Breeding status of Merlins Falco columbarius in the UK in 2008

Abstract

Capsule The third national Merlin survey estimated a UK population of 1162 breeding pairs (95% CI: 891–1462).

Aims To estimate the number of breeding Merlins (with associated 95% confidence intervals) in the UK and the four countries (Scotland, England, Wales and Northern Ireland), and to compare these with the relevant estimates from the 1993–94 Merlin survey. In addition, to calculate estimates of change for several regional populations with complete survey coverage during both national surveys.

Methods A subset of 10-km squares (Raptor Study Group squares and randomly sampled squares) was surveyed across the breeding distribution of Merlins in the UK using standardised methods devised during the 1993–94 national survey.

Results The population estimate for Merlins in the UK was 1162 breeding pairs, and in Britain was 1128 pairs (95% CI: 849–1427), which although 13% lower, was not significantly different from the British estimate of the 1993–94 survey. Scotland held the bulk (733 pairs) of the UK Merlin population, and smaller numbers of 301 pairs, 94 pairs and 32 pairs were estimated for England, Wales and Northern Ireland, respectively. The population estimate for Wales may have been biased upwards by low coverage in the south of the country. Marked declines were noted in several regional Merlin populations, particularly in areas of northern England.

Conclusions The 2008 Merlin survey suggests that the population in Britain has remained relatively stable since 1993–94, but with local declines, particularly in northern England. Currently, little is known about important drivers of regional population change in Merlins, but changes in land-use, prey populations and climate are likely to be important factors.

Merlins Falco columbarius are small, agile falcons that occur at northern latitudes in Europe, Asia and North America, where they are associated primarily with open country (Cramp & Simmons 1980). In the UK, where the species is at the southern limit of its European distribution, Merlins (subspecies F. c. aesalon) breed in a variety of upland habitats, occurring at particularly high densities in areas of heather moorland (Gibbons et al. 1993). Nests may be placed on the ground in deep vegetation, on rocky crags, or in the stick nests of other birds in trees (e.g. corvids). While some subspecies of Merlins are highly dispersive after breeding, most British birds move only short distances to low-lying inland and coastal regions, although some overwinter in continental Europe (Heavisides 2002). Resident Merlins in the UK are supplemented in winter by members of the Icelandic subspecies, F. c. subaesalon (Heavisides 2002).

Merlins are a species of long-standing conservation concern in Britain. Parslow (1967) reported that the species experienced a period of marked decline during the 1950s and early 1960s. Thereafter, regional population studies conducted during the 1970s to the mid 1980s reported further decreases in many areas, including Wales (Roberts & Green 1983, Bibby 1986), Northumbria and the Peak District (Newton et al. 1981, 1986), and Orkney (Meek 1988). These declines led to Merlins being designated a species of special concern in the British Red Data List (Batten et al. 1990), and set the backdrop for the first national Merlin survey in 1983–84, which returned a population estimate of 550–650 breeding pairs for Britain (Bibby & Natrass 1986). Further regional studies in the late 1980s to early 1990s reported contrasting population trends, with declines, increases and stable populations documented (e.g. Ellis & Okill 1990, Brown & Shepherd 1991, Rebecca et al. 1992).

A second national Merlin survey in 1993–94 estimated the British population as 1291 breeding pairs (95% CI: 1108–1500; Rebecca & Bainbridge 1998), providing evidence that the population had increased since 1983–84. Improvements in the sampling protocol and geographic coverage in 1993–94 may explain some of this apparent increase. However, in all areas where robust comparisons could be made, regional populations had either increased or remained relatively stable, indicating that some increases were genuine (Rebecca & Bainbridge 1998; see also Little et al. 1995, Brown & Stillman 1998). After the 1993–94 survey, Merlins were moved to the Amber list of Birds of Conservation Concern (Gregory et al. 2002), and they remain there following the most recent assessment (Eaton et al. 2009).

