Petras & Vrezec - Modelling Pop Dyn From Non Standardized Ringing Data - Bird Study - 2022

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 15

BIRD STUDY

https://doi.org/10.1080/00063657.2022.2026876

Modelling population dynamics and trends in migratory birds from non-


standardized multi-species ringing data: the potential of multi-model selection
Tina Petrasa and Al Vrezec b,c

a
Institute of the Republic of Slovenia for Nature Conservation, Izola, Slovenia; bNational Institute of Biology, Ljubljana, Slovenia; cSlovenian
Museum of Natural History, Ljubljana, Slovenia

ABSTRACT ARTICLE HISTORY


Capsule: Long-term and non-standardized migratory bird ringing data can be used in models Received 17 August 2020
controlling variation in bird ringing methodology for reliable population trend estimations. Accepted 25 November 2021
Aims: Bird ringing data usually cover long periods and might reflect long-term population
changes. However, they are mainly derived during non-standardized multi-species catching at
numerous sites during the autumn migration period. We searched for the best modelling
approach to determine reliable species population dynamics and trend estimation models
based on annual multi-species bird ringing data.
Methods: We used ringing data from the Slovenian Bird Ringing Scheme and selected data in
three steps according to temporal, quantitative, and qualitative data selection. Annual indices
were constructed based on two types of denominators, ringing days, and ringing totals, vs. a
robust model without a denominator. We ran 20 candidate-generalized additive models
describing migrating population dynamics for 15 bird species by combining different data
selection approaches and denominators.
Results: We found that the models were species-specific, although the universal model could also
be applied to most species. We propose a general model construction approach for population
trend assessments from non-standardized bird ringing data. The estimates obtained by this
approach were comparable to the overall European population trends derived from breeding
survey data.
Conclusions: Bird ringing data from the autumn migration period are a valuable resource for
assessing continental scale population trends taking into account the whole population (non-
breeders and juveniles included) and even some rare and endangered species, but should be
conducted according to standard protocols to ensure reliable statistical inference of population
trends.

Population monitoring should constitute an essential ringing databases are another source of long-term
part of population ecology studies and species time series data, which could be used to estimate
conservation management, since change in abundance trends (Busse 1990, Bibby et al. 1992, Peach et al.
is one of the most common responses to changes in 1999, Rintala et al. 2003, Saurola & Francis 2004,
ecosystems (Acevedo-Whitehouse & Duffus 2009). Briedis & Keišs 2016). Compared to count data,
The duration of the monitoring period is critical for ringing data offer a number of potential benefits. First,
distinguishing long-term trends from short-term there are longer runs of historical data from ringing
fluctuations (Primack 2006). To detect significant schemes, which give the opportunity to analyse species
population trends in birds, continuous monitoring population dynamics over longer time periods. Most
should be carried out for 10 years or more (White 2019). breeding bird count schemes refer to the last 30 years,
In Europe, for example, bird population trends are with the first scheme started in 1966 (Heldbjerg et al.
currently assessed using data from a well-established 2019), while most of the ringing scheme datasets cover
network of national breeding bird counts, which periods of 100 or more years, with the first scheme
provide reliable regional and continental trend started in 1899 (Preuss 2001, Greenwood 2009, Møller
estimates (Gregory et al. 2007, Heldbjerg et al. 2019). & Fiedler 2010, Saurola et al. 2013, Bairlein et al.
However, alongside breeding bird count data, bird 2014). Second, ringing data cover bird populations on

CONTACT Al Vrezec al.vrezec@nib.si, al.vrezec@gmail.com


Supplemental data for this article can be accessed at https://doi.org/10.1080/00063657.2022.2026876.
© 2022 British Trust for Ornithology
2 T. PETRAS AND A. VREZEC

a larger, continental scale, while breeding bird count analyses that increase the reliability of statistical
data are limited to local or regional populations. inference from these data are of key importance in the
During autumn migration, when most of the ringing evaluation of population trends (Mason & Hussey
is conducted, birds aggregate from larger areas thus 1984, Rintala et al. 2003, Briedis & Keišs 2016).
enabling more comprehensive insight into continental Various methods have been suggested for the
scale population trends, including birds from remote standardization of heterogeneous bird ringing data;
areas with otherwise scarce ornithological activity for example, the comparison of single-species totals
(Spina & Volponi 2008). Third, different aspects of with other species or with other parameters that can
population data are captured in ringing data. While be estimated from available data (Ginn 1969), or the
breeding bird count data consider breeding pairs or identification of indicator species with similar
territorial males only, ringing data provide ecological traits, to control for variation in bird
information on the whole population, including ringing methodology (Perrins 2003, Rintala et al.
breeders, non-breeders, and immature individuals, an 2003, Keišs et al. 2007, Briedis & Keišs 2016).
aspect that is already well implemented in the In this paper, our aim was to combine different
Constant Effort Site ringing programme (Robinson approaches to account for bird ringing data variability
et al. 2009, Kampichler & van der Jeugd 2011). and thus determine the best species population
Finally, data on ringed migratory birds also include dynamics and trend estimation models based on
those for some rare and endangered species (e.g. annual multi-species bird ringing datasets. The bird
Bluethroat Luscinia svecica and Moustached Warbler ringing datasets were mainly derived during non-
Acrocephalus melanopogon), which are not included in selective and non-standardized multi-species catching
common bird censuses but are regularly ringed during with mist-nets at numerous sites during the autumn
autumn migration (Spina & Volponi 2008, Kralj et al. migration period. For trend estimation, therefore, it
2013, Valkama et al. 2014, European Bird Census was necessary to distinguish individual species ringing
Council 2017). data from non-specific multi-species ringing data. We
The initial aim of bird ringing was to track migration attempted to explore whether one general model could
patterns and routes of birds and not to monitor be used to assess population trends, or whether the
population changes, which make some ringing data modelling of bird ringing data must be species-
overall less standardized and potentially more biased. specific. The main aims of our study were: (1) to
The main assumption when using bird ringing data in reduce bias to the greatest extent possible in the
population trend analyses is that there is a close link assessment of ringing effort, (2) to estimate the long-
between the number of birds ringed annually and term trends of the migrating populations of the target
ringing effort (Hjort & Lindholm 1978). Thus, reliable species, (3) to compare the best models for assessing
and straightforward estimations of population trends population trends, and (4) to assess model trend
are possible when the methodology is standardized. estimates against known trends derived from breeding
Such standardizations are advancing as a part of bird count data.
ringing schemes in Europe, Asia, and North America
(Baillie et al. 1986, DeSante et al. 1995, Sokolov et al.
2000, Robinson et al. 2009). However, many bird Methods
ringing schemes in Europe have been operating
Target species and selection of bird ringing data
continuously for 100 years or more, and much of the
long-term datasets were obtained in a non- We used bird ringing data from the Slovenian Bird
standardized way, which hampers analyses of Ringing Scheme and selected 15 bird species that
population trends. The source of the variability in the sufficiently describe the variability in the bird ringing
data lies in variation in the number of ringers, dataset according to several ecological and
locations, catching methods, number, and type of nets methodological criteria, as well as differences in
and/or traps, the use of playback, weather conditions, sample sizes (Table 1). The Slovenian Bird Ringing
and habitats. These variables have their own trends Scheme is run by the Slovenian Museum of Natural
and this hinders the straightforward estimation of History, and the ringing database encompasses data
relevant population trends for the species concerned on ringed bird species that have been continuously
(Busse 1990, Sutherland 1996, Rintala et al. 2003). collected since 1927 (Božič 2009). In the current
Since annual changes in ringing effort are almost analysis, we included data from the 17-year period
impossible to estimate at a long-term historical level between 2000 and 2016, during which several field
(Hjort & Lindholm 1978, Rintala et al. 2003), indirect ringing stations were operating on a more or less
BIRD STUDY 3

