Introduction
Farmland bird populations continue to decline, often despite substantial conservation efforts (Burns et al. Reference Burns, Eaton, Burfield, Klvaňová, Šilarová and Staneva2021; Rigal et al. Reference Rigal, Dakos, Alonso, Auniņš, Benkő and Brotons2023). Failing recovery may result from incomplete knowledge of the factors that limit population growth (Green Reference Green1995), such as large-scale determinants of distribution, abundance, and population connectivity, the impact of land use or weather extremes on reproduction and survival, or the local availability of food or safe nesting sites (Crick et al. Reference Crick, Dudley, Evans and Smith1994). Since at least some of this information is often unavailable, conservation programmes tend to copy or extrapolate from findings in other regions or species (Sutherland Reference Sutherland2022), which may result in locally insufficient or inappropriate measures.
The problem is increased when conservation recommendations derive from correlative evidence that is mistaken for causation (Green Reference Green1995; Josefsson et al. Reference Josefsson, Hiron, Arlt, Auffret, Berg and Chevalier2020). For example, associations of territory placement with habitat composition can be spurious when populations in suboptimal habitats carry an extinction debt (Hylander and Ehrlén Reference Hylander and Ehrlén2013; Kuussaari et al. Reference Kuussaari, Bommarco, Heikkinen, Helm, Krauss and Lindborg2009). Furthermore, territory delineation may be determined by song-post availability rather than by limitations to local productivity, e.g. nest-sites (Van Horne Reference Van Horne1983; Vickery et al. Reference Vickery, Hunter and Wells1992). Habitat association studies – on which conservation measures are often based – therefore benefit from studies investigating nest locations, hatching and fledging rates, and reproductive success (Green Reference Green1995; Sutherland Reference Sutherland2022).
An example where incomplete information on breeding ecology may impede effective conservation is the Corn Bunting Emberiza calandra. This farmland bird’s global population is concentrated in Europe (Burfield et al. Reference Burfield, Rutherford, Fernando, Grice, Piggott, Martin, Balman, Evans and Staneva2023), with on-going declines in central and north-western Europe (e.g. Anthes et al. Reference Anthes, Boschert and Daniels-Trautner2017; Comolet-Tirman et al. Reference Comolet-Tirman, Gazay, Quaintenne and Wroza2021; Crick et al. Reference Crick, Dudley, Evans and Smith1994). Many Corn Bunting conservation schemes developed from observations in large or stable populations in vast open agricultural landscapes with >10% proportions of set-aside cropland, as in eastern central Europe (Fischer Reference Fischer1999; Goławski and Dombrowski Reference Goławski and Dombrowski2002; Schmidt et al. Reference Schmidt, Fartmann, Kiehl, Kirmer and Tischew2022). Given that territory-holding males also associate with set-aside farmland in western central Europe (Burgess et al. Reference Burgess, Bright, Morris, Field, Grice and Cooke2015; Meichtry-Stier et al. Reference Meichtry-Stier, Jenny, Zellweger-Fischer and Birrer2014; Schmidt et al. Reference Schmidt, Fartmann, Kiehl, Kirmer and Tischew2022; Staggenborg et al. Reference Staggenborg, Back, Debatin, Grom, Hielscher and Schneider2024), perennial fallows and agri-environment scheme flower fields are key components of Corn Bunting conservation. However, this does not consider whether a territorial male is among the often substantial proportion of unpaired males, given widespread polygyny in Corn Bunting (Hartley et al. Reference Hartley, Shepherd and Burke1993). It further ignores nest locations, which may concentrate outside apparently “safe” habitats (Hegelbach Reference Hegelbach, Glutz von Blotzheim and Bauer1997), and breeding success, which may vary between different nest habitats.
Positive effects of set-asides are well-documented for eastern central Europe (Fischer Reference Fischer, Flade, Plachter, Schmidt and Werner2006; Fischer and Schneider Reference Fischer and Schneider1996; Goławski et al. 2001 as cited in Dombrowski and Golawski Reference Dombrowski and Golawski2002). It remains unclear, however, whether this holds for regions with more productive soils, higher crop diversity, smaller field sizes, and attractive alternative nesting habitats such as cereal fields, clover-grass leys (Setchfield et al. Reference Setchfield, Mucklow, Davey, Bradter and Anderson2012; Stein-Bachinger and Fuchs Reference Stein-Bachinger and Fuchs2012), or meadows (Broyer et al. Reference Broyer, Curtet and Chazal2014; Suter et al. Reference Suter, Rehsteiner and Zbinden2002). The latter are often cut during the breeding season and studies in France have shown that grassland populations there recovered only with late mowing on ≥50% of the local meadow area (Broyer et al. Reference Broyer, Curtet and Chazal2014, Reference Broyer, Sukhanova and Mischenko2016).
