Introduction

Urban greenspaces are an integral part of the modern urban ecosystem. In particular, woody plants not only beautify urban landscapes but also provide important ecosystem services (Jim and Chen 2009). Urban forests enhance the “livability” and human comfort of urban areas, positively affecting people’s physical and psychological health (de Vries et al. 2013; Pouso et al. 2021; Ebi and Bowen 2023). Urban forests absorb carbon dioxide and particulate pollution from air, while releasing oxygen (Hartig et al. 2014; Han et al. 2020). They moderate urban microclimates and mitigate the urban heat island effect (Iungman et al. 2023). Urban forests also provide habitat and food for wildlife (Silva and Fadini 2017). To maximize their benefits to urban ecosystems, it is crucial to understand the variety of factors affecting urban trees, including parasitic plants.

Parasitic plants are divided into semi-parasitic and holoparasitic plants based on their ability (or not) to photosynthesize (Mutlu et al. 2016). Holoparasitic plants, such as Cuscuta spp., do not photosynthesize and are entirely dependent upon their hosts for nutrition. In contrast, mistletoes, belonging to the families Loranthaceae, Misodendraceae, and Santalaceae, are semi-parasitic plants. Mistletoes survive by their own photosynthesis combined with the absorption of (some) carbohydrates, inorganic salts, and water from their hosts via specialized root-like organs called haustoria, which penetrate host xylem (Schulze and Ehleringer 1984; Gairola et al. 2013; Lim et al. 2016). Mistletoes are generally small sub-shrubs that parasitize woody plants, usually trees but sometimes also shrubs and lianas. Depending on where mistletoes first attach to their hosts, they can be further divided into root parasitic or shoot parasitic plants (Lim et al. 2016). In terms of plant physiological metabolism, mistletoe hosts can transpire at much higher rates than specimens of the same species growing nearby that are not parasitized by mistletoes. Mistletoes can also cause mineral deficiencies in their hosts and weaken the oxidative metabolism of host cells (Schulze and Ehleringer 1984). Mistletoes have also been found to reduce tree growth and reproduction for Scots pine (Pinus sylvestris L.) at the forest-stand level (Kollas et al. 2018).

Both abiotic and biotic factors affect the distribution of mistletoes within a given area. Abiotic factors are mainly humidity, temperature, sunshine, and cloud cover, as well as anthropogenic activities (Díaz-Limón et al. 2016). Biotic factors include host tree characteristics such as size, age, canopy structure, diameter at breast height (DBH), pests and disease, as well as animal pollinators and seed distributors (Watson 2001; Skrypnik et al. 2021), with additional host species-specific factors (Díaz-Limón et al. 2016). Mistletoe prevalence in urban greenspaces tends to be higher than nearby natural forests (Díaz-Limón et al. 2016). This is because urban forests are often collections of various monocultures, such as allées and groves, making it easier for mistletoes to find suitable hosts and spread between them (Maruyama et al. 2012). Yet, the parasitism rate of mistletoes at the edges of natural forests also tends to be much higher than inner forests, possibly related to the ideal mistletoe growing conditions (Maruyama et al. 2012).

We, therefore, chose to survey mistletoe distribution across Sichuan University’s Wangjiang Campus in Southwest China’s largest urban conglomeration, Chengdu.Footnote 1 As early as 2012, the phenomenon of mistletoe parasitism on this campus was investigated, finding that larger trees within a species (in terms of height and DBH) were more likely to be infected by mistletoes (Ma et al. 2020). However, with approximately nine years passing since that study’s data collection, we wanted to resurvey the entire campus to better understand the factors influencing mistletoe distribution. The trees on campus have also changed greatly over this period as the campus built-environment developed, but we observed continued mistletoe infestation. Meanwhile, although multiple articles have studied the factors influencing mistletoe distribution (Maruyama et al. 2012; Shaw et al. 2020; Skrypnik et al. 2020), we also sought to study factors influencing the prevalence and intensity of infection. Thus, to better understand possible factors influencing mistletoe parasitism in an urban ecosystem, in the present study, we sought to: (1) survey the distribution of mistletoes on Sichuan University’s Wangjiang Campus, and (2) assess the relationships between various tree-related variables (e.g., tree taxonomy, size, age class, nativity, foliage type, density, and species diversity) with mistletoe prevalence and intensity.

