Dynamics and reproduction of a nest-box breeding population of Pied Flycatcher Ficedula hypoleuca in a subalpine birch forest in Swedish Lapland during a period of 46 years

The Pied Flycatcher Ficedula hypoleuca is certainly one of the most studied European passerine bird species and the lots of publications deal with different aspects of its life (for an overview see Lundberg & Alatalo 1992). However, studies on the breeding biology of the species at the periphery of its range in Scandinavia as in subalpine habitats are few. Here environmental factors, as weather conditions, are relatively unpredictable and much variable within and between breeding seasons, and put the possibilities of the Pied Flycatcher to accomplish successful breeding to the test. Longterm studies are needed to identify influencing factors and evaluate the degree of their effect during the breeding course. Besides the present study, long term studies on breeding conditions and population dynamics of the Pied Flycatcher in subalpine habitats are performed in northern Finland and Norway. Periods of harsh weather conditions may strongly affect the breeding result of pied flycatchers (Järvinen 1983, 1986, 1989a, Järvinen &Väisänen 1984, Thingstad Dynamics and reproduction of a nest-box breeding population of Pied Flycatcher Ficedula hypoleuca in a subalpine birch forest in Swedish Lapland during a period of 46 years


Introduction
The Pied Flycatcher Ficedula hypoleuca is certainly one of the most studied European passerine bird species and the lots of publications deal with different aspects of its life (for an overview see Lundberg & Alatalo 1992).However, studies on the breeding biology of the species at the periphery of its range in Scandinavia as in subalpine habitats are few.Here environmental factors, as weather conditions, are relatively unpredictable and much variable within and between breeding seasons, and put the possibilities of the Pied Flycatcher to accomplish successful breeding to the test.Longterm studies are needed to identify influencing factors and evaluate the degree of their effect during the breeding course.
Besides the present study, long term studies on breeding conditions and population dynamics of the Pied Flycatcher in subalpine habitats are performed in northern Finland and Norway.Periods of harsh weather conditions may strongly affect the breeding result of pied flycatchers (Järvinen 1983, 1986, 1989a, Järvinen &Väisänen 1984, Thingstad Dynamics and reproduction of a nest-box breeding population of Pied Flycatcher Ficedula hypoleuca in a subalpine birch forest in Swedish Lapland during a period of 46 years Beståndsvariation och häckningsresultat hos holkhäckande svartvit flugsnappare i fjällbjörkskog i Lappland under 46 år N. ERIK I. NYHOLM A nest-box breeding population of Pied Flycatcher Ficedula hypoleuca in subalpine birch forest at Ammarnäs in northern Sweden was studied in 1965-2010.The population showed on average a significantly decreasing trend during the initial 25 years, after which it remained on a static level.The yearly variation of the population size was significantly correlated with breeding result in preceding years, which in turn was affected by e.g.nest predation, adverse climatic factors, clutch size, and incidence of defective egg shell formation.It is also suggested that the population decline during the initial 10-15 study years followed from over-establishment of the breeding habitat.The average breeding result was 2.3 fledglings per pair, which is far below the number needed for the population to be self-reproducing.The occurrence of unusually high rates of nest predation and defective egg shells seems to be passing phenomena, predicted to approach zero after about 30 and 60 years, respectively, after the start of the study.But even with these factors eliminated and with other conditions remaining unchanged, the Ammarnäs population will continue to be a sink population.

Erik Nyholm, Department of Ecology and
Environmental Science, Umeå university, SE-901 87 Umeå, Sweden.E-mail: erik.nyholm@emg.umu.se1997).In North Scandinavian breeding habitats there are cyclic populations of small rodents (e.g.Hörnfeldt 2008) and lepidopterans (e.g.autumnal moth Epirrita autumnata in the subalpine birch forest; Selås et al. 2001), which might indirectly or directly influence the life of the birds.The present study furthermore unveiled that birds breeding along lakes in the study area often were subjected to environmental problems which were suggested to be related to acidification (Nyholm & Myhrberg 1977, Nyholm 1981).