Table 1. Ecological and methodological traits for 15 migratory bird species that describe the variability in the bird ringing dataset,
with the mean number of annually ringed birds at autumn migration in the period 2000–2016 (±sd). W: woodland, F: farmland,
M: marshland, U: urban. Status – RB: regular breeder, OB: rare occasional breeder, PM: passage migrant, RW: rare wintering,
S: habitat specialist at stopover sites (wetlands), G: habitat generalist at stopover sites, NA: no data available. Data references:
Cramp & Perrins (1983), Mukhin et al. (2005), Mukhin et al. (2008), European Bird Census Council (2017), and BirdLife International
(2020).
Number of birds Annual
ringed annually at Habitat Responding to European
autumn migration in specificity at playback of the Included in population
Slovenia (2000- Breeding Breeding Status in migration Garden Warbler night playback trend estimate
Species 2016) habitat period Slovenia stopover sites song vocalisations (2000–2016)
Tree Pipit Anthus 112 ± 41 W/F IV – VIII RB, PM G No No −0.6%
trivialis
European Pied 122 ± 49 W late IV - OB, PM G No No −1.4%
Flycatcher late VI
Ficedula
hypoleuca
Icterine Warbler 345 ± 175 W late V - VII OB, PM G No No −1.2%
Hippolais
icterina
Eurasian Wryneck 252 ± 72 W/F V – VI RB, PM G No No +0.2%
Jynx torquilla
Red-backed 224 ± 85 F V - VII RB, PM G No No −0.7%
Shrike Lanius
collurio
River Warbler 17 ± 10 W late V - RB, PM S NA No −2.7%
Locustella mid-VII
fluviatilis
Common 145 ± 65 F/M late IV - RB, PM S Yes No −0.2%
Grasshopper mid-VII
Warbler
Locustella
naevia
Common 181 ± 41 W/F late IV - RB, PM G NA No +0.3%
Nightingale mid VII
Luscinia
megarhynchos
Spotted 81 ± 27 W/U mid-V - RB, PM G No No −0.6%
Flycatcher mid-VIII
Muscicapa
striata
Common Redstart 62 ± 24 W/F late IV - RB, PM G No No +2.2%
Phoenicurus mid-VII
phoenicurus
Common 2207 ± 484 W IV - early RB, PM, G NA Yes +3.1%
Chiffchaff VIII RW (background)
Phylloscopus
collybita
Eurasian Blackcap 20728 ± 5608 W/F mid-IV - RB, PM, G Yes Yes +4.6%
Sylvia atricapilla VIII RW (background)
Garden Warbler 6972 ± 2131 W/F IV - VII RB, PM G Yes Yes (main voice) −1.2%
Sylvia borin
Common 472 ± 147 F IV - VII RB, PM G No No −1.2%
Whitethroat
Sylvia
communis
Song Thrush 322 ± 71 W mid-III (IV) RB, PM, G NA No +1.6%
Turdus - mid- RW
philomelos VIII

regular annual basis during the autumn migration migration, birds were caught, mostly with mist-nets,
period (Vrezec et al. 2014; Figure 1). There were 450 and information on ring number, species, age, sex,
active ringing locations during the study period, 86% wing length, mass, date and place of ringing, and
of which were active for less than 75% of the study name of ringer was collected. Since our study focused
period (1–12 years) and 14% of which were active for on the population dynamics of migrating birds in the
more than 75% of the study period (13–17 years). autumn, we excluded from the dataset all ringed
Among the latter group, 17 ringing stations were nestlings and data from the spring period (January to
active every year and were evenly distributed June). The use of playback of a recording of birdsong
throughout Slovenia (Figure 1). During autumn significantly increases the catch rate (Mukhin et al.
4 T. PETRAS AND A. VREZEC

Figure 1. Ringing stations that were operating during the autumn migration periods between 2000 and 2016 in Slovenia. Locations
that were active at least 75% of the time during the 17-year period are marked with black spots, while locations that were active less
than 75% of the time during the 17-year period are marked with white spots.

2008). In Slovenia, this method has been used since 1995 with at least six additional species vocalizing in the
(Šere 2001). Playback was usually used during mist- background, predominantly Common Chiffchaff
netting in the autumn migration period and broadcast Phylloscopus collybita and Blackcap Sylvia atricapilla
during the night from about 01:00 to one hour after (Table 1). The recording was taken at the Ljubljana
sunrise (B. Lapanja, pers. comm.). A recording of the Marsh in central Slovenia during the breeding season
song of the Garden Warbler Sylvia borin was used by Tomi Trilar (Slovenian Museum of Natural

Table 2. Set of a priori models for assessing the population dynamics of the autumn migratory populations of 15 bird species
considering different combinations of data selection approaches and denominators for calculating indices.
No. Model Autumn migration period Ringing effort Indicator species Denominator
Species- Selected days with
General specific defined minimum No Selected days No. of
autumn autumn No ringing of ringed birds or indicator with at least ringed birds No. of
migration migration effort ringed bird species one indicator without ringing
period period selection species selection species No nestlings days
1 GEN_NO_NO_N x x x x
2 GEN_NO_NO_B x x x x
3 GEN_NO_NO_D x x x x
4 GEN_NO_IND_B x x x x
5 GEN_NO_IND_D x x x x
6 GEN_SEL_NO_N x x x x
7 GEN_SEL_NO_B x x x x
8 GEN_SEL_NO_D x x x x
9 GEN_SEL_IND_B x x x x
10 GEN_SEL_IND_D x x x x
11 SPE_NO_NO_N x x x x
12 SPE_NO_NO_B x x x x
13 SPE_NO_NO_D x x x x
14 SPE_NO_IND_B x x x x
15 SPE_NO_IND_D x x x x
16 SPE_SEL_NO_N x x x x
17 SPE_SEL_NO_B x x x x
18 SPE_SEL_NO_D x x x x
19 SPE_SEL_IND_B x x x x
20 SPE_SEL_IND_D x x x x
BIRD STUDY 5