Methods

Study area

We chose to survey mistletoe parasitism on Sichuan University’s Wangjiang Campus in Chengdu, China. As the largest university campus in downtown Chengdu, with an area of more than two square kilometers (30°37’63.2"~30°37’66.7” N and 104°4’59.5"~104°5’54.3” E), Wangjiang Campus lies between the First and Second Ring Roads (north to south), with a major thoroughfare (Kehua North Road) to the west, as well as a major river (Jinjiang) and greenspace (Wangjiang Tower Park) to the east (Fig. 1). The campus has an average altitude of 492 m above sea level and experiences a subtropical humid monsoon climate. The campus has a long history, with development first beginning in 1936, and a relatively mature, but diverse sylvan landscape.

Fig. 1
figure 1

Sichuan University Wangjiang Campus and surroundings. Map Credit: YN & BL

Full size image

Across Wangjiang Campus, there are both woody plants that have grown spontaneously and those intentionally cultivated. Some of China’s most common street trees are found growing in diverse microclimates across the campus, such as: Ginkgo biloba, Cinnamomum camphora, and Osmanthus fragrans. There are 16 trees greater than 100 years old (indicated by protected tree markers), especially on the eastern part of campus, as well as many newly-planted trees, particularly near the campus’ south gate. Wangjiang Campus also contains residential areas and affiliated kindergartens, primary, and secondary schools (Fig. 1), having high human population densities and activity levels. Thus, the campus not only represents a complex collection of plant species, but it is also representative of many other large urban greenspaces. Therefore, Wangjiang Campus represents an ideal site location to investigate the influencing factors of mistletoe distribution, prevalence, and intensity in urban and subtropical environments.

Research methods

Preliminary investigation

From March 5th -8th, 2021, a preliminary investigation was conducted at Wangjiang Campus, with photos of mistletoes taken for identification and general assessment purposes. Most host plants were in the dormant stage, which was convenient for observing mistletoe morphological characteristics (Fig. 2D). Data from this preliminary survey informed the subsequent systematic survey of all trees on campus to document the campus-wide distribution, prevalence, and intensity of mistletoes. Cycads and bamboos were excluded from further investigation as mistletoe parasitism was not observed in these functional groups.

Fig. 2
figure 2

Mistletoe parasitism on Sichuan University’s Wangjiang Campus, Chengdu. (A) Pendulous habit, evergreen foliage, and showy flowers of Taxillus sutchuenensis; (B) Brightly-colored fruit and flowers of T. sutchuenensis; (C) Flowers and immature fruit of Scurrula parasitica; (D) Heavy infestation of evergreen mistletoe clumps (T. sutchuenensis) on a large Koelreuteria paniculata tree while dormant. Root-like haustoria of two species of mistletoe invading the xylem of their hosts, (E)T. sutchuenensis and (F)Scurrula parasitica. Photo Credits: BCS

Full size image

Mistletoe survey

Between March 19th and June 1st, 2021, 2–3 days per week (usually Friday, Saturday, and/or Sunday) were chosen as investigation days. Using an Android smart phone-based APP that our research team developed to record tree survey data, we systematically surveyed all trees and tree-like shrubs on campus except for those growing within the restricted-access, on-campus staff residential areas, affiliated K-secondary schools, and student dormitories. We collected data on both the trees and the presence and abundance of mistletoe clumps. These data were either directly captured by the APP itself (e.g., photographs, GIS location, elevation, etc.) or assessed and measured by survey team members and recorded directly into the APP (e.g., tree species, DBH, crown width, mistletoe presence, mistletoe clump quantity, etc.). All data were uploaded to cloud storage through the APP for subsequent analysis. Morphological features of mistletoes were observed and documented with photographs, including their flowers, fruit, leaves, and haustoria (Fig. 2), with vouchers collected for identification purposes, but we did not differentiate mistletoe species at the tree-level in this survey.

Data analysis

After the field survey concluded, data were cleaned, with accepted Latin names verified using World Flora Online (http://www.worldfloraonline.org/). We surveyed a total of 6,312 woody trees, tree-like shrubs, and arborescent vines, but, of these, only 6,012 could be included in subsequent data analyses. Of the 300 excluded specimens, 150 were dead (140 unidentifiable and 10 identifiable species) but cause of death could not be determined, and 150 trees lacked necessary geographic information due to a technological error that essentially occurred randomly across the campus. For each tree species, we calculated the maximum, minimum, and average crown width. To rapidly survey all trees on campus within a short time window, DBH was only recorded for single-trunk trees (consequently, this variable was not used in every subsequent analysis). We summed number of mistletoe clumps per tree (Skrypnik et al. 2020).