Our study connects spatial and temporal variation of Corn Bunting nesting ecology with land-use practices as a plausible driver for on-going population declines. Covering several of the remaining marginal Corn Bunting populations in south-west Germany and Switzerland, we first characterised variation in nest-site selection along a gradient from cropland- to grassland-dominated landscapes. Secondly, we quantified Corn Bunting nesting phenology and its variation between regions, years, and habitats. Finally, we associated nest survival with nest habitat, season, and plot-specific mowing, harvesting or grazing schedules. Based on these findings, we evaluated whether Corn Bunting conservation requires habitat- or landscape-specific adjustments to improve productivity in western central Europe.
Methods
Study areas
Our seven study areas in south-west Germany and one in north Switzerland (Figure 1) comprised typical Corn Bunting nesting habitats spanning landscapes dominated by extensive grasslands in the Rhine River floodplains (four areas with n = 46 nests surveyed) to landscapes dominated by cropland (three areas with n = 38 nests). One study area (Rottenburg, n = 141 nests) had a more balanced proportion of cropland and grassland and comprised cattle pastures under rotational grazing with high stocking densities (site characterisations in Supplementary material Figure S1 and Table S1). Most nest surveys took place between 2018 and 2020 (Table S1). In each study area and year, we mapped land-use types (crop types and landscape elements) using QGIS version 3.16 (Open Source Geospatial Foundation, http://qgis.org). Between late April and mid-July, we recorded patch-specific land-use dates for mowing, grazing, and (fodder) crop harvest; cereal harvest (of winter barley) rarely commenced before early July.
Location of the seven study areas. Labels link to Supplementary material Figure S1 and Table S1. Digital elevation model from European Union Copernicus data (EU-DEM layers).
Nest habitat and clutch parameters
We attempted to locate nests in all male territories, irrespective of habitat. Our final sample comprised 225 nests with exact locations derived from repeated observations of nest building activity in identical positions (n = 66 nests), repeated provision of nestling food in identical positions (n = 60 nests), or confirmation of nest-sites through nest visits (n = 99 nests). For each nest, we recorded the following parameters: (1) nest habitat: crop or land-use category at the nest location (all nests); (2) nest height: height of the upper nest rim above the ground (cm; nests visited); (3) vegetation cover and vegetation height: horizontal cover as viewed from above (5% increments) and dominant height (10 cm increments) of the herbaceous vegetation estimated visually within 1 m of nests visited.
As applicable, we further recorded the number of eggs and the number and age of nestlings in days (day 1 = hatching day). Nestling age was estimated according to a calibrated picture series from local nestlings of known age (Figure S2).
Nesting phenology
To compare nesting phenology with land-use activities, we derived nest-specific timespans from nest building to full fledging from objective time points as available: nest building date, first egg date (FED) for incomplete clutches (backdated assuming a 1-day interval between successive eggs), and hatching date from nestlings of known age (available for n = 155 nests). We accepted less precise estimates when nests were visited only once during incubation, or when we confirmed nestling provisioning without nest visits. In those cases, we assumed that observations occurred midway in the incubation or nestling stage, respectively, and thus estimated their start 6 days before the observation, accepting a maximum estimation error of ± 6 days (n = 63 nests). From these dates, we extrapolated dates of first nest building, first egg laying, incubation, hatching, nest leaving, and full fledging based on literature data (Boschert Reference Boschert and Hölzinger1997; Gliemann Reference Gliemann1973; Hegelbach Reference Hegelbach, Glutz von Blotzheim and Bauer1997), and own observations on the mean duration of each phase (Figure S3) rounded to the next full day to add a small buffer for delays that may occur in each nesting phase:
-
(i) nest building: 3 days (typically 2–4 days)
-
(ii) egg laying: 6 days (adjusted to full clutch size where known)
-
(iii) incubation: 13 days after clutch completion (typically 11–14 days)
-
(iv) nestlings: 12 days (9–13 days)
-
(v) jumplings: 12 days. Corn Bunting chicks leave the nest before they are able to fly (Hegelbach Reference Hegelbach, Glutz von Blotzheim and Bauer1997). We term them “jumplings” until fully fledged and capable of flight. In response to perceived danger, jumplings typically remain stationary in the vegetation and may thus be killed by harvest or mowing given the low cutting and high suction of modern machines. Previous studies imply that jumplings need at least 10–15 days after leaving the nest before switching from a “freeze” to a flight escape response to approaching threats (Gliemann Reference Gliemann1973; Perkins et al. Reference Perkins, Maggs, Wilson and Watson2013).