History). The playback was used consistently ringing data collected in a non-standard way is
throughout the study period, but bias occurred due to that there are no reliable data on target species
its non-systematic use with respect to the duration of samples vs. samples in which the target species
playback, combination with other recordings during was not likely to be caught and ringed. In
the daytime and uneven use between ringing stations. previous studies, species totals as a proportion of
There were approximately 110 ringing stations (24%) totals of ecological indicator species have been
that were using night playback, while the playback proposed to overcome this problem (Ginn 1969,
broadcast during the daytime was applied regularly at Perrins 2003, Rintala et al. 2003, Keišs et al.
all ringing stations. 2007). Our approach to account for unreliable
As ringing data were collected in a non-standardized data was to first select all species that occurred in
way, without clear reference to the protocol used (e.g. the days when the target species were ringed from
the number of mist nets, duration of ringing, weather 2000 to 2016 during the species-specific autumn
data, playback used or not), we selected the data in migration period in Slovenia. Among the selected
three steps according to the following criteria before species, we identified potential indicators as
conducting analyses (Table 2): species that co-occurred with the target species in
75% or more of the ringing days during the
(1) Temporal data selection. As our study focuses on autumn migration period. The Barn Swallow
the autumn migration period, we defined two Hirundo rustica was excluded from the potential
time periods: general and species-specific. The indicator species since it was known to be ringed
general autumn migration period encompasses the in large numbers at night roosts within the
whole autumn migration period and was framework of the EURING Swallow Project which
arbitrarily defined as the second half of the year, proceeded in Slovenia until 2013 (Vrezec et al.
from 15 July to 31 December. The species-specific 2014) and therefore could potentially obscure the
time period for the target species was determined trend direction of the target species. Further, we
from ringing data. In the analysis, only the modelled the presence/absence of the target
months in which ringed individuals of the target species (response variable) with the presence/
species constituted 5% or more of the total absence of the potential indicators (predictors) for
number of ringed birds in the autumn migration each day. When selecting indicator species, we
period (across all years from 2000 to 2016) were assumed that they behaved in a similar way to the
included in the core migration period for the target species during migration and were thus
target species in Slovenia. more likely to be caught together with the target
(2) Quantitative data selection (ringing effort selection). species (Bairlein 1983, Rintala et al. 2003, Briedis
Multi-species ringing data are composed of data & Keišs 2016). This approach also acknowledged
obtained by intensive mist-netting in the the inconsistent use of playback, which might
migration period as well as data which were not attract target species in a similar way as indicator
relevant for our analysis, such as those on species. We identified such species co-occurrence
occasional ringing of nestlings or found birds and patterns and possible species groupings with a
species-specific ringing. Therefore, we selected dendrogram created in the PAST software
ringing days according to the number of program (Hammer 2019), applying Ward’s
individual birds and the number of species ringed method, which combines clusters by minimizing
per day. To define thresholds for ringing days variance, measured by the sum of squares
relevant for migration monitoring, we plotted the (Murtagh & Legendre 2014, https://www.
number of ringed birds and ringed species per statisticshowto.com/wards-method/). Logistic and
day per location and then visually defined the stepwise regression (Thompson 1989, Royle et al.
point on the curve delineating intensive vs. non- 2012, Smith 2018) were performed in the R
intensive ringing days. We used the geom_vline software program using the package MASS (R
function from the package ggplot2 in the Development Core Team 2018). Data analyses
statistical program R (R Development Core Team were based on the assumption that birds were
2018), to add a vertical line representing the trapped and ringed randomly. The result was a
threshold between intensive vs. non-intensive guild of species which best indicated the
ringing days (Teetor 2011). probability of the presence of the target species.
(3) Qualitative data selection (target species sample To obtain a finer selection of the indicator species
selection). One of the shortcomings of using guild, we set the significance level at P < 0.01. A
6 T. PETRAS AND A. VREZEC

sampling day for the target species was defined as a the 95% confidence intervals are computed as the
day in which the target species had the potential to sample mean ± 1.96 ∗ se (Banjanovic & Osborne 2016).
be caught when ringing was carried out, i.e. a day in Additionally, we determined the universal model, i.e.
which the target species or at least one of the the model which fitted best across the widest range of
indicator species was caught. species, to be served as the best overall model for
assessing population trends from multi-species bird
ringing data on autumn migrants. We thereby selected
To standardize data characterized by variable ringing all species for which the best models were not
effort, we constructed annual indices based on two types rejected. For each model in each species, we calculated
of denominators: (i) ringing days and (ii) ringing totals the proportion of DE by the best model, giving the
for all birds without nestlings. In addition, we also value 1.0 to the best model. Furthermore, we
applied a robust model without a denominator, using calculated the mean proportion of DE for each model
only (iii) absolute numbers of ringed target species per across all species and defined the universal model as
year. We ran 20 candidate-generalized additive models the model with the highest mean proportion of DE,
(GAMs) describing migrating population dynamics for which indicates the overall best fit in the greatest
15 bird species by combining different data selection number of species.
approaches and denominators (Table 2). Since the GAM does not provide population trend
estimates, we applied fitted values from the GAM in a
general linear model (GLM) for estimating the
population growth rate, as the model coefficient gives
Population dynamics modelling
the predicted change in indices for one-unit increase
The deviance explained (DE), which is a statistic that is per year (Carey 2013).
identical to pseudo-R 2, was used for model comparison
and selection and was calculated with the following
equation (Wiersma & Skinner 2011): Results

DE = 1–(residual deviance/null deviance) Examined populations

Among the candidate models, the model with the According to data from the literature and recoveries
greatest percentage of explained deviance for each stored in the Slovenian Bird Ringing Centre database,
species was considered the best model. The selected the migratory directions of the target species followed
model was chosen for further analyses of population routes from Scandinavia southwards, and from Russia
dynamics. Non-significant models and those with and Finland towards the south-west (Figure 2). Some
explained deviance under 50% were rejected and were populations of Grasshopper Warbler Locustella naevia,
not used in further trend analyses. We fitted our data Common Chiffchaff, and Song Thrush Turdus
using a GAM, which results in non-linear, more philomelos migrate in a more eastern direction, from
complex trends (Simpson 2018). In addition, the fitted Russia and Belarus, while only the Red-backed Shrike
curve of the GAM allows for comparison of the Lanius collurio has a generally south-eastern migratory
dynamics between different models (species). As a direction (Cramp & Perrins 1983, Andreotti et al.
type of smoothing, we used a thin plate spline, which 1999, Wernham et al. 2002, Cepák et al. 2008, Spina &
is more general and especially useful for noisier data Volponi 2008, Božič 2009, Korner-Nievergelt et al.
(Craven & Wahba 1979, Hastie & Tibshirani 1986). 2012, Tøttrup et al. 2012, Kralj et al. 2013, Bairlein
For model validation, we used standard errors and et al. 2014, Busse et al. 2014, Ouwehand et al. 2015,
confidence intervals. We applied the bootstrap Pedersen et al. 2018, Belgian Ringing Scheme 2019).
technique, a statistical tool that can be used to
quantify the uncertainty associated with a given
Thresholds for data selection
estimator (Varian 2005, Banjanovic & Osborne 2016,
James et al. 2017). Here, we resampled the indices In the temporal data selection step, we defined the
from the best models, using 95% confidence intervals species-specific migration period for the target species,
with the recommended number (399) of bootstrap which was used in further analyses (online Table S1).
replicates for data similar to ours (Fewster et al. 2000). Furthermore, ringing effort selection was set to
We used normal bootstrap confidence intervals, where locations with a daily minimum of 50 ringed birds
the standard error (se) is computed as the standard and five ringed species (online Figure S1). The
deviation of the bootstrap distribution. Subsequently, number of indicator species that showed the best co-
BIRD STUDY 7

Figure 2. General migration routes of the target species through Slovenia. Abbreviations: ANTTRI: Tre Pipit Anthus trivialis, FICHYP:
European Pied Flycatcher Ficedula hypoleuca, HIPICT: Icterine Warbler Hippolais icterina, JYNTOR: Wryneck Jynx torquilla, LANCOL: Red-
backed Shrike Lanius collurio, LOCFLU: River Warbler Locustella fluviatilis, LOCNAE: Grasshopper Warbler Locustella naevia, LUSMEG:
Common Nightingale Luscinia megarhynchos, MUSSTR: Spotted Flycatcher Muscicapa striata, PHOPHO: Common Redstart
Phoenicurus phoenicurus, PHYCOL: Common Chffchaff Phylloscopus collybita, SYLATR: Blackcap Sylvia atricapilla, SYLBOR: Garden
Warbler Sylvia borin, SYLCOM: Common Whitethroat Sylvia communis, TURPHI: Song Thrush Turdus philomelos.

occurrence with the target species after performing Jynx torquilla, and Common Nightingale Luscinia
logistic and stepwise regression varied between one megarhynchos; (2) Pied Flycatcher Ficedula hypoleuca,
(for River Warbler Locustella fluviatilis) and eight (for Red-backed Shrike, Garden Warbler, Icterine Warbler,
Icterine Warbler Hippolais icterina) with a median of and Spotted Flycatcher Muscicapa striata; (3)
five, and were specified as indicators (Table S1). Common Chiffchaff and Song Thrush; and (4) Tree
Hierarchical clustering distinctly separated four Pipit Anthus trivialis, River Warbler, Common
groups of species according to similarity of indicator Redstart Phoenicurus phoenicurus, Blackcap, and
species (Figure 3): (1) Grasshopper Warbler, Wryneck Common Whitethroat Sylvia communis. The first level
of selection was done according to migration period
and the second-level selection according to
preferences of habitat and migratory direction.

