Bird stomach contents analysis - final report: Goosander and Cormorant diet on four Scottish rivers 2019 to 2020

6. Discussion

Samples available

Care must be taken when interpreting the findings of the present study because of the relatively small sample sizes of both birds and study rivers. Indeed, in several instances, samples of fewer than 12 stomachs with food were available and these were likely to be too small to ensure that they were representative of general diet. Only samples of Goosanders from all four rivers during the smolt run period, for Tweed Goosanders and for Tweed and Nith Cormorants during the following autumn-winter had 12 or more stomachs that contained food.

There was a general paucity of Cormorants on rivers during the smolt run sample period and of Goosanders during the autumn-winter period, at least as indicated by the consistently small samples sizes of birds provided. The reasons for this are unclear but are likely to have been related to the numbers of birds on particular rivers at particular times of year. For instance, it is thought (DSFB pers. comm.) that relatively few Cormorants are present on the Spey (and the Dee) throughout most of the year, few are on the Tweed in spring, and few on the Dee during the smolt run. There may also have been fewer Cormorants than 'normal' on the Nith and Dee during the 2019 smolt sampling period (possibly related to local river conditions). Similarly, Goosanders may have been relatively scarce in the autumn on some rivers, and there will be always be practical difficulties in obtaining samples depending on local conditions and the availability of resources.

Whilst current information on the dietary composition of, and the length frequencies of Salmon consumed by, these birds is undoubtedly informative and the first of its kind for over 20 years, it is nevertheless limited. Future research involving larger samples of Goosanders and Cormorants at other times of year and from a larger number of rivers is required to improve levels of confidence in the evidence for predation by these species in relation to the protection of Salmon stocks in Scottish rivers.

General diet

(a) Smolt run period

During the 2019 smolt run sample period (March/May), the proportion of Salmon (50% by mass) in the diet of Goosanders on the R. Dee was higher than that for any other species in the diet and was also similar to comparable historic samples from the 1990s (Marquiss et al. 1998). The proportion of Trout (26% by mass, the second highest dietary proportion in spring 2019) was lower, and the proportion of Minnow (15% by mass) considerably higher, than these previous samples.

Goosander diet on the R. Nith during the 2019 smolt run sample period (April/May) comprised Trout (31% by mass) and Stone Loach (26% by mass) and proportions of both species were higher than recorded previously. The proportion of Salmon (16% by mass) was higher than previously recorded but only slightly. Conversely, proportions of Eel (5% by mass) and 3-Spined Stickleback (1% by mass) in 2019 were considerably lower than in previous samples.

On the R. Tweed during the smolt run sample period (Apr/May), the proportion of Salmon (15% by mass) in the diet of Goosanders was similar to historical samples (although considerably lower than estimated in April 1991). The proportion of Trout (20% by mass) was strikingly lower in 2019 than in three of the four historical samples, whereas that of Minnow (38% by mass) was some 4-13 times greater than recorded historically. Indeed, Minnow was by far the largest dietary component by mass in the 2019 Goosander smolt run sample period.

On the R. Spey, Trout (40% by mass) was the largest component in the diet of Goosanders during the smolt run sample period (March/April), very closely followed by Salmon (36% by mass) and these proportions were within historical levels. Proportions of Minnow (4% by mass) and 3-Spined Stickleback (6% by mass) were low in spring 2019 but still higher than historically recorded, whilst the proportion of Eel (14% by mass) was similar to the previous 1994 level.

The numbers of Cormorants available (and/or containing food) from the smolt run sampling period were very low (or zero) from the rivers Dee, Nith, and Spey in 2019 and so little could be determined about their diet and comparisons with earlier samples were not possible. The numbers of Cormorants sampled on the R. Tweed during the 2019 smolt run sample period were also likely to be too small to be representative of general diet (N = 6 with food). Nevertheless, Trout (45% by mass) and Salmon (40% by mass) were the largest components of the diet of Cormorants in this sample, and these were generally higher than the proportions recorded previously. The relatively low proportion of Minnow (10% by mass) recorded in 2019 was higher than that previously recorded. Conversely, the proportion of Grayling (6% by mass) was considerable lower than recorded previously, when proportions were around 40% by mass.