Material and methods
The study was performed in the subalpine birch forest on the southerly facing slopes of the mountains Gaisatj and Valletjåkke, about five to ten kilometres west of Ammarnäs, Swedish Lapland (Figure 1).The approximate position of the nest-box area was 65°58´ N, 16° E, at 500 to 650 m.a.s.l.The habitat is a predominantly rich type of forest with luxuriant undergrowth of herbs ("meadow birch forest").In dryer parts the forest turns into the heath type.The habitats of the nest-box plots Table 1.Nest-box plots and the number of nest-boxes 1965-2010.78 and 58 nest-boxes of black plastic are included in the numbers for V1 1967V1 -1981V1 and T1 1967V1 -1979, respectively. Figures , respectively. Figures   have remained unchanged during the study period, and are similar to those of the mountain slope as a whole (cf.Enemar et al. 2004).The Pied Flycatcher occurs naturally as a relatively sparse inhabitant on these mountain slopes.The nest-boxes were erected in separate plots (Table 1).Some of the plots had to be forsaken during the study period (F in 1979 and G in 1980) and others were established along Lake Stor-Tjulträsket (R in 1977 andTO in 1979).The positions of the currently studied nest-box plots K2, V2, E, TO, T2, and R are shown in Figure 1.The plots F and G were situated from about 0.3 to 1 km to the east of plot E. Plot V1 extended along the slope of Valletjåkko just to the north and east of plot V2, and plots K1 and T1 had about the same extension as K2 and T2, respectively.When estimating the sizes of the plots and the densities of nest-boxes (Table 1) the radius of the activity area of breeding Pied Flycatchers was set to 50 metres.Thus an area could also be calculated for K2, where the nestboxes are situated in one 3 km long row, as 3 km x 100 metres (30 ha).
The inner bottom area of the nest-boxes measured 9x10 cm, the diameter of the entrance hole 32 mm, and the distance from its lower rim to the box base was 16 cm.The quality of the nest-boxes in plots V and T varied in 1967-1981 and 1967-1979, respectively, when nest-boxes made of a black plastic material, with the same diameter of the entrance hole but with somewhat smaller bottom area (VB and TB), were available besides the wooden ones.Since 1982 all nest-boxes were wooden and had uniform dimensions.The nest-boxes were originally placed at the height of some meters, and were then only reached by using a ladder.To facilitate the inspections all boxes, except the K-boxes, were moved to about 1.5 m in1988.
The nest-boxes were inspected regularly each season to get information on at least the breeding number, clutch size, period of egg laying (except in 2003), hatching success, and the final breeding result.Throughout the paper a breeding attempt means that at least one egg has been laid and a successful attempt means that at least one fledgling has been produced.
Eggs were also inspected for eggshell porosity (Nyholm & Myhrberg 1977, Nyholm 1981).Incidences of nest predation and other causes of nest desertion (e.g.disturbance by the investigators, Wrynecks Jynx torquilla, voles) were registered.Breeding birds were caught to be weighed and ringed (except in1982-1984).Females were usually caught when incubating, and males when feed-ing the nestlings.Nestlings were ringed when five days or older.
The K-boxes (K for control) were inspected showing special discretion to minimise disturbance of the breeding birds.Thus, ringing and measuring of the adults was restricted to the feeding period, the breeding phase at which they are least apt to be disturbed.
Nest predation, which was predominantly by small mustelids (least weasel Mustela nivalis and stoat Mustela erminia) and probably sometimes by martens Martes martes, occurred at any breeding phase (egg-laying, incubation, or nestling period).Where nests were preyed upon, occurrences of excess feathers and wings or legs that had been bitten off were looked for within and close to the nest-box to judge whether the breeding female or male was killed or not.A few nests were destroyed by Wrynecks (1975 and1979) or small rodents.
The rate of predation by mustelids was related to the abundance of small rodents (bank vole Myodes glareolus, grey-sided vole M. rufocanus, field vole Microtus agrestis, and lemming Lemmus lemmus).These were censused during1983-1995 and 2000-2001 by means of the Small Quadrate Method (Myllymäki et al. 1971).Data from 1975-1982 (on the bank vole populations only) was obtained from Gustafsson (1983), and concerning all rodent species in 1996-1998and 2001-2010from Hörnfeldt (2011)).Abundance of small rodents is presented as number of trapped animals/100 trap nights.The small rodent populations also reached high peak values in 1973 and 1974, but no figures can be presented from these years.As census methods and species selected were not quite uniform among the contributors of data the yearly population values given are not fully comparable, but since the amplitude is so large and species often synchronous they indicate the pattern and size of variation in the small rodent population density sufficiently well for this study.
Data on temperature and precipitation were obtained from the Swedish Meteorological and Hydrological Institute (SMHI).Most years the data refers to the local situation at Ammarnäs.In years when local measurements were lacking the meteorological data from other weather stations were used without adjustments.Thus, during one week in May or June in 1992-1995 data originated from Björkheden (30 km ESE of Ammarnäs) and all data from 2001-2010 are from the Boksjö weather station (35 km SSW of Ammarnäs).Overlapping temperature data from Boksjö and Ammarnäs during 15 May to 15 July in 1999 and 2000 showed similar mean value (0.6 °C higher at Ammarnäs).During the same period the rainy days at the stations were in close agreement, and gave as a mean 0.9 mm more precipitation at Ammarnäs.