Population dynamics model selection


For each species, we selected the best GAM with the
greatest percentage of explained deviance and thus the
highest explanatory power (Table 3; online Table S2).
However, not all of the best models were actually
significant or explained at least 50% of the deviance.
Quantitative data selection and different denominators
primarily contributed to the division between accepted
and rejected best models, while the other types of data
selection (temporal and qualitative data selection) did
Figure 3. Ward’s hierarchical clustering of analysed species not have such power.
(guilds) according to common indicator species. See Figure 2 The best models for the Tree Pipit and Blackcap did
for species codes. not have denominators and were non-significant (Table
8 T. PETRAS AND A. VREZEC

Table 3. Best and universal general additive models of population dynamics for the 15 target bird species. Estimated annual growth
rate (trend) was assessed in a GLM. Significant codes: ∗∗∗ P < 0.001, ∗∗ P < 0.01, ∗ P < 0.05. Abbreviations: UM: universal model (the
general autumn migration period from 15 July to 31 December, selected days with ≥50 ringed birds and ≥5 ringed species, no
indicator species selection, number of ringing days as denominator), CI: confidence intervals, DE: deviance explained.
Parametric coefficients
Approximate Estimated
Best GAM model/ Lower Upper significance of DE Model annual growth
Species UM Estimate CI CI smooth terms (F) (%) acceptance rate (GLM) Trend
Tree Pipit GEN_SEL_NO_N 80.765 63.940 96.920 2.8 15.9 Rejected Uncertain
European Pied Flycatcher GEN_SEL_IND_D 0.427 0.361 0.492 3.6∗ 77.6 Accepted −0.012∗∗∗ Decline
UM 0.216 0.179 0.257 3.5∗ 66.3 Accepted −0.022∗∗∗ Decline
Icterine Warbler SPE_SEL_NO_B 0.006 0.005 0.008 3.4∗ 75.0 Accepted −0.042∗∗∗ Decline
UM 0.669 0.518 0.822 8.1∗∗ 47.5 Rejected Uncertain
Eurasian Wryneck SPE_SEL_NO_B 0.005 0.004 0.005 2.1 70.1 Rejected Uncertain
∗ ∗∗∗
Red-backed Shrike SPE_SEL_NO_B 0.003 0.003 0.004 4.5 84.9 Accepted −0.019 Decline
UM 0.368 0.307 0.429 7.9∗∗ 74.7 Accepted −0.052∗∗∗ Decline
River Warbler GEN_NO_NO_D 0.094 0.068 0.119 10.8∗∗ 62.0 Accepted −0.090∗∗∗ Decline
UM 0.034 0.025 0.043 7.4∗∗ 54.9 Accepted −0.082∗∗∗ Decline
Common Grasshopper Warbler GEN_SEL_NO_D 0.301 0.247 0.359 8.9∗∗ 56.5 Accepted −0.059∗∗∗ Decline
UM 0.301 0.247 0.359 8.9∗∗ 56.5 Accepted −0.059∗∗∗ Decline
Common Nightingale GEN_SEL_NO_B 0.002 0.002 0.003 9.8∗∗∗ 77.4 Accepted +0.054∗∗∗ Increase
UM 0.351 0.312 0.389 5.0∗ 64.6 Accepted +0.028∗∗∗ Increase
Spotted Flycatcher GEN_NO_IND_D 0.934 0.807 1.054 12.5∗∗∗ 80.3 Accepted −0.013∗∗∗ Decline
UM 0.146 0.124 0.166 6.5∗∗ 70.2 Accepted −0.023∗∗∗ Decline
Common Redstart SPE_SEL_IND_B 0.108 0.085 0.131 3.6∗ 56.6 Accepted +0.041∗∗∗ Increase
UM 0.110 0.088 0.132 0.5 3.1 Rejected Uncertain
Common Chiffchaff SPE_NO_NO_B 0.030 0.025 0.035 6.2∗∗ 78.2 Accepted +0.041∗∗∗ Increase
UM 3.912 3.538 4.270 3.7∗ 55.3 Accepted +0.009 Stable
Eurasian Blackcap SPE_SEL_NO_N 17711 15106 20337 2.7 33.3 Rejected Uncertain
∗∗∗
Garden Warbler SPE_NO_IND_B 0.904 0.890 0.917 24.5 96.6 Accepted −0.001 Stable
UM 14.39 12.50 16.36 9.1∗∗ 91.0 Accepted −0.039∗∗∗ Decline
Common Whitethroat SPE_SEL_IND_B 0.042 0.037 0.048 9.0∗∗∗ 90.2 Accepted −0.003 Stable
UM 0.930 0.783 1.067 17.1 ∗∗∗ 80.4 Accepted −0.050∗∗∗ Decline
Song Thrush SPE_NO_NO_B 0.004 0.003 0.005 5.2∗ 25.7 Rejected Uncertain

3). The best model for the Wryneck was also non- according to qualitative (indicator species) and
significant, while the model for the Song Thrush was quantitative criteria (ringing effort) also contributed to
significant but had low explained deviance (26%). Due model improvement.
to low explanatory power, all four of these species Model 8 (the general autumn migration period from
were excluded from further analyses since the current 15 July to 31 December, selected days with ≥50 ringed
dataset, or our set of a priori models, did not birds and ≥5 ringed species, no indicator species
sufficiently account for sampling biases in these four selection and with number of ringing days as
species. denominator) fitted best across the widest range of
The best estimated models were in most cases those species with an average proportion of DE compared to
in which the denominators were the number of ringed the best model being 0.78, which was determined as a
birds and in which migration time was specified by universal model. The second (model 10) and third
the species-specific period. All of these models were (model 18) best candidate universal models had 0.74
characterized by high explained deviance, which and 0.72 of DE, respectively.
mostly varied between 75% and 97% (Table 3).
Models with ringing days as denominators were
Estimation of population trends
determined by the general migration period and had
56–80% of explained deviance, but only in species Modelling of population dynamics was conducted only
with lower mean annual ringing totals (less than 150 for nine out of 15 species with accepted models, i.e.
ringed birds per year; Table 3). Data selection significant models with adequate explanatory power
BIRD STUDY 9

