Abundance and Behavior of Capelin Schools (Mallotus villosus) from Aerial Surveys

This paper was presented at the Forage Fishes in Marine Ecosystems Alaska Sea Grant College Program, AK-SG-97-O1, 1997

Brian S. Nakashima Department of Fisheries and Oceans. Northwest Atlantic Fisheries Centre, St. John's, Newfoundland, Canada

Gary A. Borstad G.A. Borstad Associates Ltd., Sidney, British Columbia, Canada

ABSTRACT

The school area index based on aerial surveys of capelin aggregations near spawning beaches in Newfoundland has been considered with other indices in assessments since 1985. Trends in relative stock size from the aerial survey are significantly correlated with other indices such as catch rate data. The earlier surveys used color photography until 1991 when a digital imaging technique was used. The change has increased the capability to collect data more often. The distribution of school sizes within transects suggests that survey strategy should insure all large schools are measured at the expense of missing small schools. Image enhancement makes school margins easier to identify and therefore area estimates are more objective. The increase and buildup of schools during the spawning season corresponds to estimates of egg deposition. Large schools tend to have irregular and elongated shapes which may consist of male capelin near spawning beaches. Small schools can have more diverse shapes but tend toward regular and circular shapes. These schools are probably males and females arriving in the spawning areas.

INTRODUCTION

Capelin (Mallotus villosus) is a pelagic schooling fish that makes up a significant portion of the diet of many commercially important fish species, marine mammals, and seabirds (e.g., Winters and Carscadden 1978) in the northwest Atlantic Ocean. There has been a market-driven inshore commercial fishery since the late 1970s to supply Japanese demand for frozen females with roe. The fishery is prosecuted mainly with purse seines and capelin traps on prespawning aggregations of age-3 and age-4 fish as they migrate inshore to spawn on gravel beaches along the Newfoundland coast.

Annual assessments of the Northwest Atlantic Fisheries Organization (NAFO) Div. 2J3KL capelin stock consider data from several sources to estimate stock size and year-class strength. The surface area of capelin schools near spawning beaches has been considered as a fishery-independent index of mature biomass since 1985 (Nakashima 1985). In the 1980s the annual trend in estimates of school areas in coastal areas derived from aerial photographic surveys compared favorably with annual trends in stock size derived from more familiar indices such as commercial catch rates and hydroacoustic estimates of biomass (Carscadden et al. 1994). Since 1991 there has been a divergence in the perception of the size of the stock depending on which index was accepted (Carscadden and Nakashima 1996). Recent assessments have, however, given more consideration to trends in stock size observed from inshore indices such as inshore catch rate data, egg deposition studies, and aerial surveys (Anon. 1996).

Capelin schools near the surface and close to shore are easily identified by their gray-blue color, well-defined shapes, and movements. The first aerial photographic surveys of capelin schools were conducted in 1982 using an aerial mapping camera at an altitude of 454 m along the coastline of Conception Bay and Trinity Bay in Newfoundland (Figure 1) over defined transects (Nakashima 1985). Transects were chosen because of their proximity to the St. John's airport and the majority of capelin landings were from these two bays. The timing of flights was designed to correspond to the arrival, congregation, and spawning of mature capelin on spawning beaches in June. Surveys were designed to photograph the schools as frequently as possible during a 20-day interval chosen 6-8 weeks in advance to facilitate aerial survey company preparations and aircraft work schedules. Weather determined the frequency of coverage and often limited data collection. For example, in June 1986 the transect in Trinity Bay was covered only once because of consistently poor photographic conditions compared to seven overflights of the same transect in 1984.

Figure 1. Aerial survey transects established in Conception Bay and Trinity Bay Newfoundland.

Experimental flights using a prototype Compact Airborne Spectrographic Imager (CASI) in 1988-1989 showed that capelin and herring schools can be detected in a wavelength range of 440-540 nm (Nakashima et al. 1989, Borstad et al. 1992). Estimates of the areas of 20 schools (range = 100-23,000 m) using aerial photography and digital imaging techniques were significantly correlated (r2 = 0.98; Nakashima and Borstad 1993). Both the film negatives and the hard copy have limited dynamic range. Therefore, color photography requires bright, sunny conditions to obtain clear photographs. By comparison, digital imagery can be manipulated using image processing to facilitate school recognition. For this reason, digital imagery can be collected even when the sky is overcast, a common condition in June when capelin are most likely to spawn (Templeman 1948). For these and other reasons digital imagery has replaced aerial photography as the principal method of counting schools and estimating areas since 1991.