To better understand the relationship between mistletoe infection and tree species-related traits, we compiled additional variables for each taxa including nativity (e.g., Sichuan Basin or not) and foliage type (e.g., deciduous or evergreen). For taxa with at least one mistletoe-infected specimen, we summed host tree number and mistletoe clumps per species. We then calculated three mistletoe infection indices, including prevalence rate (PR), parasitic intensity (PI), and preference ratio (PrefR). Where for a given species (sp), Nhost(sp) refers to the host tree count for a given species, Ntotal(sp) refers to total specimens for a given species on campus, and Nclumps(sp) refers to total mistletoe clumps for a given species. Across all trees on campus (ALL), ALLhost refers to total number of host trees across all species and ALLtotal refers to the total number of specimens across all species on campus:

$${\text{P}\text{R}}_{sp} = \frac{{N}_{host\left(sp\right)}}{{N}_{total\left(sp\right)}}$$
(1)
$${\text{P}\text{I}}_{sp} = \frac{ {N}_{clumps\left(sp\right)}}{{N}_{host\left(sp\right)}}$$
(2)
$${\rm{Pref}}{{\rm{R}}_{sp}} = \frac{{{N_{host\left( {sp} \right)}} \div AL{L_{host}}}}{{{N_{total\left( {sp} \right)}} \div AL{L_{total}}}}$$
(3)

To assess the impact of tree density and species diversity on mistletoes, we used ArcGIS to calibrate the latitude and longitude of each tree, then imported their coordinates into the campus basemap. The average crown width on campus was six meters, but due to the diversity of average crown widths by species, in ArcGIS we calculated buffers for each tree of three, six, and ten-meter radii, then used “spatial join” with pertinent variables for subsequent analyses. The number of tree GIS locations that fell within each buffer circle represented the tree density variable for the tree at that buffer radius. Similarly, the number of unique tree species within each buffer circle represented the diversity variable.

With the large number of zeros in our mistletoe count data (e.g., many uninfected trees), choosing an appropriate model was very important. In most count datasets with zero-inflated data, the conditional variance is greater than the conditional mean (usually much greater), known as overdispersion, which has consequences for choosing statistical models (Dean 1992; Xekalaki 2014). This can cause underestimation of standard errors and lead to overconfidence in results (e.g., incorrectly rejecting H0). When overdispersion makes Poisson regression inadequate, negative binomial regression models can be used (Bhaktha 2018). We, therefore, used a hurdle negative binomial regression model, using package “pscl” in R (version 4.2.0) (Zeileis et al. 2008; Jackman et al. 2020).

In negative binomial regression models, the object is to model the conditional mean E(Y|X):

$$\text{log}\left(E\left({Y}_{i}|{X}_{i}\right)\right)= {\beta }_{0}+{\beta }_{1}\times {X}_{1i}+\dots +{\beta }_{p}\times {X}_{pi}$$
(4)

where Y is the outcome variable and X represents the predictor variable(s) believed to be associated with Y. With log transformation, the coefficients do not have a simple linear interpretation. The outcome (e.g., response) variable in the model is conceptualized as a rate. Positive coefficients (estimates) indicate higher rates (positive associations) between the two variables and negative coefficients indicate lower rates (negative correlations).

To understand the relative importance of each predictor variable, assessing variables by odds ratios (OR) and Incidence Rate Ratios (IRR) is more intuitive (Obiegala et al. 2021). The OR and IRR as exponentiated coefficients are multiplicative, with OR and IRR indicating the effect of one unit change of the predictor on the rate of the response (Bhaktha 2018). That is, “if eβ > 1 then the rate increases and if eβ < 1 then the rate decreases for each unit change of the respective predictor” (Bhaktha 2018). The OR and IRR can be converted to percent change by using (eβ)*100, so one “unit change in the predictor causes (eβ)*100 increase or decrease in the rate” (Bhaktha 2018).