(˃50% explained deviance) and for which it was possible Model selection
to determine acceptable universal models (Table 3,
We have demonstrated that the accepted best models
Figure 4). In all of these species the universal model
of population dynamics in the studied migratory birds
showed the same direction of population trend, but in
derived from non-standardized long-term bird ringing
three species (Common Chiffchaff, Garden Warbler,
data significantly improve sampling effort control.
Common Whitethroat) the significance of the trend
Most of the species analysed are long-distance
was changed (Table 3). The universal models for the
migrants, with a migratory peak in August.
Common Nightingale showed an increasing trend
Presumably, the night playback of calls influenced
with estimated annual growth rate of 2.8%. In the
the capture rates of certain species, especially Sylvia
majority of species (European Pied Flycatcher, Red-
warblers and the Common Chiffchaff (Mukhin et al.
backed Shrike, River Warbler, Grasshopper Warbler,
2008; Table 1), and attracted species even to
Spotted Flycatcher, Garden Warbler, and Common
suboptimal habitats (Herremans 1990, Mukhin et al.
Whitethroat) there was a declining trend, with an
2008). Nevertheless, playback is often used during
annual rate of change ranging from −2.2% to −8.2%.
bird ringing sessions in autumn, as it allows larger
The trend for the Common Chiffchaff was stable
sample sizes to be obtained, but such data might be
(Table 3). With respect to population dynamics, the
biased in several ways (Mukhin et al. 2008, Arizaga
Pied Flycatcher, Spotted Flycatcher, and Garden
et al. 2015, De La Hera et al. 2017). Therefore,
Warbler showed patterns with greater fluctuations
models of species that respond to playback are
(Figure 4).
expected to have more error and presumably
In comparison with the general estimated population
reduced explanatory power.
trends in Europe based on breeding survey data for the
Four of the models were rejected (Tree Pipit,
same time period (Table 1), the estimated trends from
Blackcap, Wryneck, and Song Thrush). Within these
ringing data matched the direction in all accepted
models, only one species that responds to acoustic
models, when considering best models as well as a
signals was included (Blackcap), so response to the
universal model. However, the magnitude of trend
playback per se was not the key bias for statistical
estimates was found to be generally higher in ringing
inference. The rejected models were primarily those
than in count data.
with no denominators. They showed low explained
deviance and insufficient model fitting. Unlike the
Discussion other non-significant models, the model for the
Wryneck had ringed birds in the denominators and a
We have shown that non-standardized multi-species
high proportion of explained deviance. Further
bird ringing can provide useful long-term data for
modelling, taking into account other aspects of
statistically reliable population trend estimation. The
variability in the data, would therefore be required in
best models were found to be species-specific, but
future studies to search for other possibilities in
we have found that an universal model with the
controlling biases in non-standardized ringing data.
best fit in the greatest number of studied species
Most of the models were accepted and showed that
provided similar and well supported trend estimates
the selections proposed in the study resulted in good
for most of the species. Generally, data derived from
model fits. The most powerful models were those that
variable ringing activity are too heterogenous to
considered selected data according to quantitative
allow for a uniform modelling procedure; therefore,
(ringing effort) data selection. Through quantitative
our approach was to construct 20 a priori models
selection we essentially removed incidental findings or
for model selection, which allowed us to derive
occasional ringing sessions and thus gained more
reliable population trend estimates comparable to
homogenous samples, with mist-netting as the
overall European population trends derived from
prevailing method. This type of selection provided the
breeding survey data (European Bird Census Council
best fit in 64% of the accepted models.
2017). Although abundance estimates based on non-
The best estimated models with qualitative data
standardized methods have largely been avoided in
selection alone were found only in two species.
the past (Pollock et al. 2002), the recent
Qualitative data selection, alone or together with other
development of more sophisticated statistical
selections, provided the best fit in 45% of the accepted
modelling helps to control for different sources of
models and hence contributed to the power of models.
variability in data (Walker & Taylor 2017) and thus
The use of indicators has already been shown to be a
provides more powerful and reliable statistical
beneficial approach (Rintala et al. 2003). Our study is
inference.
10 T. PETRAS AND A. VREZEC

Figure 4. Population dynamics of nine bird species with standard errors and partial deviance residuals according to the universal
model.

based on the assumption that particular species groups at the first level of differentiation (Figure 3),
congregate in specific habitats at stopover sites, which the main separation parameter was the migration
is a well-known migratory pattern that is associated period, with the first group formed by species
with food requirements (Bairlein 1983). Considering migrating from July to September and the second
such species guilds, we defined potential sampling group formed by species migrating from August to
(ringing) days for the target species according to the October. Such separation was expected, as we chose
presence of indicator species. indicator species in the species-specific migration
Differentiation between groups of indicator species periods. In each first level species group, there were an
was based mainly on the migration period. In two additional two groups representing a second level of
BIRD STUDY 11

differentiation (altogether four) separating species by the actual population dynamics of the species. In such
habitat preferences and migratory direction. One was cases, modelling is not possible.
comprised of species with a south-western autumn
migratory direction and occurring predominantly in
Population dynamics and trends
farmland habitats: the Grasshopper Warbler, Wryneck,
and Common Nightingale (Cramp & Perrins 1983, Our results of long-term trend analyses using bird
Cepák et al. 2008, Spina & Volponi 2008). The second ringing data were in agreement with the overall
group included two species, the Common Chiffchaff directions of European trends derived from breeding
and Song Thrush, which are woodland species with a surveys (European Bird Census Council 2017).
western autumn migratory direction (Cramp & Perrins However, the magnitude of the observed deviations
1983, Andreotti et al. 1999, Busse et al. 2014). from European trends was most probably derived from
The models in which the numbers of ringed birds were the fact that ringing data addressed different aspects of
considered as the denominator typically showed higher a population than did breeding count data. The latter
explained deviance. Such models were more powerful, consider only breeding pairs or territorial males, while
as they were better at controlling for sampling effort ringing data refer to the whole population, including
variability than those denominated by ringing days. For immature individuals, and thus also reflect reproductive
instance, differences in the time and duration of the information for the population from the current year.
catch, the use of a different number of nets, the use of Additionally, the differences in trends could be
playback or methods other than mist-netting might related to the low accuracy of overall European trend
result in different numbers of birds caught per ringing estimates of some species, resulting from the rough
day. In such cases, ringing days do not adequately methodology of combining the results of existing
control for the variable sampling effort, especially for national schemes (van Strien et al. 2001). Both
more abundant species. In contrast, for species which datasets obtained from the large scale pan-European
were ringed in lower daily numbers, ringing days better breeding survey and from autumn bird ringing
account for changes in ringing activity since ringing activity include many bird populations. In this regard,
days remained more stable throughout the period than the trend estimates based on breeding surveys are
the number of ringed birds. On the contrary, the large assessed artificially, by a compilation of trends from
number of species, i.e. nine of 11 species with accepted different countries (van Strien et al. 2001). In contrast,
best models, could be modelled by a universal model bird populations from bird ringing data are combined
with ringing days as the denominator. The trends, naturally, resulting in assessment of the actual overall
obtained by the best and universal models, showed the trend of the European population, especially if the
same direction, with slight differences in estimated breeding range of autumn migrating populations is
annual growth rate between them. known for a certain area. Further studies would be
All models with denominators provided good needed to compare trends assessed from autumn bird
improvement in model fit for the majority of species. ringing data for selected species from several countries
This is in contrast to some previous studies, where to verify the coherence of the ringing data and their
data obtained in a non-standardized way with consistency with European census data.
multiple sources of bias were better suited to robust When comparing trends obtained from breeding and
modelling without denominators (Walker & Taylor migratory population surveys we should consider
2017, Boersch-Supan et al. 2019). In such instances, important distinctions between them: (i) data on
the sample size may play an important role. The large migratory populations are gathered from a wider
dataset obtained from multiple locations could geographical area and (ii) in the autumn migratory
obscure the biases and heterogeneity in the samples population, breeders, and non-breeders are combined,
(Kaplan et al. 2014) and may increase the precision in each of which might show its own trend. The
trend estimates (Gregory et al. 2004, Fink et al. 2020). generally accepted consensus that trend estimates
With respect to the rejected models, further using bird ringing data coincide with trends obtained
improvement of the models should be made to obtain by other methods (Berthold 2004) is especially true
more powerful species-specific models of population when comparing trends from the autumn migration
trends from long-term bird ringing data. It is period. At that time of the year, birds show more
important to note that it is very likely that estimating flexible behaviour in selecting habitats (Mason 1979,
population trends from past bird ringing data are not Moore et al. 1995, Parrish 2000) and are therefore
possible for all species due to a species-specific catch more likely to be caught. Moreover, such positive
rate, which might be too accidental and unrelated to correlations in trends obtained by different bird
12 T. PETRAS AND A. VREZEC