METHODS

The CASI is an imaging spectrometer manufactured by ltres Instruments Inc., which operates in a range of 423-946 nm (Borstad et al. 1992). Mounted in a small unmodified aircraft, it takes the place of an aerial film camera in our present operations. Digital imagery which is 512 pixels wide is acquired in up to 15 coregistered spectral channels and later calibrated to upwelling radiance and corrected for variations of aircraft altitude. The instrument configuration is varied between flights to optimize signal levels, which depend on ambient light levels, aircraft speed, altitude, and the number and width of bands used. Typical operating ranges are: ground speed 140-220 km/h, altitude 1,050-1,350 m, integration times of 35-50 msec, and four spectral channels.

The current capelin aerial survey using the CASI is conducted at 1,220 m altitude over three transects (Figure 1) established during the earlier years using only aerial photography. Optimal conditions of low winds and reduced glare occur early in the morning between 0700 and 1030 NDT. However, flights are conducted throughout the day as required. We collect digital data using four spectral channels: 475-500, 52 5-590, 650-6 70, and 745-75 5 nm (Figure 2) which is less than the eight bands flown during the comparative study (Nakashima et al. 1989). These bands can be widened when light conditions are poor either before or during the flight. Capelin schools are identified by experienced spotters prior to digitally recording the information. If there is any doubt as to the presence of schools, data are collected for later examination. Data can be collected continuously along a transect without having a spotter in the aircraft; however, the volume of data for post-processing would be considerable. Therefore, the present compromise combines human detection with imagery from observed schools. For each data set or flight line the ground speed, altitude, and integration times are recorded to calculate the relevant pixel size (usually less than 2m).

Figure 2. The spectrum of upwelling radiance from a capelin school (thin line) and the surrounding sea (heavy line) with spectral bands chosen to image these differences (shaded areas).

After it has been calibrated and corrected for distortions resulting from aircraft roll and pitch, the imagery is examined on an image-processing workstation and school areas are calculated. Here again we use a mix of human recognition and machine processing. The operator visually identifies each school and places a cursor on it. The software compares the digital value of the "seed" pixel with its neighbours, expanding outward. It continues to expand the neighboring area outward until it senses a large change in the histogram of the accumulated points, which occurs when it begins to add pixels outside of the school. A graphics overlay is created at the same time to give the operator visual feedback. On an 80486 processor, this operation takes a few seconds for each school. The operator can override the machine calculation, or adjust it by drawing in interpreted edges which the algorithm cannot find. In practice, the operator identifies the schools and calculates the graphics in a series of images, then runs a batch program which calculates the statistics and creates a text file which can be imported into a spreadsheet program. These text files are labelled according to the image file number, and for each school in the file contain image coordinates, school area, school perimeter, and a "roundness index" (ratio of perimeter of a school to the perimeter of a round school with similar area).

The daily estimate of area is based on the total area of all capelin schools per day detected along a transect. The annual index is the sum of the peak estimate for each transect. This method assumes fish density in each school is uniform spatially and among years. We also assume that maximum total school area is reached at peak spawning and all schools observed prior to or following the peak have already been accounted for.

RESULTS

The estimates of school areas from analysis of digital imagery in the 1990s have been combined with the data from aerial photography in the 1980s to formulate an index of relative mature stock size for 1982-1995. The aerial survey index is evaluated along with other indices in formulating annual advice for capelin assessments (Carscadden et al. 1994, Anon. 1996). The school area index is significantly correlated (r2 = 0.58, F= 12.49, n = 10) with the pattern observed in the commercial trap catch rate series from the inshore fishery (Figure 3). The aerial survey period was chosen to overlap the commercial fishery in NAFO Div. 3L on the east coast of Newfoundland, but the two indices could not be compared in 1991, 1994, and 1995. In 1991 the aerial survey was over before capelin had moved inshore to spawn and in 1994-1995 catch rate data were unavailable. The capelin fishery was very small in 1994 and did not open in 1995 due to the high proportion of small females in the spawning biomass which did not meet Japanese market expectations for large females. Data from the 1995 aerial survey provided the only estimate of year classes that would contribute to the 1996 spawning biomass (Anon. 1996).