To build our models, we chose mistletoe clumps per tree as the dependent variable. Additional covariates, including crown width (“cw”), recently planted trees (“recent”), tree foliage type (e.g., evergreen or deciduous; “foliage”), and nativity (e.g., Sichuan Basin or not; “nativity”) were treated as independent variables. Tree density (“den”) and tree species diversity (“div”) were also included as independent variables and the three buffer sizes (3 m, 6 m, and 10 m) were compared. We originally included an additional explanatory variable for recently pruned (e.g., pollarded; “prun”) trees, expecting it to have an impact on our model. However, we came to believe it was a bias rather than explanatory variable, so we chose to exclude the 590 pollarded trees from our model, with 5,422 trees retained in the final analysis. Thus, we compared three models, with the following model equations:

$$\begin{array}{l} \text{M}\text{o}\text{d}\text{e}\text{l}\,1\,=\,\text{m}\text{i}\text{s}\text{t}\text{l}\text{e}\text{t}\text{o}\text{e}\,\text{c}\text{l}\text{u}\text{m}\text{p}\text{s}\,\sim\,\text{c}\text{w}\,+\text{d}\text{i}\text{v}3\,+\text{d}\text{e}\text{n}3\,+\,\text{r}\text{e}\text{c}\text{e}\text{n}\text{t}\\+\,\text{f}\text{o}\text{l}\text{i}\text{a}\text{g}\text{e}\,+\,\text{n}\text{a}\text{t}\text{i}\text{v}\text{i}\text{t}\text{y},\,\text{d}\text{i}\text{s}\text{t}\text{r}\text{i}\text{b}\text{u}\text{t}\text{i}\text{o}\text{n}\\=\,\text{"}\text{n}\text{e}\text{g}\text{a}\text{t}\text{i}\text{v}\text{e}\,\text{b}\text{i}\text{n}\text{o}\text{m}\text{i}\text{a}\text{l}\text{"}\end{array}$$
(5)
$$\begin{array}{l} \text{M}\text{o}\text{d}\text{e}\text{l}\,2\,=\,\text{m}\text{i}\text{s}\text{t}\text{l}\text{e}\text{t}\text{o}\text{e}\,\text{c}\text{l}\text{u}\text{m}\text{p}\text{s}\,\sim\,\text{c}\text{w}\,+\text{d}\text{i}\text{v}6\,+\text{d}\text{e}\text{n}6\,+\,\text{r}\text{e}\text{c}\text{e}\text{n}\text{t}\\+\,\text{f}\text{o}\text{l}\text{i}\text{a}\text{g}\text{e}\,+\,\text{n}\text{a}\text{t}\text{i}\text{v}\text{i}\text{t}\text{y},\,\text{d}\text{i}\text{s}\text{t}\text{r}\text{i}\text{b}\text{u}\text{t}\text{i}\text{o}\text{n}\\=\,\text{"}\text{n}\text{e}\text{g}\text{a}\text{t}\text{i}\text{v}\text{e}\,\text{b}\text{i}\text{n}\text{o}\text{m}\text{i}\text{a}\text{l}\text{"}\end{array}$$
(6)
$$\begin{array}{l}\text{M}\text{o}\text{d}\text{e}\text{l}\,3\,=\,\text{m}\text{i}\text{s}\text{t}\text{l}\text{e}\text{t}\text{o}\text{e}\,\text{c}\text{l}\text{u}\text{m}\text{p}\text{s}\,\sim\,\text{c}\text{w}\,+\text{d}\text{i}\text{v}10\,+\text{d}\text{e}\text{n}10\,+\,\text{r}\text{e}\text{c}\text{e}\text{n}\text{t}\\+\,\text{f}\text{o}\text{l}\text{i}\text{a}\text{g}\text{e}\,+\,\text{n}\text{a}\text{t}\text{i}\text{v}\text{i}\text{t}\text{y},\,\text{d}\text{i}\text{s}\text{t}\text{r}\text{i}\text{b}\text{u}\text{t}\text{i}\text{o}\text{n}\\=\,\text{"}\text{n}\text{e}\text{g}\text{a}\text{t}\text{i}\text{v}\text{e}\,\text{b}\text{i}\text{n}\text{o}\text{m}\text{i}\text{a}\text{l}\text{"}\end{array}$$
(7)

Akaike information criterion (AIC) (Akaike 1973) was used to select the best fit model (Feng 2021). AIC is computed as AIC = 2k–2L, where L is the maximum value of the likelihood function for the model and k is the number of estimated parameters in the model (Bolker 2022). AIC scores are represented as ∆AIC scores, or the difference between the best model (smallest AIC) and each model’s AIC (so the best model has a ∆AIC of zero). R package “sjPlot” was used to visualize factors influencing mistletoe prevalence and intensity using OR and IRR (Lüdecke et al. 2021).

