLC 200 PRO and Avtech Avm543.
11
Regarding visibility in darkness, the CCTV camera performed considerably better than the time-lapse camera, for which it was barely possible to distinguish the different pairs from the time of 22:30 and forward (figure 12).
Figure 12. The A1 ledge. Time: 22:50.
3.3 Study protocols
When comparing the results between field protocols and camera, and then comparing data between the two cameras the following results was obtained.
For the year of 2018 and the time-lapse camera, 79 feeding events were observed, for 2019 ant the CCTV camera, 72 feeding events were observed, i.e. there was 7 more events that occurred during 2018. Table 5-9 shows the number of times a certain parameter was noted out of the total amount of feeding events.
Table 5 shows how many complete matches between the camera protocol and the field study protocol there was. Avtech Avm543 (CCTV) had the highest number of matches with 58 % and Brinno TLC 200 PRO (time-lapse) had the lowest with 37%.
Table 5. Number of complete matches between camera and field protocol for each year and camera type, time not included.
Camera
Complete protocol match, time not included
Time-lapse 29/79
CCTV 42/72
Table 6 shows the number of matches between field and camera protocol regarding failed foraging (if a parent arrives on the ledge without bringing a fish for the chick), and failed feeding (if the chick fails at eating the fish). A match is noted as “Yes”. If one protocol, e.g., the camera protocol, has recorded a failed feeding or foraging, and the corresponding field protocol have noted it as successful or anything else that do not match the camera protocol, it is marked as “No”. None of the cameras had any failed feeding matches. Brinno TLC 200 PRO (Time-lapse) had 3% failed feeding “non-matches”, 9% failed foraging matches and 6%
failed foraging non-matches. Avtech Avm543 (CCTV) had 1% failed feeding “non-matches”, 6% failed foraging matches and no failed foraging “non-matches”. Data not available
represents all the feeding events that were not a failed feeding or foraging (i.e. fish brought to the ledge and species noted and feeding successful).
12
Table 6. Number of matches between field and camera protocols regarding failed feeding and failed foraging.
Camera
Failed feeding match - Yes
Failed feeding match - No
Failed foraging match - Yes
Failed foraging
match - No Data not available
Time-lapse 0/79 2/79 7/79 5/79 65/79
CCTV 0/72 1/72 4/72 0/72 67/72
Sometimes the pair number got confused between field and camera protocol. This happened 3 times (4%), for Brinno TLC 200 PRO (Time-lapse) and none for Avtech Avm543 (CCTV).
All 3 times the correct pair was noted in the video protocol and wrong pair noted in the field protocol.
Table 7. Shows if pair number have been confused between camera and field protocol.
Camera Pair confused? Yes Pair confused? No
Time-lapse 3/79 76/79
CCTV 0/72 72/72
Species of fish matched 70% of times for Brinno TLC 200 PRO (Time-lapse) and 89% of times for Avtech Avm543 (CCTV). The remaining times there were no data available, this happens when fish species only is noted in one protocol or when failed foraging is noted. It occurred 30% of times for Brinno TLC 200 PRO and 11% for Avtech Avm543.
Table 8. Shows how many times species of fish matched between protocols and how many times data were not available.
Camera
Species of fish match?
Yes
Species of fish match?
No
Species of fish match?
Data not available
Time-lapse 55/79 0/79 24/79
CCTV 64/72 0/72 9/72
Feeding opportunity was missed 2 times (3%) for Brinno TLC 200 PRO (Time-lapse) (the feeding opportunities was noted in the field protocol but not in the camera protocol). No feeding opportunities were missed for Avtech Avm543 (CCTV).
Table 9. Number of feeding opportunities that were missed for each camera.
Camera
Feeding opportunity missing in video?
Yes
Feeding opportunity missing in video?
No
Time-lapse 2/79 77/79
CCTV 0/72 72/72
13
Fish size as estimated in the field and from the camera footage are shown in figure 13. Size 2 is the most common size across both field and camera studies. Fish size is categorized 1-4 (1=
less than 2 bill lengths, 2=2 1/3 bill lengths, 3= 2 2/3 bill length, 4= 3 or more bill length (Kadin et al 2012)). Errors in fish size determination is shown in figure 14.
Figure 13. Frequencies of fish size categories noted in field and camera study protocols.
Mean values of fish size error for the Time-lapse and CCTV camera did not differ much and were close to zero, variance did however show bigger differences, see table 10.
Table 10. Mean value and variance of fish size error for Brinno TLC 200 PRO (time-lapse) and Avtech Avm543 (CCTV)
Camera Mean Variance
Time-lapse -0.03278689 0.9322404
CCTV 0.04615385 0.4197115
The permutation test showed no significant difference in mean values between cameras. A test of variance did however show that Brinno TLC 200 PRO has a larger variance than Avtech Avm534 and the difference is statistically significant. Figure 14 display how frequent both cameras had the wrong size of fish noted compared to the field protocols (Fish size error; -1=one size too small in camera protocol compared to field protocol, 1=one size too big, etc.).