We analysed variation in nesting phenology with FED as the response variable with linear mixed effects models in the glmmTMB package (version 1.1.8; Brooks et al. Reference Brooks, Kristensen, van Benthem, Magnusson, Berg and Nielsen2017) in R (version 4.3.1; R Core Team 2023). We first compared FED between the four study regions (Table S1). Second, we compared FED among nest habitat types (up to nine levels, see Figure 3) separately for two regions with sufficiently large sample sizes, i.e. Rottenburg and the Rhine River valley. These and all further models contained study year as random intercept to account for between-year variation in mean FED.
For model assessment according to Santon et al. (Reference Santon, Korner-Nievergelt, Michiels and Anthes2023), we inspected standardized residuals for independence from fitted values and homogeneity across predictor variables. We further conducted posterior predictive checks on model-simulated data for dispersion, zero inflation, and distribution relative to observed data. FED model assessment favoured a Gaussian distribution, after excluding nests for which observations implied second or replacement broods because these generated extreme values that were insufficiently captured by our models.
In the Statistical Supplement, we report fixed and random coefficient and effect size estimates with their compatibility intervals (CI) but refrain from presenting P values and their associated evaluation of binary null hypotheses in accordance with recommendations for unbiased statistical reporting (Berner and Amrhein Reference Berner and Amrhein2022; Halsey et al. Reference Halsey, Curran-Everett, Vowler and Drummond2015).
Apparent survival with or without nest protection
We classified nests as “successful” when adults were observed feeding at least one jumpling clearly outside, but in close vicinity of a nest previously found active. We routinely protected nests from agricultural operations that remove nest-holding vegetation through contractual agreements for postponed mowing, harvesting or grazing. Therefore, our raw nest success data only reflect failures due to “natural” causes such as adverse weather or predation. Apparent survival of protected nests (ASprotected) thus quantifies the proportion of nests where chicks reached the jumpling stage, given our protection from land-use operations.
In the absence of nest protection, nests additionally risk destruction by land use, with close-to-zero nest survival if mowing, harvest or intense grazing occur during nesting (Stein-Bachinger et al. Reference Stein-Bachinger, Fuchs, Gottwald, Helmecke, Grimm and Zander2010; authors’ unpublished data). To approximate such “anthropogenic” losses and correct our ASprotected estimates, we calculated land-use conflict probabilities (LCP) as the fraction of study years in which nest-specific breeding schedules overlapped with patch-specific land-use dates, excluding years where contractual conservation agreements postponed land use on the same patch. We then calculated the apparent survival of unprotected nests as ASunprotected = ASprotected * (1 - LCP). Confidence intervals for this combined estimate were obtained based on the delta-method (Buehler Reference Buehler1957).
Both AS-values may overestimate true nest survival because early nest failures may be underrepresented if nest detection is biased towards later nesting phases (Mayfield Reference Mayfield1975). We assumed similar nest detection rates (and thus similar bias in survival estimates) across nest habitats and regions, allowing for relative comparisons among nest habitats in the maximum possible nest sample (Table S1).
Statistical analyses followed the modelling approach outlined above, with nest habitat (nine levels) as the predictor variable and study year as random intercept. Model assessment favoured a logit-link binomial error distribution for ASprotected and a logit-link betabinomial error distribution for the mildly overdispersed ASunprotected.
Mayfield survival with or without nest protection
Mayfield estimates of daily nest survival rates (DSR) remove the bias in apparent nest survival because they integrate the time each nest has been under observation (exposure duration; Johnson Reference Johnson1979; Mayfield Reference Mayfield1975) and allow for variation in survival rates with nest age (Rotella et al. Reference Rotella, Dinsmore and Shaffer2004; Weiser Reference Weiser2021). We implemented Mayfield logistic regression in MARK (White and Burnham Reference White and Burnham1999), accessed through the R package RMark (Laake Reference Laake2013), with nest outcome as the binary response variable (Dinsmore and Dinsmore Reference Dinsmore and Dinsmore2007).
