A high-precision dataset of breeding bird distributions in forested landscapes in Estonia

Abstract

Diversity and abundance of breeding birds are frequently reported and analysed as indicators of environmental change. However, such data available for forests typically contain either relative abundances based on snapshot observations or have been collected in small sample plots, which limit their distributional and ecological analysis across landscapes. I present a spatial dataset from three adjacent landscapes in Estonia (hemiboreal Europe), which has been obtained by standard multiple-visit mapping of nesting territories in 2020–2022. The records constitute the most likely centroids of distinct nesting territories of all 98 breeding species detected; these have been extracted and interpreted based on observations from an average 7–8 visits per season, and quality-assessed for three levels of spatial accuracy. One landscape was mapped in all three years, the others in either 2021 or 2022. The total area mapped was 14.3 km², including 86 % woodlands of diverse types and origins; a woodland characteristics dataset accompanies the bird data to facilitate habitat analysis. The paper describes the study plots; technical protocols of fieldwork and record interpretation; limitations (notably the likely missing of 10–20 % of pairs in most species); and possibilities to use the data in basic and applied ecological research. The main values of the dataset are that (i) it provides landscape-scale distribution map for the whole breeding assemblage of birds at high spatial precision, (ii) has accompanying woodland habitat data, and (iii) it also includes a repeatedly mapped landscape for detecting temporal variation in bird distributions.

Specifications Table

Subject	Biological Sciences
Specific subject area	Landscape-scale distribution of breeding bird assemblages in heterogeneous forested landscapes.
Data format	Raw data (quality assessed)
Type of data	Spatial datasets with exportable Data Tables (MapInfo; ESRI Shape; csv).
Data collection	Data collection followed a standard protocol of bird territory mapping based on repeated local observations that refer to nesting and are distinct from neighbouring locations. On average, each location was visited 7–8 times per season by the same person in the daylight (usually early morning but at least two evening visits) plus special visits (early season; summer nights). Standard manual procedures were used to distinguish nesting territories and establish their likely centres (data points). The accompanying habitat data of woodland patches was compiled by critically reviewing and amending official forestry database records, in order to represent habitat conditions during the bird surveys.
Data source location	Field data were collected in East Estonia, in three adjacent landscapes (total area 14.3 km2) between 58.22–58.29 N and 27.12–27.19 E. All the data are geopositioned, and accompanied with additional spatial datasets depicting woodland habitat patches in the study years and the studied landscape borders.
Data accessibility	Repository name: Zenodo Data identification number (DOI): 10.5281/zenodo.12747761 Direct URL to data: https://zenodo.org/records/12747762 Instructions for accessing these data: Given proper citation, the data are freely accessible to scholarly and educational work but not for commercial purposes.
Related research article	Lõhmus, A., Absolute densities of breeding birds in Estonian forests: a synthesis, Acta Ornithol. 57 (2022) 29–47. https://doi.org/10.3161/00016454AO2022.57.1.003

1. Value of the Data

• •Distribution of nesting birds on landscapes constitutes a major research issue: it has been a model system in basic population and community ecology, and serves as an indicator of the wider environmental quality. Most woodland birds maintain seasonal nesting territories that provide them access to resources and are critical for population recruitment. The resulting seasonally stable assemblage distributions can be mapped but, due to detectability limits, it is a highly laborious task over large areas.
• •The dataset introduced here provides a globally unique documentation of how pairs of all breeding birds are distributed across forested landscapes (>14 km² of total area; 86 % woodland cover), based on multiple-visit mapping of a hemiboreal bird assemblage in Estonia. One of the three adjacent landscapes was mapped during three subsequent years. Each data point is the most likely centroid of a bird territory, as interpreted from field records according to a standard protocol. The bird data are accompanied with spatially explicit data on local woodland characteristics.
• •The data can be analysed for bird assemblage patterns, intraspecific and interspecific spacing, species-habitat relationships, and landscape ecological phenomena such as land productivity, biodiversity hotspots and spatial arrangement of ecological structures. Through the standard technical protocols, similar data can be collected elsewhere for widening the geographic scope of the study or be repeated in the same landscapes for temporal trends.
• •The data constitute a transparent space-for-time substitution record of how birds are distributed in relation to landscape change, notably to clearcutting-based forest management. The background dataset specifies the origin, age and recent forestry operation history of all woodland patches (in total, 960 forest stands) in these landscapes at the time of the study.