Factors driving historical changes in UK Merlin populations are poorly understood, but probably include organo-chlorine pesticide contamination and land-use change. High pesticide burdens were associated with eggshell thinning (Newton 1973, Newton et al. 1982) and reduced reproductive success (Fox 1971, Fyfe et al. 1976). Merlins still demonstrate the highest pesticide residue concentrations of any UK raptor (Shore et al. 2006), but levels have fallen in recent decades, and the population recovery between 1983–84 and 1993–94 may be partly attributable to this (Rebecca & Bainbridge 1998, Newton et al. 1999). Furthermore, in some regions Merlins have been shown to abandon territories following the establishment of commercial forest plantations, which cover their former foraging areas (Rebecca & Cosnette 2003, Rebecca 2006, but see Little & Davison 1992, Parr 1994).

This article presents the results of the third national Merlin survey (the first to include Northern Ireland) undertaken in 2008, as part of the UK SCARABBS (Statutory Conservation Agencies/RSPB Annual Breeding Bird Scheme) monitoring programme. The main objectives of the survey were to estimate the number of breeding pairs of Merlins in the UK, and the four constituent countries of England, Scotland, Wales and Northern Ireland, and to compare these with the results of the 1993–94 Merlin survey.

METHODS

Survey design

We defined the UK breeding range of Merlins as all 10-km national grid squares reported as occupied in the 1968–72 and 1988–91 British and Irish Breeding Bird Atlases (Sharrock 1976, Gibbons et al. 1993), supplemented by records provided by Raptor Study Groups (RSGs) and from the RSPB species database (Fig. 1). The sampling strategy comprised two main components. First, in some regions, RSGs have accumulated an extensive local knowledge of breeding Merlins, and members of these groups aimed to provided complete coverage of self-selected 10-km squares (RSG squares). Second, for the remainder of the range, a subset of 10-km squares was selected randomly for survey by RSPB staff and volunteers (random squares). Following Rebecca & Bainbridge (1998), the random squares were drawn from the grid of key squares devised by the Botanical Society of the British Isles (BSBI; a grid constituted by every third 10-km square along a north–south and east–west axis). This grid was superimposed on the Merlin breeding distribution, and squares were randomly selected from the BSBI subset that fell within the defined range. The counts from the surveyed squares were then used to derive an extrapolated population estimate for areas with no coverage. Slight variations in this general design were implemented in the four UK countries, and a more detailed description of the different protocols follows.

Figure 1. Distribution of 10-km squares occupied by Merlins in the UK between 1968 and 2007. Black squares indicate those surveyed by RSG members, and grey squares indicate those selected randomly for coverage by RSPB staff and volunteers in 2008. The white squares were not covered during the survey.

Scotland

Merlins are widespread in Scotland with a breeding distribution covering 690 10-km squares (62% of UK range). Before the start of the survey, Scottish RSGs predicted that they could provide full coverage of 139 10-km squares. This included a full survey of Orkney and the Uists, and significant coverage of northeast Scotland and Tayside. From the remainder of the distribution a random sample of 46 BSBI squares was selected for coverage by RSPB staff and volunteers.

England

The breeding range of Merlins in England comprised 203 10-km squares (18% of UK range), of which 83 were selected for survey by RSG members, with particular concentrations in Northumbria, the North York Moors, the Peak District and County Durham. From the remainder of the distribution, 11 BSBI squares were randomly selected for coverage by RSPB staff.

Wales

The breeding distribution of Merlins in Wales covered 125 10-km squares (11% of UK range). As there had been relatively little regular coverage of Merlins in Wales by RSG members, the sampling strategy differed slightly to that for Scotland, England and Northern Ireland, with the Welsh range divided into two sampling strata. Stratum one consisted of 42 squares that were considered important for Merlins, as they either (1) encompassed Special Protection Areas (SPAs) for which Merlins are a qualifying feature; (2) were known to hold relatively high densities of Merlins (i.e. squares with two or more breeding pairs in any year between 1993 and 2006); or (3) had regular occupancy (i.e. squares occupied during 3 or more years between 1993 and 2006). Eighteen of these were randomly selected for coverage. Stratum two comprised the remaining 83 squares from within the known range, of which a further 12 were randomly selected for coverage. Both strata were surveyed by RSPB staff and volunteers.