censuses and survey techniques are especially common species ringing using mist-nets only, but in the future
in those species that exhibit the same population trend assessments from multi-species historical data
trends over a large area (Berthold 2004). We argue based on other catching methods should be tested in
that ringing data on migrants can provide more order to obtain patterns of population dynamics over
accurate and representative trend estimates at the longer time periods that might reveal population
continental scale, but have low power in assessing statuses of bird species over 100 or more years.
population change at the local or regional scale
compared to the breeding population censuses.
In conclusion, when variability in ringing effort is
Acknowledgements
taken into account, long-term bird ringing data are a We are particularly grateful to the volunteer ringers from the
valuable source of information for describing patterns Slovenian Bird Ringing Centre (Slovenian Museum of
Natural History) who provided data on birds in the period
of long-term population dynamics. Firstly, there is great
2000–2016: Dušan Belingar, Franc Bolta, Darjo Bon, Ivo
potential for performing historical analyses when other Božič, Franc Bračko, Igor Brajnik, Jože Bricelj, ml., Jože
sources of data on birds are not available. Secondly, the Bricelj, st., Alfonz Colnar, Stane Černalogar, Marjan Debelič,
approach could serve as a beneficial supplement to Dušan Dimnik, Jože Dolinšek, Stanko Drašček, Dare Fekonja,
other bird monitoring programmes (Rintala et al. 2003). Jernej Figelj, Jože Geiser, Marjan Gobec, Jože Gračner, Dejan
For example, the monitoring of local breeding birds Grohar, Peter Grošelj, Vojko Havliček, Ludvik Jakopin,
Marko Jankovič, Tone Jankovič, Leon Kebe, Milovan Keber,
could assess population dynamics on a limited number Brane Koren, Stane Kos, Jelko Kozjak, Brane Lapanja, Ivan
of sites at a local or regional scale, while monitoring in Lipar, Anton Lisec, Zvonko Lončarevič, Tone Macele, Tomaž
the migration period could provide population data Mihelič, Jurij Mikuletič, Jože Nered, Žan Pečar, Miro
over extensive geographical areas, especially if ringing Perušek, Dušan Petkovšek, Rajko Piciga, Zdravko Podhraški,
recoveries reveal the geographic origin of the migrating Dušan Pogačar, Milan Pustoslemšek, Aljaž Rijavec, Miran
Romšak, Luka Simončič, Branko Slabanja, Andrej Sovinc,
population. In future studies we should combine trend
Željko Šalamun, Dare Šere, Iztok Škornik, Pavle Štirn, Vlado
estimates of autumn migratory populations across Štolfa, Polde Štricelj, Rudolf Tekavčič, Tomi Trilar, Andrej
Europe to fully understand the dynamics of migratory Trontelj, Miro Vamberger, Lojza Vesel, Bogdan Vidic, Milan
birds and their cross-continental connectivity. Vogrin, Iztok Vreš, Davorin Vrhovnik, Eva Vukelič, and Ivan
The main issue in trend assessments derived from Zlobko. Special thanks go to Dare Šere, who coordinated bird
ringing data is variable ringing effort, which produces ringing in Slovenia from 1987 to 2011, and to Dare Fekonja,
who has coordinated bird ringing in Slovenia since 2012. We
data samples that are difficult to compare. Within are very grateful to Przemysław Busse for helpful discussions
ringing sites, ringing effort, and sampling about the evaluation of bird ringing data. Dare Šere, Tomi
methodology should be as consistent as possible, Trilar, and Brane Lapanja kindly provided useful information
particularly in regard to the length of mist-nets used, on bird ringing methodology. We would like to thank Andrej
their position, time of netting, and the use of playback Kapla for map preparation. We acknowledge using data
source on European trends of common birds by EBCC/
calls. The use of playback can significantly increase the
BirdLife/RSPB/CSO.
catching rate of some target species (Mukhin et al.
2008) and might greatly influence the unexplained
variability in ringing data if not used consistently, Funding
despite it resulting in larger samples. When playback Bird ringing in Slovenia is financially supported by the
is used, for example in monitoring of specific target Ministry of Culture of the Republic of Slovenia. The study is
species, a strict protocol should be developed, a part of TP’s PhD work, which is financially supported by
including type of recording, broadcast time, duration, the Ministry of Education, Science and Sport. AV was
and loudness, to assure its consistent use and to avoid supported by the Ministry of Culture of the Republic of
Slovenia and research core funding No. P1-0255 by the
data biases. On the other hand, for assessing trends
Slovenian Research Agency.
from long-term historical multi-species ringing data,
we recommend using a universal model approach
with data obtained from the whole autumn migration ORCID
period by selecting the most intensive ringing days Al Vrezec http://orcid.org/0000-0002-4699-6451
(i.e. ≥50 ringed birds and ≥5 ringed species) and by
using ringing days as denominator. However, attention
should be given to the significance of the trend References
estimates, since in some species biases in the data Acevedo-Whitehouse, K. & Duffus, A.L. 2009. Effects of
cannot be accounted for with this approach, as shown environmental change on wildlife health. Philos.
in this study. Our models were derived from multi- Trans. R. Soc. Lond. B Biol. Sci. 364: 3429–3438.
BIRD STUDY 13

