Habitat quality, urbanisation & pesticides influence bird abundance and richness in gardens

habitat quality. We show that there was an interaction between the habitat quality of the surrounding area and pesticide use: negative effects of pesticides on species richness were more pronounced in gardens in areas of high habitatqualitycomparedtothosesurroundedbypoorhabitat.Wefoundthatpesticideuse,andparticularlyglyphosate and metaldehyde, negatively predicted the abundance of house sparrows, a fast-declining bird species. The average house sparrow abundance was 12.1 % lower in gardens applying any pesticide, 24.9 % lower with glyphosate, and 38.6 % lower with metaldehyde. Overall, our study shows that garden bird abundance and richness is strongly in ﬂ u- enced by both extrinsic and intrinsic factors, and suggests that garden management, particularly regarding pesticide use, has a signi ﬁ cant effect on bird life.


H I G H L I G H T S G R A P H I C A L A B S T R A C T
• All biodiversity estimates were lower in suburban gardens compared to rural gardens. • Provision of bird-friendly habitats positively influenced biodiversity estimates. • Positive effect of local habitat was lowered in gardens where pesticides were used. • 32 % of gardens used pesticides, with glyphosate comprising 53 % of applications. • House sparrow number was lower in gardens using pesticide, glyphosate & metaldehyde.

A B S T R A C T A R T I C L E I N F O Editor: Rafael Mateo Soria
Keywords: Glyphosate Metaldehyde Acetamiprid House sparrow Herbicide Gardens are regularly portrayed as green oases, refuges for wildlife that has been displaced from the countryside by intensive farming practices which have reduced habitat availability. Pesticides are also commonly used in urban areas, but few studies have investigated their impacts. In this study, we explored how bird richness and abundance in gardens across the UK are influenced by habitat quality, urbanisation level and pesticide practices. To achieve this, we collaborated with the British Trust for Ornithology (BTO) which runs Garden Birdwatch, a citizen-sciencebased garden bird recording scheme. Participants in the study were asked to complete a questionnaire about their pesticide practice. From the 615 gardens that provided useful data, we found that 32.2 % applied pesticides in their gardens and that glyphosate comprised 53.3 % of these applications. We found that bird abundance and species richness was lower in suburban compared to rural gardens but positively influenced by measures of garden quality and by surrounding habitat quality. We show that there was an interaction between the habitat quality of the surrounding area and pesticide use: negative effects of pesticides on species richness were more pronounced in gardens in areas of high habitat quality compared to those surrounded by poor habitat. We found that pesticide use, and particularly glyphosate and metaldehyde, negatively predicted the abundance of house sparrows, a fast-declining bird species. The average house sparrow abundance was 12.1 % lower in gardens applying any pesticide, 24.9 % lower with glyphosate, and 38.6 % lower with metaldehyde. Overall, our study shows that garden bird abundance and richness is strongly influenced by both extrinsic and intrinsic factors, and suggests that garden management, particularly regarding pesticide use, has a significant effect on bird life.

