Increasing urbanisation, and the resulting habitat loss, is a major driver of the current biodiversity crisis (Czech et al. Reference Czech, Krausman and Devers2000, Mcdonald et al. Reference Mcdonald, Kareiva and Forman2008, Laurance Reference Laurance, Sodhi and Ehrlich2010). However, human settlements also provide suitable habitat for animals. For example, many synanthropic bird and mammal species find shelters and breeding sites in or on buildings. Whether these can serve as an important resource for animals depends on certain characteristics of the buildings, such as the availability of accessible cavities. In the course of building renovations and demolitions, such structures often vanish. This phenomenon is readily observable in Europe, where modernisation and thermal insulation of buildings is currently happening at a high rate in order to improve energy efficiency (BMWi 2011, EU 2012). This development has the potential to lead to a strong decline of building-dependent species if it is not accompanied by appropriate mitigation measures.
In Germany, all naturally occurring bird species are protected under the Federal Nature Conservation Act, which implies a general prohibition on destroying their breeding sites, even outside the breeding season (Deutscher Bundestag 2009). If it is impossible to maintain such sites during building renovations or demolitions, the responsible authorities usually mandate compensatory measures. Similar regulations exist in Poland, for example (Luniak and Grzeniewski Reference Luniak and Grzeniewski2011). However, studies on the efficacy of compensatory measures such as the installation of nest-boxes remain sparse. Here, we present a study on the occupancy of nest-boxes for the Common Swift Apus apus (hereafter Swift). The boxes had been mounted to mitigate the loss of nesting sites during the renovation or demolition of buildings in Greifswald, north-eastern Germany. As in many other cities of the former Eastern Bloc, a large number of prefabricated buildings have been renovated in Greifswald during the last 20 years, resulting in a substantial loss of cavities suitable as breeding sites for Swifts.
Swifts are aerial insectivores that inhabit a large Palearctic breeding range, covering the Mediterranean, temperate, and boreal zones (Cramp Reference Cramp1985). The wintering grounds are mainly located in central and southern Africa (Cramp Reference Cramp1985, Åkesson et al. Reference Åkesson, Klaassen, Holmgren, Fox and Hedenström2012). In most parts of the range, nesting sites in buildings nowadays outnumber the original nesting sites in rock crevices and tree cavities (Cramp Reference Cramp1985), which is partly due to the rarity of the latter structures in managed forests (Günther et al. Reference Günther, Hellmann and Nicolai2004). In buildings, Swifts find suitable cavities under the eaves, in gaps in walls, under roof tiles (Weitnauer and Scherner Reference Weitnauer, Scherner, Glutz von Blotzheim and Bauer1980), and in the horizontal interstices between the concrete slabs of prefabricated buildings (Lehmann Reference Lehmann2005).
Changes in the construction of buildings are considered to be a major threat for Swifts (Gory Reference Gory, Hagemeijer and Blair1997, Bauer et al. Reference Bauer, Bezzel and Fiedler2005, Crowe et al. Reference Crowe, Coombes, Lysaght, O'Brian, Choudhury, Walsh, Wilson and O'Halloran2010), and it is known that renovations can cause considerable local declines in Swift populations (Braun Reference Braun1999). However, data on population trends at a larger scale are ambiguous. In Germany, Südbeck et al. (Reference Südbeck, Bauer, Boschert, Boye and Knief2007) judged the Swift population to be stable and not threatened, whereas Sudfeldt et al. (Reference Sudfeldt, Bairlein, Dröschmeister, König, Langgemach and Wahl2012) estimated a decline of 1–3% per year between 1991 and 2010. Population declines were reported from other European countries (BirdLife International 2004, Eaton et al. Reference Eaton, Brown, Noble, Musgrove, Hearn, Aebischer, Gibbons, Evans and Gregory2009, Crowe et al. Reference Crowe, Coombes, Lysaght, O'Brian, Choudhury, Walsh, Wilson and O'Halloran2010), but the overall European population was considered to be stable between 1980 and 2010 (PECBMS 2012).
