SEARCH

SEARCH BY CITATION

Keywords:

  • diet;
  • seal;
  • California sea lion;
  • Zalophus californianus;
  • Australian fur seal;
  • Arctocephalus pusillus doriferus;
  • fecal analysis;
  • near infrared spectroscopy

Abstract

  1. Top of page
  2. Methods
  3. Results
  4. Discussion
  5. Acknowledgments
  6. Literature Cited

We investigated the utility of Near Infrared Spectroscopy as a means to quantify the diet of seals via analysis of feces. A pilot study showed that we could accurately determine the proportion of each species in a mixture of flesh of two species of fish and one squid. Having done so, we then assessed whether the same capacity existed for fecal material. Again we used two species of fish and one squid as the diet components offered to two captive seals (a California sea lion and an Australian fur seal). Five of the six calibrations could accurately and precisely quantify how much of a given diet component the seal had eaten the previous day from an NIR scan of the feces. NIR spectroscopy is theoretically a viable way to quantify seal diets. We discuss the logistical requirements for development of calibration equations for application to a field study. These may be prohibitive in many cases, but in others may be offset by the particular circumstances or objectives of the study. Adoption of NIRS may confer significant benefits in these circumstances.

Information on diets of marine animals is vital to understanding their ecosystem roles, yet extremely difficult to acquire. It is difficult, if not impossible, to observe marine mammals feeding as they move more quickly and dive deeper than is feasible for humans to follow. As a result, a range of alternative approaches has been used to determine the organisms on which these animals feed. These include analysis of indigestible hard parts from feces such as otoliths or bone fragments (Marcus et al. 1998), fatty acid profiles from the blubber or milk samples (Iverson et al. 1997), or stable carbon and nitrogen isotope ratios in animal tissues. Each of these has associated with it both advantages and disadvantages, including biases due to differential digestion or inability to resolve material to species level. Additionally, current methods are often either expensive or time consuming, and this may limit the capacity to examine large sample sets.

In this study, we have assessed the potential use of an alternative approach to those used to date to assess marine mammal diets via fecal analysis. We acknowledge that there are concerns about the application of information derived from scat analysis, primarily that they represent only the last meal consumed and may not elucidate what happens at foraging grounds farther away (Iverson et al. 2002). However, in many cases scat analysis will remain one of the only ways in which to get any data on diet, so methods that improve the quality of the information or reduce the costs of collecting it will be valuable. This will be particularly so for attempts to quantify the relative contribution of different components to the overall diet.

Near Infrared Spectroscopy (NIRS) potentially offers significant advantages over other methods in terms of reduced bias and savings in laboratory and reagent costs. In most cases, NIRS analysis requires only that samples be dried and ground, using no reagents once calibrations have been developed (see below). Because of this limited preparation, and rapid analysis, once a calibration equation is available for a given component, in excess of 100 samples can be analyzed per day at low cost (we currently pay approximately US$2/sample).

We briefly outline the basis of NIRS here, but for more detail the reader should refer to the extensive review of Foley et al. (1998) and references therein. NIRS is a technique whereby the composition of an organic sample is captured in the spectrum of light reflected (or transmitted, depending on the machine) when the sample is irradiated with light in the NIR spectrum. Organic material absorbs NIR light at wavelengths characteristic of particular bonds. Therefore, if samples are irradiated with light of a standardized spectrum, the amount of light reflected at a given wavelength is indicative of the concentration of compounds with that particular bond. While single bonds can be relatively easily quantified, organic material is more complicated and requires a statistical approach to relate the spectra to the characteristic of interest. That is, a set of spectra must be calibrated against known values. Once the calibration equation is developed, the characteristic can then be estimated for additional samples on the basis of their spectra alone, provided they come from the same spectral population.