Introduction
As the human population expands, the landscape is becoming increasingly urbanised (Grimm et al., 2008), with currently 55 % of the global Science of the Total Environment 870 (2023) 161916 human population living in our growing cities and an estimate that this will reach 68 % by 2050 (United Nations, 2018). Spreading urbanisation results in habitat destruction and fragmentation (Enoksson et al., 1995;Mörtberg, 2001;McKinney, 2002) and tends to homogenise biodiversity by facilitating some adaptable species over more specialised ones (Crooks et al., 2004;Kark et al., 2007;Ortega-Álvarez and MacGregor-Fors, 2009;Mayorga et al., 2020).
Gardens are described as green sanctuaries within these spreading urban areas, and they do support a diversity of bird species (Chamberlain et al., 2004(Chamberlain et al., , 2007Gaston et al., 2005;Cannon et al., 2005). In the UK alone, there are an estimated 22.7 million gardens, which when combined with the median garden size of 190m 2 gives a crude estimate of 432,924 ha (Davies et al., 2009). Although gardens taken individually may be of minor biological significance, collectively they represent an important potential habitat and food resource for urban wildlife. Over recent decades many studies have considered factors that influence their importance for conservation (e.g., Gaston et al., 2005;Smith et al., 2006;Loram et al., 2008;Stewart et al., 2009;Hill et al., 2021;Segar et al., 2022, Negret et al., 2022. For example, studies have looked at the effects of garden vegetation type and structure on bird richness, abundance, occurrence, and communities (Jokimäki, 1999;Daniels and Kirkpatrick, 2006;Paker et al., 2014;Campos-Silva and Piratelli, 2021;Gonçalves et al., 2021). Some of the findings were that tree composition affected bird species richness in gardens, with deciduous trees supporting higher richness than coniferous trees in European gardens (Thompson et al., 1993), or that the surrounding landscape seemed to be the main driver for the occurrence of most bird species in gardens (Chamberlain et al., 2004), and more recently, that the number of trees in a neighbourhood and the garden size were both positively linked to bird richness in a Brazilian city (Gonçalves et al., 2021).
While the impacts on biodiversity of pesticides used in farming systems has been the subject of considerable research and debate (e.g. Boatman et al., 2004;Lopez-Antia et al., 2018;Addy-Orduna et al., 2019;Eng et al., 2019;Fernández-Vizcaíno et al., 2020), the use of pesticides in gardens has received scant attention. While agricultural pesticide applications are documented and reported in many countries, in comparison domestic pesticide applications are generally not quantified. When investigating children's exposure to pesticides within households, Grey et al. (2006) found that 93 % of UK parents applied a pesticide in their home or garden over a year, with 76 % of pesticides used in the garden. Additionally, in 2019 the UK's Health and Safety Executive (HSE) conducted an online survey to identify UK gardeners' habits of use of plant protection products (usage and disposal) and found that, out of the surveyed gardeners, 42.8 % (495/1155) used pesticides. Furthermore, they established that the majority of products purchased were weedkillers, molluscicides and insecticides (HSE, 2019).
There is clear potential for garden bird species to be exposed to pesticides, directly or indirectly. Direct exposure, to both agricultural and domestic pesticides, could take place by consuming contaminated food (seeds, fruits, insects etc.) and/or water. Subsequent to a spray event, dermal exposure could also occur; pesticides may be assimilated through the skin or ingested during preening behaviour (Vyas et al., 2007). An indirect effect of pesticides could also occur through the decrease in insect prey or availability of weed seeds (Geiger et al., 2010;Hallmann et al., 2014).
To our knowledge, only one previous study has investigated pesticide practices and how they may influence garden wildlife, in their case butterflies. In French private gardens Muratet and Fontaine (2015) found a negative correlation between both butterfly and bumblebee abundance and the application of insecticides and herbicides. They also showed that the impact of pesticides differed with the urbanisation level, with highly urbanised areas showing a stronger negative effect of insecticides.
In this study, we investigate whether domestic pesticide use predicts bird richness and abundance in gardens across three urbanisation levels, and how extrinsic and intrinsic measures of habitat quality may influence this relationship. We hypothesize that environmental factors such as habitat quality and surrounding landscape quality would positively influence bird richness and abundance, while urbanisation and the use of pesticides would negatively affect bird richness and abundance. This has, to our current knowledge, not previously been investigated in birds.