Overall, for Goosanders, the proportions of Salmon recorded in the diet during the 2019 smolt run sample period were broadly similar to historical levels for the rivers Dee and Tweed, and possibly lower for the R. Spey and slightly higher for the R. Nith than previous samples. Few Cormorants were sampled during the 2019 smolt run period and historical comparisons were only possible for the R. Tweed, albeit limited by small sample size in 2019. Nevertheless, comparisons do suggest a reduction in the proportion of Grayling recorded in the diet, and a coincidental increase in the proportions of Salmon, Trout, and Minnow.

Statistical comparisons of general diet assessments could only be made for a restricted number of samples, namely those of Goosanders from the Tweed (see Section 4). Data on the mass of each prey species in each bird stomach was only available for these samples. Statistical comparisons were further confounded by limitations caused by the absence of particular fish species in some samples and by the need to establish a threshold for the number of values (mass of fish/stomach, including zero values). The latter was necessary in order to reduce the 'uncertainty' of diet estimates for samples where a particular fish species is only recorded in a small number of stomachs and to increase confidence in the interpretation of the statistical test. Whilst, there is a future research requirement for simulation studies to understand what sample size, and what number of positive values within a sample, might be most appropriate for statistical comparisons, the restricted statistical tests possible here confirmed the broader observations of variation in general diet assessments described in Section 4.

Sampling and the timing of the smolt run

There are a number of sampling issues that should be remembered when considering how best to interpret the findings produced from the stomach contents analysis used in the present study, where the sampling unit is an individual stomach and not the number of fish recorded it contains. At its simplest, a relatively small sample of stomachs is examined, the findings from any sample of stomachs (realistically, this is necessarily always going to be a relatively small number) should not be considered 'representative' subsequently nor applied to a wide area such as a river catchment, or to a different year or time of year.

Marquiss & Carss (1997) showed that the prevalence of smaller fish species is underestimated, and the average size of juvenile Salmon eaten overestimated by about 20% by considering only the intact fish in stomachs (as smaller individuals could be too digested to be recognisable), and concluded that measurements of uneroded bony material in stomachs (which would also include that from smaller fish) gave more accurate diet estimates in terms of the numbers and sizes of fishes consumed. These authors also explored the effect of sample size on general assessments of Goosander and Cormorant diet in terms of the fish community represented in stomachs (see Methods re. 'adequate' estimates of diet from samples of 12-15 stomachs containing food).

There are also potential temporal issues with sampling, and the issue of the level of synchrony between bird sampling and the smolt run period is discussed below. A broader issue, concerns the 'patchiness' of sampling and its potential influence on the degree of confidence in, and wider applicability of, the findings of stomach contents analysis. Here, such interpretation could be compromised by the fact that birds are often aggregated within samples, generally coming from particular locations and particular days (or times of day). Furthermore, birds might only be collected from locations where, and when, it is possible to get close to birds and so sampling opportunities are most appropriate and favourable.

The timing of the sampling of birds in relation to the timing of the smolt run is an important issue to consider in the present study because the synchrony of these two events is not clear. This means that sampling (and hence diet assessments) might not be 'representative' of the peak of the smolt run, and that comparisons with previous samples may be difficult because they rarely included birds from May for instance. Nevertheless, the smolt run period birds included in the present study (and those spring birds in Marquiss et al. 1998) were sampled at times and places when predation on smolts was thought to be an important issue.