2. Background

Due to their mobility and living-space requirements, bird populations and assemblages respond to environmental change most informatively at broad spatial scales [[1], [2], [3], [4]]. Since the costs of obtaining precise broad-scale data on land birds are considerable, such research is mostly based on snapshot surveys or other methods (e.g., soundscape recording) that produce indices of relative abundance [3,5,6].

This dataset was collected as a part of an alternative approach, which claims that many basic and applied questions on avian populations and their environmental relationships cannot be answered based on relative surveys [7,8]. The main standard method available for the alternative approach in heterogeneous landscapes is bird territory mapping [9,10]. That method is based on the biological facts that, during the nesting season, most birds that can contribute to the recruitment of the next generation are relatively sedentary around their nests – a distribution that can be detected through repeated observations and, specifically, by territorial behaviour. The mapping can provide spatial datasets of the territory locations of all bird species nesting in an area, which can then be analysed for many ecological relationships and compared with the trends reported using the relative abundance metrics.

3. Data Description

The paper describes one spatial dataset with bird territory centroids as records (point objects; Birddata.zip) and its supplementary datasets with areal objects as records: the study area borders (Bird_area_borders.zip) and woodland patches (Bird_woodland_data) in three adjacent forested landscapes in East Estonia (Fig. 1). The data have been created using MapInfo Professional Version 10.5 (manufactured by Pitney Bowes Inc.) in L-EST97 (EUREF89) projection, with one alternative format provided along: ESRI Shape File (for polygons) or .csv with coordinate fields (for point objects). These formats also allow extracting simple data tables, and the dataset [11] includes an explanations file that describes all the fields and codes used.

Fig. 1.

Map of the three adjacent study landscapes (Plots A, B and C) in East Estonia. Forests (three age classes, as of 2020) and scrub areas within the landscapes are coloured (greenish outside). Darker shade denotes bog areas, revealing covariation with forest age structure (e.g., concentration of clear-cuts to non-bog areas).(For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

The map layer in Birddata.zip is based on bird-mapping fieldwork by the author in the area in 2020–2022 (Plot A in all three years; Plot B in 2021; Plot C in 2022). It is a full-area dataset, i.e., in a study year all bird territories within a plot have been mapped as precisely as possible (but see Limitations). The records also include nesting territories in the vicinity of the plot borders, which enable some analyses (e.g., nearest-neighbour) but are not limited to certain distance and are based on lower number of observations (not comprehensive). Of the total of 5398 bird records (representing 98 breeding species), 5079 records are within the three landscape plots and 319 are outside. Sixty-one bird species are represented with at least 10 records over the years (Table 1).

Table 1.

List of bird species that are present with at least 10 records in the dataset.

No. of records	Species (no. of records)
>100	Fringilla coelebs (737), Erithacus rubecula (382), Phylloscopus collybita (369), P. trochilus (355), Anthus trivialis (344), Phylloscopus sibilatrix (318), Troglodytes troglodytes (194), Turdus merula (193), T. philomelos (185), Parus major (174), Sylvia borin (161), S. communis (146), S. atricapilla (105)
51–100	Regulus regulus (93), Spinus spinus (88), Emberiza citrinella (80), Columba palumbus (79), Dendrocopos major (77), Poecile montanus (69), Lophophanes cristatus (61), Caprimulgus europaeus (56), Ficedula hypoleuca (56)
10–50	Pyrrhula pyrrhula (50), Acrocephalus schoenobaenus (48), Garrulus glandarius (46), Cuculus canorus (42), Prunella modularis (40), Turdus viscivorus (40), Carpodacus erythrinus (37), Sitta europaea (35), Scolopax rusticola (32), Motacilla alba (31), Periparus ater (30), Cyanistes caeruleus (28), Certhia familiaris (27), Tringa ochropus (26), Loxia curvirostra (23), Luscinia luscinia (23), Phoenicurus phoenicurus (23), Muscicapa striata (22), Emberiza schoeniclus (19), Sylvia curruca (19), Tetrastes bonasia (17), Sturnus vulgaris (17), Buteo buteo (16), Carduelis chloris (16), Crex crex (16), Gallinago gallinago (16), Oriolus oriolus (16), Anas platyrhynchos (15), Acrocephalus dumetorum (14), Dryocopus martius (14), Ficedula parva (14), Locustella fluviatilis (14), Saxicola rubetra (14), Acrocephalus palustris (13), Hippolais icterina (13), Hirundo rustica (13), Alauda arvensis (12), Corvus corax (10), Passer montanus (10)