Data collection
For this study, we collaborated with the British Trust for Ornithology (BTO) and the volunteers participating in the Garden BirdWatch Scheme (GBW). GBW is a structured, citizen-science monitoring programme, established in 1995 to monitor how birds and certain non-avian taxa utilise UK gardens . Participants submit weekly presence and maximum counts of target taxa from their gardens, the latter defined as the maximum number of individuals of a species observed at any one time in a given week, throughout the years (all season included). While participants differ in the amount of time that they devote to monitoring their garden each week, they are instructed to maintain a consistent level of monitoring effort from one week to the next, and to discard data from weeks that are under-or over-recorded. Recording takes place throughout the year and most of the observations are collected through a dedicated JAVA-based web application, with built-in validation procedures. A small proportion of participants use paper recording forms, which are scanned, validated and loaded into the cloud-based ORACLE database that also hosts those data submitted through the web application.
GBW currently has some 24,000 participants across the UK, the distribution of which is closely linked to the pattern of human population density. Participation levels are greater in the south-east of England than in northern Scotland, but the levels of coverage achieved are sufficient to produce robust measures of garden use by birds at both national and regional levels . Although there will inevitably be some degree of variation in the abilities of participants to identify the species they record, and in the amount of time spent recording, the use of site-effects in analyses of GBW data, coupled with the very large sample sizes involved, support the production of robust metrics.
Participants also provide additional information on their garden and its surroundings (the latter defined as within 100 m of the garden boundaries), providing data on a total of 56 individual variables which include the urbanisation level, defined as "urban: densely built-up areas and town centres with very few natural or near-natural bird feeding sites", suburban as "inhabited areas on the outside of built-up areas, near countryside or with large gardens, municipal parks or recreational areas", or rural as "areas away from towns, with just a few scattered houses, farms or other isolated buildings". This also include if their garden has coniferous trees, evergreen hedgerows, pond, or shrubberies and if it is surrounded by woodland, wetland, or river for instance, or the presence of other particular features (Chamberlain et al., 2004). The complete method of bird and garden data collection is available on the BTO website (https://www.bto.org/ourscience/projects/gbw).
We sent the volunteers a questionnaire about their pesticide practices between 2020 and 2021 (Appendix A). The questionnaire included queries such as: did they apply pesticides during the previous year; the brand name of all pesticides applied; during which season the pesticide applications mostly occurred. We subsequently identified the active substances linked to each pesticide brand. Volunteers also indicated if pets were present in their garden (cats and dogs) and their use of ferric iron pellets.
We received a total of 866 responses to our questionnaire, out of which some were not affiliated with the GBW scheme or had not provided data on garden characteristics or on bird observations. Previous studies showed that diversity estimates are influenced by sampling efforts (Moreno and Halffter, 2000;Willott, 2001). Therefore, we conducted a species accumulation curve analysis using the vegan package (v2.5.7; Oksanen et al., 2020) to determine the minimum number of weeks of participation needed to produce reliable species richness (Plummer et al., 2019). The species accumulation curves averaged across gardens reached an asymptote with 21 weeks of observations (Appendix B). This represented 85 % of the total number of bird species across all gardens being accounted for at 21 weeks of observations. Following Plummer et al. (2019), we used general linear mixed models (GLMMs), including garden and year as random effects, to verify that increasing sampling effort over 21 weeks had no significant effect on bird abundance and richness.
After removing gardens with incomplete information or with less than 21 weeks of participation, we ended up with a total of 615 gardens in this study, distributed across the UK (Fig. 1).

Creating garden quality and surrounding quality indices
The habitats present in a garden and in the surrounding area are very likely to influence their suitability for birds of different species. From the garden characteristics and the surrounding characteristics (within 100 m of the garden) collected by the GBW volunteers, we created a garden quality index (GQI) and a surrounding quality index (SQI), respectively.
We created the two indices by including variables that we expect to have a positive impact on local bird communities (directly or indirectly), through the increase of food availability, shelter, and nesting habitat (Appendix C). In our analysis, the higher the score for SQI and GQI, the more bird-friendly the garden is estimated to be. The GQI included 17 variables such as the estimated proportion of vegetable patches, the number of small coniferous trees, or the proportion of flowerbeds. Likewise, the SQI included 13 variables such as if there was broadleaved woodland, scrubland, or a stream within 100 m (no = 0, yes = 1 for each) (Appendix C).
Correlograms were conducted, to assess the correlation between all variables within and between indices. The correlations were found to be either weak or moderate with a maximum correlation score of 0.54, therefore no Principal Component Analyses (PCA) were conducted.
Garden size and urbanisation level were highly correlated, meaning that large gardens almost exclusively occurred in rural habitat. Therefore, we chose to exclude garden size (large/medium/small) and only use urbanisation level (rural/suburban/urban) in our analysis.