Factors such as increasing day length and water temperature, and varying flow rates are likely to influence biological events like smolt migration, although the details have only recently begun to be quantified through smolt catching and tagging research. Appendix 5 thus provides a preliminary exploration of the sampling dates for Goosanders and Cormorants on the river Tweed in relation to information on the timing of the smolt run there in spring 2019. This Appendix does not explore issues of spatial abundance (of birds or fishes) which might also be important in understanding and interpreting dietary findings - as birds were sampled from a limited number of locations in the absence of information on the distribution of smolts within any of the study river systems.

Consideration of the possibility of 'mismatch' between the timing of bird sampling and the timing of the smolt run raises similar issues over the comparability of samples in relation to the timing of their collection. For instance, the majority of historical smolt run samples were collected in March and April whilst those in the present study also covered these months but were also extended into May. Theoretically, there could thus be up to a 13 week difference in timing between birds sampled in the first week of March and those from the last week of May. The implications of such a temporal mismatch in sampling dates are currently poorly understood, as are the likely levels of inter-annual variation in diet assessments from samples from the same location taken in different years.

Knowledge of the temporal and spatial dynamics of the smolt run is advancing, and so exploring the implications for sampling regimes, and the timing thereof, to assess fish-eating bird diet requires further investigation. Future research requirements will also include fine-scale analysis of temporal dietary composition (e.g. to better understand over what time-periods dietary estimates are representative), examination of the influence of sampling location on diet assessments and of both sample location and sample size on subsequent length frequency estimation for prey fishes, particularly Salmon. As birds can be highly mobile, it is also vital to know where birds are actually foraging rather than where they were sampled and so further research (including tagging and tracking) is needed to better understand the spatial and temporal distribution of birds within river systems, their movements and behaviour, and levels of fidelity to different locations (e.g. foraging sites, roosts) at different times of year.

(b) Autumn-winter period

During the autumn-winter 2019/20 period (November), only two Goosanders were sampled on the R. Dee. Both contained Salmon and Trout, and one contained Minnow. Due to the constraints of small samples (current and historical), no comparison was possible with previous data, including that from the 2019 smolt run period.

The sample of R. Nith Goosanders from the autumn-winter (Oct 2019-Feb2020) period was too small (10 birds, all with food) to be confident it was representative of general diet. Nevertheless, by far the highest dietary proportions in this sample comprised Trout (41% by mass) and Grayling (34%). The proportion of Trout had increased since the smolt run period whilst Grayling was not recorded at all earlier in the year. The proportion of Salmon (7% by mass) in the autumn-winter was low and much lower than reported in the preceding smolt run period (16%). The same trend of lower proportions in the autumn-winter was also apparent for Stone Loach and Minnow.

On the R. Tweed during the autumn 2019 period (Sep/Nov), the proportion of Salmon (13% by mass) in the diet of Goosanders was roughly similar to that in November 1992 but considerably smaller than was estimated in November 1991. The proportion of Trout (40% by mass) in autumn 2019 was higher than any other species in this sample and also the two historical samples. About a third of the autumn 2019 diet was Grayling (34% by mass), considerably higher than most other species but a little lower than recorded in the November 1993 sample. The proportions of Minnow, 3-Spined Stickleback, and Stone Loach were relatively small in autumn 2019 but tended to be higher that levels recorded historically. Proportions of Salmon recorded in the autumn samples were similar to those earlier in the year during the smolt run period, although they were lower than for many other prey fish in both samples. Comparing the 2019/20 sampling periods, there was a striking increase in the proportions of Trout and Grayling in the autumn sample, the latter species being absent in the spring. The proportion of Minnow declined dramatically later in the year in comparison to the level in the smolt run period. This pattern was also seen within the autumn sample itself, with increased Trout and Grayling proportions later in the autumn, and lower proportions of Minnow, 3-spined Stickleback and Stone Loach later in the season.