Each bird record contains the species name (six-letter abbreviation of the Latin name), year, and spatial precision class based on original observation quality: 0 = low accuracy, either few or diffuse observations attributable to this pair; 1 = standard territory-mapping accuracy, see below; 2 = high accuracy, either based on nest found or the bird behaviour indicating nest vicinity (typically within ca. 10 m). The latter can be used as a filter for habitat analyses that require high spatial precision [7].

The woodland patch datasets contain field-checked local information and are meant to facilitate habitat analyses. The aim of field checking was to improve official woodland data with regard to actual habitat conditions during the bird surveys. It included recording recent cuttings and the stand characteristics of those forests not included in the official forestry databases (specifications available in the dataset [11] file ‘Birddata_explanations.docx’). The woodland data are divided between two map layers containing spatial polygons:

• (i)Forest stands (Bird_forestdata), which are delineated based on relatively homogeneous tree-layer structure and constitute units of management in Estonia. The forest stand database is comprehensive for the plots and depicts the situation as of 2020, with changes affecting the 2021–2022 nesting seasons indicated in the field “OtherCut”. For each stand, it provides data on broad forest type (as shown in Table 2), the stand origin (nine classes from forest cutover origin to planting on former fields), the start year of the dominant tree generation (cf. age estimates for stands on Fig. 2A–E), forest site type (21 types found in the area), productivity (seven classes), ownership (state- or private-owned), data on the most recent partial cuttings, and stand area;
• (ii)Non-forest scrub (Bird_scrubarea), which in this area is formed by tall deciduous shrubs and trees such as willow thickets in wet areas (Fig. 2F), overgrown arable land, or strips of willows and Prunus padus along roads. No further descriptive data is available for these polygons.

Table 2.

Area cover of broad woodland types in the studied landscapes (data in [11]).

Woodland typea	Cover (%) by stand age classes			Total cover (%)
	0–10 years	11–60 years	>60 years
Pinus sylvestris dominated	3.5	10.8	29.5	43.8
Picea abies dominated	1.2	5.0	1.3	7.4
Mixed conifer forest (≥80 % in total)	0.1	2.4	0.9	3.4
Conifer-deciduous mixture (neither ≥80 %)	2.1	4.4	6.4	12.9
Betula sp. dominated	1.0	6.5	1.3	8.8
Alnus incana dominated	0.1	0.3	0.0	0.4
A. glutinosa dominated	0.0	0.1	0.0	0.1
Populus tremula dominated	0.0	0.8	0.0	0.8
Other deciduous forest (≥80 % in total)	1.0	5.8	1.1	8.0
Scrub (non-forest)				0.5
Total woodland cover	9.1	36.0	40.5	86.2

^adominance refers to ≥80 % of overstorey composition.

Fig. 2.

Common woodland habitats for birds in the study area. A – 71 yr-old planted and managed Pinus forest on Myrtillus site (16 May 2021; 58.275 N 27.165E). B – 10 yr-old oligotrophic paludifying clear-cut with solitary retention trees (16 May 2020; 58.268 N 27.147E). C – 94 yr-old Pinus stand on drained mixotrophic bog; nest of Turdus viscivorus in the middle (2 June 2022; 58.258 N 27.139E). D – 44 yr-old intensively managed Picea stand on afforested Oxalis site (8 May 2021; 58.261 N 27.132E). E – 58 yr-old Betula-Pinus stand on overgrown fen (8 May 2021; 58.289 N 27.158E). F – tall Salix scrub on floodplain fen (13 August 2021; 58.274 N 27.144E). Photos by A. Lõhmus.