Nest-boxes are widely used in many European countries to support Swift populations, e.g. in Germany (Kaiser Reference Kaiser1993, Wortha and Arndt Reference Wortha and Arndt2004, Arens Reference Arens2011), Poland (Luniak and Grzeniewski Reference Luniak and Grzeniewski2011), Switzerland (Weitnauer Reference Weitnauer1980) and the UK (Lack and Lack Reference Lack and Lack1951, Wilson Reference Wilson2011). Recommendations on how to install nest-boxes for Swifts can be found in various brochures for building owners and architects. These agree that Swifts require an unobstructed flight path in front of the boxes (Kaiser Reference Kaiser1993, Du Feu Reference Du Feu2003, Scholl Reference Scholl2004). Moreover, nest-boxes should have a certain minimum size and should be placed at a minimum height above ground level (5–7m), with little exposure to direct sunlight (Kaiser Reference Kaiser1993, Scholl Reference Scholl2004, Mayer Reference Mayer2013). As Swifts are gregarious (Cramp Reference Cramp1985), it is generally recommended to install several boxes close to each other (Kaiser Reference Kaiser1993, Du Feu Reference Du Feu2003). It is also possible to attract Swifts to new potential breeding sites by playing Swift calls. This can considerably shorten the time Swifts need to detect the new sites, especially when existing colonies are relatively far away (Arens Reference Arens2004, Tigges and Mayer Reference Tigges and Mayer2011). Furthermore, because of the Swift’s pronounced nest-site fidelity (Cramp Reference Cramp1985), it is suggested to carry out renovation work outside the breeding season and either to preserve the original cavities or to install compensatory nest-boxes in exactly the same places (Scholl Reference Scholl2004).
Anecdotal reports of nest-boxes occupied by Swifts are not rare (e.g. Lack and Lack Reference Lack and Lack1951, Weitnauer Reference Weitnauer1980, Kaiser Reference Kaiser1993, Arens Reference Arens2011), but to the best of our knowledge, there is only one published and detailed empirical study on how frequently nest-boxes are used by Swifts. Wortha and Arndt (Reference Wortha and Arndt2004) monitored c.1,200 compensatory nest-boxes for Swifts in Berlin, Germany, and found the overall proportion of occupied boxes to be only about 10%. Moreover, at 15 out of 16 buildings, fewer Swifts were breeding in the boxes than in the original cavities before the renovation work. However, this study lacked both a clearly defined survey method and a multi-factorial statistical analysis showing the influence of nest-box properties on box occupancy.
Against the background of indications of a population decline in Swifts, ongoing renovations of buildings in many countries, and uncertainty regarding the efficacy of mitigation practices, we aimed to answer the following questions, using detailed field observations and multi-factorial statistics: 1) what proportion of compensatory nest-boxes are occupied by Swifts? 2) How do the numbers of occupied boxes relate to the numbers of breeding sites before the renovations? 3) Which properties of the boxes influence the occupation probability?
Our study was conducted in Greifswald, Mecklenburg-Western Pomerania, Germany (54°05'N, 13°24'E), a city with approximately 56,000 inhabitants (Statistisches Amt Mecklenburg-Vorpommern 2013). In spring 2013, there were nearly 800 nest-boxes for Swifts on facades of buildings throughout the city. Out of those, we selected 477 boxes installed on 23 buildings. Boxes were selected in order to ensure a sufficient sample size for several different box properties (see below). Twenty-two of the studied houses were prefabricated apartment buildings. The number of storeys ranged between two and six. All boxes had been mounted between 2004 and 2012. No attraction calls were used on any of the sites. Most of the buildings had been renovated and equipped with a thermal insulation shell (Figure 1), but at three buildings, the boxes had been installed as compensation for the demolition of another building nearby. Contrary to the general advice, all buildings except one had been renovated during the Swift breeding season. The time of year when renovation is performed is supposed to have a large effect on occupation probability: if it takes place during the Swifts’ stay in Africa, the birds which used to breed on the building will probably return afterwards and readily occupy the compensatory boxes. If renovation is carried out during the breeding season, the Swifts that find their former breeding sites inaccessible will look out for nest sites elsewhere and will normally not return the year after renovation (Scholl Reference Scholl2004). Consequently, the building needs to be rediscovered as a breeding site by different individuals. To ensure comparability, the four nest-boxes at the single building that had been renovated outside the breeding season were excluded from the analysis concerning the box properties.
It was possible to obtain data on the number of Swift breeding sites prior to renovation for 10 buildings or sections of buildings. These data were taken from expert assessments that had been conducted to determine the required amount of compensatory measures. When counts were available from two different years prior to renovation, we calculated the arithmetic mean. On all those buildings, more nest-boxes were installed than breeding sites had been detected prior to renovation (on average 6.7:1; Table 1). Details on the survey method were not available in the expert assessments.