Results

Mistletoe distribution

We surveyed 6,012 living trees, shrubs, and arborescent vines with accurate GIS locations, representing 96 species, 76 genera, and 44 families on Wangjiang Campus. Average crown width was 6.32 m (min: 0.09 m; max: 26.0 m). Spatially, mistletoe host trees tended to be concentrated on the eastern part of campus and closer to waterbodies (Figs. 1 and 3). Note, morphologically similar infraspecific taxa were lumped together into their respective species for subsequent analyses (Supplementary Materials, Table SM1).

Fig. 3
figure 3

Spatial distribution of trees with (red) and without mistletoes (teal) on Sichuan University’s Wangjiang Campus (A). Wangjiang Campus’s North Gate with lotus ponds and Salix babylonica behind (B). Lotus (Nelumbo nucifera) in flower near the North Gate (C). Map credit: YN & BL; Photo Credits: BCS

Full size image

Species with the highest abundance on campus (e.g., greatest exposure to potential mistletoe infection) were Cinnamomum camphora (n = 729), Ficus microcarpa (506), Ginkgo biloba (440), Platanus acerifolia (432), Cinnamomum burmanni (356), Ficus virens (337), Osmanthus fragrans (289), Lindera megaphylla (222), Chaenomeles speciosa (210), and Metasequoia glyptostroboides (197). Overall, 353 trees hosted mistletoe, that is 58.72 trees per thousand (Prevalence: 5.87%). Species with highest mistletoe prevalence for samples over 10 specimens (Table 1), were Citrus maxima (44.4%), Salix babylonica (44.0%), Robinia pseudoacacia (42.7%), and Magnolia liliflora (35.7%). Fifty species on campus (52.1%) had no mistletoe at all (Table SM1). We documented two mistletoe species on campus, Scurrula parasitica L. and Taxillus sutchuenensis (Lecomte) Danser. (Fig. 2). Yet, of the 353 host trees, few hosted both mistletoe species, one Ginkgo biloba, one Prunus cerasifera, and one Robinia pseudoacacia.

Table 1 Mistletoe host tree species listed in descending order by mistletoe preference ratio. Dashed line separates top 23 from bottom 23 host tree species. Nativity refers to species native to the Sichuan Basin
Full size table

Of the 19 mistletoe host species with highest exposure (> 50 specimens), seven averaged between one to three mistletoe clumps per infected tree (Parasitic Intensity, PI; Table 1), in ascending order: Chaenomeles speciosa (1 host tree; PI: 1.0); Platycladus orientalis (1; 1.0); Acer palmatum (3; 2.0); Osmanthus fragrans (9; 2.0); Platanus acerifolia (20; 2.45); Prunus cerasifera (32; 2.563); and Ficus virens (5; 2.6). Nine species had PI ranging between 3 and 5 clumps per infected tree: Cedrus deodara (1 host tree; PI: 3.0); Lagerstroemia indica (2; 3.0); Metasequoia glyptostroboides (34; 3.206); Cinnamomum septentrionale (6; 3.5); Chimonanthus praecox (4; 3.75); Ligustrum lucidum (28; 3.786); Cinnamomum burmanni (7; 3.857); Ginkgo biloba (30; 3.933); and Magnolia grandiflora (19; 4.737). Three species had PI greater than five clumps per infected tree: Cinnamomum camphora (39; 5.41); Robinia pseudoacacia (32; 6.625); and Lindera megaphylla (1; 11.0). Regardless of exposure level, the five species with the highest PI overall in descending order were: Euonymus maackii (2 host trees; PI: 11.5); Lindera megaphylla (1; 11.0); Celtis sinensis (1; 10.0); Jacaranda mimosifolia (1; 8.0); and Broussonetia papyrifera (2; 7.5).

Model analysis and selection

For the three models (excluding the “recently pruned” variable), the ∆AIC were 53.0 (Model 1), 24.7 (Model 2), and 0.0 (Model 3), indicating Model 3 was best. The ∆logLik values, which represent the difference between each likelihood and the maximum, were 0.0, 14.1, and 26.5. When comparing the best fit model (Table 2; Fig. 4) with the best fit model when including the “recently pruned” variable (Table SM2; Fig SM1), the results are very similar.