Figure 14. Frequencies of fish size error for each camera and year.
0 5 10 15 20 25 30 35 40
-3 -2 -1 0 1 2 3
Fish size error
2018 time-lapse 2019 CCTV 0
5 10 15 20 25 30 35 40 45 50
1 2 3 4
FREQUENCY
FISH SIZE CATEGORIES 1-4
Frequencies of fish size
2018 Field study 2018 Time-lapse 2019 Field Study 2019 CCTV
14
4 Discussion
The broad goal for this thesis is to assess the potential in using cameras for monitoring feeding habits of the common murre, and to conclude which of the two cameras used in the study that was more suitable for this purpose. It is clear that the results from the CCTV camera, Avtech Avm543, are closest to the results from the field protocol. There were no apparent practical disadvantages with using cameras instead of field studies.
It has to be mentioned again, before discussing the results obtained, that this thesis relies on the field protocol as the correct “answers” for comparison between protocols. Since the field study is used actively for research it is considered to deliver reliable results. However, one has to be aware that all the results from the field study might not always be correct. The
assessment of fish size for example, is made in a second or two in the field, and different people are taking part in the study that might make different assessments. This leaves room for misjudgment and sometimes not noticing a feeding event at all. This thesis gives insight in whether a camera-based feeding study could be at least as reliable as the field studies conducted regularly on Stora Karlsö.
The use of photography for monitoring sea birds and feeding habits are supported by results from several studies such as Sothwell and Emmerson (2015) who utilized a remotely operated camera to monitor breeding success among Adèle penguins in Antarctica. The results from the camera-based study correlated strongly to field observations. De Pascalis et al. (2018) demonstrated that time lapse photography can be a useful and flexible tool in seabird ecology by using time-lapse photography for monitoring black-legged kittiwakes foraging trip
duration, predation, attendance patterns and assessing effects of disturbance. Minimizing disturbance on bird colonies while doing research is of great importance (Carney and
Sydeman 1999; Hentati-Sundberg et al. 2012) and during the feeding study the birds showed no behavioral changes except for when installing the time-lapse camera, with some minor reactions such as warning calls and cautious looks, but no birds left the ledge. During the feeding study the birds showed no behavioral changes.
The permutation test performed on fish size error demonstrates a lower variance in fish size error for the CCTV camera than the time lapse camera. This indicates that even if the number of times a fish got noted as the wrong size was similar between cameras, the error was
generally smaller for CCTV. This implies that it is easier to assess fish size from the CCTV camera. Since the assessment of fish size in the field is done quickly and from a distance of 7 meters there is no guarantee that size is noted accurately each time in the field either. Hence a complete match between field and camera protocols regarding fish size was not expected.
The CCTV camera has a considerably higher percentage of complete matches in protocols with 58% than the Time-lapse camera with 37%. The percentage of non-matches does not necessarily mean that the camera fell short, it also depends on events that are wrongly recorded in the field protocol. Pair confusion only occurred three times in total, all of them for the 2018 feeding study. The right pair was noted in the camera protocols all three times, showing the advantage of being able to rewind and re-watch the footage. Species of fish matched 89% of the times for CCTV and 70% of the times for time-lapse. The time-lapse camera had 2 missing feeding opportunities from the video protocol and the CCTV hade none. These results mentioned above suggests that the CCTV Camera is more suitable for monitoring than the time lapse camera.
When taking technical performance into consideration as well as results from study protocols it is clear that the CCTV camera excels at data storage capacity, power supply (no batteries that can run out of power) and picture quality. It is also clear that the use of the CCTV camera is more resource efficient than the field study (in terms of people needed) and more time efficient than the time-lapse camera. The CCTV camera study requires 1-2 people on the island for installing equipment and then 72 hours to go through the footage for a three-day feeding study. The time-lapse camera study requires one or two people on the island for installing equipment and then 102 hours to go through the footage. The field study requires
15
six people and 60 hours to perform the study, this means that 4-5 more persons than for a filmed study has to be transported to and from the island and take part in the study.
Transportation and personnel costs money and it might not always be possible to recruit enough volunteers and/or personnel. Of course, there is an initial cost when investing in camera technology, but depending on how high the cost is, it still might be less expensive over time. This should be investigated further to get a clear view of potential economical benefits of using a camera instead of field studies.
The fact that the CCTV camera performs better seems to depend on technical aspects such as storage, picture resolution, power supply etc., rather than the type of camera, i.e. CCTV or time-lapse.