The sample of R. Spey Goosanders from the autumn-winter 2019/20 period was too small (8 birds, 7 with food) to be confident it was representative of general diet. Nevertheless, the diet of this sample was dominated by Salmon (64% by biomass), followed by Trout (34%) and Minnow (2%). Due to the constraints of small samples (current and historical), no comparison was possible with previous data. Comparisons with findings from the 2019 smolt run period may be compromised by small sample size, but it is interesting to note that the proportion of Salmon increased later in the year compared to the smolt run sample period (from 36% to 64% by mass). At the same time, the dietary proportion of Trout decreased somewhat (from 40% by mass to 34%).

The numbers of Cormorants available (and/or containing food) from the autumn-winter 2019/20 sampling period were higher than those sampled in the spring, but were sometimes still small. For instance, only three of the four autumn-winter Cormorants from the R. Dee contained food. One contained a large adult Salmon and the other two each contained a single Pike. Due to the constraints of small samples (current and historical), no comparison was possible with previous data, including that from the 2019 smolt run period when no Cormorants were available.

The small number of Cormorants from the R.Nith during the smolt run period meant that comparisons were not possible with the autumn-winter sample of 12 stomachs with food where the diet was dominated by adult Salmon (37% by mass), Grayling (34%), and Trout (16%). The proportion of Salmon in the autumn-winter diet was low (2% by mass), as it was for several other species that were recorded in small (sometimes single) numbers. Due to a lack of historical samples, comparisons were not possible with historical data.

Good samples of Cormorants were collected during autumn 2019 on the R. Tweed (24 stomachs with food) and the highest proportion of the diet comprised Trout (26% by mass). This was considerably lower than the proportion recorded during the smolt run period, although the sample of Cormorants in the spring was very small. The same trend was apparent for Salmon with the proportion by mass falling between the smolt run and autumn periods from 40% to 8%. Apart from Trout, much of the autumn diet (44%) comprised single large (adult) Salmonids in each of three stomachs. Several Grayling were taken in the autumn (11% by mass), higher than the level recorded in spring (6%). Within the autumn period itself, dietary proportions of Trout and Salmon increased a little, whilst the proportion of Grayling increased considerably. More large Salmonids (presumably breeding fish) might have been taken later in the autumn (one fish early vs two late) but the sample is not large enough to certain of this.

With only a single Cormorant available from the R. Spey during the smolt run period, no comparison was possible with the autumn-winter where samples were also small (only 8 stomachs with food). Later in the year, the diet was dominated by large (presumably breeding) fish: seven large Trout, two adult Salmon and an unidentified Salmonid. The dietary proportions of Salmon and Minnow were exceptionally low in the autumn-winter sample – with only one individual of each species being found in those stomachs with food. Due to a lack of historical samples, no other comparisons were possible with previous data.

Overall, for both Goosander and Cormorant diet in the autumn-winter 2019/20, there were very few available historical comparisons. The exception was for Tweed Goosanders where, broadly similar to both historical samples, diet was dominated by Trout and Grayling, proportions of Salmon were lower, and those of other fish species were lower still. However, as noted above for the smolt run period, care has to be taken with such comparisons because of the possible temporal mismatch between sampling dates.

Samples from both the smolt run and autumn-winter periods in 2019/20 were, again, generally too small to compare - either in autumn-winter (Dee and Nith Goosanders and Cormorants) or in both periods (Spey Cormorants). For the three paired samples where dietary comparisons were possible, there appeared to be broad patterns. For Nith Goosanders and Tweed Goosanders and Cormorants, the proportions of Salmon in the diet (by mass) were lower in the autumn-winter period than in the smolt run samples. Proportions of Minnow were also higher in the diet of Goosanders earlier in the year on both rivers, whilst the proportions of Trout, and particularly Grayling, were higher in the autumn-winter. A similar but more modest increase in the proportion of Grayling in the diet of Tweed Cormorants was also recorded later in the year.