Table 2 Parameters affecting the prevalence (A, zero count process) and the intensity (B, positive count process) of mistletoe infestation on Wangjiang Campus with the Hurdle model and negative binomial distribution
Full size table
Fig. 4
figure 4

Parameters affecting the prevalence (A, zero-inflated model) and intensity (B, conditional/count model) of mistletoes using a hurdle model. Results are shown with 95% confidence interval. The vertical black line (1.0) indicates no association. Significance: ***: P-value < 0.001, **: P-value < 0.01, *: P-value < 0.05

Full size image

Factors affecting mistletoe prevalence and intensity

Our model showed four independent variables significantly affected mistletoe prevalence on campus (Table 2A). Three, including crown width (β = 0.06121 ± 0.01494; p < 0.001), density (β = 0.03022 ± 0.01132; p < 0.01), and diversity (β = 0.22163 ± 0.03337; p < 0.001), positively correlated with mistletoe prevalence. In contrast, evergreen foliage (β= -0.65874 ± 0.12137; p < 0.001) had a significantly negative correlation. Thus, deciduous trees were more likely to become infected by mistletoes (Table 2A). The second part of the hurdle model only used positive count data of mistletoe clumps (Table 2B), representing mistletoe infection intensity. Results showed that only one factor, crown width (β = 0.076558 ± 0.018606; p < 0.001), significantly affected mistletoe infection intensity. That is, trees with larger crown widths are likely to have higher mistletoe infection intensities (e.g., more mistletoe clumps per infected tree).

Based on Fig. 4A, the factors that are positively or negatively correlated with mistletoe infection rate (e.g., prevalence) are the same, but the relative importance of each is more evident. Factors significantly influencing mistletoe prevalence on campus were: (1) tree foliage (OR 0.52), meaning evergreen trees were 52% less likely to be infected, or deciduous trees were 48% more likely; (2) tree species diversity (OR 1.25), meaning a tree is 25% more likely to be infected for every increase of its surrounding tree species diversity by one species; (3) crown width (OR 1.06), meaning that a tree is 6% more likely to be infected for every increase in crown width by one meter; and (4) tree density (OR 1.03), meaning a tree is 3% more likely to be infected for every increase of tree density by one. Based on Fig. 4B, for every increase in an infected tree’s crown width by one meter (IRR 1.08), mistletoe clumps increase by 8%.

In total, of the 6,012 living trees, shrubs, and arborescent vines with accurate GIS locations in our study area, 590 had recently undergone major mechanical cutback (pruning), essentially being pollarded. A small number were scattered across several species. That is, for most species with pollarded specimens on campus, this treatment was only given to a relatively small percentage of the overall specimens: Cinnamomum camphora (3/729; 0.4%), Erythrina variegata (2/8; 25%), Ficus microcarpa (1/506; 0.2%), Ficus virens (4/337; 1.2%), Ginkgo biloba (2/440; 0.5%), Ligustrum lucidum (19/145; 13.1%), and Prunus cerasifera (1/142; 0.7%). However, pollarding was the preferred management method for two species: Lagerstroemia indica (178/191; 93.2%) and Platanus acerifolia (380/432; 88.0%). Although we did include the “recently pruned” variable in our original model (Table SM2; Fig. SM1), we came to believe this was a bias, rather than an explanatory variable. This was confirmed in that pruned trees had significantly smaller crown widths and significantly less mistletoes than unpruned trees (Fig. 5). The results of our original model showed values less than one for the “recently pruned” variable, for both prevalence and intensity, indicating that pruning did have a negative impact on both measures of mistletoe infection.

Fig. 5
figure 5

Linear regression between DBH and crown width (A) and comparisons of crown width (B), DBH (C), and mistletoe clumps (D) for pruned and unpruned trees

Full size image

Discussion

Globally, according to the Angiosperm Phylogeny Group (Stevens 2017), 1,478 species of mistletoes representing 85 genera have been identified across three families: Loranthaceae (950 species; 77 genera); Misodendraceae (8 species; 1 genus), and Santalaceae (520 species; 7 genera). They occur on all continents and major island groups except Antarctica, with most being found in tropical and subtropical regions. Two mistletoe genera, Scurrula (Loranthaceae) and Viscum (Santalaceae), are common across Eurasia (Lim et al. 2016). There are 51 mistletoe species in China of which 18 species are endemic (Qiu and Gilbert 2003). Although we did not assess mistletoe species at the tree level in our survey, we did document two species on Sichuan University’s Wangjiang Campus, Scurrula parasitica and Taxillus sutchuenensis (both Loranthaceae).