Northern Ireland

The breeding distribution of Merlins in Northern Ireland covered 98 10-km squares (9% of UK range). The Northern Ireland RSG selected 21 squares to be surveyed by its members, and a further 10 BSBI squares were randomly selected for coverage also by RSG members.

Fieldwork methods and recording

RSPB fieldworkers followed generic guidelines, as in the 1993–94 survey (Rebecca & Bainbridge 1998). Random squares were surveyed in April, May, June and July. During each month, fieldworkers passed within about 500 m of all suitable Merlin breeding habitat, defined as grass and heather moorland, young plantation, the edges and rides of mature plantation, and open woodlands. Urban and suburban areas, arable farmland, enclosed pastures and land above 600 m were regarded as unsuitable for breeding Merlins and excluded. Areas of grass moorland found to have no potential Merlin nest sites (i.e. absence of deep heather, trees, crags, or stream banks) during early visits were also discounted from further searching.

RSPB Fieldworkers were instructed to search squares on foot, and also watch from good vantage points to check for Merlins. They were asked to be vigilant for displaying Merlins, and to focus on other raptors, corvids and herons, as these larger species can provoke a territorial response from Merlins. Prominent perches, such as large rocks, fence lines, hummocks, isolated trees, and stone walls/dykes, were identified and searched for prey remains, pellets, whitewash and moulted Merlin feathers. Particular attention was also paid to potential nest sites for Merlins, including crags, steep stream banks, and deep heather banks. In areas with plantation forest, all forest edges, including rides up to 100 m from the forest boundary, were walked in search of signs of Merlins.

In contrast to the methods described above, RSG fieldworkers used their ‘traditional’ survey methods, which focused largely on assessing occupancy at previously documented breeding sites. Rebecca & Bainbridge (1998) compared the efficacy of RSG field methods with the more prescribed approach used by RSPB fieldworkers during the 1993–94 national survey, finding no difference in the number of pairs found. We assumed in this survey that the efficacy of both methods would continue to be equivalent, thereby avoiding the need for further time-consuming validation work.

Where Merlins were seen or signs found, fieldworkers provided a six-figure grid reference, and information on the number, sex and behaviour of birds. At the end of the survey period, an annotated 1 : 25 000 Ordnance Survey map was completed for each 10-km square detailing the location of Merlin sightings and signs, the extent of survey effort, and areas of the square that were unsuitable or from which fieldworkers were denied access.

Calculation of population estimates and population change

Merlin records during the national survey were categorised either as an occupied territory or a breeding pair according to the criteria outlined in Table 1. These classifications are equivalent to those used in the 1993–94 survey. Country and national population estimates were derived by adding the counts from areas of complete coverage to the extrapolated estimates from the remainder of the breeding distribution. Extrapolated estimates were calculated as (n ₁/p ₁) (or [(n ₁/p ₁) + (n2/p2)] if estimates required to be combined across two strata), where n _i was the total count from stratum i, and p _i was the proportion of squares in stratum i that were surveyed. Note that the method used to calculate individual square totals during 2008 differed slightly to that used in RSG areas in 1993–94 due to the contrasting length of the fieldwork period. Counts in 2008 were derived by summing the number of observations that fulfilled the different categories of breeding evidence (i.e. occupied territory, breeding pairs), whereas in RSG areas in 1993–94 square totals were calculated by combining individual year counts (after assessing for alternate breeding sites within a territory) to give the maximum observed number of pairs over the survey period (Rebecca & Bainbridge 1998).

Table 1. Classification criteria for Merlin survey records.