To determine the occupancy of the boxes, all were visited once between 25 June and 9 July 2013. Boxes that were not found to be occupied by Swifts or other bird species during this first survey were monitored for a second time between 10 and 24 July 2013. These two periods correspond to the nestling period (Lack and Lack Reference Lack and Lack1951, Weitnauer and Scherner Reference Weitnauer, Scherner, Glutz von Blotzheim and Bauer1980), when Swifts visit their nests most frequently (Arens Reference Arens2011). The observations were carried out from 09h00 to 15h30 (CEST) and from 20h00 to 45 minutes after sunset to cover the times with the highest feeding frequency (Lack and Owen Reference Lack and Owen1955, Arens Reference Arens2011). We performed the observations only during good weather (air temperature > 12°C, average wind speed ≤ 30 km/h, no heavy or persistent rain). Each box was observed for 1 h per observation, but when average wind speed was more than 20 km/h, the observations lasted 1.5 h to correct for a presumed lower feeding activity in these conditions. In most cases, it was possible to observe several boxes at the same time. The total observation time was 270.5 h. We defined a box to be occupied if at least once a Swift was observed entering or leaving it or if an individual (nestling or adult) was visible at the entrance. The same criteria were used for passerine species occupying the boxes.
For Ostseeviertel Parkseite, a 20 ha large section of the district Ostseeviertel (see below), it was possible to calculate a minimum population density of Swifts. Nearly all buildings in this area had been renovated since 2004, and therefore it was very unlikely that breeding sites outside the nest-boxes existed. We surveyed about 75% of all boxes in that area.
We included eight different box properties as predictor variables in our analysis: box type, number of neighbouring boxes, absolute height, height relative to the roof edge, manner of installation, orientation, box age, and city district. Our sample included seven box types produced by two manufacturing companies (Schwegler, Schorndorf, Baden-Württemberg, Germany and Strobel, Schmölln, Thuringia, Germany): Schwegler 17, 17A, 17B, 17C, 1MF and Strobel 416 and 430 (Figure 2). Because of their similarity, the different variants of the Schwegler 17 group were pooled for the statistical analysis. None of the Strobel 416 boxes had been installed in the way intended by the manufacturer. The designated bottom side with the entrance hole had been turned to the front. We differentiated between Strobel 416 boxes with the entrance hole at the top (416t) and at the bottom (416b) of the front side of the box (Figure 2d, 2e). For further information on the box types, see Table S1 in the online Supplementary Materials.
Boxes were treated as a “group” if the distances between the entrances of adjacent boxes were less than 1 m. We counted the number of boxes with which a given box was associated in a group (“number of neighbours”). The distances between the entrances were measured on photographs, where the known exterior dimensions of the boxes served as reference lengths. We grouped the boxes into the classes 0, 1, 2, 3-5, and 8-14 neighbouring boxes to ensure a sufficient sample size in each category.
The absolute height (height above ground level) and the distance to the roof edge (“relative height”) were determined using a laser rangefinder (Bosch GLM 80). The absolute height ranged between 5.15 and 18.15 m. Here, we chose the classes 5-7.9, 8-10.9, 11-13.9, and 14-19 m, again to ensure sufficient sample sizes in the categories. The relative height was expressed in two categories, “at roof edge” and “not at roof edge”, with a threshold distance to the roof edge of 2 m.
Concerning the manner of installation, we distinguished between external and internal boxes, the latter being integrated in the thermal insulation. Box orientation was subdivided into four classes based on the cardinal orientation of the facade. We calculated the age of the boxes as the number of possible breeding periods (May to August) for Swifts from box installation up to (but not including) the 2013 breeding season. Finally, we considered the city district as a predictor variable with three categories: City centre, Ostseeviertel and Schönwalde. The former comprises a mixture of prefabricated and older houses and is characterised by a higher density of buildings than the two other districts, which almost exclusively consist of large prefabricated apartment complexes.