Most of the calculations are based on time series, which means that the variables are not always independent because a value any given year is partly dependent on its value the preceding year.Correlation and regression tests (linear, quadratic, cubic, and exponential models) have nevertheless been applied as they are considered to be efficient and

Occupancy of nest-boxes
Six bird species used the nest-boxes for breeding, among which the Pied Flycatcher constituted about 95% of the occupants (6162 cases).The other five species were Great Tit Parus major (250 cases), Redstart Phoenicurus phoenicurus (89), Wryneck (18), Siberian Tit Parus cinctus (6), and Blue Tit Parus caeruleus (1).The first two were breeders in all and forty of the forty-six years, respectively.The yearly rate of nest-boxes occupied by breeding Pied Flycatchers varied between 24 and 86 percent, being highest in the first season, 1965, and lowest in the 1990s and 2000s.Occupancy showed a significantly declining trend during the study period (linear regression: r 2 = 0.57; b = -0.83;p<0,001) (Figure 2).Applying the cubic regression model, which indicated a stabilisation of the nestbox breeding population from early 1990s, gave, as did an exponential model, a somewhat better fit (r 2 = 0.59) than the linear model.
The yearly occupancy of the various nest-box plots co-varied in 1983-2010 (Table 2).
The three plots V2, E, and K2 were supplied with nest-boxes at different densities: plot V2 contained 5.3, E 3.3, and K2 1.0 nest-boxes/ha.When newly established, the densities of breeding Pied Flycatchers differed between the plots.Thus, in 1983 to 1987 on average 2.1 pairs/ha bred in plot V2, 1.6 in E, and 0.8 in K2.The trends of the population density over time varied between the plots, however, so that V2, with the highest initial breeding density, showed the most pronounced decline (Figure 3).The density in plot E likewise showed a significant decrease, while the density in plot K2 showed no trend.The best fit to describe the trends of the population densities in plot V2 and plot E  was obtained by means of the cubic regression model (V2: r 2 =0.54;E: r 2 =0.49).This model indicates that the decreases in the plots V2 and E were most pronounced during about the first ten years.Later, in 1993-2010, the mean population densities in the two plots were similar (mean ± s.d.: 1.1 ± 0.21 and 0.9 ± 0.17 pairs/ha, respectively,).During the same period did on average 0.6 ± 0.10 pairs/ha breed in K2.

Breeding performance
Breeding was interrupted in as much as 49.7% of the totally 6270 nests with complete or incomplete egg clutches in 1965-2010, for known reasons (nest predation 16%, disturbance by the researchers 1.9%, and female found dead 0.8%) or unconfirmed reasons (31%).The variation between years of the latter category was unrelated to predation rate, but strongly correlated with the mean daily air temperature in June (Figure 4).The number of eggs in complete clutches varied between 1 and 10 eggs.Ninety-five percent of the clutches (without defect eggshells; see further below) contained four to seven eggs (Figure 5).A single clutch with ten eggs occurred.Such an extreme clutch size could be suspected to emanate from more than one contributing female.As, however, only one egg was laid per day (3-12 June 1974), one and the same female was observed to incubate the actual clutch, and hatching (6 young) occurred 15 days after the last egg was laid, all eggs were probably produced by that single female.
Clutch size was related to the date of laying, so that females that started later on average laid fewer eggs, that is the "calendar effect" (v.Haartman 1967).In 1965-2010 the laying started earliest at 22 May and latest at 4 July.Ninety-five percent of the clutches were initiated from 28 May to 18 June (Figure 7).The reduction rate of clutch size was 0.075 eggs for each successive day of delayed start according to the linear regression formula "Clutch size = 8.35-0.075*Date(from May 1 st )" (r 2 =0.95; p<0.001) (Figure 8).The yearly mean date of the start of egg-laying varied from 30 May (year 2000) to 15 June (1977) (Figure 9).Other years with early egg-laying were 1981, 1983, and 1984, while egglaying was relatively late in 1991 and 1995.The yearly mean date of laying start was on average 7 June and showed no trend over the study period.
Variation of clutch size with age was demonstrated by females that returned to breed one and/ or two years after having been ringed as nestlings.Thirty-nine females at the age of one year started breeding on average on 9 June, one day later than the average date of the total population (8 June) and produced clutches of similar size (mean 5.33 eggs) as population average (5.40 eggs) the corresponding seasons.The mean laying date of the two years old females was on average four days earlier (2 June) than the population mean the corresponding seasons (6 June).The average clutch size of two years old females was 5.82 eggs (n=22) and the average for the total population was 5.52 eggs in the corresponding seasons.When standardised by the laying date, the mean clutch size of two years old females was similar to that of the total population (5,82-4*0.075= 5.52), as was the average clutch size of the one year old females (5.33+7*0.075= 5.86) to that of the two years old females.1965-2010 (utom 2003) (1 = 1 maj).