These relationships can be developed, and are more intuitively acceptable, for defined aspects of composition, such as nitrogen or specific sugars (Foley et al. 1998). However, with appropriate caution, statistical techniques can be used to relate NIR spectra to characteristics of organic materials that are less well defined as long as there is expected to be an organic chemical basis driving the relationship. Such attributes to which NIRS has been successfully applied include estimation of the digestibility and palatability of herbivore diets (Coleman and Murray 1993, McIlwee et al. 2001). Recent studies have even shown that NIRS can be used to resolve the species composition of a mixed sample of plants (Volesky and Coleman 1996) and that this can be further extended to determining the composition of the diet of a herbivore from stomach contents (André and Lawler 2003) or even from fecal samples (Lyons et al. 1995).

Studies of the capacity of NIRS to resolve diets of free-ranging animals, as mentioned above, have largely concentrated on herbivores (Leite and Stuth 1994). To date there are no published studies in which calibration of NIRS spectral composition of feces against food intake has been attempted for a carnivore. One study has shown that NIRS can be used to predict digestibility of diets offered to minks (Dahl et al. 2000). Given that NIRS has provided a powerful tool for the analysis of foraging ecology of vertebrate herbivores, it is reasonable to ask whether it can do the same for carnivores. Because there are many difficulties inherent in existing methods of diet analysis of seals, applying the significant advantages of NIRS analysis to this group may be particularly valuable.

This project is the first such study to assess whether seal diets can be assessed with NIRS. Our aim was simply to establish whether the NIRS spectra of the feces of a seal can be calibrated against its known diet to produce an equation that enables the diet to be determined from the spectra of new fecal samples. As this is the first attempt at such a study, the particular species of either diet or seal were not important. Both were determined by convenience as the work was conducted in association with an oceanarium (Sea World, Queensland, Australia), which allowed us to work with its nondisplay animals. We discuss our findings in terms of limitations of the technique and logistical and theoretical requirements for development of functional sets of calibrations for the study of foraging by wild seals.

Methods

  1. Top of page
  2. Methods
  3. Results
  4. Discussion
  5. Acknowledgments
  6. Literature Cited

Pilot Study

As the NIRS approach had not previously been used in this way for carnivores, we first wanted to know whether the proposed diet components could be distinguished via NIRS prior to digestion by an animal. That is, it may be expected that some of the chemical differences between the diet components would be altered, or reduced, by digestion. Thus, this pilot study was an attempt to first establish that different species of animal could be distinguished spectrally, before going to the significant logistical difficulties of carrying out the subsequent experiments on seals.

We used a mixture of two fish and a squid species. This was on the basis that we might expect squid and either of the two fish species to be more chemically distinct than the two species of fish from each other. Thus, this gave us the capacity to assess the resolving capacity of NIRS at two levels. We chose squid (Photololigo spp.), pilchard (Sardinops neopichardus), and mullet (Mugil cephalus) as these were readily available commercially. As we were interested in quantifying the amount of each component, rather than simply identifying their presence we used a range of treatments in which the concentration of each component was varied. There were 45 treatments with the composition of each component ranging from 0% to 100% (wet weight) across treatments. We finely minced and mixed samples of whole individuals of each species and selected from this to develop each treatment.

Collection of Fecal Samples from Seals

This study was conducted at Sea World on the Gold Coast, Queensland, Australia. Two male adult seals, a California sea lion (Zalophus californianus) and an Australian fur seal (Arctocephalus pusillus doriferus), were used. The choice of animals was made by Sea World as these two individuals were considered not suitable for display and were housed in separate pens. There were two sets of feeding trials. The first one was carried out from 30 April to 17 May 2004 and the second trial was conducted from 21 June to 10 July 2004.

Each seal was kept in a separate pen during the feeding trials to ensure that fecal output could unequivocally be attributed to that animal. There were 20 different treatments in each trial with varying proportions of three kinds of diet: squid (Photololigo spp.), whiting (Sillago ciliate), and mullet (Mugil cephalus) (Table 1). The types of diet component were selected under advice from the Sea World curators. There was some concern that excess amounts of squid in particular may cause digestive upsets leading the seals to go off their food. Consequently, we limited the maximum amount of any diet component to 80% (wet weight). Furthermore, as squid was the item of most concern, we alternated the treatments so that the seals did not receive high amounts of squid on successive days. In addition, the 20 treatments were slightly reordered in the second trial to focus on the treatments for which we could not obtain samples from the first trial (e.g., because feces were lost in the filtration system).