Statistical analysis
We used as dependent variables: species richness, defined as the total number of bird species recorded per garden, bird abundance, defined as the total average number of individuals per garden (all 141 species pooled, see appendix D), and common bird abundance defined as the average number of individuals per garden of 10 common garden bird species combined (see list of common garden bird on BTO and Royal Society of the Protection of Birds-RSPB websites (https://www.bto.org/; https://www.rspb.org.uk/). The 10 species used in the common bird abundance included: the blue tit (Cyanistes caeruleus), great tit (Parus major), common blackbird (Turdus merula), European robin (Erithacus rubecula), common chaffinch (Fringilla coelebs), greenfinch (Chloris chloris), collared dove (Streptopelia decaocto), dunnock (Prunella modularis), starling (Turnus vulgaris) and house sparrow (Passer domesticus).
We later used the abundance of the above-mentioned 10 species separately as dependent variables, in order to assess what species-specific relationship occurred.
We focused our analysis on the use of all pesticide in gardens and the use of the 6 most used active substances in the studied gardens (glyphosate, acetamiprid, mecoprop.p, metaldehyde, deltamethrin and dicamba). We did not include any analysis of the use of pesticide mixtures or interactions between pesticides in our study.
The time of year was not included in our analysis. After correcting for sampling effort, most of the studied gardens provided much more than the minimum 21 weeks of observations. The gardens used in this analysis had an average participation of 47.8 weeks, with 540 gardens recording data for over 40 weeks.
The use of bird feeder and presence of nest boxes in gardens were not included in the analysis, as most of the surveyed gardeners owned a bird feeder and at least one nest box (respectively 98 % and 77.9 % of gardens), limiting us from conducting meaningful analysis.
These variables were tested, using linear models (LMs), against several independent variables, along with all relevant interactions (see appendix E). The full model for each dependent variables included all independent variables and interactions listed in appendix E. Both the bird abundance and common bird abundance were log-transformed before analysis.
We established the most suitable model for each response variable by successively removing explanatory variables from the complete model, starting with interactions and the least significant terms, with the best model being the one with the lowest Akaikes's Information Criterion (AIC based model selection). Normality and homoscedasticity of residuals were established using quantile-quantile plots and the residuals were plotted against simulated values using the DHARMa package (v0.4.4; Hartig, 2021). We performed post-hoc Tukey test to compare urbanisation levels and post-hoc interaction test using the Phia package (v0.2-1; De Rosario-Martinez, 2015) to compare interactions and reported them when necessary. All statistical analyses were carried out using R v. 4.0.3 5 (R Core Team, 2020).

Result
Of the 615 gardens, we ended up with a total of 579,280 bird observations. The majority of participants were female (58.4 %) and over 65 years of age (57.2 %).

Effect of urbanisation and latitude
The bulk of the surveyed gardens were suburban gardens (49.3 %), followed by rural (43.4 %) and urban gardens (7.3 %).
The most important effects were that bird richness, total bird abundance and common bird abundance were all higher in rural gardens (bird richness rural-suburban p < 0.001; rural-urban p < 0.001; bird abundance rural- suburban p < 0.01; common bird abundance rural-suburban p < 0.001; Table 1). Post-hoc test revealed that, for bird abundance and common bird abundance, Suburban and Rural were significantly different (respectively, bird abundance: p < 0.01, 95 % C.I. = −0.15, −0.03; common bird abundance: p < 0.001, 95 % C.I. = −0.2, −0.06). More specifically, the average species richness in rural gardens was 43.9 % higher than for urban gardens. Additionally, bird abundance and common bird abundance were respectively 6.7 % and 10.4 % higher in rural vs urban gardens (Fig. 2). Similar results were found for species-specific abundance, with all species' abundance being lower in suburban and/or urban gardens compared to rural gardens (ranging from p < 0.001 to p < 0.05), except for starlings which did not present any significant relationship (Appendix F). Post-hoc test revealed that for all bird species except house sparrow and starling, Suburban-Rural and Urban-Rural were significantly different (p value ranging from p < 0.05 to p < 0.001).
We found a positive effect of the garden's latitude on the total bird abundance (Table 1), and when looking at species independently, 6 out of 10 bird species seem to show some influence of the garden latitude, either positive (blackbird, chaffinch, house sparrow) or negative (blue tit, great tit, robin) ( Table 2).

