2.1 Study sites and lockdown restrictions in Poland
Avian life-history and reproductive data were collected from 2017 to 2020 across seven study sites set in a gradient of urbanisation in the capital city of Warsaw, Poland. Each study site is characterised by an assigned number of Schwegler woodcrete nestboxes (type 1b, with a 32 mm entrance hole and erected in a 50m-distance grid) suitable for great tits and blue tits. The study system here described aims to accurately reflect the urban matrix, as it comprises a wide range of diverse and contrasted habitat patches 41.
While a state of epidemic was officially declared in Poland on March 20th, a series of increasingly restrictive measures limiting human presence outdoors were subsequently introduced. A strict lockdown period forbidding the use of urban green areas was introduced between the 1st of April and 20th of April included. During this time, city dwellers were not allowed to access urban green areas, recreational locations, natural reserves or protected areas within and outside city borders. The only allowed activities outside of homes included the purchase of food supplies and other essentials items, caring duties and work, which enabled the authors of this study to access green areas within the remit of their work.
Study areas in our urban study system were thus subjected to contrasted levels of access restrictions during the SARS-CoV-2 lockdown. We consequently assigned each study site to the following categories:
- “Lockdown - Entrance Allowed” (LEA) – pertains to 4 sites and a total of 173 nestboxes; included streets and residential areas where residents were allowed outdoors to fulfil their essential needs during the pandemic
- “Lockdown – Entrance Not Allowed” (LENA) – pertains to 3 sites and a total of 236 nestboxes; included parks, woodlands and forest reserves, all of which were closed to the public during the strict lockdown period. All these sites re-opened to the public on the 20th of April 2020.
We provide a brief description with lockdown information (as “LEA” or “LENA”) below; sites are listed from the most distant to the closest to Warsaw city centre. More details on each study site can be found in Corsini et al 21,42 and Szulkin et al. 41.
A. Suburban village (n=47 nestboxes, LEA). Palmiry village (20º46’48.9748’’E - 52º22’11.3382’’N) is located c. 20 km away from Warsaw city centre and borders Kampinos National Park (Site B). Palmiry is a typical suburban village, where residential homes with gardens are interconnected by tree-lined avenues.
B. Natural forest (n=110, LENA). Kampinos National Park (20º47’14.3867’’E - 52º21’22.5409’’N) is a large forest located c. 20 km from Warsaw city center. The area is characterised by pine and mixed oak-pine forest habitats.
C. Residential area II (n=52, LEA). Osiedle Olszyna neighbourhood (20º57’39.37097’’E - 52º16’23.71883’’N) is a block of flats intermixed with green spaces and recreational facilities. It borders with the urban woodland “Las Olszyna” (site D).
D. Urban woodland (n=21, LENA). Las Olszyna (20º57’33.93652’’E - 52º16’10.55093’’N) is a green space that includes a deciduous, wet alder forest and an open space with an adjacent playground.
E. Office area (n=28, LEA). The Warsaw University “Ochota” Campus (20º59’8.85224’’E - 52º12’43.77676’’N) is located next to the urban park Pole Mokotowskie (site G) and belongs to one of the central districts of the city. Buildings consist of university offices, laboratories and other student facilities.
F. Residential area I (n=46, LEA). The “Muranow” neighbourhood (20º59’5.74332’’E - 52º14’52.17925’’N) is a residential area, similar in design to Residential area II (site C).
G. Urban park (n=105, LENA). Pole Mokotowskie (21º0’6.98321’’E - 52º12’46.66874’’N) is an extensive urban green area located close to the city center. With its alternation of meadows, tree-covered areas and recreational structures (i.e. playgrounds and sport facilities), it provides a centrally-located recreational area for city dwellers.
2.2 Avian life-history traits data collection
From the end of March, we checked nestboxes weekly to identify those occupied by great tits and blue tits. A nestbox was considered as “occupied” when at least one egg was laid on a completed nest. Weekly checks allowed to record the date of the first egg laid (e.g. laying date recorded from the 1st of April, corresponding to the value of 1), incubation duration (given in days and calculated as: hatch date – first egg laid date – clutch size – 1, 43, though incubation occasionally starts earlier or later than clutch completion in tits 44) and clutch size (total number of eggs in the nest). Only first broods were included in the analyses 45.