Table 1.  Composition of diets offered to seals. Figures under each diet component indicate the percentage wet weight of each diet.
Treatment/daySquidWhitingMulletTotal
 103070100
 28012.57.5100
 32.577.520100
 4757.517.5100
 551580100
 6702010100
 77.542.550100
 867.532.50100
 9106525100
1060040100
1112.522.565100
12502.547.5100
13157015100
14452530100
1520575100
164047.512.5100
172517.557.5100
1837.5602.5100
19301060100
203527.537.5100

Each diet was divided equally into three batches to feed to seals three times a day: 0900–∼0930, 1200–∼1300, and 1400–∼1530. The variability in the feeding times was a function of curatorial staff being more likely to have other activities to accommodate as the day progressed. Meals were offered over the course of the day because the seals sometimes would not eat their full allowance in a single meal and thus may have led to unintended variation in diet composition within days. In this way, the proportion of each food item remained constant throughout the batches. Each fish or squid was chopped into pieces (∼30 cm3) and mixed to ensure that proportions remained similar within meals. That means it could be assumed that if digestive efficiency was equivalent among the three diet components, the feces defecated by the seals in any time of the day were likely to contain the same proportion of the diets. When seals refused to eat some food items, those were weighed and subtracted from the total amount of the diet so that percentage of each diet for that treatment was recalculated.

Pens were checked and feces collected three times a day (at the same time as feeding). When feces were found in the pool, they were scooped out using a mesh net. All samples were immediately frozen and stored at −20°C.

Because the digesta passage time of seals has been generally estimated to be less than 9 h (Lawson et al. 1997), the feces defecated overnight were assumed to result from the diet of the previous day and thus reflect the diet composition of the previous treatment. When animals defecated more than once a day, these were weighed and stored separately.

Sample Preparation and Scanning

Samples from both the pilot study and the feeding trials were put into a drying oven at 65°C for three days until they reached constant weight. The dry samples were then ground in a Cyclotec 1093 sample mill (Foss Tecator, Hilleroed, Denmark) to pass through a 1-mm sieve. Ground samples were scanned on an NIRS system 6500 spectrophotometer equipped with a spinning cup module (Foss NIR-systems). Samples were packed into a 50-mm diameter cup and scanned 32 times per sample.

Development of the Calibration Equation

For both the pilot study and fecal analyses, scanning, mathematical processing, calibration, and statistical analysis were conducted using the software ISI II V 1.02a (Infrasoft International LLC., State College, PA). Laboratory reference values were calibrated against NIRS spectra using modified partial least-squares (MPLS) regression (Shenk and Westerhaus 1993). The full range of wavelengths recorded was used in the calibration (every 2 nm between 408.0 and 2,492.8 nm). All available samples were used in the development of the calibration equation.

Four mathematical treatments were compared to obtain best calibrations. There is no scope here to discuss the basis for these treatments, but generally they are used to reduce “noise” in the data. The combinations we adopted are widely used in a range of applications and the interested reader should refer to Anon (1995). Those used were: 1.4.4.1, 2.8.6.1, 3.10.10.1, and 4.10.10.1 where the first digit indicates derivatives, the second digit indicates the range of wavelengths (in nanometers) over which the derivative is to be calculated, the third digit describes the range of wavelengths over which a smoothing function is applied, and the fourth is a second smooth, which is almost never used (1 = no smooth).

MPLS regression requires cross validation to avoid over fitting of the model (Shenk and Westerhaus 1993). Cross validation was calculated by splitting the calibration group into a number of small groups (Foley et al. 1998). The calibration is then run a number of times with each group omitted sequentially. The optimum equation is obtained using the standard error of cross validation (SECV), which is defined as an error due to differences between laboratory reference values and NIRS spectra values within the cross validation set. The process is repeated until all groups have been used for validation at least once and the minimum SECV in cross validation was taken to be the best model (Shenk and Westerhaus 1993, Anon 1995).