It is worth noting that the CCTV camera is positioned 7 meters above the A1 ledge as opposed to the time lapse camera that was lowered to 2.3 meters above the ledge. The CCTV camera is hence less likely to disturb the birds when installing the equipment, doing maintenance etc. It seems that the birds accept both the cameras when installed though. Since the CCTV camera has continuous power supply through the ethernet cable there is no need to move it more than occasionally to do maintenance if needed, hence minimizing human activity close to the ledge. If the feeding study is camera based the total amount of time spent by people by the ledge will be considerably lower, especially if the CCTV camera is used, minimizing
disturbance throughout the field season compared to if the study is done in the field.
Sea birds are already considered to have a high conservational value and easily gain public support (Hentati-Sundberg et al. 2012; Carney and Sydeman 1999) and it is important that we maintain this status. With camera technology it is easy to collect footage for websites and social media. With the CCTV camera it is possible to follow the birds live and engaging people in the research to create a stronger connection between the public and the common murre.
Studying the common murre and other sea birds that can be used as indicators of the health status of the Baltic sea is an important task. There is still much that we do not know about these relationships due to the challenging nature of measuring foraging behavior and food supplies amongst marine predators (Harding et al. 2007). This implies that further
development of monitoring is required to fully take advantage of the potential in sea bird monitoring. Kadin et al. (2016) showed that the number of fishes brought to the chicks partly depended on the age of the chick. They also showed that the Murres at Stora Karlsö adjusts provisioning rates according to fish size. This makes a case for further improvement of picture quality when conducting feeding studies. Fish species and its size play an important role in research of the common murre and therefore it is not suitable to use camera technique that cannot measure these factors properly. It was quite hard to get the right fish size for both cameras. The CCTV camera however, had sharper photos and more detail. This is reflected in the results where the CCTV camera has a higher number of correctly noted fish size and lower variance in fish size error. Even so, there is room for improvement such as the ability to zoom in at a specific target with high resolution.
4.1 Conclusions
The results from this study are rather straightforward. It seems clear that a camera study very well could replace the feeding field study without any larger concerns. The difference in results between the CCTV camera and the field study are in my opinion acceptable but shows that improvement is needed. There are some factors that need to be considered if the study is to maintain a high quality resulting in reliable results. The choice of camera is important in order to obtain reliable data. A camera should be used with at least the same performance level as Avtech Avm543 to monitor the feeding study on the A1 ledge, since the ledge hosts many bird pairs and there is a high level of activity. I would suggest the use of a camera with features such as: better picture quality (higher resolution, better autofocus), more powerful zoom and a joystick to target certain birds. If one would be able to zoom in and see ring
16
numbers, this would be an important improvement since both cameras fell short on this function. It seems however that the concept of using a CCTV camera is a good option since the technology is available and well developed due to its usage in regular surveillance for many years. There is no need to use time lapse photography since storage space is not a problem and it did not take less time to process the footage from the time-lapse camera. If suitable adaptations are made, I see no obstacles with replacing the feeding field study with a CCTV camera or similar technology. This also applies to other studies and monitoring on Stora Karlsö, as well as seabird monitoring around the world. In light of recent events with the ongoing pandemic, the benefits of remotely operating cameras become especially clear. It allows research to go on even when there is limited possibility to have people on site.
5 Acknowledgements
First and foremost, I would like to thank my mentor Jonas Hentati Sundberg for great patience and guidance through the writing process. I also want to thank Aron Hejdström for all the help with the practical aspects of the field work. It has been inspiring to work closely with two key persons in the Baltic Seabird project. A larger interest of the Baltic sea and its inhabitants have been born and I look forward to engaging further in the subject. At last I would like to thank my family and friends that happily have listened to me talking about subjects they know nothing about and helped when feeling stuck.
6 References
Baltic seabird Project. 2021. Study species. http://www.balticseabird.com/studiearter/
(Downloaded 20/3/2021).
BirdLife International (2021) Species factsheet: Uria aalge.
http://datazone.birdlife.org/species/factsheet/common-murre-uria-aalge (Downloaded 15/03/2021)
Carney, K. M., and Sydeman, W. J. (1999). A Review of Human Disturbance Effects on Nesting Colonial Waterbirds. Waterbirds: The International Journal of Waterbird Biology, 22(1): 68–79. https://doi.org/10.2307/1521995
Cury, P. M., Boyd, I. L., Bonhommeau, S., Anker-nilssen, T., Crawford, R. J. M., Furness, R.
W., Mills, J. A., Murphy, E. J., Österblom, H., Paleczny, M., Piatt, J. F., Roux, J-P., Shannon, L. and Sydeman W. J. (2011). Global Seabird response to Forage Fish Depletion - One third for the birds. Science, 334(Dec): 1703-1706.