In the nest-boxes located nearest to Lake Stor-Tjul träsket (plots R, T2, and TO), and only occasionally in other boxes, a considerable number of the females produced eggs with defective shells (See further below!), usually combined with reduced clutch size.On average, 24% (N = 686) of the clutches in these boxes contained one to six defective eggs.The size of such clutches averaged 4.2 eggs (range 1-7 eggs).Eggs with defective shell were found in only nine (0.2%) of the clutches in the nest-box plots located distant from the lake.Also the mean size of the clutches with all eggs normal, was smaller in the plots along Stor-Tjulträsket than in the other nest-box plots, 5.19 eggs/clutch (1-8 Mean hatching rate of the eggs in complete clutches, 1965-2010, all plots included, was only 62.8% (Table 3).Hatching failed totally in 27% and partial hatching failure struck 18% of the clutches.Identified proximate causes for partial hatching failure were occurrence of unfertilised eggs, embryo mortality, and desiccation of the egg contents due to defective shell formation.
Nest predation by mustelids and air temperature during the incubation period were the main factors that affected hatching rate.Together these factors accounted for 71% of the variation.Predation alone accounted for 62% of the yearly variation in hatching rate (p<0.001)(Figure 11).The best fit to describe the yearly variation in the hatching failure rate, as well as the predation rate, was obtained by the cubic regression model (r 2 = 0.34 and r 2 = 0.29, respectively).This model indicates that the rates of predation and hatching failure culminated in the 1990s.
Hatching success was about equally affected by predation in all nest-box plots.Typically the depredated clutches were totally destroyed.The yearly variation in hatching rate of the partially hatched clutches was insignificantly related to predation rate.The significant negative influence of low daily mean temperature during the incubation period on hatching success is illustrated in Figure 12.
The hatching rate varied between the nest-box plots if depredated clutches were excluded.The lowest hatching success was then in the plots located along Lake Stor-Tjulträsket, where eggs with abnormally porous shells were frequent.This was the main reason why only 26% of the eggs hatched in the clutches with eggshell defects (depredated clutches excluded) (Table 4).
Nest-box plots K1 and K2 showed the highest hatching rate.When losses due to depredation were excluded, the hatching rate of complete clutches in these plots amounted to 84.9% compared to 76.3% in the other plots distant from Lake Stor-Tjulträsket (F, G, E, V1, and V2) and 62.9% in the plots T1, T2, TO, and R. (Table 5).
The yearly nestling survival rate 1965-2010 varied considerably (range 27 to 98%; CV = 26%), and was on average 77.6%.Predation, air temperature, and outbreaks of autumnal moth larvae were factors studied for their influence on the nestling survival.
Nest predation caused the loss of about 6.0% of the nestlings in 1965 to 2010, and the yearly variations in nestling survival and predation rate were significantly correlated (r = -0.30,p < 0.05, N = 46).The nestling survival rate in 1965-2010 was on average 82% when depredated clutches were excluded from the calculations (Figure 13).
The yearly variation in mean temperature during the nestling period was not correlated to the nestling survival 1965 to 2010 (r = 0.14, p=0.34).In single years, however, bad weather caused reduced survival, e.g. in 1981, when low temperature (mean daily mean 10.6 °C) and continuous rain during   eleven days at the beginning of the nestling period was probably the main reason for the extremely low nestling survival rate that year.Also the low nestling survival in 1987 and 1990 coincided with periods of rainy weather combined with low mean temperatures (10.4 and 10.9 °C, respectively).The survival rate in 1988 was relatively high though the nestling period was equally rainy, but then the mean temperature was higher (15.3 °C).The low survival rate in 1977 was mainly due to high mortality during 12-15 July, which was a period with low air temperature (daily mean 6.6-10.7 °C) and predominantly young nestlings.Only a few clutches hatched before 5 July that season, due to late breeding start (Figure 9).
The yearly variation in nestling survival rate (after exclusion of depredated clutches) was correlated with the abundance (log) of autumnal moth larvae (r 1968-2010 = 0.36, p<0.02).Abundance of autumnal moth larvae has varied a lot in the subalpine birch forest at Ammarnäs since 1968, when a standardised census program was started (Andersson & Jonasson 1980, Selås et al. 2001).Four outbreaks have occurred, 1973-1975, 1984-1986, 1993-1995 (relatively week), and 2003-2005 (http://www.luvre.org/insekter).In the years with outbreaks, especially high numbers of larvae were available as prey items for several bird species.As the periods with larvae usually lasted to about the first week of July these could be available as food items for the pied flycatcher nestlings.At three of the four outbreaks the nestling survival rates were    1973-1975, 1985-1986, 1993-1995 (weak outbreak), and 2003-2005; from http://www.luvre.org/insekter.htm.)Depredated clutches excluded.
higher than average (82%) (with depredated clutches excluded from the calculations) (Figure13).Average nestling survival rate in 1969-2010 in nest-box plots K1 and K2 was higher (91.8%) than in the other plots distant from Lake Stor-Tjulträsk (79.8%), and in the broods along the lake (84.5%), when clutches being depredated and containing defect eggs were excluded from the calculations (Table 5).The survival of the nestlings from clutches which contained defective eggs was 78.4% (Table 4), i.e. at the same level as in the plots F, G, E, V1, and V2 (Table 5).