Other terms have to be considered to measure the quality of the relationship between the NIRS spectra values and the laboratory reference values in the regression equations. These are:

  • 1
    Standard error of calibration (SEC)—the error due to the differences between laboratory values and NIRS-predicted values within the calibration. This is always lower than the SECV because of the larger number of samples
  • 2
    Bias—the difference between the mean actual value and the mean predicted values for a component.
  • 3
    Assessment of the relationship between the NIRS-estimated values and the laboratory values. Both the coefficient of determination (r2) and the slope should approach one in a high-quality calibration.

We should note that the standard practice for assessing the utility of the calibration also includes comparisons of NIRS-estimated values and laboratory values for samples outside of the calibration set (Shenk and Westerhaus 1993, Anon 1995). We had intended to do that in this case, with each of the two trial periods listed above acting as an independent validation set for the other. Unfortunately we were not able to collect feces for all treatment combinations (see Results) and thus we had too few samples to do this with appropriate statistical rigor. This is not an impediment to the aims of this study, but would be necessary for development of field applications.

Three calibration equations were developed for the pilot study: one each for squid, pilchard, and mullet. Similarly, for fecal analysis three equations were developed for each seal species, combined across the two trial periods: squid, whiting, and mullet.

Results

  1. Top of page
  2. Methods
  3. Results
  4. Discussion
  5. Acknowledgments
  6. Literature Cited

Pilot Study

There were a total of 45 samples to develop the calibration for each diet. Good calibrations were obtained from all three diets (Table 2), with r2 values ranging from 0.974 to 0.999 for the regression of the relationship between the NIRS-estimated values and the laboratory values. In particular, excellent calibration was obtained for squid (Fig. 1a), while only one sample was an inexplicable outlier in the other two calibrations (Fig. 1b, c). Similarly, the slopes of all these relationships were very close to 1. That is, there was little bias in the calibration such that the NIRS-estimated value was close to the true value across the range of diet proportions.

Table 2.  Validation statistics of pilot study to determine ability of Near Infrared Spectroscopy of discriminate flesh of potential marine prey species.
DietnBiasSECSECVSlopeInterceptr2Math treatment
Squid45 0.1941.2962.0340.990.070.999  2,8,6,1
Pilchard45 0.4935.2987.9210.960.880.97 4,10,10,1
Mullet45−1.0317.1179.01 0.972.260.9593,10,10,1
image

Figure 1. Results of pilot study to determine whether NIRS could discriminate flesh of different potential marine prey species, (a) squid, (b) pilchards, and (c) mullet. Values are expressed in terms of percentage of wet weight.

Download figure to PowerPoint

Near Infrared Spectroscopic Determination of Diet from Fecal Samples from Seals

California sea lion— There were 30 samples (14 samples from the first trial, 16 samples from the second trial). The relationships between the actual values and the NIRS-estimated values were excellent for whiting and mullet, with r2 values that were both above 0.95 (Fig. 2b, c, Table 3). The same relationship for squid was not quite as effective but was still of high quality (Fig. 2a, Table 3). In calibrations of all diets, samples from the first trial and the second trial distributed evenly along the regression line indicating that pooling the samples between trials was appropriate. SEC values were acceptable, while SECV values were higher, probably reflecting the relatively small number of samples in the calibration set.

image

Figure 2. Ability of Near Infrared Spectroscopy to determine the diet of Z. californianus via fecal samples. Species were (a) squid, (b) whiting, and (c) mullet. Solid squares are the first trial, open squares are the second trial, solid line is regression for all values. Values are expressed in terms of percentage of wet weight.