10.1126/science.1212928
De Pascalis, F., Collins, P. M., and Green, J. A. (2018). Utility of time-lapse photography in studies of seabird ecology. PLoS ONE, 13(12): 1–17.
https://doi.org/10.1371/journal.pone.0208995
Røjbek, M. C., Tomkiewicz, J., Jacobsen, C., and Støttrup, J. G. (2014). Forage fish quality:
seasonal lipid dynamics of herring (Clupea harengus l.) and sprat (Sprattus sprattus L.) in the baltic sea. Marine Science. 71: 56–71. doi:10.1093/icesjms/fst106
Harding, A. M. A., Piatt, J. F., Schmutz, J. A., Shultz, M. T., Van Pelt, T. I., Kettle, A. B., and Speckman, S. G. (2007). Prey density and the behavioural flexibility of a marine predator: The common murre (Uria aalge). Ecology, 88(8): 2024–2033.
https://doi.org/10.1890/06-1695.1
Hentati-Sundberg, J., Österblom, H., Kadin, M., Jansson, Å., and Olsson, O. (2012). The Karlsö murre lab methodology can stimulate innovative seabird research. Marine Ornithology, 40(1): 11–16.
Wanless, S., Harris, M. P., Redman, P., and Speakman, J. R. (2005). Low energy values of fish as a probable cause of a major seabird breeding failure in the North Sea. Marine ecology progress series 294: 1–8.
Kadin, M., Österblom, H., Hentati-Sundberg, J., and Olsson, O. (2012). Contrasting effects of
17
food quality and quantity on a marine top predator. Marine Ecology Progress Series, 444(January): 239–249. https://doi.org/10.3354/meps09417
Kadin, M., Olsson, O., Hentati-Sundberg, J., Ehrning, E. W., and Blenckner, T. (2016).
Common Guillemot Uria aalge parents adjust provisioning rates to compensate for low food quality. Ibis, 158(1): 167–178. https://doi.org/10.1111/ibi.12335
Koskimies, P. (1989). Birds as a tool in environmental monitoring. Annales Zoologici Fennici 26(3), 153–166.
Lucia, M., Strøm, H., Bustamante, P., Herzke, D., and Gabrielsen, G. W. (2017).
Contamination of ivory gulls (Pagophila eburnea) at four colonies in Svalbard in relation to their trophic behaviour. Polar Biology, 40(4): 917–929.
https://doi.org/10.1007/s00300-016-2018-7
Lynch, T. P., Alderman, R., and Hobday, A. J. (2015). A high-resolution panorama camera system for monitoring colony-wide seabird nesting behaviour. Methods in Ecology and Evolution, 6(5): 491–499. https://doi.org/10.1111/2041-210X.12339
McMahon, C. R., Bester, M. N., Hindell, M. A., Brook, B. W., and Bradshaw, C. J. A. (2009).
Shifting trends: Detecting environmentally mediated regulation in long-lived marine vertebrates using time-series data. Oecologia, 159(1): 69–82.
https://doi.org/10.1007/s00442-008-1205-9
Olsson, O., and Hentati-sundberg, J. (2017). Population trends and status of four seabird species (Uria aalge, Alca torda , Larus fuscus , Larus argentatus) at Stora Karlsö in the Baltic Sea. Ornis Svecia 27: 64–93. https://doi.org/10.34080/os.v27.19558
Pirie-Hay, D., and Bond, A. L. (2014). Thickness of Common murre (Uria aalge) eggshells in Atlantic Canada. Canadian Field-Naturalist, 128(1): 72–76.
https://doi.org/10.22621/cfn.v128i1.1553
Faulhaber, C., Schwarzer, A., Malachowski K., Rizkalla, C. E., and Cox, A. (2017). Effects of human disturbance on shorebirds , seabirds , and wading birds : Implications for Critical Wildlife Areas. August. Florida Fish and Wildlife Conservation Commission.
IHR 2016-003.
SEAPOP. Om SEAPOP. https://seapop.no/om-oss/ (Downloaded 19/3/2021).
SLU Artdatabanken. Artbestämning - Sillgrissla.
https://artfakta.se/artbestamning/taxon/uria-aalge-100144 (Downloaded 15/3/2021).
Southwell, C., and Emmerson, L. (2015). Remotely-operating camera network expands Antarctic seabird observations of key breeding parameters for ecosystem monitoring and management. Journal for Nature Conservation, 23: 1–8.
https://doi.org/10.1016/j.jnc.2014.11.002
Ransgart, M. (red.). 2009 .Fågelguiden: Europas och Medelhavsområdets fåglar i fält. 2.
edition. Bonnier fakta
Österblom, H., Casini, M., Olsson, O., and Bignert, A. (2006). Fish, seabirds and trophic cascades in the Baltic Sea. Marine Ecology Progress Series, 323: 233–238.
https://doi.org/10.3354/meps323233