For the statistical analysis of the box properties’ influence on occupation probability, we used boosted regression trees (BRT; Friedman Reference Friedman2001, Elith et al. Reference Elith, Leathwick and Hastie2008). This technique combines several advantages compared to parametric regression models: it can fit complex non-linear relationships, interaction effects between the predictor variables are automatically considered, and correlations between predictors, as well as missing data and outliers, do not pose numerical problems. Additionally, the resulting models show a higher predictive performance than conventional modelling methods, such as generalised linear models (Friedman and Meulman Reference Friedman and Meulman2003, Elith et al. Reference Elith, Leathwick and Hastie2008). Instead of delivering p-values, BRT use internal validation processes, in our case 10-fold cross-validation. The predictive performance of BRT models is described by the proportion of explained deviance, i.e. the variability explained by the model, and the cross-validated AUC of the ROC (area under curve of the receiver operation characteristic) that quantifies, applied to our case, how well the model is able to discriminate between occupied and non-occupied nest-boxes. Three parameters can be adjusted within such models: 1) tree complexity, defining the order of interactions considered in the model, 2) learning rate, controlling the contribution of each regression tree to the final model, and 3) bag fraction, determining the proportion of data which is held back at each step. Tree complexity was set to 5, and subsequently, learning rate was adjusted to 0.001 so that at least 1,000 trees would be drawn (following Elith et al. Reference Elith, Leathwick and Hastie2008). We varied the bag fraction in order to maximise predictive performance; for the final analysis, a bag fraction of 0.7 was used. As box occupancy constitutes a binary response variable, a Bernoulli distribution was fixed in the model settings. We used the Software R (version 3.0.1; R Core Team 2013), with the package gbm (version 2.1; Ridgeway Reference Ridgeway2013). A spreadsheet with our raw data can be found in Table S2.
In total, 116 boxes (24.3%) were found to be occupied by Swifts, 102 (21.4%) by House Sparrows Passer domesticus, and one each (0.2%) by Black Redstarts Phoenicurus ochruros and Great Tits Parus major. We found loam traces at the entrances of 20 boxes (4.2%), indicating previous use by Northern House-martins Delichon urbicum (Figure 2d). In Ostseeviertel Parkseite, we found 68 nest-boxes that were occupied by Swifts. Equating an occupied box with a breeding pair leads to a minimum density of 34 breeding pairs per 10 ha in that district.
For eight out of 10 buildings with data available for comparison, the number of occupied nest-boxes in 2013 was greater than, or equal to, the number of breeding sites that had been detected in the original cavities prior to renovation (Table 1). In two cases, the number of breeding sites had declined. In total, the number of breeding sites on those 10 buildings increased from 29.5 to at least 34.
Influence of box properties on the occupation probability
The BRT model explained 27.3% of the total deviance with a cross-validated AUC of 0.765 (Table 2). The number of neighbouring boxes proved to have the highest relative influence on occupation probability (30.5%). Boxes without neighbouring boxes within 1 m had the highest probability of being occupied by Swifts, and the occupation probability decreased with increasing number of neighbours (Figure 3a). None of the 60 boxes with at least eight neighbours was occupied (Table 3f). The second most influential predictor variable was box age (24.4%). Older boxes tended to have higher occupation probabilities than younger ones (Figure 3b). Further important predictor variables were orientation (16.3%) and city district (10.1%). The Swifts slightly preferred northward facing boxes (Figure 3c) and the boxes in Greifswald city centre had a markedly higher occupation probability than those in the two other districts (Figure 3d). The remaining box properties (relative absolute height, absolute height, manner of installation, and box type), had relative influences of 6.0, 5.9, 3.9, and 2.9%, respectively. Boxes at the roof edge and those placed higher than 11 m above ground had a higher occupation probability (Figure 3e, 3f). The 116 occupied boxes had a mean absolute height (± SD) of 13.49 ± 2.95 m. Furthermore, external boxes had a higher occupation probability than internal ones (Figure 3g). There were only negligible differences in occupation probability between the different box types (Figure 3h). However, within the Schwegler 17 group, single boxes (17 and 17/17B) were much more frequently occupied than triple boxes (17A; Table 3e).
The overall proportion of nest-boxes occupied by Swifts in Greifswald (24.3%) was more than twice as high as that found in Berlin by Wortha and Arndt (Reference Wortha and Arndt2004). This discrepancy may be partly due to methodological differences between the studies. In the study by Wortha and Arndt, the observations were mostly conducted in the morning and early evening, while the Swift’s feeding frequency peaks around noon in good weather, and again in the later evening (Lack and Owen Reference Lack and Owen1955, Arens Reference Arens2011). Additionally, the time when breeding Swifts return to their nests for the night (in the latitudes of northern Germany normally about 15–45 minutes after sunset; Church Reference Church1956, Arens Reference Arens2011) was also not considered. In our study, older boxes showed higher occupation probabilities than recently installed ones. Hence, the overall occupancy rate of the examined boxes in Greifswald can be expected to grow even further in the near future.