Importantly, narrow (<10 m wide) forest tracks, ditches, and other linear objects (such as powerline corridors) are usually included in the woodland polygons, particularly within large forest patches. Depending on the ecological question, these objects may need to be distinguished if the aim is to account for within-forest fragmentation or to analyse only areas under tree canopies. In such cases, the woodland cover estimates in the whole landscape also become lower than reported here. Both the thematic spatial data and orthophotos to identify the linear elements are openly accessible at the Estonian Land Board Geoportal.

4. Experimental Design, Materials and Methods

4.1. Study area

The study was carried out in three adjacent forested landscapes (hereafter: Plots A, B, and C) in eastern Estonia, at altitudinal range 31–39 m a.s.l. (Fig. 1). Estonia is a lowland country in the European hemiboreal vegetation zone. Currently, forest land covers 51 % of the country, most of which is managed by clear-cutting based forestry. Since bird assemblages vary along forest successional stages, with additional modifying impacts of site type (also on management intensities), the country's bird populations are profoundly influenced by the forestry practices [8]. The annual mean temperature in the study area is +6 °C (–4.5 °C in February, +18 °C in July; 1991–2020 norms as reported by the Estonian Weather Service).

The plots were delineated without prior knowledge of their bird distributions. The goal was to include replicated managed-forest landscapes in full, which would include country-typical settings with fields, settlements, and waterbodies adjacent to forests. Thus the three plots have a similar basic organization: bordering with the Ahja River in the west and having central bog areas surrounded by managed forests on increasingly productive soils (Fig. 1). The size of the areas that could be covered by territory mapping in a single season determined the range of options available, including a decision to only include small human settlements (villages and individual farmsteads). Eventually, Plots A, B and C encompassed a total of 14.3 km² of the Estonian forested countryside, which includes a heterogeneity of ‘moderate’ management approaches, but does not contain extremes of the land-use gradients in the country (e.g., no nature reserves, urban areas, large agricultural or mining areas). According to archaeological findings, human habitation in these forested areas can be reliably dated back to at least the 3rd century AD.

Table 2 provides a summary of the area's woodlands at the time of the study, as recorded in the Supplementary dataset [11]. Overall, 40.5 % of the landscapes was covered by >60 year-old forest stands, 36.0 % by forest stands 11–60 years of age, and 9.6 % by younger clear-cuts or non-forest scrub. Based on overstorey (1st layer) composition, 53 % of the forest stand area was Scots pine (Pinus sylvestris) dominated and 12 % was either planted Norway spruce (Picea abies) stands or conifer mixtures. Deciduous-conifer mixtures formed 14 % of stands, birch (Betula sp.) dominated stands 10 % and other deciduous stands 11 %. Images of typical woodlands in the area are given in Fig. 2; these also indicate that most forests in this region only have a single tree layer developed after some form of former clearance (either as clear-cutting or agricultural use).

Of the forest land, 88 % was state-owned and managed by the State Forest Management Centre, which holds a Forest Stewardship Council (FSC) certificate since 2002. During the study, 33 ha (2.7 % of forest land) were clear-cut and 52 ha were thinned; its immediate impacts on bird densities have been described in [12]. Most of the area has been drained for forestry in the mid-20th century.

4.2. Field survey methods

In each year and plot, primary field data were collected by multiple-survey mapping of all breeding birds based on conventional techniques [10,13], with an average 7–8 visits from mid-April to late June. This covers the breeding seasons of most species in the area. The visits were undertaken in suitable weather (no rain or strong wind) and all the area was eventually covered by walking slowly along spaced-out trajectories (ca. 50 m distances in forests). Depending on bird abundance, ca. 1 km² can be usually mapped in one day, so each round of visits took ca. one week. For example, in 2021, the fieldwork took place in 11 days in April, 22 days in May, and 26 days in June.