Effect of habitat quality and surrounding gardens
In our surveyed gardens, the GQI ranged from 0.25 to 9.35 and the SQI ranged from 0 to 7. Bird richness, bird abundance and common bird abundance were found to be positively correlated with GQI (species richness: p < 0.001, bird abundance: p < 0.05, and common bird abundance: p < 0.05; Table 1). In particular, the bird richness, bird abundance, and common bird abundance were all higher by 26.1 %, 8 % and 9.7 %, respectively, in gardens with a GQI above 3.85 (the median; Fig. 2). However, only the bird richness was found to be positively correlated with SQI (Table 1) with an average bird richness greater by 16.1 % when gardens have a SQI above 1 (the median).
When looking at the species separately, our results showed that the garden quality seems to positively impact the abundance of 4 bird species in particular: blue tit (p < 0.001), great tit (p < 0.001), common blackbird (p < 0.001), and European robin (p < 0.001; Table 2).
Note that the garden and surrounding quality index are also involved in interactions which are described below.
A total of 29 different active substances were used throughout the participating gardens, with an average of 3 yearly applications per garden, with one garden exceeding 13 applications. But only 3 out of the 6 substances analysed were associated with effects, and at a lower level of significance compared to the habitat parameters.
Regarding bird richness, we found a significant interaction between the SQI and pesticide applications. This indicates that the species richness increases with the surrounding quality, both for gardens that do not use pesticides and for gardens that applied pesticides, but this effect is significantly less strong when pesticides are applied (Table 1 & Fig. 3a). We then looked at active substances separately, and found that glyphosate, acetamiprid and deltamethrin could explain the general results as they were also negatively affecting bird species' richness in gardens in interaction with the SQI (glyphosate: p < 0.01, acetamiprid p < 0.05, deltamethrin p < 0.01; Table 1; Fig. 3b, c & d). Additionally, for bird abundance we found a significant interaction between the use of metaldehyde and urbanisation levels (urban level, p < 0.01). Post-hoc tests revealed that bird abundance was significantly lower in the presence of metaldehyde in rural and suburban gardens, but significantly higher in the presence of metaldehyde in urban gardens (p < 0,01; Table 2).
Subsequently, when we looked at the species-specific abundances, we found that overall pesticide use was significantly negatively correlated with the abundance of only house sparrows (Table 2). Two out of the 10 most common species (robin and house sparrow) showed a negative association with the use of glyphosate (p < 0.05), while metaldehyde was significantly negatively associated with house sparrow abundance (p < 0.05) ( Table 2; appendix G). More specifically, house sparrow average abundance was 12.1 % lower in gardens applying pesticides in general, 24.9 % lower when glyphosate was used and 38.6 % lower when metaldehyde was applied (appendix G).
Altogether, we found various results for the seven active substances we looked at more closely, which can be found in Table 1 and Table 2. Of 70 possible bird abundance x pesticide combinations (10 bird species, 7 pesticides), negative main effects were detected in 10 (great titglyphosate; great titmecoprop p; chaffinchmecoprop p; chaffinchdicamba; blackbirdmecoprop p; robinglyphosate; robindicamba; dunnockmecoprop p; house sparrowglyphosate; house sparrowdicamba). Positive effects were found for two (great titdicamba; green finchmetaldehyde).

Table 1
Summary of the models on bird species richness (total number of bird species recorded per garden), total bird abundance (average number of individuals per garden) and common bird abundance (average number of individuals per garden of 10 common garden bird species combined) from the 615 gardens included in the analysis'. (n: sample size, SE: standard-error, the asterisks represent the associated p value, *p < 0.05, **p < 0.01, ***p < 0.001).

Other variable effects
Additionally, 33 % of gardeners reported having pets (14.3 % dogs and 20.9 % cats) and we found no significant effect of pet presence, cats or dogs, on bird abundance or richness (Table 1). Nonetheless, when analysing bird species individually, dunnock and greenfinch were negatively impacted by the presence of cats. In contrast, we found that collared doves seemed to be positively influenced by the presence of cats (Table 2).
In our analysis, the use of ferric iron pellets had no apparent effects, both for the general bird richness, abundance and common abundance, as well as for species-specific estimates.

Discussion
We investigated bird species richness, bird abundance and common garden bird abundance, as well as the abundance of 10 bird species separately, in gardens across various habitat quality, urbanisation levels and pesticide practices. Our results showed that the relationship between garden practices, bird richness and abundance is complex.

Environmental effects: urbanisation and habitat quality
We found that bird richness, abundance, and common bird abundance were significantly reduced in gardens in more urbanised areas (suburban and urban when compared to rural gardens), with this being the strongest predictor of all the factors considered in this study. This is in accordance with what can be found in the literature, for birds but also other taxa such as butterflies (Chamberlain et al., 2004;Di Mauro et al., 2007;Luck and Smallbone, 2010;Muratet and Fontaine, 2015;Silva et al., 2015;Olivier et al., 2016;Fontaine et al., 2016;Mayorga et al., 2020). Previous research found that this negative influence of urbanisation could be mitigated with other garden practices, as Fontaine et al. (2016) showed that butterfly richness and total abundance of butterfly communities across French gardens were negatively correlated with urbanisation but that this effect could be decreased by providing rich nectar sources. Our results are similar, in that within-garden habitat quality was also found to be a powerful driver of bird richness and abundance. Local habitat qualities (GQI and SQI) were found to be the next most important factors to positively influence bird richness and abundance. Similar findings were reported by Olivier et al. (2016) in their butterfly study. Although, in their study the

Table 2
Summary of the models on bird species specific abundances (average number of individuals per garden per species) from the 615 gardens included in the analysis'. (n: sample size, SE: standard-error, the asterisks represent the associated p value, *p < 0.05, **p < 0.01, ***p < 0.001).  decline of habitat quality was linked to urbanisation, and this was not our case. Habitat quality and overall habitat characteristics have been shown to be important in shaping bird communities and assemblage (Daniels and Kirkpatrick, 2006).