The 116 occupied boxes in our study represented about a quarter of Greifswald’s estimated Swift population (450–500 breeding pairs; Starke et al. Reference Starke, Bendt, Köhler, Luhn and Sellin2010). The (minimum) population density of Swifts breeding in nest-boxes in the district Ostseeviertel Parkseite (34 breeding pairs per 10 ha) falls within the highest values of Swift densities given by Bauer et al. (Reference Bauer, Bezzel and Fiedler2005). It is highly likely that fewer Swifts were breeding in this district before the onset of the extensive renovation works in 2004 (see Starke et al. Reference Starke, Bendt, Köhler, Luhn and Sellin2010).
Our comparison of the number of breeding sites prior to renovation and in 2013 gave no indication of an overall decline. In contrast, Wortha and Arndt (Reference Wortha and Arndt2004) found the aggregate number of breeding pairs on 16 renovated buildings to be only about one third of the original number. On the other hand, results similar to ours were achieved in a project with nest-boxes as mitigation measures in England (D. Newell in litt. 2014). When interpreting our findings on breeding sites before renovation and in 2013, it should be noted that it was not possible to figure out details of the survey method (e.g. observation duration) used for the expert assessments prior to renovation. Possibly, less effort was expended in these assessments to determine the number of breeding sites than in our investigation. Thus, the number of breeding sites prior to renovation might have been underestimated. On the other hand, the nest-boxes on the 10 buildings regarded here had been installed only relatively recently (Table 1), and hence, the number of occupied boxes on these buildings will probably increase in the next years.
Our results confirm that nest-boxes for Swifts are frequently used by other cave-nesting bird species (e.g. Wortha and Arndt Reference Wortha and Arndt2004), notably by House Sparrows that occupied nearly as many boxes than Swifts. Hence, installing surplus Swift boxes is probably an effective measure to support populations of the House Sparrow, another synanthropic species that showed declines during the last decades in Europe (PECBMS 2012) and that is red-listed in certain countries (Eaton et al. Reference Eaton, Brown, Noble, Musgrove, Hearn, Aebischer, Gibbons, Evans and Gregory2009).
Influence of box properties on occupation probability
The finding that boxes with no other boxes within 1 m had the highest occupation probability suggests that Swifts avoid to some degree breeding in the immediate vicinity of conspecifics. This is somewhat surprising, as it contradicts statements on the usual distances between Swift breeding - sites in buildings (Weitnauer and Scherner Reference Weitnauer, Scherner, Glutz von Blotzheim and Bauer1980). Colombo and Galeotti (Reference Colombo and Galeotti1993) even reported that Swifts favoured breeding sites within 0.5 m of conspecifics. There are also anecdotal reports of large multi-compartment boxes where many cavities were simultaneously occupied (e.g. Kaiser Reference Kaiser1993). In contrast, Thurston (Reference Thurston1991) recommended installing nest-boxes with at least 0.5 m between the entrance holes. Possible disadvantages of nesting in close proximity to each other include an enhanced movement of parasites (such as the Common Swift Louse Fly Crataerina pallida; see Walker and Rotherham Reference Walker and Rotherham2011) between nests, as well as an increased risk that Swifts enter the wrong nest site by mistake, which can lead to violent fights with the resident pair (Lack Reference Lack1956). It is important to note that our results do not question that Swifts are gregarious breeders, but indicate that breeding pairs within a “colony” prefer distances of more than 1 m between nest sites if they have the opportunity to choose.
In our study, Swifts showed a preference for boxes situated higher than 11 m. Other studies reported preferences for rather lower heights (Colombo and Galeotti Reference Colombo and Galeotti1993, Wortha and Arndt Reference Wortha and Arndt2004). Corrales et al. (Reference Corrales, Bautista, Santamaría and Mas2013) found no preferences within 3.5–12.2m height in a medieval city wall. However, in the latter study, higher cavities situated less than 1.7 m below the upper edge of the wall were avoided, probably because they were accessible to terrestrial predators. In contrast, we observed a higher occupation probability for nest-boxes located closer than 2 m to the roof edge. Theoretically, this could be explained by a shade-giving roof overhang, but most of the houses we investigated did not have such structures.