Classification	Criteria
Breeding pair	Direct or indirect evidence of a breeding attempt. - nest with eggs or young found. - recently fledged young observed or heard. - used nest, eggshell fragments found. - adult leaving or returning to a nest site. A pair of Merlin engaging in courtship behaviour (copulation, food delivery, nest-site prospecting), or at least one Merlin showing territorial behaviour (repeatedly alarm calling, mobbing potential predator). A pair of Merlin seen during the breeding season at a potential breeding site. At least one Merlin holding territory (bird seen or heard at the same location on at least two visits separated by more than 1 week). Presence of accumulated signs (pellets, prey remains, whitewash, moulted feathers) indicating occurrence of breeding attempt.
Occupied territory	A single bird seen during the breeding season in apparently suitable habitat. Signs of Merlin occupation (but no birds sighted) at a location on two or more visits.

Confidence intervals for the extrapolated estimate were generated by bootstrapping, implemented in SAS 9.2 using the jackboot macro (Sarle 1997). This macro randomly selected sampled squares, with replacement, until a new data set was created with a sample size equal to that of the original data set. This procedure was then repeated 10 000 times, resulting in a sampling distribution for the population estimate, and 95% confidence intervals calculated as the 2.5 and 97.5 percentiles of the bootstrap distribution (Manly 2006).

We examined country and national population changes by calculating the percentage change in the extrapolated estimates between 1993–94 and 2008. The statistical significance of these changes was assessed using bootstrap tests (Manly 2006). Specifically, the observed difference D in the population estimates between the two surveys was compared against a bootstrap distribution of differences (d ₁ … d ₉₉₉₉), which was derived by generating 9999 bootstrap population estimates for each survey, and subtracting the corresponding estimates. The population estimates from the two surveys were considered to be significantly different if fewer than 5% of the d values had an absolute value of greater than D (i.e. P < 0.05; Manly 2006).

Estimates of population change were also derived for seven regional populations of Merlins (outlined in Fig. 2) that were surveyed comprehensively in both 1993–94 and 2008 (largely by RSG members). These regions do not follow particular political or administrative boundaries, but were selected on the basis that they represent geographically discrete, relatively contiguous areas with full coverage during both national surveys. For each region, an estimate of change was calculated by comparing the total number of pairs in fully covered squares. Two regions, the North York Moors and South Pennine Moors, are SPAs with Merlin as a qualifying feature. As both SPAs received almost complete coverage in 2008 (>80% coverage), the number of breeding pairs in each SPA was compared with the expected SPA site total.

Figure 2. Populations of Merlins in regions with high levels of survey coverage in 1993–94 and 2008. The grey squares indicate the 10-km squares that received full coverage in both national surveys.

Finally, where random and RSG squares were targeted to be fully surveyed, but received no or only partial coverage, they were included in the group of unsurveyed squares for which an extrapolated population estimate was calculated.

RESULTS

Survey coverage and observed numbers of Merlins

In 2008, fieldworkers surveyed 211 RSG and 82 random 10-km squares, approximately one-quarter (26.8%) of the species' UK range (Fig. 1). In England, realised coverage largely matched what had been targeted prior to the beginning of the survey (Table 2). Coverage in Scotland and Northern Ireland was less complete, with 14% and 29%, respectively, of targeted 10-km squares receiving only partial or no coverage. In Wales, three (17%) random squares in stratum one and three (25%) in stratum two were unsurveyed. There was some bias in the Welsh squares that received no coverage, resulting in southern parts of the Merlin's range being under-represented (Fig. 1).

Table 2. Number of Raptor Study Group (RSG) and random 10-km squares covered and counts of occupied territories and breeding pairs in each country during the 2008 Merlin survey. Note that no RSG squares were surveyed in Wales in 2008.