The breeding success, expressed as the yearly mean number of fledglings produced per breeding pair (i.e.all initiated clutches taken into account), was on average 2.26 fledglings (range 0.34-4.71) in 1965-2010 (Figure 14).The variation in the breeding success was best described by a cubic regression model (r 2 = 0.27) (Figure 15).This model indicates that the decreasing trend was broken in the 1990s and was followed by an increasing trend.
When all attempts were taken into account, breeding success in nest-box plot K2 (3.0 fledglings per pair) was better than in the other plots (1.8) in 1983 to 2010.Referring to successful broods only, the values were 4.2 and 3.1, respectively.

Eggshell defects
Each year, 1965-2010, a number of Pied Flycatcher females breeding near the shore of Lake Stor-Tjulträsket laid eggs with defective shells.Typically, the shells were thinner than normal and more or less porous, or even, the eggs totally lacked a calcified shell.This anomaly was described by Nyholm & Myhrberg (1977).Nyholm (1981) suggested that the abnormality was associated with increased exposure of aluminium obtained by eating insects at the shore line.It was also proposed that the amount of aluminium available was enhanced due to increased release from the bedrock following environmental acidification.
The shell defects caused abnormal evaporation from the egg contents, which rapidly caused death of the embryos during incubation, and reduced hatching rate.However, the survival rate of nestlings that hatched in clutches with one or more defective eggs was not affected (cf.Tables 4 and 5).
The rate of females producing defective egg-shells in the shore zone of Lake Stor-Tjulträsket increased during the study period (linear regression: r 2 =0.22,p<0.03).A quadratic regression model, which gives a better fit (r 2 = 0.47, p<0.001), indicates that the rate increased during the first 25 years and then declined from the end of the 1990s (Figure 16).
A two year study, 1975 and 1976, of the Pied Flycatcher clutches along the shore of Bissan, another lake in the Ammarnäs area, showed that 29% and 21% of the clutches, respectively, contained similarly defective eggs.Such eggs also occurred in nests of several other bird species breeding in the shore zone of Lake Stor-Tjulträsket: Common Sandpiper Actitits hypoleucos, Dunnock Prunella modularis, Bluethroat Luscinia svecica, Songt hrush Turdus philomelos, Redwing Turdus iliacus, Fieldfare Turdus pilaris, Willow Warbler Phylloscopus trochilus, Great Tit, Brambling Fringilla montifringilla, and Reed Bunting Emberiza schoeniclus.
The production of defective eggshells was only one of the obvious anomalies shown by Pied Flycatchers breeding along Lake Stor-Tjulträsket.Also a disproportionate part, 78% (39 out of 50), of the females found dead on their nests in 1965-2010, occurred in the nest-box plots by the lake, which constituted 41% of the total number of nests.

Nest predation
Rate of nest predation varied significantly during the study period (range 0-56%).Small mustelids (stoats and least weasels) were by far the most frequent predators.
Nest predation by mustelids was almost absent during the initial twelve years, 1965-1986, and high only in 1979 in connection with collapses of the small rodent populations (Figure 17).Thereafter, however, the predation rate was frequently high and occurred yearly, i.e. during all phases of the small rodent cycles.Only in 2007 and 2010 it was again very low at dense small rodent populations.A cubic model of regression seems to show the overall pattern with a decreasing trend during the 2000s (Figure 18).

Population dynamics
Pied Flycatchers are highly attracted to use suitable nest-boxes for breeding, as has been obvious in several studies carried out in a broad variety of forest habitats throughout the range of the species.In North-European habitats the introduction of nest-boxes may cause considerably increased local populations of different hole-nesting species, and predominantly Pied Flycatcher (e.g.von Haartman 1951, Meidell 1961, Campbell, 1968, Järvinen 1983, Sörensen et al. 1990, Nilsson 2008).This was true also in the sub-alpine birch forest at Ammarnäs, where on average 95% of the nest-box occupants were Pied Flycatchers and the remaining share was represented by five species.
The introduction of 116 nest-boxes in 1965, in the previously nest-box free forest, attracted 100 Pied Flycatcher pairs to breed within about 65 hectares, i.e. 155 pairs/km 2 .That density was at that time more than ten times that in the surrounding 9 km 2 of similar forest surveyed by Enemar et al. (2004).The number of territories is known to closely equal the number of breeding pairs (Enemar et al. 1976).Thus, the introduction of the nest-boxes at once more than doubled the breeding Pied Flycatcher population on the mountain slopes.This suggests that the density of breeding pied flycatchers in the nest-box free part of the forest was restricted by lack of natural cavities suitable for breeding.