Although it is known that overheating of the nesting site can cause losses of Swift nestlings (Weitnauer Reference Weitnauer1980), we found only a slight preference for north-facing and no particular avoidance of south-facing boxes. It is worth noting, however, that we did not consider shade-giving structures (trees, other buildings, roof overhang) that may have reduced the risk of overheating in some cases. In other studies, no clear avoidance of south-facing breeding sites was found either (Colombo and Galeotti Reference Colombo and Galeotti1993, Wortha and Arndt Reference Wortha and Arndt2004). In this context, it is also interesting that our analysis did not reveal a preference for internal boxes, which presumably protect better against overheating. In contrast, a survey in England found internal boxes to be largely preferred by Swifts (D. Newell in litt. 2014), whereas Wortha and Arndt (Reference Wortha and Arndt2004) reported no significant difference between internal and external boxes. In conclusion, the general recommendation not to mount boxes on south-facing aspects (Mayer Reference Mayer2013) cannot be supported. But, despite the lack of clear empirical evidence, it is probably advisable not to install nest-boxes (especially thin-walled models) externally on south-facing facades without shade-giving structures to avoid the risk of overheating.
We observed large differences in box - occupation probability across the three investigated districts of Greifswald. A reason for preferences for certain breeding areas could be shorter commuting distances to good feeding sites. However, this scenario probably does not apply to our case, as all three districts are not far from rural areas and from water bodies. A more plausible reason for the observed differences could be the very high supply of potential breeding sites in Ostseeviertel Parkseite (367 nest-boxes on 20 ha), where most boxes of the district Ostseeviertel were located. That may have led to a lower occupation probability per box. On the other hand, additional building renovations carried out during spring and summer 2013 in Greifswald city centre may have increased the demand of nesting sites in this district (Du Feu Reference Du Feu2003).
Between the different nest-box types, we observed only small differences in occupation probability. This suggests that all box types considered in our study are generally suitable for Swifts. This also applies to the boxes installed in the “wrong” way (Strobel 416t and 416b), which consequently had a smaller interior depth (8.5 and 4 cm, respectively), and in the case of 416t, a higher threshold at the entrance hole (c.8 cm) than recommended (Kaiser Reference Kaiser1993, Scholl Reference Scholl2004, Mayer Reference Mayer2013). The large difference in the rate of occupancy between the single and triple variants of the Schwegler 17 group can be explained by the avoidance of sites with many neighbouring boxes.
The moderate predictive performance of the model indicates that Swifts were relatively flexible with respect to the box properties studied. In addition, factors not considered in the present study may also have influenced the occupation probability. For example, the degree to which boxes were exposed to direct sunlight was only indirectly included in our analysis via facade orientation, and the total number of potential breeding sites in the surroundings was also not considered. The partially unbalanced study design (Table 3), which was unavoidable, may also have influenced some of our results. We did not measure breeding success, and some of the box properties (e.g. orientation) that apparently did not strongly affect the Swifts’ choice of a box might still have influenced the reproductive outcome. Future investigations on that matter would thus be especially worthwhile. Finally, we want to reiterate that all buildings considered in the statistical analysis were renovated during the Swift breeding season. When renovation can be done outside the breeding season, the boxes should probably at best be installed exactly at the places of the original breeding sites to facilitate the boxes’ discovery by the returning Swifts (Scholl Reference Scholl2004).
Our study provides several lines of evidence that nest-boxes for Common Swifts can be successful compensatory measures in the context of renovation and demolition works. First, the overall rate of box occupancy was relatively high. Secondly, the total number of occupied boxes was higher than the assumed number of breeding sites before the renovations. Thirdly, we recorded a remarkably high population density of Swifts breeding in nest-boxes in a district where nearly all buildings had been renovated in the past 10 years. If renovation has to be carried out during the Swift breeding season, our results suggest that nest-boxes should be installed a few metres apart from each other and close to the roof edge. Furthermore, it is probably somewhat preferable to mount the boxes externally, on north-facing facades and above 11 m height. In part, our findings do not conform to results of some other studies and anecdotal reports of Swift breeding-behaviour. Therefore, further investigations should be carried out to assess whether our results and conclusions can be confirmed in other situations.
The supplementary materials for this article can be found at journals.cambridge.org/bci
We are grateful to the numerous observers who took part in the nest-box survey; to the Abteilung Umwelt- und Naturschutz, Stadtbauamt, Universitäts- und Hansestadt Greifswald for financial support; to the Landesamt für Umwelt, Naturschutz und Geologie Mecklenburg-Vorpommern (LUNG) and to J. Berg for providing the expert assessments; to the housing corporations WVG and WGG Greifswald for additional information on the examined buildings; and to C. Schöner, J. van Schaik, P. Atkinson and D. Newell for helpful comments on the manuscript.