Country	Number of RSG squares	No. of breeding pairs (and occupied territories) in RSG squares	No. of random squares	No. of breeding pairs (and occupied territories) in random squares
Scotland	118	161 (180)	41	41 (55)
England	78	101 (122)	10	16 (16)
Wales – Stratum 1	–	–	15	17 (21)
Wales – Stratum 2	–	–	9	5 (7)
Northern Ireland	15	10 (14)	7	2 (3)
UK	211	272 (316)	82	81 (102)

Surveyors observed 316 occupied territories in RSG squares (mean = 1.50 square⁻¹) and 102 territories in random squares (mean = 1.24 square⁻¹). Fewer breeding pairs were documented than occupied territories, with 272 breeding pairs of Merlins in RSG 10-km squares (mean = 1.29 square⁻¹) and 81 pairs in random squares (mean = 0.99 square⁻¹) (Table 2). The majority of breeding pairs were found in Scotland (202; 57%) and England (117; 33%), with fewer recorded in Wales (22; 6%) and Northern Ireland (12; 3%).

Merlin population estimates and recent changes

The UK Merlin population was estimated at 1162 breeding pairs (95% CI: 891–1462) in 2008, with 1128 pairs (95% CI: 849–1427) in Britain (Table 3). While no comparable estimate is available for the UK from previous national surveys (surveys were not undertaken in Northern Ireland), Rebecca & Bainbridge (1998) derived a population estimate of 1291 pairs (95% CI: 1108–1500) in Britain in 1993–94. The 2008 estimate is 13% lower than that from 1993–94, but this difference was not statistically significant.

Table 3. National population estimates for Merlins derived from the 1993–94 and 2008 surveys, and percentage change across the two surveys. Numbers in parentheses represent the 95% confidence intervals of the estimates. Where estimates are not accompanied by confidence intervals, a complete survey was undertaken. In the ‘Change in population estimate (%)’ column, (n.s.) indicates that a bootstrap test showed the 1993–94 and 2008 did not differ significantly from one another at P = 0.05 (see methods).

Country	Occupied territories in 2008	Breeding pairs in 2008	Breeding pairs in 1993-94	Change in population estimate (%)
Scotland	947 (688–1238)	733 (512–979)	785 (602–994)	–7 (n.s.)
England	322 (173–518)	301 (151–496)	401	–25 (n.s.)
Wales	123 (76–178)	94 (62–130)	81	16 (n.s.)
Northern Ireland	49 (27–81)	34 (12–66)	–	–
Britain	1392 (1081–1720)	1128 (849–1427)	1291 (1108–1500)^a	–13 (n.s.)
UK	1441 (1116–1823)	1162 (891–1462)	–	–
^aThis total includes a minimum of 24 pairs estimated for areas where access was denied to fieldworkers (see Rebecca & Bainbridge 1998).

The Merlin population estimate for Scotland in 2008 was 733 breeding pairs, constituting 63% of the total UK population (Table 3). Overall, there was no clear change in the Scottish population between 1993–94 and 2008, with only a small, non-significant decrease (–7%) recorded across the two surveys. An estimate of 301 breeding pairs was calculated for England, while in Wales a population of 94 pairs was estimated, representing a 25% decline and 16% increase, respectively, in the English and Welsh Merlin population estimates since 1993–94. However, neither of these population changes was significant. The population estimate for Merlins in Northern Ireland was 32 pairs.

While there was no overwhelming evidence of important changes in Merlin populations between 1993–94 and 2008 at a national scale, counts in areas with high levels of survey coverage suggested that some regional populations, particularly in England, may have experienced relatively marked declines (Table 4). The most dramatic was observed in southwest England, where the population dropped from six breeding pairs in 1993–94 to only one in 2008 (–83%), but marked regional decreases were also documented in Northumbria (–69%), the North York Moors (–47%), the South Pennine Moors (–47%), Wales (–38%; but also see the change in national estimates in Table 3) and Northeast Scotland and east Tayside (–27%). The only regional population that remained stable or increased was on Orkney, which increased by a single pair. Numbers of Merlins in the two well surveyed SPAs (North York Moors and South Pennine Moors) were well below expected site totals.