Mistletoes often exhibit long, overlapping flowering and fruiting seasons, with abundant supplies of nectar and fleshy fruit, thereby attracting avian pollinators and seed dispersers and directly contributing to the alpha diversity of individual ecosystems (Watson 2001). Human-modified urban landscapes exhibit a difficult suite of environmental stressors favoring certain tree species over others, and monocultured tree-lined street canyons are by themselves very inhospitable to faunal diversity. Yet, mistletoes have been found to increase avian biodiversity in human-modified monoculture landscapes like rubber tree plantations (Sreekar et al. 2016). Mistletoe clumps provide nesting and roosting sites for bird species representing at least 43 families around the world, and insectivorous birds frequently forage for insects on mistletoe clumps (Watson 2001). A South African study found exotic tree species were more likely than native species to be parasitized by mistletoes, but the exotic trees were less favored by bird species overall (Shackleton 2016). Thus, native mistletoe species may actually increase the suitability of exotic trees to meet the foraging needs of native bird species. For example, we found many mistletoe clumps colonizing trees that otherwise lacked fleshy, edible fruit (e.g., Ginkgo, Koelreuteria, Metasequoia, Platanus, Populus, Salix). This indicates mistletoes may increase avian carrying capacity in otherwise harsh urban environments, especially for the passerine bird species that frequently consume mistletoe fruits (Whelan et al. 2008).

General understanding of mistletoe distribution on Wangjiang Campus

According to our results, mistletoe host trees were largely concentrated on the eastern part of campus, with only a few scattered host populations elsewhere (Fig. 3A). There are several possible reasons for the greater mistletoe infestation in the eastern part of campus:

1) This area is very close to Wangjiang Tower Park (Fig. 1). Wangjiang Tower Park is a large greenspace, surrounded on three sides (north, east, and south) by a bend in the Jinjiang River, and adjacent to our study area on its western side. It is a relatively mature landscape with rich vegetation, attracting many kinds of birds to roost and nest. This may also contribute to the large number of birds we observed in our study area as well, including Chinese bulbul (Pycnonotus sinensis), red-billed leothrix (Leothrix lutea), and fire-breasted flowerpecker (Dicaeum ignipectus). Moreover, due to its proximity to the Jinjiang River, the humidity in the park is relatively high, providing good moisture conditions for mistletoe seeds to germinate (Skrypnik et al. 2020). We have also found mistletoes heavily infesting the trees in Wangjiang Tower Park (unpublished data), possibly being the original mistletoe infection source.

2) Mistletoe hosts also tended to cluster near waterbodies, including the waterlily pool on east campus and the Jinjiang River that flows around the eastern campus boundary (Figs. 1 and 3). This may be due to the importance of water for mistletoe seed germination, with the higher ambient humidity and the waterbodies themselves providing ideal conditions for both seed germination and habitat for avian mistletoe dispersers (Skrypnik et al. 2020). Moreover, two large, man-made lotus ponds are situated near the campus’ North Gate in the northwest part of our study area (Figs. 1 and 3). Around these pools grow many Salix babylonica (19 specimens), and about half of these (9/19) were infested with mistletoes. In contrast, relatively few mistletoe host trees were in the southern and western parts of campus.

3) The trees on the eastern part of our study area were found to be growing more densely and with older, more mature specimens (including 16 specimens > 100 years old). Many trees there had overlapping crowns, so their close physical association likely facilitated mistletoe spread (Gairola et al. 2013). The greater tree size and density in the eastern part of campus may also make this area more hospitable to birds, thereby increasing mistletoe spread (Maruyama et al. 2012).

4) Many trees growing on the eastern part of campus were distributed in green patchy areas, relatively farther away from roads with less overall human interference, possibly being more conducive to mistletoe spread. This is also because trees in green patches on campus are less frequently pruned than trees growing near roads and walkways, meaning that mistletoes there were less likely to be mechanically removed. At the same time, avian mistletoe dispersers and pollinators may also prefer nesting in the green patch areas with less overall human intervention.