All the birds seen or heard were marked on maps, using specific symbols to denote their behaviour, movement (if considerable) and simultaneous observations of all neighbouring individuals or pairs. All the fieldwork, data recording, and later interpretation were carried out by the author. A single-expert approach generally reduces the observer bias (e.g., [14]) and I had nearly 30 years of prior forest-bird mapping experience in Estonia [7]. Specifically, five main approaches to increase the comprehensiveness of the standard survey were added.

• (i)Distant birds were included and marked by directions (preferably from multiple points) as recommended by Brauze [15]; this appeared an important source of information for some noisy, but rare or highly mobile, species [16].
• (ii)While most visits were in mornings, the whole set also included at least two evening visits to detect some species that are more effectively mapped in that time [10].
• (iii)To each plot, at least four additional visits with modified methods were undertaken outside the main season: one visit earlier in spring (notably to detect woodpeckers and some vocal early-breeders), another in July (mostly to places that appeared less covered), and at least two nocturnal surveys in late June-early July (notably for detecting Caprimulgus europaeus, Crex crex, some Acrocephalus species, and owl broods). The walking distances and speed were typically higher in these visits due to their specific targets.
• (iv)In 2020, I estimated the effectiveness of the standard survey in three thrush (Turdus) species by nest searching after the breeding season [7]. This test showed ca. 10–30 % underestimation of true densities and frequent misinterpretation of the territory centres (birds were often active at territory borders). For that reason, nest searching was included among the field methods for the current study (also the 2020 data were reinterpreted). Almost all territories of diurnal raptors, owls, woodpeckers, and swallows have been assigned based on nests. Even in some difficult species, such as Turdus philomelos [17], up to one-third of the nests were eventually found.
• (v)In the case of birds nesting in gardens and houses, I obtained additional data from local people, which provided improved data particularly for swallows, S. vulgaris, Motacilla alba and Muscicapa striata.

A seasonal problem, most notable in common species, was to separate potentially nesting birds from migrants. As a rough rule, I did not map migrant species during that round of visits when the species was first encountered (I kept track of all arrival dates). The observations from the next rounds were recorded but, for the time period depending on species, filtered to use only records indicative of nesting (e.g., singing throughout the day or a pair together). For example, the mean arrival time of Chaffinch (Fringilla coelebs) was 29 March (2018–2024 data); yet, I did not map this species in April and only considered the nesting-indicative observations between 1 and 10 May. Those passerines arriving in April were typically mapped only from mid-May onward, etc. (further details available upon request). Such omission rules were not used for late-arriving species (such as Acrocephalus dumetorum) that tend to establish nesting territories, form pairs and cease singing rapidly [18].

4.3. Deriving territory centroids from the observations

All field observations were digitalized and interpreted after the field season for each species separately. There were two tasks for the interpretation: (i) to distinguish individual nesting territories, and (ii) to provide the best location estimate for the centre of the territory (ideally a nest). The final product of the two steps constitutes the dataset [11] presented in this paper.

For the task (i), I followed the standard protocols of the territory mapping method [10,13]. The first step was to fix the locations of clear or obvious nests, which are given the highest of three accuracy classes of the records. In the second step, all other observations were considered in the following sequence: simultaneous records of neighbouring birds; observations of territorial behaviour; other spatially accurate observations; other observations. In this step, four conditions must have been met to identify a nesting territory: (a) at least two observations that were (b) separated by at least two-day interval, (c) were distinct from other territories (either by simultaneous recording or long distance), and (d) included at least one observation of either territorial behaviour, pair together, or other indication of nesting. In the final step, I considered all the remaining observations together, in order to identify a minimum number of territories outside those already assigned. This step was most important for poorly detectable highly mobile species (e.g., Scolopax rusticola, Garrulus glandarius).

There were two species-specific modifications to this general protocol. First, some species do not have stable nesting territories or are otherwise better identified by nest locations. Of such species, C. canorus was mapped based on territorial males [16]. For S. rusticola, I first considered nests, broods, and pairs seen together on ground before considering display flight observations. In lekking grouse (Tetrao tetrix and T. urogallus) I prioritized broods and only in their absence assigned ‘territories’ based on lekking males.