Download figure to PowerPoint

Table 3.  Validation statistics of first and second trials obtained from assessment of ability of Near Infrared Spectroscopy to determine the diet of Z. californianus.
DietnBiasSECSECVSlopeInterceptr2Math treatment
Squid3007.44412.2460.892.910.893,10,10,1
Whiting3003.48715.9470.980.670.984,10,10,1
Mullet3003.75110.02 0.980.960.98  2,8,6,1

Australian fur seal— A total of 26 samples were obtained (13 samples from the first trial and 13 samples from the second trial). The values for r2 for the calibrations of squid and mullet were high, being above 0.95 (Fig. 3a, c, Table 4). However, there was a considerable decrease in the quality of the calibration for whiting (Fig. 3b, Table 4). When we compared the plots of the first and second trials, they distributed evenly along the regression line in the calibrations of squid and mullet. For whiting the data from the two trials were also consistent, even though the relationship overall was poor. Thus, pooling across trials was appropriate. The SECV values were relatively higher than for Z. californianus probably as a result of the smaller data set.

image

Figure 3. Ability of Near Infrared Spectroscopy to determine the diet of A. pusillus via fecal samples. (a) squid, (b) whiting, and (c) mullet. Solid squares are the first trial, open squares are the second trial, solid line is regression for all values. Values are expressed in terms of percentage of wet weight.

Download figure to PowerPoint

Table 4.  Validation statistics of first and second trials obtained from assessment of ability of Near Infrared Spectroscopy to determine the diet of A. pusillus doriferus.
DietnBiasSECSECVSlopeInterceptr2Math treatment
Squid260 5.04520.9410.96 1.320.964,10,10,1
Whiting26020.44329.6320.2526.180.253,10,10,1
Mullet260 5.72617.9740.95 1.590.953,10,10,1

Discussion

  1. Top of page
  2. Methods
  3. Results
  4. Discussion
  5. Acknowledgments
  6. Literature Cited

Our pilot study successfully showed that NIRS is able to resolve a mixture of flesh of fish and squid. In addition to simply identifying the presence of each species, it was able to accurately and precisely quantify the amount of each in different mixtures. That it was able to do so is perhaps not surprising, as a number of researchers have shown that NIRS has been a useful technique to assess qualities such as moisture, oil, crude protein, salt, or texture of raw fish, which can be applied to monitor the conditions in the fish industry (Isaksson et al. 2001; Cozzolino et al. 2002a, b; Uddin et al. 2002). Although the latter types of approaches led researchers in herbivory to extend NIRS analysis to resolution of plant mixtures with some success (Coleman and Murray 1993, Stuth and Tolleson 2000, André and Lawler 2003) we understand this to be the first such application to animal flesh.

Diet Analysis via Scanning of Seal Feces

From the above we then turned to the more relevant question of whether NIRS is feasible to quantify the amount of diet components after digestion (i.e., feces) although we may expect the chemical changes resulting from digestion to alter the chemical distinctness of each diet component. Overall this study has shown that NIRS can be used to quantify the diet of a carnivore. For two species of seals, most of the values estimated by NIRS showed close agreement with the known composition of the diet offered. However, in reaching this conclusion, we must acknowledge that this preliminary study has dealt with a limited set of circumstances and cannot immediately be applied to the field situation. Below we discuss first the merits of adopting NIRS, but then follow with consideration of the steps necessary to develop calibrations that could be applied to a field study.

Fecal analysis via NIRS is a relatively low-cost way to investigate the foraging ecology of seals in the wild, compared to other methods such as using fatty acid compositions or isotope ratio of carbon and nitrogen. In addition, NIRS examines the chemical composition of the feces via their spectra, which requires less time to analyze the diet compositions than identifying each diet item via bone fragments. Moreover, the NIRS technique can identify the diet composition to species level and quantify the amounts consumed.

NIRS requires only drying, grinding, and scanning and low cost (currently US$2/sample, but this is variable depending on leasing and servicing arrangements for the instrument). In excess of 100 samples can be analyzed per day using NIRS. Quantifying diet from fecal samples via otoliths costs US$2 to clean per sample and from US$15 to have bones identified.2 Some consultants will identify the bones at a cost of US$60 per hour.3 It takes approximately 6 min per sample to clean with an elutriator and identification of otoliths and other bones fragments takes an average of 20–30 min.3 Therefore, compared to those techniques, NIRS can deal with a large number of samples much more efficiently and economically. However, this is conditional on having first developed appropriate calibrations. This study highlighted some of the issues in the development of such calibrations if NIRS is intended to be used in a field application.