Andreotti, A., Bendini, L., Piacentini, D. & Spina, F. 1999. De La Hera, I., Fontanilles, P., Delalande, L., Glad, A. &
The role of Italy within the Song Thrush (Turdus Sarraude, T. 2017. Attraction of other species by
philomelos) migratory system analysed on the basis of Bluethroat Luscinia svecica song playback during autumn
ringing-recovery data. Vogelwarte 40: 28–51. migration: an experimental test using bird ringing data.
Arizaga, J., Musseau, R., Laso, M., Esparza, X., Ardeola 64: 5–13.
Unamuno, E., Azkona, A. & Fontanilles, P. 2015. DeSante, D.F., Burton, K.M., Saracco, J.F. & Walker, B.L.
Biases associated with the use of a playback in 1995. Productivity indices and survival rate estimates
stopover ecology studies of Bluethroats Luscinia svecica. from MAPS, a continent-wide programme of constant-
Bird Study 62: 280–284. effort mist netting in North America. J. Appl. Stat. 22:
Baillie, S.R., Green, R.E., Boddy, M. & Buckland, S.T. 1986. 935–947.
An evaluation of the Constant Effort Sites Scheme. BTO European Bird Census Council. 2017. https://www.ebcc.
Research Report, 21. BTO Thetford, Norfolk, UK. info/trends-of-common-birds-in-europe-2017-update/.
Bairlein, F. 1983. Habitat selection and associations of species Assessed 7 February 2020.
in European passerine birds during southward, post- Fewster, R.M., Buckland, S.T., Siriwardena, G.M., Baillie,
breeding migrations. Ornis. Scand. 14: 239–245. S.R. & Wilson, J.D. 2000. Analysis of population trends
Bairlein, F., Dierschke, J., Dierschke, V., Salewski, V., for farmland birds using generalized additive models.
Geiter, O., Hüppop, K., Köppen, U. & Fiedsler, W. Ecology 81: 1970–1984.
2014. Atlas des Vogelzugs: Ringfunde deutscher Brut- und Fink, D., Auer, T., Johnston, A., Ruiz-Gutierrez, V.,
Gastvögel. AULA-Verlag, Wiebelsheim. Hochachka, W.M. & Kelling, S. 2020. Modelling avian
Banjanovic, E.S. & Osborne, J.W. 2016. Confidence intervals full annual cycle distribution and population trends with
for effect sizes: applying bootstrap resampling. Pract. citizen science data. Ecol. Appl. 30: e02056.
Assess. Res. Eval. 21: 1–20. Ginn, H.B. 1969. The use of annual ringing and nest record
Belgian Ringing Scheme. 2019. Scientific web sites and card totals as indicators of bird population levels. Bird
applications, Royal Belgian Institute of Natural Science. Study 16: 210–248.
http://odnature.naturalsciences.be/bebirds/en/ring- Greenwood, J.J.D. 2009. 100 years of ringing in Britain and
recoveries. Assessed 12 January 2019. Ireland. Ring. Migr. 24: 147–153.
Berthold, P. 2004. Use of mist nets for monitoring landbird Gregory, R.D., Gibbons, D.W. & Donald, P.F. 2004. Bird
autumn population trends, and comparison with other census and survey techniques. In Sutherland, W.J.,
methods. Stud. Avian Biol. 29: 112–115. Newton, I. & Green, R. (ed) Bird Ecology and
Bibby, C.J., Burgess, N.D. & Hill, D.A. 1992. Bird Census Conservation: a handbook of techniques, 17–55. Oxford
Techniques. Academic Press Limited, London. University Press, Oxford.
BirdLife International. 2020. IUCN Red List for birds. http:// Gregory, R.D., Voříšek, P., van Strien, A., Gmelig Meyling,
www.birdlife.org. Assessed 7 February 2020. A.W., Jiguet, F., Fornasari, L., Reif, J., Chylarecki, P. &
Boersch-Supan, P.H., Trask, A.E. & Baillie, S.R. 2019. Burfield, I. 2007. Population trends of widespread
Robustness of simple avian population trend models for woodland birds in Europe. Ibis 149: 78–97.
semi-structured citizen science data is species-dependent. Hastie, T. & Tibshirani, R. 1986. Generalized additive
Biol. Conserv. 240: 1–27. models. Stat. Sci. 1: 297–310.
Božič, I.A. 2009. Rezultati obročkanja ptičev v Sloveniji: Heldbjerg, H., Fox, A.D., Lehikoinen, A., Sunde, P.,
1926-1982. Scopolia 4: 23–110. Aunins, A., Balmer, D.E., Calvi, G., Chodkiewicz, T.,
Briedis, M. & Keišs, O. 2016. Extracting historical population Chylarecki, P., Escandell, V., Foppen, R., Gamero, A.,
trends using archival ringing data – an example: the Hristov, I., Husby, M., Jiguet, F., Kmecl, P., Kålås, J.A.,
globally threatened Aquatic Warbler. J. Ornithol. 157: Lewis, L.J., Lindström, Å, Moshøj, C., Nellis, R.,
419–425. Paquet, J.Y., Portolou, D., Ridzoò, J., Schmid, H.,
Busse, P. 1990. Studies of long-term population dynamics Skorpilová, J., Szabó, Z.D., Szép, T., Teufelbauer, N.,
based on ringing data. Ring 13: 221–234. Trautmann, S., van Turnhout, C., Vermouzek, Z.,
Busse, P., Zaniewicz, G. & Cofta, T. 2014. Evolution of the Voøíšek, P. & Weiserbs, A. 2019. Contrasting population
Western Palaearctic passerine migration pattern trends of Common starlings (Sturnus vulgaris) across
presentation style. Ring 36: 3–21. Europe. Ornis Fenn. 96: 153–168.
Carey, G. 2013. The general linear model (GLM): a gentle Herremans, M. 1990. Body-moult and migration overlap in
introduction. In Carey, G. (ed) Quantitative Methods in Reed Warblers (Acrocephalus scirpaceus) trapped during
Neuroscience, 129–142. Department of Psychology and nocturnal migration. Le Gerfaut 80: 149–158.
Neuroscience, Institute for Behavioral Genetics, Hjort, C. & Lindholm, C.-G. 1978. Annual bird ringing totals
University of Colorado, Boulder. and population fluctuations. Current Bird Migration
Cepák, J., Klvaňa, P., Škopek, J., Schröpfer, L., Jelínek, M., Research. Proceedings of a Symposium at Falsterbo,
Hořák, D., Formánek, J. & Zárybnický, J. 2008. Atlas Sweden, 3–8 October, 1977. Oikos 30: 387–392.
migrace ptáků České a Slovenské republiky. Aventinum, James, G., Witten, D., Hastie, T. & Tibshirani, R. 2017. An
Praha. Introduction to Statistical Learning: with applications in R.
Cramp, S. & Perrins, C.M. 1983. Handbook of the Birds of Springer, New York.
Europe, the Middle East and North Africa. The Birds of Kampichler, C. & van der Jeugd, H.P. 2011. Monitoring
the Western Palearctic. Oxford University Press, Oxford. passerine reproduction by constant effort ringing:
Craven, P. & Wahba, G. 1979. Smoothing noisy data with evaluation of the efficiency of trend detection. Ardea 99:
spline functions. Numer. Math. 31: 377–403. 129–136.
14 T. PETRAS AND A. VREZEC