To minimise predictor collinearity, we conducted hierarchical modelling following Rotella et al. (Reference Rotella, Dinsmore and Shaffer2004) (Table S2). First, we screened for informative predictor variables among time and region predictors (set 1) and among nest and brood characteristics (set 2). Set 1 contained Hatching Day (days after the earliest hatching day observed in the data set), Time (days since our earliest nest recording), Nest Age (days after incubation start), Study Year (nine years), and Study Region (two categories: Rottenburg and others). Set 2 contained Nest Habitat (simplified to seven levels to avoid small sample sizes: cereal fields combined with other crops, pasture combined with fallow grassland), First Brood (first or later broods), Clutch Size (n eggs per clutch), Nest Height, Vegetation Cover, and Vegetation Height. Second, we combined the most informative predictor(s) per subset in a final model set, selecting predictors within 2 ΔAICc of the best performing model and retaining nesting habitat as our core predictor.
We derived mean DSR estimates per nest habitat at median covariate levels and from these estimated Mayfield nest survival across the entire incubation (13 days) and nestling (12 days) periods as DSR raised to the power of 25 (Johnson Reference Johnson1979). DSR analysis was conducted for 93 nests visited at least twice during incubation or nestling phase. DSR estimates again included nest protection in meadows, alfalfa/clover-grass leys, and pasture. To obtain realistic survival estimates, Mayfield nest survival was thus multiplied with LCP-values as detailed for apparent survival above.
Results
Nest habitat and clutch parameters
Nests were predominately placed in cultivated land (77%), but habitats varied greatly among study regions (Figure 2a). In the cropland-dominated Creglingen region, most nests were in conventional winter wheat and barley. Mixed farmland landscapes had nests primarily in perennial agri-environment scheme flower fields in Ertingen/Klettgau, but showed a more even spread across meadows, alfalfa/clover-grass leys, cereals, and pastures in Rottenburg. Nests in the grassland-dominated Rhine River valley were mostly in fallow grassland and meadows. The largest sample of nests, from Rottenburg, revealed between-year variation in nest habitats (Figure 2b), with a dominance of clover-grass versus cereal fields versus meadows alternating between successive years.
Corn Bunting nest habitats per study region (left panel) and per year at Rottenburg (right panel). Values show the proportion of nests per region or year. For sample sizes see Table S1.
Average clutch size was 4.49 eggs (95% CI: 4.08–4.95, n = 93 full clutches), without indications for variation with FED, study region or nest habitat (Statistical Supplement A). Hatching rate, i.e. the proportion of eggs that hatched, was 0.88 on average (95% CI: 0.84–0.91, n = 70 clutches with hatching information). We found no variation in hatching rates with FED, clutch size or study region, but hatching rates in agri-environment scheme flower fields tended to be lower than in other nesting habitats (Statistical Supplement A).
Nesting phenology
FED was spread between 22 April and 13 July and varied between study regions (Figure 3a). Earliest egg laying occurred in the Rhine valley lowlands (predicted mean FED: 20 May; 95% compatibility interval (CI): 14–27 May), followed by Rottenburg (25 May, CI: 21–30 May), Ertingen/Klettgau (5 June, CI: 25 May–16 June), and Creglingen (7 June, CI: 31 May–13 June).
Variation in Corn Bunting first egg dates (FED) between regions (a) and between nest habitats in Rottenburg (b) and the Rhine valley (c). Grey dots are raw data, large blue dots and flags are model-predicted mean values with 95% compatibility intervals. Estimated pairwise differences are given in Statistical Supplement B.
In Rottenburg, egg laying was earliest in “grassland-like” habitats (meadows and pastures), but some early FEDs also occurred in clover-grass leys (Figure 3b). Late egg laying occurred in cereal fields, semi-natural margin habitats, and perennial agri-environment scheme flower fields. Mean FED in flower fields was 19.3 (CI: 8.9–29.6) days later than in meadows (Statistical Supplement B). In the Rhine valley, egg-laying dates were rather similar in all grassland nest habitats (Figure 3c and Statistical Supplement B).
Apparent survival with or without nest protection
ASprotected varied substantially between nest habitats (Figure 4a and Table 1). It was highest in fallow grassland, followed by oilseed rape and root crops, and agri-environment scheme flower fields ≥3 years after sowing. Lowest survival (<60%) occurred in flower fields in the second year after sowing, and in alfalfa/clover-grass leys. These estimates already contain the benefit of nest protection from mowing, harvesting or grazing, initiated for 43 out of 74 nests in meadows, 25 out of 30 nests in alfalfa/clover-grass leys, and 19 out of 22 nests in pasture.