Table 4. Counts and estimates of population change for seven regional Merlin populations with high levels of survey coverage during the 1993–94 and 2008 national surveys. See Fig. 2 for information on the location of the regional populations. An asterisk (*) indicates the region is a Special Protection Area (SPA) with Merlin as a qualifying feature. Figures in parentheses in the ‘Breeding pairs in 2008’ column represent the expected site total for each SPA.

Regional population	Number of squares	Breeding pairs in 1993–94	Breeding pairs in 2008	% change in pairs
Orkney	13	16	17	6
Northeast Scotland	42	90	66	–27
Northumbria	17	39	12	–69
North York Moors*	13	36	19 (c. 35)	–47
South Pennine Moors*	14	55	29 (c. 58)	–47
Wales	17	29	18	–38
Southwest England	3	6	1	–83

DISCUSSION

Population estimates – precision, bias and comparability

The 2008 Merlin survey calculated a population estimate of 1162 breeding pairs in the UK (95% CI: 891–1462), and 1128 breeding pairs in Britain (95% CI: 849–1427). Comparing the population estimates from 2008 with those of the 1993–94 Merlin survey, updated estimates for Britain, Scotland and England were 13%, 7% and 25% lower, respectively. However, while these findings hint at modest declines, no statistically significant population change was detected between 1993–94 and 2008.

Modest populations changes are unlikely to be detected with high statistical confidence by national surveys of relatively widespread species which rely on sample-based survey designs. However, our ability to discern population change in Merlins was also hampered by the imprecision of the population estimates, indicated by the breadth of the confidence intervals around the estimates (Table 3). For example, the confidence intervals around the English Merlin population estimate indicate that population changes ranging from a 62% decline to a 23% increase between 1993–94 and 2008 were compatible with observed data.

According to the estimate for Wales, Merlins increased by 16% between 1993–94 and 2008, but this is at odds with the estimate of population change (i.e. –38%) from counts in areas with high levels of survey coverage. This disagreement probably reflects geographical bias in the completion of selected survey squares, with unsurveyed squares mostly in the southern portion of the Merlin's distribution, where independent data sets suggest that Merlin densities are lower than further north (RSPB unpublished data). It is thus likely that the Welsh Merlin population estimate is inflated, and that the true estimate probably lies at the lower end of the calculated confidence interval. Nonetheless, the bias is likely to be relatively small (about 10–20 pairs), and does not undermine the UK or British population estimates.

Methods used to calculate counts of Merlins in individual 10-km squares in RSG areas differed slightly between the 1993–94 and 2008 national surveys as a necessary consequence of the contrasting durations of the two survey periods. In particular, the approach adopted by Rebecca & Bainbridge (1998) of pooling counts across two years to give maximum observed numbers in a square risks overestimating the number of Merlins present in any one year if individual pairs occupy alternate sites in successive years, and these are treated separately. However, Rebecca & Bainbridge (1998) assessed for potential alternate sites, and where there were concerns over the possibility of double-counting the lower figure was used. Therefore, while counts in some RSG squares from the 1993–94 survey may be slightly inflated (if pairs moved further afield in successive years) compared to the situation where the survey had been undertaken in a single year, the extent of any bias is again likely to be small, and should not affect the general patterns of observed population changes.

Regional estimates of population change and potential drivers

This survey detected population change in several regions with complete survey coverage. (Note that as there was no sampling component to the surveys of these areas, we assume that there are no restrictions on our ability to discriminate population change.) Important Merlin populations in Northumbria, the North York Moors and South Pennine Moors declined by 69%, 47% and 47%, respectively, between 1993–94 and 2008. The latter two sites are SPAs with Merlin as a qualifying feature, and the decline means that current numbers of breeding Merlins are well below those present at the time of SPA designation (Table 4), suggesting that both should be regarded as being in unfavourable condition. Furthermore, well-surveyed populations in southwest England and northeast Scotland and east Tayside also declined by 83% and 27%, respectively.