These birds contained considerably lower proportions of Salmon and Trout in the autumn than during the spring, as well as many large (adult) Salmonids, presumably reflecting the abundance and activity of adult spawners later in the year. Comparing Goosander and Cormorant diets in early and late autumn for the Tweed showed a general pattern of similar proportions of Salmon but higher proportions of Trout and Grayling later in the autumn for both bird species. There was also a suggestion that Cormorants might consume more adult Salmonids as the autumn progressed but further samples would be needed to confirm this, both on the Tweed and other rivers.

There is thus a clear future research requirement for further dietary data for Goosanders and Cormorants from times of year outside the smolt run period in the spring. For both bird species, this would improve our understanding seasonal variations in the relative proportions (and sizes, see below) of juvenile Salmon throughout the year as well as putting predation during the spring into broader context and improving our understanding of fine-scale temporal variation in diet and any implications this has for the degree of mismatch between comparative samples (see above).

Proportions of Salmon in the diet: broad summary

Where samples were adequate (12 or more stomachs with food), the proportions of Salmon (by mass) in Goosander stomach contents varied between each of the four rivers sampled. For the smolt run period samples, these were 50% (Dee), 36% (Spey), 16% (Nith), and 15% (Tweed). The only adequate sample in the autumn-winter was from the Tweed where the dietary proportion of Salmon was 13%. The broad pattern from the present study was thus that Salmon comprised higher proportions of the diet in northern rivers than in southern ones, a finding consistent with earlier work (Marquiss et al. 1998) that showed a similar latitudinal trend. Salmon comprised the highest dietary proportion only for the Dee Goosander sample in spring. For the remaining samples, 1-3 other fish species were consumed in larger proportions by mass, invariably Trout and/or Minnow and up to one other species. Again, this is consistent with Marquiss et al. (1998, and references therein) that Goosanders and Cormorants are generalist predators.

For Cormorants sampled in the smolt run period and for all other autumn-winter samples, similar dietary trends were apparent, despite the small sample sizes involved (10 or fewer stomachs with food).

Estimated Salmon length frequencies

Salmon length frequencies were generally bimodal, sometimes multimodal, and occasionally unimodal distributions and were often skewed to some degree. The differences in Salmon length frequencies between samples were compared using a statistical test applicable to data with such distributions. Furthermore, the test had no restrictions on sample size and so small sample sizes were acceptable.

Nevertheless, as before, care had to be taken when interpreting these findings and making comparisons between samples, primarily because of the incidence of small samples (fewer than 12 stomachs with food) which might not be representative of general diet (and similarly particular prey species within it). However, because even a sample of more than 12 birds is still relatively small, there may also be an - untested - influence of sampling location and/or of restricted sampling period on dietary assessments. For example, samples may come from a relatively small number of beats and/or from a short, restricted sampling period (e.g. a high proportion of birds sampled in particular weeks during the season) - see Section 2 and Appendix 3 for details. Further research is thus required to explore the scale of any potential influence of sampling location and sampling period on the results of stomach contents analysis, including Salmon length frequencies. Similarly, any individual bird may select smaller (or larger) than average fish when foraging (or, alternatively, be feeding in a location where such fish occur). As such, the fish recorded in the stomach contents are not an independent sample. Larger datasets than were available in the present study should thus be used to explore whether statistical analyses need to take account of the stomach as a 'sampling unit' (see similar approach for general diet assessment in Methods), perhaps comparing the 'average' size of fish per stomach.

(a) smolt run period

Estimated Salmon length frequencies for Dee Goosanders in spring 2019 were significantly smaller than those recorded from birds sampled in Mar-Apr 1995 and spring 1996. Length frequency distributions suggested fewer large individuals over 100mm and more smaller ones, especially in the 41-70mm length range. Salmon from R. Tweed Goosanders in 2019 were significantly shorter than those recorded in Apr 1991 and in April 1995. Length frequency distributions suggested fewer large fish over 100mm and more smaller ones, especially in the 61- 70mm length category.