Variation in Corn Bunting nest survival. Panel (a) shows apparent survival with nest protection ASprotected per nest habitat (blue circles), complemented with corrected ASunprotected values (red diamonds) for the three habitats given in (b). Panel (b) shows land-use conflict probabilities (LCP), i.e. probabilities that a given nesting period conflicts with harvest, mowing or grazing on the same patch (cf. Figure 5). Panel (c) compares apparent nest survival between visited and unvisited nests. Grey dots display raw data, large blue dots and flags model-predicted means with their 95% compatibility interval. Full model outputs including estimated pairwise differences are given in Statistical Supplement C.
LCP, i.e. the fraction of nests that would conflicts with regular land-use operations on the same patch in the absence of nest protection (Figure 5), were 64% for nests in alfalfa/clover-grass leys, 56% in meadows, and 38% in pastures (Figure 4b and Table 1). Cereal harvest almost never overlapped with active Corn Bunting broods and was never postponed for nest protection. Thus, the ASprotected estimate for cereal fields (62%) (Figure 4a and Table 1) already includes potential harvesting effects. After correcting apparent survival for land-use conflict, we found particularly low survival estimates for meadows and alfalfa/clover-grass (ASunprotected in Table 1).
Nesting periods of individual Corn Bunting nests (bars) in meadows and alfalfa/clover-grass leys across the season (x-axis). Blue dots indicate observed dates of land use (mowing or harvest) per individual field. Typical spans per breeding phase are given in Figure S2.
Apparent nest survival (AS) per habitat. Values are observed apparent survival of nests protected from land-use operations (ASprotected; see Figure 4a), land-use conflict probabilities (LCP; see Figure 4b), and apparent survival after accounting for land-use conflicts, with ASunprotected = ASprotected* (1 - LCP). For sample sizes see Figure 4
* LCP set to zero for habitats without dedicated nest protection, i.e. ASunprotected = ASprotected.
We found no indication that nest visits affected nest success (Figure 4c).
Mayfield nest survival
Mean DSR with nest protection (DSRprotected) was 0.960 (95% CI: 0.0.943–0.971), implying that nests had an average chance to survive the 25-day incubation and nestling period of 0.357 (CI: 0.220–0.485). DSR tended to decline with hatching date (Figure 6a) and nest height above ground (Figure 6b). Small sample sizes prevented a robust differentiation of DSR between nest habitats or regions (Table S2). Yet, best available estimates suggest low DSR for agri-environment scheme flower fields, particularly in their second year after sowing, and highest DSR in semi-natural margin habitats and meadows (Table 2 and Figure 6c and d).
Mayfield estimates of daily survival rates of protected nests (DSRprotected) for hatching date (a), days since earliest recorded hatching, nest height (b), and nest habitat (c), and the estimated total survival during the 25-day nesting period per nest habitat (d) derived from (c). Dotted lines show overall mean survival rates in (c) and (d). In (d), red diamonds show survival rates after correction for land-use conflict probabilities (LCP) as given in Figure 4b for clover-grass, meadows, and pasture.
Mayfield nest survival estimates per habitat. Daily nest survival rates (DSR; Figure 6c) were predicted from Mayfield models at median hatching date per habitat. From those we calculated Mayfield survival (MS) for the entire nesting period as DSR^25 (Figure 6d). For pasture, meadows, and alfalfa/clover-grass, we corrected these survival estimates with land-use conflict probabilities (LCP) as given in Table 1. For sample sizes see Figures 6c
* LCP set to zero for habitats without dedicated nest protection, i.e. MSunprotected = MSprotected.
To estimate survival rates in the absence of postponed mowing, harvesting or grazing, we multiplied our DSR values with 1 - LCP (Table 1). This suggested that just 13.2% of nests in alfalfa/clover-grass leys, 21.3% in meadows, and 20.3% in pastures would have been successful, on average, in the absence of individual nest protection (Table 2).
Discussion
Our study revealed several challenges for effective conservation of a ground-nesting farmland bird in cultivated land. It highlighted the importance of local knowledge on nesting ecology before extrapolating or copying conservation measures from other regions.
First, nesting phenology spread widely between study sites and nest habitats. Earliest nests were built in April and the last chicks fledged in late August. Nesting commenced earlier at lower altitudes and – within study sites – progressed from grassland habitats (meadows and pastures) to arable crops and agri-environment scheme flower fields. This sequence matches findings in British and French populations (Brickle and Harper Reference Brickle and Harper2002; Perkins et al. Reference Perkins, Maggs, Wilson and Watson2013, Reference Perkins, Maggs and Wilson2015; Setchfield et al. Reference Setchfield, Mucklow, Davey, Bradter and Anderson2012) and likely arises from a preference for dense and medium-tall sward structures that provide optimal nest concealment (Perkins et al. Reference Perkins, Maggs and Wilson2015).