2.3 Tree-cover measurements
We measured the percentage of tree cover in a 100m radius around each nestbox following Szulkin et al. 41. Briefly, after downloading a raster layer from Copernicus Land Monitoring Services (https://land.copernicus.eu/ sitemap;Forests/Tree Cover Density), we processed the data in qGIS (v.2.18.25). The map of tree cover was generated in 2015 and contained a 20m-pixel resolution layer. After creating a 100m radius buffer around each nestbox, we obtained the averaged value of tree cover (in %) at the nestbox level using the function Zonal Statistics in qGIS.
2.4 Statistical analyses
Statistical analyses were performed within the computing environment R (v.3.6.2), separately for great tits and blue tits, in order to directly assess species-specific trait variation.
2.4.1 Association between avian life-history traits and lockdown restrictions
To test associations between avian life-history traits and lockdown restrictions, all tests were run in a model averaging framework 46. To test the effect of lockdown on avian traits investigated in this study, we specifically focused on the interaction between year and lockdown status (LEA - Lockdown Entrance Allowed vs. LENA - Lockdown Entrance Not Allowed sites), the latter explicitly reflecting a lack of outdoors human activity in LENA sites in 2020.
To model nestbox occupancy, we fitted generalised linear models (GLMs) with binomial distribution (“glm” function in the R-package “lme4” v.1.1-21-47). A nestbox was considered occupied (1) only if a great tit or a blue tit (analysed separately) was breeding in the nestbox. Nestbox occupancy (0/1) was fitted as binomial-response variable in each model, while the interaction between the two categorical variables year (four levels: 2017, 2018, 2019 and 2020) and lockdown status (two levels: LEA and LENA study sites) were fitted as predictors.
To model variation in egg laying date (“Lay date”; the egg laying date of a nest where the first egg was laid on the 1st of April would be coded with the value of 1), we fitted Linear Mixed Effects models (LMMs) with Gaussian distribution (“lmer” function in the R-package “lme4”). As for the analysis of nestboxes occupancy, the interaction between the two categorical variables year and lockdown status were fitted as explanatory variable. To control for variation associated with site specificity, the categorical variable study site (sites A-G) was fitted as random effect. The same model structure was used to model variation in clutch size, where we additionally fitted lay date as explanatory variable to control for the fact that earlier clutches in the season are often larger than later ones 48.
For incubation duration, we ran Generalised Linear Models with Gaussian distribution (“glm” function in R; the random effect of study site was not added here due to singularity problems in the model and a lack of model convergence when the random effect was added). Incubation duration was fitted as response variable while the interaction between year and lockdown status, and the continuous-variable lay date (to control for seasonal differences in each breeding event, as incubation duration decreases later in the season 21) were fitted as predictors.
2.4.2 Association between avian life-history traits and tree cover
To test whether tree cover in a 100 m radius around each nestbox (in %) covaries with avian life-history traits, the following models were run:
For nestbox occupancy, we used the same structure as described in 2.4.1, but for the variable lockdown status, which was replaced by the continuous variable tree cover.
In lay date and clutch size models, we used Generalised Linear Models (GLM) fitting each response with a Gaussian distribution. Similarly to the models ran on occupancy detailed above, the interaction between year and tree cover were fitted as predictors. Additionally, the explanatory variable lay date was added to clutch size and incubation duration analyses to mirror analyses performed on the same response variables as detailed in 2.4.1. In contrast to analyses described in the section 2.4.1, site was not included as random effect, as it covaries with the variable tree cover (here fitted as key explanatory variable).
2.4.3 Does lockdown influence tree cover preferences in occupied nestboxes?
To test whether tit tree cover preferences in occupying specific nestboxes changed due to the reduced human presence that occurred in 2020, we performed a one-way ANOVA test to model tree cover as a function of year among occupied great tit and blue tit nestboxes..