Comparison Between Seal Species

It is important that any NIRS calibration be specific to both the prey and predator species of interest. While the results presented here are similarly strong for both species of seal, the calibration from one does not necessarily apply to the other. The weakest calibration was that for whiting eaten by A. pusillus doriferus. Given that the calibration for whiting worked for Z. californianus it may be argued that there is some inherent difference between the two seal species that caused this effect. However, since we were able to test only one individual from each species, it is not possible to discern whether the difference in result is due to the difference in the seal species or simply the difference of individual animals. While we think it is likely that NIRS will be applicable to most, if not all, species of seal, this issue clearly needs to be addressed in any future larger study. Without such further replication of individuals, we cannot offer any mechanistic explanation for the weak calibration for whiting in A. pusillus doriferus.

Limitations in Replication

An issue of this project was lack of replication within species as we tested only one individual of each seal species. This constraint was imposed by the fact that we were working with Sea World on an informal basis and thus could use only those animals that were not on display and were housed in such a manner as to allow sample collection. As our main objective was simply to determine whether NIRS could work, rather than to develop a fully functional calibration, this is not a serious weakness of this project. However, it does serve to highlight an issue that should be addressed in attempts to develop functional calibrations. That is, a calibration obtained from a single individual can probably not be generalized to apply to other animals. Moreover, because the two seals we tested were both male adults, we cannot say whether the NIRS technique is applicable to individuals of the same species but of different age or sex.

Effects of Exposure on Sample Quality

For practical reasons, feces are most likely to be collected on land (i.e., at haul-out sites). If so, they are likely to be restricted to the last meal from a limited area around the haul-out site (Iverson et al. 1997) because of the rapid digesta passage times relative to the time taken to travel back to the haul-out site from more distant foraging grounds. That means the feces are only a representation of the last meal consumed (Iverson et al. 2002).

Chemical composition of the feces from goats, measured as crude protein and digestible organic matter content, were not affected by duration of exposure after defecation for up to 7 d, or seasonal environmental shifts (Leite and Stuth 1994). However, in a separate study cattle feces collected after 48 and 72 h of exposure were decimated by insect activity during summer (Hinnant and Kothmann 1988). This is assumed to be due to low moisture content and fast-drying characteristics of goat pellets (relative to cattle feces), that may prevent the damage caused to cattle feces by biotic and abiotic factors (Leite and Stuth 1994).

Even though most of the samples in this study were collected from the water, the maximum time feces spent in the water was 20 h, from the last check of the pen in the afternoon until the first feeding time in the next morning. This is still much shorter than the experiments conducted by Hinnant and Kothmann (1988). If NIRS is to be extended to field conditions for studies of seal foraging, some examination of the effect of exposure on fecal composition, and thus calibration performance, must be made.

Recommendations for Development of Calibrations for Field Studies

Overall this study has shown that NIRS can be a useful tool for the analysis of diet based on fecal samples. However, there are significant constraints that must be considered before the application of NIRS to a true field study in a broader context:

  • 1
    The calibration must be developed from captive individuals of the same species (and preferably the same population). It is unlikely that calibrations developed for one species will be directly applicable to another species.
  • 2
    Calibrations should be developed with multiple individuals (preferably different age and sex classes). We have only tested single adult male seals and it is unknown whether results from these individuals are consistent across age or sex. Future studies would have to test whether these variables have significant effects, and if so, develop calibrations separately for each category.
  • 3
    Use prey species likely to be encountered in the wild. The types of diet components that we used in this project were selected from among the food items routinely offered to seals at Sea World. While the diets offered to seals at Sea World are varied and suitably balanced to keep the seals healthy, they are not an attempt to recreate the wild diet. Because the NIRS calibration is fundamentally based on the particular chemical composition of individual prey species, development of a functional calibration for application to a field study must use samples of those species the seal is considered to be preying upon. It should also, ideally, use samples taken from the region or population of prey found where the seals are foraging.
  • 4
    Repeat across seasons. Seasonal physiological or chemical variation in both seals and prey may cause calibrations developed in one season to fail when used in a different season (Kaneko 2004). Future studies will have to explicitly test for these effects, and if significant, build season-specific calibrations. Alternatively, it may be possible to build larger, more robust calibration sets that are applicable across seasons.
  • 5
    Increasing diet complexity. In this study we examined only three diet components. We did not have the opportunity to examine whether additional species, not in the calibration, would return spurious results. That is, the calibrations for any one species are based on that set of chemical components that distinguish them from the other species present. There is some possibility that an additional species may bear sufficient chemical similarity to one of the species for which a calibration exists to return an overestimate of the amount of that species. This could be relatively easily tested.