Drivers of regional population change between the two national surveys are poorly understood. One mechanism that could partially underlie some of the declines is change in breeding season prey availability. Merlins feed largely on abundant moorland passerines during the breeding season, including Meadow Pipits Anthus pratensis, Eurasian Sky Larks Alauda arvensis and Northern Wheatears Oenanthe oenanthe (e.g. Newton et al. 1984, Meek 1988, Ellis & Okill 1990, Rebecca 2006). Populations of these species have decreased markedly in recent decades, most notably in parts of Northern England (e.g. North Yorkshire – Meadow Pipits: –47%, Eurasian Sky Larks: –73%; South Pennines – Meadow Pipits: –29%, Northern Wheatears: –46%; Sim et al. 2005). Wright (2005) speculated that a reduction in the abundance of Meadow Pipits had contributed to the recent decline (–50% between 1996 and 2002) of Merlins in his Yorkshire Dales study area, and a reduction in average clutch size at one site. Moreover, reductions of bird prey on low-lying farmland and coastal wintering areas may also have reduced the over winter survival of Merlins.

Several regional populations where declines have been noted breed largely on moorland managed for driven Willow Ptarmigan (Red Grouse) Lagopus lagopus scotica shooting (e.g. South Pennine Moors, North York Moors, Northeast Scotland and east Tayside), and changes in habitat condition mediated by grouse moor management may have contributed directly to declines of Merlins. In particular, the intensity of heather burning has increased in some of these areas; for example, Yallop et al. (2006) showed that the extent of new burns has nearly doubled in parts of northern England since 1970 (but especially during the 1990s), with a significant reduction in the median period for rotational burning. Rotational burning is an important management technique that provides the mixed mosaic of young and old heather stands preferred by Red Grouse (Sotherton et al. 2009), but widespread increases in the frequency of burning could reduce the availability of deep heather banks which provide preferred nest-sites for Merlins, potentially leading to displacement and local population declines (Rebecca & Cosnette 2003). Anecdotal evidence also indicates that targeted burning of deep heather is used to discourage the establishment of breeding Hen Harriers Circus cyaneus on grouse moors, and this may inadvertently reduce nest-site availability for Merlins as well. Intensive burning could have further effects on Merlins, via reductions in the abundance of key songbird prey, particularly Meadow Pipits (Smith et al. 2001, Pearce-Higgins & Grant 2006).

On grouse moors, some raptors (e.g. Golden Eagles Aquila chrysaetos and Hen Harriers) are heavily persecuted as predators of Red Grouse, such that their populations are limited by illegal killing in these areas (Etheridge et al. 1997, Whitfield et al. 2008, Anderson et al. 2009, Redpath et al. 2010). While confirmed cases of persecution of Merlins are rare (RSPB 2005, 2010), anecdotal evidence indicates that it does occur, although it is unclear what the population implications of illegal killing are in regional or national terms.

Regional population declines were not limited to areas dominated by grouse moors. Merlins in Northumbria breed extensively in old crow nests in plantation forests (e.g. Kielder Forest, Little & Davison 1992), yet this population decreased by 69% between 1993–94 and 2008. The decline of another small forest-breeding raptor, the Common Kestrel Falco tinnunculus, in this area has been attributed to an increased incidence of predation by Northern Goshawks Accipiter gentilis (Petty et al. 2003). In contrast, Merlins constituted a far smaller proportion of the diet of Northern Goshawks than Common Kestrels, and it is unlikely that intra-guild predation has played an important role in Merlin population changes. Alternatively, plantation forests in Northumbria may have become less suitable for Merlins, if only certain growth stages of plantation are preferred as breeding habitat, and much of the forest stock is similarly aged. Finally, the Merlin population in southwest England decreased from seven pairs in 1993–94 to one in 2008. Few successful breeding attempts have been documented in the Exmoor population in recent breeding seasons (RSPB unpubl. data), and a lack of local recruitment may have led to the population being unable to sustain itself.