Salmon length frequencies estimated from Tweed Cormorants in 2019 were also significantly shorter than those recorded in the Mar-Apr 1995 sample. The length frequency distribution suggested that numbers increased with size between 61-100mm length range and that almost no fish over 100mm were recorded.

Length frequency distributions of Salmon taken by Goosanders in 2019 were not significantly different to that recorded in the 1990s for the Nith (Mar-Apr 1995) or the Spey (April 1996), and all other samples were too small for any comparisons to be made.

Whilst the proportions of Salmon (by mass) in the diet of birds appeared to be fairly similar to those recorded previously (see Section 4), their size was often significantly smaller in most 2019 smolt run period samples than was recorded historically. This was the case in five of the seven possible comparisons where median Salmon lengths in these 2019 smolt run period samples were 62mm (Dee Goosanders), 70mm (Nith Goosanders), 79mm (Tweed and Spey Goosanders), and 88mm (Tweed Cormorants).

Smoltification is a growth- and size-related developmental event in the life-history of juvenile Salmon with photoperiod as the major cue determining development, modulated by both water temperature and water flow. These are likely to vary in importance (and stimulate migration differently), probably reflecting adaptations to ensure optimal timing of migration in relation to subsequent survival and growth at sea (Todd et al. 2011). Although the major movement of juvenile Salmon is generally considered to be the spring smolt run, juveniles can also make major downstream movements in autumn prior to their outmigration as smolts in the following spring (McCormick et al. 1998). Such autumn movements are discussed further below.

For many fish species, the length frequency distribution of individuals of each given year-class (i.e. all approximately the same age) typically approximates to a Normal distribution about some mean length, often with an appreciable standard deviation (Gulland & Rosenberg 1992). These length-at-age distributions are a striking feature of juvenile Atlantic Salmon growth variation and there is a strong size and growth bimodality between early and late smolts (Thorpe et al. 1982). This size (i.e. length) bimodality can be established as early as age 0+ (young-of-the-year) for Salmon in rivers with relatively fast growth (Letcher & Gries 2003). As a consequence, fish smolting in a particular spring are considerably larger than those that delay smolting for one or more years.

Across its natural range in the temperate and subarctic regions of the North Atlantic Ocean, the average total body length of wild Atlantic Salmon smolts is usually 10-20 cm, with minimum- maximum values from about 7cm to 30 cm (Thorstad et al. 2011), although these values are for Total Length (TL) and so likely to be a few millimetres longer than the Fork Lengths (FL) reported here.

As yet there is no way that Salmon smolts can be differentiated from parr in the partially digested remains of most fish in stomach contents of birds. Marquiss et al. (1998) considered that "larger" Salmon greater than 89mm in length were "of most interest to fisheries managers". Although not explicitly categorising individuals as either smolts or parr, it was presumably considered that individuals over 89mm were more likely to be either smolts or parr prior to smoltification.

Identifying the precise nature of Salmon length frequency distributions in the 2019 smolt run period samples was not straightforward. Nevertheless, those from each sample of Goosanders seemed to show (at least) a bimodal size distribution of Salmon less than 80mm long and of larger individuals (Dee, Tweed, Spey), and of less or more than 90mm long (Nith). However, in the absence of assessments of the length frequency distribution of smolts in each of the four study rivers, and given that those determined from samples in the present study were found to be smaller (median length 63-88mm) than those reported from historical samples, it is worth noting that these fish are smaller than those categorised as "larger salmon" by Marquiss et al. (1998).

Consistent findings in the present study of relatively small Salmon in the diet of Goosanders and Cormorants does not necessarily mean that at least some of the individuals consumed were not smolts. Indeed there is evidence of a steady, long-term (1963-2003) reduction in mean river age (i.e. the period spent in freshwater), and hence size, for juvenile Salmon in the North Esk, Scotland (data presented by Todd et al. 2011). This reduction may be the response of juveniles to recent milder winters in freshwater or to seasonal changes of return migration timing of returning (1SW) adults of contrasting river age, although clear interpretation was hampered by the possible influence of a legal netting season. Nevertheless, this reduction in mean river age for juveniles is not the result of an increase in their smolting at only one year of age, but of a clear shift from three- to two-year old smolts (Todd et al. 2011), presumably with a concomitant reduction in smolt size.