The choice of using the total area of all schools on the day of peak spawning as an index assumes there is one major spawning peak. In some years (1984, 1987, 1992, 1993, 1995) it was possible to identify a second spawning mode during the survey period which occurred at least 7 days after the first spawning run. Incorporating these with the single peak estimates raised overall annual estimates; however, the year-to-year trends did not change substantially (Figure 4). A typical pattern of increasing area of capelin schools during the spawning season in coastal waters is shown for the transect in Trinity Bay for July 7-19, 1994 (Figures 5 and 6). The main aggregation of mature fish was observed July 13-15, which corresponds to the highest egg deposition observed on Bellevue Beach, Trinity Bay, in July 1994 (Figure 5). In this case one peak in spawning was observed.

The highest priority of the aerial survey is to locate large schools rather than to observe all schools regardless of size. The cumulative total school area per survey day in Trinity Bay demonstrates the importance of large schools to the overall estimate (Figure 6). For most of the survey days flown in Trinity Bay in 1994 a small proportion of schools encountered contributed most of the estimated total area (Table 1). Thus a survey flight that missed many small schools (<500 m2) would not be compromised as much as one that missed one or more large schools (>25,000 m2). A similar finding was reported by Funk et al. (1995) for herring schools in the Bering Sea.

Digital imagery can be collected at lower light conditions than was possible with aerial photography and this has increased the opportunity to collect data during the short time that capelin schools are available close to shore. On average the number of flying days for each of the three transects has increased by 35% from the color photographic method to the digital imaging technique. With this capability we are able to increase the frequency of flights within the same survey period, which has reduced the probability of missing the peak spawning period during adverse weather conditions.

Figure 3. Relationship between capelin trap catch rates and school areas derived from aerial surveys from 1982 to 1993, excluding 1991.

Figure 4. Annual estimates of total observed school areas assuming one () and two spawning runs ().

Figure 5. Relationship between school area () along the Trinity Bay transect and egg deposition () on Bellevue Beach, Trinity Bay (Unpubl. data, Nakashima and Winters).

Figure 6. Cumulative school area for July 9 (), July 13(), July 15(), July 17(), and July 19(), 1994, in Trinity Bay

Table 1. Number and area of capelin schools in Trinity Bay, Newfoundland, from aerial surveys in July 1994.

The roundness index (RI) for capelin schools confirms visual observations that large schools are irregular and have elongated shapes (RI range = 0.1-0.4) compared to small schools, which have a wider range of shapes (RI range = 0.2-09) with a tendency toward more regular and round schools (Figure 7). We hypothesize that large schools are predominantly spent and partially spent males which remain in aggregations during the spawning season, whereas small schools are maturing fish of both sexes moving toward spawning beaches. This explanation is compatible with earlier observations on capelin spawning behavior (Templeman 1948).

DISCUSSION

The aerial survey index has provided capelin assessments with a fishery-independent estimate of mature stock size. An experienced spotter is used to detect schools and to guide data collection. Image enhancement techniques are then employed to estimate the areas. This is an effective combination to provide timely and reliable estimates of mature capelin schools. Experienced spotters detect and estimate the biomass of schools of herring in Alaska (Funk et aI. 1995) and northern anchovy off California (Lo et al. 1992). However, Funk et al. (1995) found that experienced spotters underestimated school areas compared to estimates using the same digital imaging methods as employed for capelin.

Figure 7. Roundness index in relation to school area July 9 ( ), July 13 (), July 15 (), July 17(), and July 19 (), 1994, in Trinity Bay.

Despite the flexibility of the instrumentation and techniques, factors such as weather and our inability to accurately predict the major spawning run mean that in some years the aerial survey index will not reflect inshore abundance. For example, in 1991 after two extensions to the survey period into late July we assumed spawning was complete. As much as 6 weeks later capelin arrived to spawn in late August (Nakashima 1996).