Conclusion

This study has shown that NIRS may be applicable to studies of the diet of seals in the wild. However, collecting the necessary data from seals to calibrate the technique has substantial logistical limitations. Seals are difficult and expensive to keep in captivity, and there are also ethical issues to consider in collecting animals solely for this purpose, especially if the impetus for the study is vulnerable conservation status. Any attempt to develop a set of calibrations applicable to a field study will have to address these constraints and at the least will be a major undertaking requiring careful planning and substantial financial support. The initial investment of time and money spent on development of calibrations may only be recouped if numbers of samples to be analyzed are in the hundreds or thousands.

There may be, however, circumstances specific to particular situations that allay some of these concerns. For example, if, as in our case, the animals are being housed for other purposes, then husbandry costs specifically associated with calibration development may be minimized. Calibrations can be developed using diets within the normal range of composition and thus need not necessarily interfere with other concurrent research or uses for the animals. Similarly, if the focus of the study were, for example, assessing potential competition with a commercial fishery, then there may only be a requirement for the development of a single calibration for the fish species of interest. Collection of fish with which to develop such a calibration would also then pose much lower logistical and financial limitations. Ultimately, NIRS analysis of fecal material may not be viable for all studies, but if the research project is to be large scale and some of the limitations of calibration development can be reduced, then we would recommend to researchers to at least consider the potential benefits that can be gained in reduced bias and efficiency. This will continue to accrue as sample numbers increase.

Footnotes
  • 1

    Corresponding author.

  • 2

    Personal communication from Andrew Trites, Marine Mammal Research Unit, Room 247, AERL, 2202 Main Mall, University of British Columbia, Vancouver, BC V6T 1Z4, 12 September 2005.

  • 3

    Personal communication from Peter F. Olesiuk, Department of Fisheries and Oceans, Pacific Biological Station, 3190 Hammond Bay Road, Nanaimo, BC V9T 6N7, 6 September 2005.

Acknowledgments

  1. Top of page
  2. Methods
  3. Results
  4. Discussion
  5. Acknowledgments
  6. Literature Cited

This project would not have been possible without the support of Sea World on the Gold Coast, Australia. In particular, we wish to thank their mammal curators Rob Clapp, Mitchell Leroy, and assistants Melinda, Brent, Daniel, Tammy, Belinda, Becky, and Natalie for allowing access to the animals, and providing the materials from which we made the diets. Kylee Verry and David Coates at CSIRO facilitated use of NIRS. Emma Gyuris commented on drafts of the manuscript. The work was approved under James Cook University Animal Ethics permit no. A895_04.