It is currently not possible to determine how much of the Goosander (or Cormorant) diet in the 2019 spring samples comprised (smaller) resident parr compared to larger (migrating) smolts. A future research requirement is thus to improve our understanding of the relationship between the length-at-age distributions of juvenile Salmon in river systems and the length frequency distributions of those individuals consumed by Goosanders and Cormorants. Inevitably, this will be linked to improved understanding of the both the timing and length of the smolt run period and to the temporal dynamics and spatial distribution of juvenile Salmon in the mainstem and tributaries over this period, but also at other times of year, particularly the period of downstream movements in autumn.

(b) autumn-winter period

Autumn-winter 2019/20 samples could not be compared with any historical ones, but only with those from the preceding smolt run period. However, these comparisons should be treated with caution because of the sample size issues described above for both the number of stomachs containing Salmon and the numbers of Salmon recorded therein (see Section 5b and Table 16). Nevertheless, in three of the four pairs of samples that could be tested statistically, the sizes of Salmon recorded in autumn-winter 2019/20 were larger than those recorded in the previous smolt run period.

Estimated Salmon length frequencies from Dee Goosanders in November 2019 were significantly larger than those recorded the previous Mar-May, with more larger individuals over 100mm and fewer smaller fish 60mm or less. Similarly, Salmon from Spey Goosanders in autumn-winter 2019/20 were significantly larger than those recorded in the previous Apr-May with more larger fish over 90mm and fewer smaller than this. For Tweed Cormorants, Salmon recorded in autumn 2019 were again significantly larger than those recorded in the preceding Apr-May, with no fish over 120mm being recorded in the spring. No significant difference was found between the lengths of Salmon taken by Tweed Goosanders in the autumn and smolt run period samples in 2019.

Relatively large numbers of birds were available from the R. Tweed, and, whilst acknowledging that the numbers of stomachs containing Salmon were small (N = 5-9), it was possible to compare Salmon lengths in 'early' and 'late' autumn 2019. Salmon taken by Tweed Cormorants later in the autumn (November) were significantly larger than those taken by birds earlier that season (Sep- Oct). The same pattern was found for Tweed Goosanders too, although the differences were not significant.

For the autumn-winter samples, Salmon from Dee Goosanders fell into two groups: those less than 70mm and those over 91mm. Corresponding groups for Spey Goosanders, were less than 70mm or over 80mm, and for Tweed Cormorants were less than 100mm or over 111mm. In these samples, Salmon consumed in the autumn-winter were larger than those consumed in the previous spring. Similarly, on a finer temporal scale, Tweed Cormorants appeared to take larger Salmon as autumn advanced, with the median length of 133mm in November of 133mm (cf. Sep- Oct = 88mm) being the largest recorded in the present study. There was thus a clear bimodality in Salmon length frequencies in the autumn-winter samples (as described above for the smolt run period) and evidence that both Goosanders and Cormorants later in the year were taking larger Salmon than they were recorded consuming in the spring.

It is possible that at least some of the predation of Salmon in autumn could be related to the downstream movement of juveniles at that time of year, prior to their outmigration as smolts the following spring (see McCormick et al. 1998). An important future research requirement is thus to collect more information on the dietary proportions of likely smaller ('resident') and larger ('mobile') parr at times of year outside the smolt run period, particularly during the downstream movement phase of larger pre-smolt fish in the autumn. As with the smolt run period itself, this will inevitably be linked to improved understanding of both the timing and length of autumn parr movements and to the temporal dynamics and spatial distribution of juvenile Salmon in the mainstem and tributaries over this period.