Merlins occur at the southern-most edge of their European breeding distribution in the UK (Cramp & Simmons 1980). Bioclimatic models predict that the Merlins' European breeding distribution will shift substantially northwards in the face of climate change, with the species' current climate envelope retreating from much of the currently occupied UK range south of the Scottish Highlands (Huntley et al. 2007). It is notable that the survey estimates of change are more negative for regional populations at southern latitudes (e.g. Southwest England, Northumbria, North York and South Pennine Moors) than those of northern populations (e.g. Orkney, Northeast Scotland and east Tayside). This pattern is consistent with climate change drivers impacting regional Merlin population dynamics, but further research is needed to identify whether this is indeed the case, and through which mechanism(s) the species may be affected.

CONCLUSIONS AND RECOMMENDATIONS FOR FUTURE MERLIN SURVEYS

The 2008 Merlin survey suggests that the British population has remained largely stable since 1993–1994, but with some local declines, particularly in England. However, the national survey is not the only source of trend information for Merlins in the UK, with several long-running regional population studies at sites spread across the species distribution. To complement and build on the population estimates and changes identified by the national survey, we would encourage Raptors Study Groups to publish the updated population trends of their regional surveys. This would provide an assessment of the long-term patterns underpinning population change, rather than simply the snapshot changes between two years afforded by the national survey.

Future national surveys would benefit from improved statistical power. The precision of estimates, and the power to detect population change, depends largely on the sample size of random squares, and is less influenced by the number of RSG survey squares. We suggest that greater survey effort should be allocated to the random component of the sampling design, perhaps even to the extent that the next national Merlin survey is a fully randomised survey. This would require more funds to be made available through the SCARABBS programme for additional field staff to survey the larger cohort of randomly selected squares, and/or for RSG members to cover random squares within their study areas. In addition, future surveys would benefit from using a refined distribution map, which will be available following the publication of the updated UK breeding bird atlas, based upon field surveys covering 2007–2011.

Finally, our current understanding of the causes of Merlin population change is poor. We suggest that an integrated and comprehensive analysis of the combined regional data sets may offer a means by which to investigate the key demographic parameters or ecological drivers underlying population change, in much the same way as has been done recently for Golden Eagles (Whitfield et al. 2008). Furthermore, new field research may be needed to elucidate the role of potentially important drivers of decline, such as the effects of changing heather burning regimes, other land-use changes and climate change.

ACKNOWLEDGEMENTS

The 2008 Merlin survey was funded by RSPB, Scottish National Heritage, Natural England, Countryside Council for Wales, and Northern Ireland Environment Agency as part of the SCARABBS monitoring programme. We thank the large number (too many to mention individually) of RSG, ringing and upland bird study group members and RSPB staff and volunteers who contributed to the survey. In particular, we thank Eric Meek and Jim Williams (Orkney RSG), Brian Cosnette (Northeast RSG), Wendy Mattingley and Ron Downing (Tayside RSG), Bob Stakim (South Strathclyde RSG), Chris Rollie (Dumfries and Galloway RSG), Martin Davison (Northumbria Ringing Group), David Raw (Durham Upland Bird Study Group), Colin Dilcock (North York Moors RSG), Steve Davies (Peak District RSG), and Mick Taylor (South Peak RSG) for helping to collect and collate regional records. We also thank Paul Haworth and Robin Reid for providing information on Uist Merlins. The efforts of the RSPB fieldworkers Trevor Smith, Ian MacPherson, Ben Hayes, Shelley Barbour, Jon Brain, Geoff Wallace, Richard Storton, Tom Hall and Richard Mearns are appreciated. Jerry Wilson, Richard Gregory, and Megan Davies provided valuable comments on previous drafts of this article. Finally, thanks are due to the many landowners and their agents who cooperated with our access requests.