Literature Cited

  1. Top of page
  2. Methods
  3. Results
  4. Discussion
  5. Acknowledgments
  6. Literature Cited
  • André, J., and I. R. Lawler. 2003. Near infrared spectroscopy as a rapid and inexpensive means of dietary analysis for a marine herbivore, dugong Dugong dugon. Marine Ecology Progress Series 257: 259266.
  • Anon. 1995. Standard practices for infrared multivariate quantitative analysis (designation E1655–00). American Society for Testing and Materials , West Conshohocken , PA .
  • Coleman, S. W., and I. Murray. 1993. The use of near-infrared reflectance spectroscopy to define nutrient digestion of hay cattle. Animal Feed Science and Technology 44: 237249.
  • Cozzolino, D., A. Chree, I. Murray and J. R. Scaife. 2002a. The assessment of the chemical composition of fish meal by near infrared reflectance spectroscopy. Aquaculture Nutrition 8: 149155.
  • Cozzolino, D., I. Murray and J. R. Scaife. 2002b. Near infrared reflectance spectroscopy in the prediction of chemical characteristics of minced raw fish. Aquaculture Nutrition 8: 17.
  • Dahl, P. L., B. M. Christensen, L. Munck, E. P. Larsen and S. B. Engelsen. 2000. Can spectroscopy in combination with chemometrics replace minks in digestibility tests? Journal of the Science of Food and Agriculture 80: 365374.
  • Foley, W. J., A. P. McIlwee, I. R. Lawler, L. Aragones, A. P. Woolnough and N. Berding. 1998. Ecological applications of near infrared spectroscopy—a tool for rapid, cost-effective prediction of the composition of plant and animal tissues and aspects of animal performance. Oecologia 116: 293305.
  • Hinnant, R. T., and M. M. Kothmann. 1988. Collecting, drying and preserving for chemical and microhistological analysis. Journal of Range Management 41: 168171.
  • Isaksson, T., L. P. Swensen, R. G. Taylor, S. O. Fjaera and P. O. Skjervold. 2001. Non-destructive texture analysis of farmed Atlantic salmon using visual/near infrared reflectance spectroscopy. Journal of the Science of Food and Agriculture 82: 5360.
  • Iverson, S. J., J. P. Y. Arnould and I. L. Boyd. 1997. Milk fatty acid signatures indicate both major and minor shifts in the diet of lactating Antarctic fur seals. Canadian Journal of Zoology 75: 188197.
  • Iverson, S. J., K. J. Frost and S. L. C. Lang. 2002. Fat content and fatty acid composition of forage fish and invertebrates in Prince William Sound, Alaska: Factors contributing to among and within species variability. Marine Ecology Progress Series 241: 161181.
  • Kaneko, H. 2004. Analysis of seal diet using Near Infrared Refrectance Spectroscopy (NIRS). Honours thesis, School of Tropical Environment Studies and Geography, James Cook University , Townsville , Australia . 71 pp.
  • Lawson, J. W., E. H. Millaer and E. Noseworthy. 1997. Variation in assimilation efficiency and digestive efficiency of captive harp seals (Phoca groenlandica) on different diets. Canadian Journal of Zoology 75: 12851291.
  • Leite, E. R., and J. W. Stuth. 1994. Influence of duration of exposure to field conditions on viability of fecal samples for NIRS analysis. Journal of Range Management 47: 312314.
  • Lyons, R. K., J. W. Stuth and J. P. Angerer. 1995. Fecal NIRS equation field validation. Journal of Range Management 48: 380382.
  • Marcus, J., W. D. Bowen and J. D. Eddington. 1998. Effects of meal size on otolith recovery from fecal sample of gray and harbor seal pups. Marine Mammal Science 14: 789802.
  • McIlwee, A. M., I. R. Lawler, S. J. Cork and W. J. Foley. 2001. Coping with chemical complexity in mammal-plant interactions: Near-infrared spectroscopy as a predictor of Eucalyptus foliar nutrients and of the feeding rates of folivorous marsupials. Oecologia 128: 539548.
  • Shenk, J. S., and M. O. Westerhaus. 1993. Analysis of agriculture and food products by near infrared reflectance spectroscopy. Infrasoft International , Port Matilda , PA .
  • Stuth, J. W., and D. R. Tolleson. 2000. Monitoring the nutritional status of grazing animals using near-infrared spectroscopy. Compendium on Continuing Education for the Practicing Veterinarian 22: S108S115.
  • Uddin, M., S. Ishizaki, E. Okazaki and M. Tanaka. 2002. Near-infrared reflectance spectroscopy for determining end-point temperature of heated fish and shellfish meats. Journal of the Science of Food and Agriculture 82: 286292.
  • Volesky, J. D., and S. W. Coleman. 1996. Estimation of botanical composition of esophogeal extrusa samples using near infrared reflectance spectroscopy. Journal of Range Management 49: 163166.