The yearly nest-box occupancy co-varied with  Nest-box occupancy as well as territory density showed significant declining trends over the 46 years (Figure 2).The total population of Pied Flycatchers on the mountain slopes did thus decrease significantly during the study period.The indication that the decline of the nest-box breeding population was arrested from the 1990s (1990-2010: linear regression coefficient r 2 = 0.00; n.s.) was not found in the surrounding forest where the decline continued (1990-2010: linear regression coefficient r 2 = 0.32; p<0.005; Figure 2).The territory density of pied flycatchers outside the nestbox plots showed a relatively stronger decline than the population density in the nest-box plots.From 1983 to 2010, when the number and location of the nest-boxes were static (Table 1), an increasing share of the Pied Flycatchers on the whole mountain slope bred in nest-boxes (Figure 19).Thus, an allocation of breeding flycatchers to the nest-box plots from the surrounding forest took place without increasing the occupancy of the nest-boxes, but leaving potential breeding locations unoccupied in the surroundings.
The decreasing trend of the Pied Flycatcher population at Ammarnäs was syncronous with the large scale trend of other Swedish and European populations.These populations decreased 20-30% from the 1980s to 2009 (www.ebcc.info;Lindström et al. 2011;Ottvall et al. 2009).Several nest-box breeding Pied Flycatcher populations on the European continent showed decreasing trends since the 1970s (e.g.Coppack & Both 2002, Both et al. 2006), and in UK since 1990 (Goodenough et al. 2009), but others have reported increased populations (e.g.Winkel & Winkel 1998).
The declining trend of the nest-box breeding population at Ammarnäs was, however, also related to factors of more local character, of which the declining breeding success was the most significant proximate factor during the study period as a whole.The breeding success was significantly correlated to the nest-box occupancy in the two preceeding seasons (r= 0.55, p < 0.001 and r = 0.53, p<0.001, respectively).The rationale to test the influence of breeding success over two seasons is that a significant share of Pied Flycatchers does not breed until the age of 3K (Curio 1959, Harvey et al. 1985, Nyholm 1986).In turn, the yearly varia-tion in breeding success was highly correlated with the rate of nest predation by mustelids.The variation in rate of nest predation 1965-2010 accounted for 45% of the variation in breeding success, and 37% of the variation in occupancy the subsequent year.The variation in breeding success was also significantly correlated with clutch size and June temperature.Together with nest predation rate they accounted for 74% of the variation.The variation in predation rate was the only of these factors that showed any significant trend over the years.The density of territories outside the nest-box plots was also significantly related to the predator activity (as judged from the predation rates in the nest-boxes) (linear regression: r 2 = 0.13, b = -0.09,F = 6.1, p<0.02), indicating that the predators operate similarly in the whole nine square kilometre study area.
On average about 92% of the breeding females were yearly recruited from outside the nest-box plots; only about 8% returned in the subsequent year (range 0-26%; N = 3775; 1966-2010) with no trend over time (r 2 = 0.04, p = 0.19).The mean return rate of fledglings was 1.1% (Nyholm 1986) and of breeding males at about 24% (Nyholm & Myhrberg 1983).The return rate of the females was negatively influenced by nest predation the preceding breeding season (Figure 20), which consequently also had some impact on the breeding population size.Thus, to have a significant impact on the nestbox occupancy at Ammarnäs the affecting factors could not be only local but must have a concurrent impact in a larger area.That area could be very wide as judged from recoveries of ringed individuals.For example, two nestlings (females) ringed at Ammarnäs were recovered when breeding (2K) at Kilpisjärvi in northern Finland (about 400 km NNE), another nestling bred later at Tärnaby (46 km SW), and two individuals which were ringed at Gauto (50 km NNE) as nestling (male) and 2K female, respectively, bred at Ammarnäs the subsequent year.All these nestlings were born and recovered as breeders in subalpine habitats which indicates faithfulness to their birth habitat.The yearly rate of nest predation varied in relation to the phase of the small rodent population cycles, which, in turn, was synchronized over a wide subalpine area, extending hundreds of kilometres southwards and northwards of Ammarnäs (Hörnfeldt 2010).
In many long term nest-box studies (Table 6) it has been observed that the number of breeding Pied Flycatchers decreases during the initial one to two decades.In the present study, that was obvious in most of the study plots which were established in 1977-1983, i.e. plots E, V2, R, and TO.The decreasing trends in V2 and E (Figure 3) appeared to be density dependent as the decline rate was greater in V2, the initially most densely inhabited plot.In both plots the decreasing trends were discontinued after some ten years to get stabilised when the population densities had reached about 1 pair/ha (Figure 3).In comparison, the population density in plot K2 (on average 0.6 pairs/ha) did not show any trend during 1983-2010.A pattern similar to that shown by plots E and V2 can be visualised from data presented by Virolainen (1984).In two nest-box plots with an about three-fold difference in population density, the density decreased in the densest plot from about 2 pairs to about 1 pair per hectare during the twelve study years (r 2 = 0.73, b= -0.087; p<0.001) whereas in the less dense plot the trend was less accentuated (from about 0.7 to 0.5 pairs per hectare (r 2 = 0.24, b= -0.036; p<0.05).