Second, we found strong variation in nest habitats between landscape types and study years. Cereals, fodder crops, and perennial flower fields and fallows dominated in cropland-dominated landscapes, and extensive or abandoned grassy habitats in grassland-dominated landscapes. In mixed landscapes, we found a rather homogeneous spread of nests among several habitats. These patterns are consistent with the findings of 16 studies on Corn Bunting nest habitats and nest parameters from across central and western Europe that we collated in a comprehensive quantitative literature overview in Table S3.
We further document substantial between-year variation in the fraction of nests placed in “low-risk” habitats such as cereal fields versus “high-risk” habitats, in which the nest-holding vegetation is removed during the Corn Bunting nesting season. Between-year variation in habitat availability alone cannot explain annual fluctuations in nest habitat because within-site proportions of most habitat types were rather stable across years. We speculate on two main reasons. First, Corn Bunting nest locations tend to be clustered (Perkins et al. Reference Perkins, Maggs, Wilson and Watson2013), so that single or a few habitat patches may attract multiple nests. Second, annually and seasonally varying weather conditions influence vegetation phenology, vegetation structure, and land management, which in turn may cause variation in habitat-specific nest-site suitability. Our data, however, do not allow the investigation of these aspects in detail.
Our data imply that Mayfield survival of protected nests in high-risk habitats (36.6%, 48.6%, and 32.6% for alfalfa/clover-grass leys, meadows, and pasture, respectively) declined to just 13.2%, 21.3%, and 20.3% when accounting for LCP. This stresses the need for, and effectiveness of, nest protection through contracts for postponed land use, which approximately doubled nest success in three core nest habitats, especially in mixed landscapes (Figure 2).
Our overall mean Mayfield DSR (0.960) falls within the range (0.894–0.979) reported across crop types and regions in Britain (Aebischer Reference Aebischer1999; Brickle et al. Reference Brickle, Harper, Aebischer and Cockayne2000; Perkins et al. Reference Perkins, Maggs, Wilson and Watson2013; Setchfield et al. Reference Setchfield, Mucklow, Davey, Bradter and Anderson2012). However, our habitat-specific nest survival estimates for cereals and fallow grassland (0.32) were well below those reported for cereals and rough grass in Scotland (0.45–0.5; Perkins et al. Reference Perkins, Maggs, Wilson and Watson2013). Similar to the findings from Scotland (Perkins et al. Reference Perkins, Maggs, Wilson and Watson2013), we found that DSR declined moderately as the season progressed. Late broods more likely conflicts with mowing, grazing or harvest, but these have minor impact in our data set given the protection of almost all known nests. We inferred plausible causes for 24 out of the 30 nest failures observed among the 90 nests visited, with all except the first category dominating in late season nests: predation (12 cases); severe rain or drought (7 cases), which are increasing during central European summers (Hänsel et al. Reference Hänsel, Hoy, Brendel and Maugeri2022); land use (3 cases of failed conservation intervention); tall vegetation collapsing on the nest (1 case); nestling starvation (1 case), possibly driven by a lack of insect food availability (Grames et al. Reference Grames, Montgomery, Youngflesh, Tingley and Elphick2023) during drought. DSR also tended to decline for above-ground nests, which in our study were restricted to tall vegetation on rather open soils, such as agri-environment scheme flower fields within the first two years of sowing. Such elevated nests tended to tip over or showed signs of predation by mammals, which can more easily access such habitats than those with denser ground swards.
We found comparably large clutches and hatching success (cf. Hartley and Shepherd Reference Hartley and Shepherd1994) (Table S3). Also, the mean size of successful broods (3.85 jumplings, n = 60 visited successful nests) was at the upper end of values reported in the literature (Perkins et al. Reference Perkins, Maggs, Wilson and Watson2013) (Table S3). We therefore argue that low population productivity in our study areas is driven less by hatching success or brood size than by nest failures (as discussed above), a low fraction of females that raise a second brood (cf. Brickle and Harper Reference Brickle and Harper2002; Setchfield et al. Reference Setchfield, Mucklow, Davey, Bradter and Anderson2012; Siriwardena et al. Reference Siriwardena, Baillie, Crick and Wilson2000), and habitat loss. As a crude indicator of productivity, we estimated the mean number of successful broods per male territory and year required for population stability. We base this on three scenarios for adult survival and juvenile recruitment as observed in (scenario 1) Switzerland (Hegelbach Reference Hegelbach, Glutz von Blotzheim and Bauer1997) and (2) Scotland (Perkins et al. Reference Perkins, Maggs, Wilson and Watson2013), complemented with (3) an “optimistic upper limit”. The three scenarios used mean adult survival rates of 0.59, 0.58, and 0.65, and juvenile recruitment rates of 0.107, 0.29, and 0.35, respectively. Coupled with our mean output of 3.85 jumplings per successful nest, these values suggest that at least 1.99 (scenario 1), 0.75 (2) or 0.52 (3) successful broods per male territory were required for population stability. Observed values in our core study site Rottenburg, however, span between 0.27 and 0.71 in 2014–2024 (mean: 0.44; unpublished data), crossing calculated minimum values never for the two data-based scenarios (1) and (2) and in just 4 out of 11 years for the “optimistic” scenario (3).