Effects of pesticide applications
From our questionnaire to the GBW volunteers, we found that 34.1 % of the surveyed gardens had received at least one pesticide in the last year. Our surveyed population, being part of the Garden BirdWatch scheme, no doubt represents a non-random sample in terms of awareness and appreciation for wildlife, and hence their use of plant protection products is likely to be lower than the national private usage. Indeed, in 2019, the Health and Safety Executive found that 495 (42.8 %) of 1155 surveyed gardeners across the UK used pesticides (HSE, 2019).
Overall, we found that the majority of participants applying pesticides applied herbicides (61.9 % of pesticide applications) and most pesticides were applied in spring (46.6 %), which coincides with breeding, brooding, hatching, and fledging period for most bird species. In addition, 77.9 % of our respondents installed at least one nest box in their garden, and nationwide there are an estimated 4.7 million nest boxes in gardens (Davies et al., 2009), potentially increasing the likelihood that young birds and breeding adults may come into contact with pesticides applied in gardens, but this would need to be tested.
We found that bird species richness was highest in gardens with high surrounding habitat quality and without pesticides, while species richness in gardens surrounded by low quality habitat tended to be low regardless of pesticide use. This relationship seems to be explained mainly by three pesticides: glyphosate, acetamiprid and deltamethrin, all of which showed a negative correlation with species richness in interaction with SQI. We are conscious that gardeners cannot easily influence what is in the surrounding landscape of their garden, but our results indicate that stopping pesticide use is likely to have a greater positive effect on bird richness if the surrounding habitat quality is high.
When analysing the abundance of specific garden bird species, house sparrows were shown to be negatively impacted by pesticide usage in general, an effect driven largely by a negative relationship with the application of glyphosate and metaldehyde. Glyphosate also negatively influenced abundance of European robin and great tit, while metaldehyde seems to negatively influence greenfinch and to positively influence blackbirds. Moreover, out of the 10 bird species analysed separately, 7 showed multiple relationships with various pesticides, mostly positive when included in an interaction and negative when on its own (Table 2). In contrast, blue tit, collared dove and starling did not show any apparent relationship with the use of pesticide in gardens. The effects of pesticide shown in this analysis could come from direct or indirect routes and we were not able to distinguish between these. Birds may come into topical contact with pesticides following their application by the homeowner or consume contaminated plant material and/or insect prey (Millot et al., 2017;Tassin de Montaigu and Goulson, 2022). In addition, indirect effects of pesticide applications, via the decrease of food availability (seeds and prey) could also contribute to the observed results (Boatman et al., 2004). This indirect effect of pesticides could be mitigated by the presence of bird feeders, as it was found that the presence of bird feeders in gardens greatly increases visitation and therefore general bird abundance in gardens and modifies the bird community of the area (Brittingham and Temple, 1992;Wilson, 1994;Wells et al., 1998;Davies et al., 2009;Plummer et al., 2015;Hanmer et al., 2017) even favouring more competitive and aggressive species over others (Francis et al., 2018). Unfortunately, the large majority (98.7 %) of our surveyed population provided supplementary feed for birds, preventing us from conducting meaningful analysis since so few gardens are without bird feeders.
In our analysis, the pesticide showing the most effects on bird abundance and richness belonged to diverse pesticide families, including an organophosphate herbicide (glyphosate), a neonicotinoid insecticide (acetamiprid), a pyrethroid insecticide (deltamethrin) as well as a tetroxocane molluscicide (metaldehyde).
Very few studies have examined the direct toxicity of glyphosate to birds. Recent research on Japanese quail (Coturnix japonica) found that parental exposure to glyphosate-based herbicides resulted in lower embryonic development in the eggs and delayed plumage development in females (Ruuskanen et al., 2019;Ruuskanen et al., 2020). In 2019, Hussain et al., looked at the exposure to sub-acute concentration of glyphosate in cockerels (Gallus gallus domestica) and found that bird exposed to the highest concentration (125 mg/kg of body weight) presented tremor, depression, lower food consumption and lower body mass. Some studies looked at the vegetation changes induced by the application of glyphosate and its implication for bird population. For instance, in 1989, Santillo et al., found that the abundance of the alder flycatcher (Empidonax alnorum) and Lincoln's sparrow (Melospiza lincolnii) were lower in treated compared to untreated clearcuts, over their three years experiment. MacKinnon and Freedman (1993) found that bird density dropped and remained lower in plots sprayed with glyphosate than controlled plots after the second post-spray year. Morrison and Meslow (1984) found that, even though no apparent density differences were detected between treated and non-treated sites, they found that one year after glyphosate application, multiple bird species lowered their use of shrub cover, and the use of deciduous trees increased. For the compound metaldehyde, very few studies have focussed on the potential impact on birds, nonetheless, poisoning of birds with metaldehyde were previously reported (Reece et al., 1985;Kidd and James, 1991). Birds ingesting metaldehyde were found to experience impaired breathing, tremors, excitability, muscle spasms and diarrhoea (Gupta, 2012). When it comes to the other active substances included in our study, the toxicological effects found ranged from 50 % decline sperm density in house sparrow (Passer domesticus) for acetamiprid (Humann-Guilleminot et al., 2019); lower body weight for chicken exposed to deltamethrin compared to controls after 5 weeks (Chandra et al., 2013). For mecoprop.p and dicamba, to our knowledge, no research has been conducted on their potential sublethal effect on birds.
It is important to note that the correlations we found between pesticide use and bird abundance need not reflect causative relationships. For example, the use of pesticides may be an indicator of a neater garden, lower tolerance of pests and weeds, or less interest in encouraging wildlife, and hence may correlate with other gardening behaviours that may influence bird abundance.