This initial decrease of nest-box breeding popu-  First 20 years lations of Pied Flycatcher observed in many studies could tentatively be related to an over-establishment in relation to the carrying capacity of the habitat.The finding by Schmidt (1986), that Pied Flycatcher populations decreased more rapidly in pessimal than in optimal habitats, may support the idea that the initially decreasing density trends may be associated with depletion of food resources.That may also be supported by the observation in this study, that clutch size was smaller at higher population densities than at lower (V2; Figure 10), and that the nestling survival in the nest-plots E+V2 was lower during the first ten seasons, 1983-1992: on average 80% (n = 327 broods), when the population density was decreasing, than when the density had got stabilised at a lower population density, 1993-2010: 93% (n=296 broods).In plot K2, with stable low population density, nestling survival was 91% (n=131) and 92% (n=232), respectively, in the same periods.In all these cases depredated broods were excluded from the calculations.
As suggested from the rather stable population densities that the nest-box plots E and V2 reached after some ten seasons, a long term population density limit could be set to about one breeding pair/ ha in the actual subalpine habitat (Figure 3).This density was about ten times higher than the mean territory density in the surrounding forest.Different habitats may carry different densities of breeding Pied Flycatchers, e.g.Winkel (1989) reports no trend  for a nest-box breeding population where the mean density was about 2 pairs/ha.

Breeding performance
The primary ultimate factor that influenced clutch size was the time schedule of egg laying.The decrease rate of the clutch size with the progress of the breeding season at Ammarnäs was 0.075 eggs per day of delay, which agrees with that found in other studies on the Pied Flycatcher, on the European continent (Creutz 1955, Berndt & Winkel 1967, Zang 1975), in Britain (Lack 1966), and Fenno-Scandinavia (v. Haartman 1967, Järvinen 1980, Virolainen 1984, Lundberg & Alatalo 1992).But, the British and Fenno-Scandinavian populations lay larger clutches than the continental in relation to laying date (cf also v. Haartman 1967) (Figure 21).
The northern position and relatively high altitude of the study area, and accordingly an increased risk of hazardous weather in early spring, contributed to delayed breeding start (cf. Järvinen 1989b).This narrowed the margin between clutch size and the number of fledglings needed for self-maintenance of the population.That margin was on average less than one egg in more than half the number of the seasons (Figure 6) if on average 4.4 fledglings per pair are needed (Järvinen 1983).
Late female arrival and/or daily mean temperature below 5-6°C at end of May and beginning of June delayed the breeding start (Figure 22).The examples showing the situations in 1977 and 1991 illustrate delays caused by long periods with low temperatures.In 1975, some of the earliest arriving females were obviously triggered to start breeding during a mild period in the second half of May.A following period of colder weather probably temporarily hampered the egg laying and delayed the mean breeding start of the population.In 1995, the breeding start was late despite an extended mild period at the end of May and beginning of June.In that year the delay must be ascribed to late female arrival rather than unfavourable local weather.Early start of egg-laying (as in 1981, 1984, 1986, and 2000) occurred when the females arrived at a relatively early date and the weather conditions were favourable.
The average clutch size of one year old females was about 0.5 eggs smaller than that of two year old females, a difference which disappeared if clutch size was standardised by the laying date.Berndt & Winkel (1967) found the same difference but in contrast to the finding in the present study the difference still remained after taking account of a few days later breeding start of the one-year old females.
The breeding success varied much between years, but the number of fledglings was generally far below that needed for the population to be selfreproducing.That production was reached in only two of the forty-six seasons, and on average did only 2.26 fledglings leave the nests (Figure 14).In three of the seasons the production was less than one potential recruit per nest.During the study period, the nest-box breeding population at Ammarnäs has thus been a typical sink population, which had to be supplied with breeders from source populations elsewhere (Thingstad et al. 2006).
A number of factors contributed to low recruitment rate.A substantial factor was the high rate of total failures, almost 50% already during the laying phase, primarily due to nest predation, and abandoned nests during periods of low air temperature.
Abandoned nests were especially frequent in the nest-box plots along Lake Stor-Tjulträsket.Other factors were poor hatching success and poor nestling survival.The variation in hatching rate was significantly correlated with nest depredation rate and mean air temperature during the incubation period, and these two factors explained 72% of the variation.In the nest-box plots R, T and TO, situated along Lake Stor-Tjulträsket, reduced hatching was also due to the occurrence of defective eggshells.
