Fish Eating Birds and Salmonids in Scotland

The populations of sawbills on the Rivers Dee and North Esk in North East Scotland

STUDY AREAS

River Dee

The River Dee study area (Figure 2.1) comprised the 2000 km2 catchment in Grampian Region, Scotland. The biology, geography and land use of the Dee watershed has been comprehensively reviewed (Jenkins 1985). The river, about 140 km long, drains parts of the East Central Highlands, falling from an elevation of ca. 1200 m above sea level to the sea. Its tributaries are much divided so that altogether there are about 390 km of waterway of width exceeding five metres from bank to bank, and many narrower streams. The river is steep in profile compared with other large British rivers, much of it runs over gravel or cobbles, and there is virtually no lowland 'depositing' section or estuary. For much of the year the river receives snowmelt. It drains predominantly granite hill ground and its waters are nutrient-poor, but pollution is rare. The river has 11 species of freshwater fishes, but only five (Atlantic salmon, trout, minnow, eel and brook lamprey), are widespread and abundant. For the Latin names of fishes, see Appendix 1.

Figure 2.1. Map of the River Dee watershed, showing the river, main tributaries, standing waters and towns.

There are few substantial standing waters in the catchment. Lochs in the hills freeze in most winters; they are deep, cold and oligotrophic, and support populations of small trout. Lowland lochs (less than 300 m elevation), particularly those draining arable land, are rich with a relatively luxuriant macrophyte flora, and support populations of perch, pike and eel.

River North Esk

The North Esk study area (Figure 2.2) comprised the 732 km2 catchment in Angus Region, Northeast Scotland. The river is about 65 km long and drains parts of the East Central Highlands, falling from an elevation of about 800m above sea level to the sea. Its upper tributaries drop steeply through heather moorland and sheepwalk to an elevation of about 100m asl, where both the main river and its biggest tributary, the West Water, pass through deep gorges into flatter, largely arable, farmland. The upper sections are fast-flowing over cobbled substrates but, in the lowland section, river gradient decreases to 2-7 m per km and the width increases to 30-40 m as the river runs slower with deep pools and mainly finer substrates. The lower 2.5 km of river is tidal and estuarine as the river mouth is protected by a shifting sand bar. Thirteen species of fishes have been recorded (Maitland & Campbell 1992), but only seven (Atlantic salmon, trout, minnow, stone loach, three-spined stickleback, eel and brook lamprey) are widespread and abundant above the tidal limit. The fish community of the North Esk, at least in the river downstream of the gorge, is more diverse than that of the Dee.

Figure 2.2. Map of the River North Esk watershed, showing the river, main tributaries and Loch Lee.

SEASONAL DISTRIBUTION OF SAWBILLS ON THESE RIVERS

Previous studies had shown that goosanders are most abundant on the upper sections of the Dee and North Esk in spring and summer, on the lower sections in late summer, autumn and winter, and on lochs in late winter and spring (Marquiss & Duncan 1994a). Many birds are in pairs in winter and spring, but adult males are absent from June to October, when the population is successively dominated by females, then females with broods, then juveniles. Most females nest far up tributaries and move downstream with small ducklings to nursery areas on the main stem. Once grown, juveniles probably move downstream again before they disperse from the watershed. Yearling males are present on the mainstem of the lower river in late April and May, and yearling females are present on the mainstem of the upper river and the nesting areas in late May, June and July.

In summer, goosanders fish chiefly on the main stem of the river, whereas in winter most birds use only the lower 20 km (Marquiss & Duncan 1994a). From October to March, the three main standing waters used by foraging goosanders are the Lochs of Kinord, Davan and Aboyne. From September to March, the birds commute to overnight communal roosts on standing waters (Marquiss & Duncan 1994b). All birds leave the roost before sunrise and return near sunset. Goosanders do not roost communally from April to August, except in May in an area where males gather prior to their leaving to moult.

Previous studies of red-breasted mergansers on the North Esk (Marquiss & Duncan 1993) had shown that they are largely absent from the river in late autumn and winter, returning in April and May as goosanders move upstream. There is relatively little overlap in their breeding distribution as most red-breasted mergansers breed on the lower sections, nesting in rank riparian vegetation. Males abandon the river by July, leaving females to incubate eggs in June and July, and rear young in July and August.