Conservation and management implications
Meadows in grassland-dominated and mixed landscapes attracted a substantial fraction of nesting attempts. These were exposed to a high failure risk from mowing, even in late-mown hay meadows. Perkins et al. (Reference Perkins, Maggs, Wilson and Watson2013) also reported lower nest survival rates in meadows under conventional mowing (DSR 0.89) compared with delayed mowing (DSR 0.96). The difference was largely caused by mowing-induced nest failures of 66%, approximately matching our estimated LCP for meadows of 56%. Delayed mowing can restore population productivity when implemented on 50–75% of the meadow habitat suitable for the local Corn Bunting population (Broyer et al. Reference Broyer, Sukhanova and Mischenko2016; Perkins et al. Reference Perkins, Maggs, Wilson and Watson2013). Fractions of <20% only suffice when detailed prior knowledge allows targeting meadow patches with proven clustering of Corn Bunting nests (Perkins et al. Reference Perkins, Maggs, Wilson and Watson2013). Non-targeted delayed mowing on such small land fractions likely fails because birds cannot distinguish early from late-mown grassland at the onset of nesting and thus may not focus nest placement on protected patches (Broyer et al. Reference Broyer, Curtet and Boissenin2012).
Mowing dates safe for Corn Bunting depend on local nesting schedules, with 15 July or 1 August plausible options in regions with regular (Rottenburg; Figure 2) or late breeding (Scotland, southern England; Brickle and Harper Reference Brickle and Harper2002; Perkins et al. Reference Perkins, Maggs, Wilson and Watson2013), respectively. While late mowing can enhance the chances for nest survival, it may locally compromise yield or meadow habitat quality, potentially conflicting with regulations to protect lowland hay meadows (habitat type 6510) under the EU-habitats directive. Conservation contracts should thus encourage early grazing or mowing before the onset of the local farmland bird breeding season, which may control weed development (e.g. autumn crocus Colchicum autumnale), favour herbs over grasses, maintain yield (cf. Angerer et al. Reference Angerer, Katzenmayer, Hölzl, Eberle and Habel2023), and prevent late-mown tall grass swards from collapsing due to wind or rain (unpublished data). Such a regime partially mimics traditional land use where winter or spring pasture precedes the meadow period (Kapfer Reference Kapfer2010), also to the benefit of insect diversity and abundance (Bruppacher et al. Reference Bruppacher, Pellet, Arlettaz and Humbert2016; Humbert et al. Reference Humbert, Pellet, Buri and Arlettaz2012).
We recommend late mowing as a complementary measure to landscape-scale restoration of safe nest habitats that are more attractive for Corn Buntings than meadows. Prime options include long-term set-aside (Staggenborg and Anthes Reference Staggenborg and Anthes2022) or the transformation of intensely mown or grazed grassland into year-round or seasonal cattle or horse pasture with low stocking densities (Handschuh and Klamm Reference Handschuh and Klamm2022). Both habitats provide favourable nest-sites and maintain a rather stable and predictable vegetation structure throughout the Corn Burning breeding season.
Our findings confirm that alfalfa/clover-grass leys can act as an ecological trap. They attract nesting Corn Buntings and other farmland birds (González del Portillo et al. Reference González del Portillo, Arroyo and Morales2022; Stein-Bachinger and Fuchs Reference Stein-Bachinger and Fuchs2012) but are associated with high risks of nest failure due to repeated harvest during the breeding season. Clover-grass is a key element of organic crop rotation for nitrogen fixation, weed suppression, fresh or silage livestock fodder, and bioenergy, typically covering 20–30% of organic farmland. Its relevance will increase given EU aims to expand organic farming to 25% by 2030 (Fetting Reference Fetting2020). We suggest contractual Corn Bunting conservation to ban clover-grass harvest between at least 1 May and the Corn Bunting-safe mowing dates proposed above, or to counsel farmers to integrate clover-grass leys into set-aside requirements under EU regulations for good agricultural and environmental conditions (GAEC).