Other variable effects
We included in our analysis the presence of pets (cats and dogs separately) reported by the respondents of the questionnaire as this has been shown to negatively influence local bird abundance by increasing the pressure of predation (Beckerman et al., 2007;Belaire et al., 2014;MacGregor-Fors and Schondube, 2011;van Heezik et al., 2010;Loss et al., 2013;Gonçalves et al., 2021). Overall, we found no significant effect of pet ownership on bird abundance or richness. However, it should be noted that our analysis did not include pets of neighbours, which may often enter gardens, and which may obscure any pattern. When analysing bird species individually, dunnock and greenfinch seemed to be impacted negatively by the presence of cats. In contrast, we found that collared doves seemed to be positively influenced by the presence of cats. While negative relationships might be easily explaineddogs might disturb and chase birds, while cats may kill themthe positive relationship for collared doves could potentially be explained by the presence of cat food left outside on which they occasionally feed.
The use of ferric iron molluscicide pellets in gardens were not found to influence bird richness, abundance or any of the specific bird species.
Our results showed that general bird abundance was influenced by the garden's latitude with 6 out the 10 bird species found to be influenced either positively or negatively. This finding emphasises even more the many variables that affect bird abundance.

Conclusion
Our study shows the importance of a wildlife-friendly garden in supporting bird populations, even in an urbanised environment. Urban gardens recorded up to 40 bird species if suitably managed, demonstrating that gardeners can mitigate the strong negative influence of urbanisation on bird diversity. Further research is needed to reveal the mechanisms by which urbanisation, habitat quality and pesticide use influence bird populations, such as via effects on food availability or habitats providing nest sites or shelter from predators, or via toxic effects of pesticides. Manipulative experiments, for example by asking some gardeners to implement specific changes to management, would be particularly valuable though challenging to perform.
It is likely that members of the BTO and GBW scheme are a biased sample with regard to garden management and pesticide use compared to the average gardener, potentially showing a lower impact of pesticides, so expanding this study to other gardens would be informative.
Overall, it is clear that the actions of gardeners, in terms of which habitats they create in their garden, and whether or not they apply pesticides, has a significant influence on the avian biodiversity of their garden, and simple guidelines could be developed to promote these actions.

Credit authorship contribution statement
Cannelle Tassin de Montaigu conceptualised and designed the study, created the methodology and partial data collection of the data, analysed the data, prepared figures, and tables, authored and reviewed drafts of the paper, and approved the final draft.
Prof. Dave Goulson supervised and validated the study, reviewed and edited drafts of the paper, and approved the final draft.

Data availability
The authors do not have permission to share data.

Declaration of competing interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.