Nest predation was also the primary single factor affecting nestling survival.More than a thousand nestlings (6%) were lost in that way.Depredation of nest-box populations by small mustelids, and its relation to density of small rodent populations, has since long been recognised in several areas in Europe (Dunn 1977, Järvinen 1990).The greatest variation in mustelid predation rate can be expected in geographic areas with cyclic populations of small rodents, as in northern Scandinavia, where mustelid populations can be built up during the high density rodent phases which typically appear every three to four years.The increased rate of nest-predation which occurred in the study area from mid-1980s to 2000s was probably related to increased mustelid populations as a consequence of reduced fox populations.An epizootic mange among foxes became prevalent in north-western Sweden in the early 1980s (Lindström et al. 1994).Increased mustelid (Martes) populations have been observed in consequence of decreased fox populations (Storch et al. 1990).Reduced fox populations and occurrence of cyclic small rodent populations, as in northern Scandinavia, should favour the build up of abnormally dense mustelid populations during the high rodent population phases.In these circumstances, subsequent crashes of the rodent populations may lead to abnormally high predation pressure on alternative prey species, and cause the extensive nest predation on Pied Flycatchers which occurred from the mid-1980s.Probably the observed "outbreak" of nest-predation by mustelids should be regarded as an episodic phenomenon, though long lasting, and that the relatively low predation rates in 2007-2010 as a return to the normal predation situation for the subalpine habitat (Figure 18).This implies that nest predation by mustelids is important only during the first breeding season after crashes of the cyclic small rodent populations, as seen in the present study in  and in the sub-alpine habitat at Kilpisjärvi 1966-1987(Järvinen 1990).In a study of nest-box breeding flycatchers performed in a similar habitat about 20 km SE of Ammarnäs was however no predation recorded during 1971-1985(Svensson 1987)).
Climatic factors, as low temperature and rainfall, contributed more or less to restrict the breeding success of the Pied Flycatchers by influencing breeding start, hatching success and nestling survival.The temperatures during female arrival, breeding start, incubation, and the nestling period, have varied between years without showing any trend over the study period.Järvinen (1983) states that egg formation is triggered at Kilisjärvi, in northernmost Finland, only after the daily mean temperature has reached about 6 °C.That limit seems to be valid also at Ammarnäs, where the first eggs of the year were laid about five days after that temperature limit was reached.These five days equals the time its takes for an egg follicle to ripen, ovulate, and the completed egg to be laid (von Haartman 1990).Relatively low temperatures, especially when combined with long rainy periods, could strongly affect the breeding result of Pied Flycatcher by forcing incubating females to leave their breeding cares for long periods (days), and even to abandon their nests, or by making it harder for the parents to feed their young properly.
The nest predation and the production of defective eggshells, two factors which had a strong negative impact on the breeding result of the Pied Flycatcher, occurred in the study area with the highest rates so far reported.The proximate cause why Pied Flycatchers and birds of other species breeding in the vicinity to Lake Stor-Tjulträsket produce defective eggshell was suggested to be environmental acidification (Nyholm 1981).An increas- ing share of the pied flycatcher females which bred along the lake became affected until mid-1980s, a share that subsequently decreased significantly during the following twenty years (Figure 16).This time schedule fits well with that of the variation in the sulphur deposition in Sweden, which has decreased to less than half of the levels recorded in 1970 (Fölster & Wilander 2002).Anonymous (2007) showed that the deposition of acidifying agents in 1980 strongly exceeded the critical load for lakes in northern mountain areas of Sweden, including the Ammarnäs area (>700 ekv/ha/year).The decreased deposition rate since then resulted in that no excess of the critical load for the lakes occurred in 2002-2004.This admits recovery from acidification in the lakes, which, as suggested by the decreasing rate of defective eggshell production observed in birds, also favours the terrestrial ecosystem in the vicinity of the lakes.With the indicated rate of recovery the incidences of defective eggs will disappear, i.e. return to the normal level, in some 15-20 years.
Besides in the shore zone along Lake Stor-Tjulträsket, occurrence of naturally breeding pied flycatchers which regularly produced defective eggshells have only been observed in severely heavy metal polluted environments (Nyholm 1995).
The tendency of improved breeding result of the pied flycatchers in the study area which was initiated in the 1990s (Figure 15) was strongly related to decreased predation pressure.In perspectives of a continuously declining predation rate to the situation before 1987 one could expect a significant concomitant improvement of the breeding performance of the pied flycatchers in the study area.

Figure 23 .
Figure 23.The variation in the numbers of fledglings per breeding pair 1965-2010, if clutches with defective eggshells and depredated clutches are excluded from the calculations.Broken line shows the total mean and the red line indicates the number of fledglings needed to make the population self-reproducing.Variationen av antalet flygga ungar per häckande par 1965-2010 om kullar med defekta äggskal och kullar rövade av mårdjur ej var med i beräkningarna.Streckad linje visar det totala medeltalet flygga ungar och röd linje anger det antal som behövs för att populationen skall vara självreproducerande.

Table 5 .
Rates of hatched eggs and surviving nestlings in three different categories of clutches in the nest-box plots K1, K2, plots F,G, E, V1, V2, and plots T1,T2,TO,R 1983-2010.Depredated clutches and those which contained defective eggshells are excluded.

Table 6 .
Linear regression of nest-box occupancy by Pied Flycatchers over years from the start of different studies: r 2 is the linear regression coefficient and b the regression slope.