Finally, we found that early successional agri-environment scheme flower fields (second year after sowing) attracted only a few Corn Buntings nests, then with low success rates. Tall vegetation coupled with low sward density near the ground can increase nest predation rates (Perkins et al. Reference Perkins, Maggs and Wilson2015). In contrast, perennial flower fields – three or more years after sowing – not only represent preferred territory centres (Zollinger et al. Reference Zollinger, Birrer, Zbinden and Korner‐Nievergelt2013) and an all-year foraging habitat for Corn Buntings (Rieger et al. Reference Rieger, Mailänder, Stie, Santon, Staggenborg and Anthes2022), but also provide safe nesting structures in patches with dense vegetation at ground level. Thus, depending on the seed mix and the local vegetation structure, late successional flower fields can qualify as a targeted Corn Bunting conservation measure (Staggenborg and Anthes Reference Staggenborg and Anthes2022).
Conclusions
Our study highlights that effective farmland bird conservation requires solid knowledge of local nesting ecology. Without such knowledge, conservation measures cannot necessarily be transferred between habitats, landscape types or regions. Its long breeding season and variable nest habitat selection between regions and years challenge Corn Bunting conservation in intensive agricultural landscapes.
We derived two key recommendations. First, given a small fraction of nests in set-aside or non-productive fields (Table S4), effective Corn Bunting conservation must stabilise and improve nest survival on regularly cultivated land. Here, we suggest refined conservation schemes for alfalfa/clover-grass leys, meadows, and pastures to enhance local Corn Bunting productivity. Future studies should evaluate the success of these measures and elucidate the currently understudied conflict of the Corn Bunting nesting period with cereal harvest and mechanical weed control in organic crops.
Second, non- or low-productive areas (in a land sparing sense; Grass et al. Reference Grass, Batáry, Tscharntke, Bohan and Vanbergen2021) may boost Corn Bunting productivity only when they become a dominant feature, e.g. in landscapes with extensive permanent pasture or substantial fractions of abandoned land (Handschuh and Klamm Reference Handschuh and Klamm2022). We suggest the development of such a model “Corn Bunting Landscapes”, ideally associated with agricultural rewilding (Corson et al. Reference Corson, Mondière, Morel and van der Werf2022), where large fractions of land may provide attractive nesting and feeding habitat for farmland birds.
A two-tier approach that increases the reproductive output of farmland birds within productive farmland while providing strategically placed species strongholds with prolific source populations may constitute the most promising approach to a recovery of farmland bird populations, and is also in line with the novel EU Nature Restoration Law.
Supplementary material
The supplementary material for this article can be found at http://doi.org/10.1017/S0959270925000048.
Acknowledgements
This research received administrative support through the regional councils of Freiburg, Karlsruhe, and Stuttgart as well as local district administrations. Thorsten Teichert and ‘Verein Vielfalt e. V. Tübingen’ coordinated nest protection measures at our core study site Rottenburg. We thank Clarissa Anders, Martin Boschert, Santiago Cruzes Vallo, Sabine Geissler-Strobel, Heiner Götz, Valentin Grom, Annika Hammerschmidt, Alexandra Ickes, Markus Jenny, Ariane Koch, Mathias Kramer, Franziska Lehle, Miriam Plappert, Timo Reischmann, Benjamin Reichelt, Mirjam Schöckle, and Caroline Schuck for general support or contributions to fieldwork. Urs Kormann and one anonymous referee provided very insightful suggestions on an earlier manuscript draft.
Data availability statement
Raw data for all reported analysis are available via Zenodo (https://doi.org/10.5281/zenodo.12799271).
Author contribution
study conception (NA, MH); data collection (MH, NA, JS); data analysis lead (NA); data analysis support (MH, JS); manuscript lead (NA); manuscript support (MH, JS).
Funding
This work was funded by Stiftung Naturschutzfonds Baden-Württemberg (NA, JS, 2017–2021, Az. 73-8831.21/546 91-1749L); Tübingen district administration (NA and Sabine Geissler-Strobel, 2014–2018); Tübingen regional council (NA, MH, 2019–2023). Nest surveys were conducted under ringing permit to NA (ringer ID 0674 at Vogelwarte Radolfzell) issued within conservation permit Az. 55-8841.03; 8853.17 by Regierungspräsidium Karlsruhe (25.4.2016) and Corn Bunting conservation mandates of the Federal State Agencies of Tübingen, Stuttgart, Karlsruhe, and Freiburg.
Open access