One important assumption to be addressed is that density is equal among schools when using surface area as a biomass indicator. From visual observations schools change in "color," some of which can be attributed to contrasting backgrounds. However, it is conceivable that color differences, i.e., digital values within and among fish schools, may be related to density differences. A study was initiated in 1996 to develop a relationship between school area and color from digital imagery and capelin density estimated with hydroacoustics.

Direct observations of schools can enhance our knowledge of school distribution and behavior (e.g., Hara 1985). For example, the number of mature capelin schools increased in a few days to a maximum along the coastline in Trinity Bay (Figure 6) which corresponded to the pattern of spawning from egg deposition collected from Bellevue Beach, a major spawning beach (Figure 5). The rapid aggregation of schools near beach es prior to spawning matches anecdotal observations. The presence of large, irregular, elongated schools near spawning beaches (Figure 7) is attributed to mature males which remain near beaches throughout the spawning period whereas smaller, more mobile schools of females move in to spawn then quickly leave an area.

Employing a digital imaging technique for capelin aerial surveys has increased the amount and quality of data collected during the same period used for photographic surveys. As instruments improve and post-processing software is developed, further benefits will be realized.

REFERENCES

Anonymous. 1996. Capelin in SA2 + Div. 3KL. DFO Atl. Fish. Res. Doc. 96/90, 269 pp.

Borstad, G.A., D.A. Hill, R.C. Kerr, and B. Nakashima. 1992. Direct digital remote sensing of herring schools. Int. J. Remote Sensing 13:2191-2198.

Carscadden, J., and B.S. Nakashima. 1997. Abundance and changes in distribution, biology, and behavior of capelin in response to cooler waters of the 1990s. In: Forage fishes in marine ecosystems. Alaska Sea Grant College Program, University of Alaska Fairbanks, AK-SG-97-01 (this volume).

Carscadden, J., B. Nakashima, and D.S. Miller. 1994. An evaluation of trends in abundance of capelin (Mallotus villosus) from acoustics, aerial surveys, and catch rates in NAFO Division 3L, 1982-89.J. Northw. Atl. Fish. Sci. 17:45-57.

Funk, F.C., G.A. Borstad, and S.A. Akenhead. 1995. Imaging spectrometer detects and measures the surface area of Pacific herring schools in the Bering Sea. In: Proceedings of the Third Conference on Remote Sensing of Marine and Coastal Environments, Seattle, September 18-20, 1995. Volume 2, pp. 833-844.

Hara, I. 1985. Shape and size of Japanese sardine schools in the waters off the southeastern Hokkaido on the basis of acoustic and aerial surveys. Bull. Jpn. Soc. Sci. Fish. 51:41-46.

Lo, N.C.H., l.D. Jacobson, and J.L. Squires. 1992. Indices of relative abundance from fish spotter data based on delta-lognormal models. Can. J. Fish. Aquat. Sci. 49:2515-2526.

Nakashima, B.S. 1985. The design and application of aerial surveys to estimate inshore distribution and relative abundance of capelin. NAFO SCR Doc. 85/ 84, Ser. No. N1058, 11 pp.

Nakashima, B.S. 1996. The relationship between oceanographic conditions in the 1990s and changes in spawning behavior, growth and early life history of capelin (Mallotus villosus). NAFO Sci. Coun. Stud. 24:55-68.

Nakashima, B.S., and G.A. Borstad. 1993. Detecting and measuring pelagic fish schools using remote sensing techniques. ICES C.M. 1993/B:7, 18 pp.

Nakashima, B.S., G.A. Borstad, D.A. Hill, and R.C. Kerr. 1989. Remote sensing of fish schools: Early results from a digital imaging spectrometer. In: IGARSS '89, 12th Canadian Symposium on Remote Sensing, Vancouver, B.C., July 10-14, 1989, pp. 2044-2047.

Templeman, W. 1948. The life history of the capelin (Mallotus villosus 0. F Muller) in Newfoundland waters. Bull. Newfoundland Gov. Lab. 17, 151 pp.

Winters, CH., and J.E. Carscadden. 1978. Review of capelin ecology and estimation of surplus yield from predator dynamics. ICNAF Res. Bull. 13:21-30.

Click HERE to return to Papers on Aquatic and Coastal

   
Last update 12 March 2007
The contents of this site are © copyright 1993 - 2009, ASL Borstad Remote Sensing Inc..