METHODS

The timing of our counts was devised to annually estimate the breeding population, the production of ducklings and the midwinter populations of sawbills on both rivers. On the North Esk, counts spanned the period from April 1987 to August 1993. Previous counts of goosanders and red-breasted mergansers by Carter & Evans (1986) were done at different times of year so, with the exception of their goosander duckling count in July 1986, their figures were not comparable with those from the present study. Counts of goosanders on the Dee started in December 1987 and finished in July 1994. Goosanders use 'traditional' cavity nest sites and we attempted to find these to monitor nesting performance.

Counts of goosanders on the River Dee

Goosanders were counted in three censuses. The numbers of breeding pairs on the Dee were estimated from weekly counts on sample stretches of river between Banchory and the Clunie (Figure 2.1) in April and early May. Duckling production was estimated from a count of well-grown ducklings in the brood areas in late July. The midwinter population was estimated from a count on the river from Aberdeen upstream as far as Aboyne in December.

None of the three censuses covered the whole river system, although the December and July counts probably covered all parts of the river used by goosanders at those times of year. In April, the goosander population is in flux and there is insufficient time to search all the breeding and foraging habitat so an index was derived from counts of adult drakes (Marquiss & Duncan 1994a). Five counts were made from the second week of April until the first week of May, on the stretches of river visible from 42 roadside viewpoints. Counts were done in the early morning within 2 hours of sunrise, a time of day when goosanders were actively foraging and before people walking the banksides disturbed them. From a more comprehensive census in 1988, there was estimated to be 61 pairs of goosanders (95% confidence limits, 59-63) in the Dee catchment. In that year, April and early May counts from the 42 viewpoints averaged 6.4 adult males per week; approximately 10.5% of the whole population.

Counts of goosanders and red-breasted mergansers on the River North Esk

The North Esk was sufficiently short to census breeding ducks by walking the whole length of the mainstem, and the main tributaries each spring. Goosanders were counted in mid-April to estimate the breeding population, in July to count ducklings, and January to count the midwinter population. Red-breasted mergansers were also recorded in mid-April counts, but they breed three to four weeks later than goosanders so their breeding population was estimated from whole river counts in mid-May. Their ducklings were counted in late July and early August, and winter numbers were recorded during the midwinter goosander count in January.

The nesting performance of goosanders

From 1987 to 1993, goosander nests were located by watching birds and examining potential cavity sites in trees and rocks. Most nests were found during laying and a few during incubation and at hatch. The sample of nests monitored varied according to our experience and effort. As we became proficient at finding nests the sample increased from 4 in 1987 to 17 in 1989 but then, to maintain the sample, search effort was increased as the goosander breeding population declined and occupied nest sites became scarce.

The laying date of the first egg was back-calculated by assuming intervals of 1.8 days between eggs and an incubation period of 34 days (Marquiss & Cook unpublished). Clutch size was the number of eggs in a nest where incubation had commenced. Five nests contained very large clutches of 15 to 19 eggs, and two of these nests were attended by two females, suggesting nest parasitism. However such nests were infrequent and there was no objective way in which parasitised nests could be distinguished, so all clutches were used in estimates of mean clutch size. Nests were revisited after hatch to count the number of hatched shells and record unhatched eggs. In three years, 1991, 1992 and 1993, some of the incubating females were given coloured leg rings and dyed, to distinguish individuals once they had left the nest with their brood.

Data analysis

Previous analyses (Marquiss & Duncan 1993, 1994a) had shown that sawbill numbers declined with distance upstream so that very low densities were recorded at higher elevations as streams narrowed. We therefore used the numbers of birds within river sections of 10 km length (measured from the river or tributary mouth on Ordnance Survey maps, scale 1:50 000) as our basic count unit, and elevation (m asl) and stream width (m, measured in the field) as covariates.

Counts were log transformed (adding 1 to accommodate zeros), before entering them into analyses of variance using General Linear Models (GLMs, computed with Minitab statistical package 'release 11', 1996). Count variation was examined with reference to 5 factors. These were 2 sawbill species, 2 rivers, 48 river sections (9 stretches on the North Esk, and 39 on the Dee), 9 years, and 4 times of year (April and May counts of adult ducks, summer duckling counts, and midwinter counts of full-grown birds). The standardised residuals from the linear model were checked for normality and further transformation of the data proved unnecessary. Duck densities were expressed as birds per hectare within 10km sections and these values were transformed using an optimally derived exponent (Box-Cox transformation, Minitab 1996), prior to entering them in GLMs.