Kaplan, R.M., Chambers, D.A. & Glasgow, R.E. 2014. Big Pedersen, L., Jackson, K., Thorup, K. & Tøttrup, A.P. 2018.
data and large sample size: a cautionary note on the Full-year tracking suggests endogenous control of
potential for bias. Clin. Transl. Sci. 7: 342–346. migration timing in a long-distance migratory songbird.
Keišs, O., Granāts, J. & Mednis, A. 2007. Estimated Behav. Ecol. Sociobiol. 72: 1–10.
population dynamics of the Corncrake Crex crex in Latvia Perrins, C. 2003. The status of Marsh and Willow Tits in the
and Europe in the 20th century by ringing data analysis. UK. Br. Birds 96: 418–442.
Acta Univ. Latv. 723: 71–97. Pollock, K.H., Nichols, J.D., Simons, T.R., Farnsworth,
Korner-Nievergelt, F., Jenni, L., Tøttrup, A.P. & Pasinelli, G.L., Bailey, L. & Sauer, J.R. 2002. Large scale wildlife
G. 2012. Departure directions, migratory timing and monitoring studies: statistical methods for design and
non-breeding distribution of the Red-backed Shrike analysis. Environmetrics 13: 105–119.
Lanius collurio: do ring re-encounters and light-based Preuss, N.O. 2001. Hans Christian Cornelius Mortensen:
geolocator data tell the same story? Ring. Migr. 27: aspects of his life and of the history of bird ringing.
83–93. Ardea 89: 1–6.
Kralj, J., Barišić, S., Tutiš, V. & Ćiković, D., eds. 2013. Atlas Primack, R.B. 2006. Essentials of Conservation Biology.
selidbe ptice Hrvatske. Hrvatska akademija znanosti i Sinauer Associates, Inc., Publishers, Sunderland, MA.
umjetnosti, Zagreb. R Development Core Team. 2018. R: A Language and
Mason, W. 1979. Habitat selection by the Parulidae during Environment for Statistical Computing. R Foundation for
spring migration along the South Fork Creek in Glasgow, Statistical Computing, Vienna, Austria.
KY. Kentucky Warbler 55: 39–42. Rintala, J., Tiainen, J. & Pakkala, T. 2003. Population trends of
Mason, C.F. & Hussey, A. 1984. Bird population trends as the Finnish Starling Sturnus vulgaris, 1952–1998, as inferred
shown by chick ringing data. Ringing & Migration 5: from annual ringing totals. Ann. Zool. Fenn. 40: 365–385.
113–120. Robinson, R.A., Julliard, R. & Saracco, J.F. 2009. Constant
Moore, F.R., Gauthreaux, S.A., Kerlinger, P. & Simons, effort: studying avian population processes using
T.R. 1995. Habitat requirements during migration: standardized ringing. Ring. Migr. 24: 199–204.
important link in conservation. In Martin, T.E. & Finch, Royle, J.A., Chandler, R.B., Yackulic, C. & Nichols, J.D.
D.M. (ed) Ecology and Management of Neotropical 2012. Likelihood analysis of species occurrence
Migratory Birds: a synthesis and review, 121–144. Oxford probability from presence-only data for modelling species
University Press, Oxford. distributions. Methods Ecol. Evol. 3: 545–554.
Møller, A.P. & Fiedler, W. 2010. Long-term time series of Saurola, P. & Francis, C.M. 2004. Estimating population
ornithological data. In Møller, A.P., Fiedler, W. & dynamics and dispersal distances of owls from nationally
Berthold, P. (ed) Effects of Climate Change on Birds, 33– coordinated ringing data in Finland. Anim. Biodiv.
38. Oxford University Press, Oxford. Conserv. 27: 403–415.
Mukhin, A., Chernetsov, N. & Kishkinev, D. 2008. Acoustic Saurola, P., Valkama, J. & Velmala, W. 2013. The Finnish
information as a distant cue for habitat recognition by Bird Ringing Atlas Vol. I. Finnish Museum of Natural
nocturnally migrating passerines during landfall. Behav. History and Ministry of Environment, Helsinki.
Ecol. 19: 716–723. Šere, D. 2001. 75 years of the bird ringing scheme Ljubljana.
Mukhin, A.L., Chernetsov, N.S. & Kishkinev, D.A. 2005. EURING Newsletter 3: 70–71.
Reed Warbler, Acrocephalus scirpaceus (Aves, Simpson, G.L. 2018. Modelling palaeoecological time series
Sylviidae), song as an acoustic marker of wetland using generalised additive models. Front. Ecol. Evol. 6: 1–21.
biotope during migration. Zool. Zhurnal. 84: 995–1002. Smith, G. 2018. Step away from stepwise. J. Big. Data. 5: 1–12.
(in Russian). Sokolov, L.V., Yefremov, V.D., Markovets, M.Y., Shapoval,
Murtagh, F. & Legendre, P. 2014. Ward’s Hierarchical A.P. & Shumakov, M.E. 2000. Monitoring of numbers of
clustering method: clustering criterion and agglomerative passage populations of passerines over 42 years (1958-
algorithm. J. Classif. 31: 274–295. 1999) on the Courish Spit of the Baltic Sea. Avian Ecol.
Obročkovalska baza in arhiv podatkov. 2016. Slovenski Behav. 4: 31–53.
center za obročkanje ptičev, Prirodoslovni muzej Spina, F. & Volponi, S. 2008. Atlante della Migrazione degli
Slovenije, Ljubljana, Slovenia. Uccelli in Italia. Volume II: Passeriformi. Ministero
Ouwehand, J., Ahola, M.P., Ausems, A.N.M.A., Bridge, dell’Ambiente e della Tutela del Territorio e del Mare,
E.S., Burgess, M., Hahn, S., Hewson, C.M., Instituto Superiore per la Protezione e la Ricerca
Klaassen, R.H., Laaksonen, T., Lampe, H.M., Ambientale (ISPRA). Tipografia SCR, Roma, Italy.
Velmala, W. & Both, C. 2015. Light-level geolocators Sutherland, W.J. 1996. Ecological Census Techniques: a
reveal migratory connectivity in European populations handbook. Cambridge University Press, Cambridge.
of Pied Flycatchers Ficedula hypoleuca. J. Avian Biol. Teetor, P. 2011. R Cookbook: Proven Recipes for Data Analysis,
47: 69–83. Statistics, and Graphics. O’Reilly Media, Sebastopol, CA.
Parrish, J.D. 2000. Behavioral energetic and conservation Thompson, B. 1989. Why won’t stepwise methods die? Meas.
implications of foraging plasticity during migration. Stud. Eval. Couns. Dev. 21: 146–148.
Avian Biol. 20: 53–70. Tøttrup, A.P., Klaassen, R.H.G., Strandberg, R., Thorup, K.,
PAST. 1999–2019. Hammer, Natural History Museum. Kristensen, M.W., Jørgensen, P.S., Fox, J., Afanasyev, V.,
University of Oslo, Oslo. Rahbek, C. & Alerstam, T. 2012. The annual cycle of a
Peach, W.J., Furness, R.W. & Brenchley, A. 1999. The use of trans-equatorial Eurasian-African passerine migrant:
ringing to monitor changes in the numbers and different spatio-temporal strategies for autumn and spring
demography of birds. Ring. Migr. 19: 57–66. migration. Proc. R Soc. Lond. Biol. 279: 1008–1016.
BIRD STUDY 15

Valkama, J., Saurola, P., Lehikoinen, A., Lehikoinen, E., Walker, J. & Taylor, P.D. 2017. Using eBird data to model
Piha, M., Sola, P., Velmala, W., Forsman, D. & population change of migratory bird species. Avian
Sulonen, H. 2014. The Finnish Bird Ringing Atlas Volume Conserv. Ecol. 12: 1–24.
II / Suomen Rengastusatlas II. LUOMUS - Finnish Wernham, C., Toms, M., Marchant, J., Clark, J.,
Museum of Natural History, Helsinki. Siriwardena, G. & Baillie, S. 2002. The Migration Atlas:
van Strien, A.J., Pannekoek, J. & Gibbons, D.W. 2001. movements of the birds of Britain and Ireland. T and A D
Indexing European bird population trends using results POYSER, London.
of national monitoring schemes: a trial of a new method. White, E.R. 2019. Minimum time required to detect
Bird Study 48: 200–213. population trends: the needs for long-term monitoring
Varian, H. 2005. Bootstrap tutorial. Math. J. 9: 768–775. programs. BioScience 69: 40–46.
Vrezec, A., Fekonja, D. & Šere, D. 2014. Obročkovalna Wiersma, Y.F. & Skinner, R. 2011. Predictive distribution
dejavnost in pregled najdb obročkanih ptic v Sloveniji v model for the boreal felt lichen Erioderma pedicellatum
letu 2013. Acrocephalus 35: 25–58. in Newfoundland, Canada. Endang. Species Res. 15: 115–127.

You might also like