The second interpretational problem was to reduce double-counting of replacement clutches in new locations and ‘floater’ individuals entering the plots after failed nesting elsewhere. This was done by ignoring summer observations in a set of early breeders, such as tits (Paridae), thrushes (Turdus spp.), and S. rusticola. In these species, the observations from mid-June onward were only considered when clearly related to the first brood (e.g., fledged broods or abandoned nests). The earliest-breeding species, Loxia curvirostra was in flocks since June, and only observations until 20 May were considered.

Focusing on the likely nest location in the task (ii) was a conceptual decision based on the features of the field data. Alternatives could have been delineating the whole activity range of a territory holder or depicting an observational ‘activity centre’, but these are dynamic in time, require more data than nest locations, and can differ among territories (e.g., singe territorial males vs. pairs; one vs. multiple activity centres). The protocol following the nest-focused approach and the techniques are illustrated in Fig. 3; all work with digitalized records was done on top of the aerial photos provided by the Estonian Land Board as shown in the figure:

• (a)Nest or obvious nest (based on aggressive or extremely agitated adults; carrying of nest material or food; or poorly moving young) was always prioritized as the centroid location.
• (b)The default solution for all other cases was the geometric mean position of the records, which was, however, shifted according to two criteria to avoid unlikely nest locations:
• (c)If the geometric centre appeared in a forest stand or habitat where no actual observations had been made – it was shifted to the stand or habitat where most observations were made; and
• (d)If the geometric centroid appeared closer to abrupt habitat edge (such as forest-clearcut) than any actual observations – it was shifted to the closest distance from the edge observed.

Fig. 3.

The protocol of identifying nesting territory centroids (yellow circles) from field observations, as exemplified by Turdus merula observations in 2022 (black symbols; the Pair ID references as in [11]). A – if found, nest (asterisk) always becomes the centroid, regardless of the distribution of other observations (Pair ID = 1078); B – geometric-mean based centroid when other considerations do not apply (Pair ID = 1036); C – centroid is not allowed outside the main habitat of observations and not closer to edge than any observation (Pair ID 1056); D – centroid is not allowed in the central stand if that has no actual observations (Pair ID 1083). Black asterisk denotes the nest; black circles are other observations considered. Background aerial photos: Estonian Land Board.(For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

In several elusive and potentially wide-ranging passerines (such as Aegithalos caudatus, Poecile montanus, Lophophanes cristatus, Pyrrhula pyrrhula, Garrulus glandarius), I additionally weighed the observations of pairs seen together in the egg-laying or incubation period twice as much as later observations. For a hypothetical case of two observations it can be imagined as the centroid being at 1/3 on the straight line from the pair location toward the other location. Similar weighing was used when one observation was accurate, while the other was not (e.g., distant hearing).

Limitations

Apart from the general problems of territory mapping [19], the main limitation of my dataset is a likely missing of nesting territories despite massive effort. The preliminary tests based on intensive searching for nests [7] have suggested that the underestimation is <10 % in well detectable species and 10–20 % in moderately detectable (typical) species, but may reach 20–30 % in some poorly detectable and highly mobile species. In the latter case, two reasons can combine: there are scarce records over a wide area, and each of these provides little ground for reliably distinguishing a nesting territory. A special case is Loxia curvirostra that starts breeding in late winter and has been mainly recorded based on fledged broods and pairs that are still together in spring; in this species, the spatial accuracy of territory centroids is unknown.

Another limitation is that the dataset only depicts the first clutches of multiple-breeders. Without tracking particular individuals, their later nesting and replacement clutches simply cannot be distinguished. Facing the dilemma of underestimating true population sizes versus multiple counting of the same individuals, I have preferred the first (conservative) approach. It may be particularly important in those species that switch habitats for the second broods [20].

Ethics Statement

The author has read and followed the ethical requirements for publication in Data in Brief and confirms that the current work does not involve human subjects, animal experiments, or any data collected from social media platforms.

CRediT Author Statement

Asko Lõhmus: all the research that provided this dataset.

Data Availability

• ZenodoSpatial dataset of breeding bird territories in East-Estonian forested landscapes, 2020–2022 (Original data).