These results are similar to those from a study in Russia’s Kaliningrad, which found most mistletoe host trees were in greenspaces (urban parks), historic housing complexes, and greenspaces along natural or man-made lakes, ponds, and waterways (Skrypnik et al. 2020). Similarly, a study in New South Wales, Australia, found that mistletoes are more abundant in riparian areas and floodplain woodlands (Watson and Herring 2012).

Factors affecting mistletoe prevalence

In terms of mistletoe infection potential, 24 species representing 16 families were highly exposed on campus (e.g., abundance > 50 specimens; Table 1 & SM1). Of these, five species, Ficus microcarpa (n = 506), Grevillea robusta (100), Livistona chinensis (110), Podocarpus macrophyllus (55), Trachycarpus fortunei (108) and one family (Arecaceae) had no mistletoes at all. Two of these species were palms, all were evergreen, and 60% were non-native. In contrast, the five species that were highly exposed with high mistletoe prevalence rates were Ligustrum lucidum (145; 19.3%), Magnolia grandiflora (94; 20.2%), Metasequoia glyptostroboides (197; 17.3%), Prunus cerasifera (142; 22.5%), and Robinia pseudoacacia (75; 42.7%). Of these, 60% were deciduous and 60% were non-native (Table 1). Although studies in South Africa (Shackleton 2016) and Brazil (Silva and Fadini 2017) found exotic trees were more likely to be parasitized by mistletoes than native species, nativity did not prove significant in our study (Table 2A).

According to Fig. 4A, the most important factors influencing the prevalence of mistletoe infection on campus in order of importance, were: (1) tree foliage, (2) tree species diversity, (3) crown width, and (4) tree density. The most influential variable affecting mistletoe prevalence displayed a negative effect (e.g., evergreen foliage). Even though 60 of the 96 tree species on campus were deciduous (62.5%), only 29 of the 50 non-host species (58.0%) were deciduous (Table SM1). Moreover, 80% (12) of the fifteen mistletoe host tree species with highest mistletoe prevalence rates were deciduous, but for the bottom fifteen species, only 46.7% (7) were deciduous (Table 1). Studies have found mistletoes require high light levels to germinate and successfully establish on new hosts (Watson 2001). Thus, our findings that evergreen trees are less likely to become infected by mistletoes may be because the dense shade found in evergreen canopies inhibit mistletoe establishment.

Studies in and outside of China have found larger trees are more likely to be infected by mistletoes (Maruyama et al. 2012; Gairola et al. 2013; Ren 2015; Díaz-Limón et al. 2016; Sreekar et al. 2016; Ma et al. 2020). This is likely because wide, airy canopy structures are best for both mistletoe establishment (Watson 2001) and the nesting, roosting, and foraging preferences of their avian dispersers (Whelan et al. 2008). We found similar results with the crown width variable, in that tree species diversity, crown width, and tree density each displayed a significant, positive effect on mistletoe prevalence (Table 2A, Fig. 4A). For larger trees with wider crown widths, when the density of trees growing around them is higher, there is more overlap between trees, with more opportunities for mistletoes to spread between them. But, comparing the tree density and diversity variables, tree diversity proved to be more important for mistletoe prevalence, possibly due to the greater faunal diversity supported by more diverse urban forests, particularly avian diversity (Whelan et al. 2008; Watson and Herring 2012).

Mistletoe parasitic intensity and preference

According to our model results (Table 2B; Fig. 4B), only one variable significantly influenced mistletoe infection intensity. Crown width positively affected infection intensity, meaning larger trees were not only more likely to be infected by mistletoes (prevalence), but also more likely to have more intense infections. Moreover, when looking at host species ranked by mistletoe Preference Ratio (Table 1), any result greater than 1.0 means the mistletoe infection is greater than would be expected by chance, while those less than 1.0 represent active avoidance, and results close to 1.0 indicates infection is random (Shackleton 2016). Preference Ratio results indicate very few species could be understood as randomly infected. Even ignoring the 50 species with no mistletoes at all, 14 of the 46 mistletoe host species have Preference Ratios less than 0.9 (12 were less than 0.6), indicating these species were actively avoided by mistletoes. That is, these species were not ideal for successful mistletoe establishment or were largely inhospitable to their avian dispersers (Whelan et al. 2008). In contrast, 54.3% of mistletoe host tree species on campus had Preference Ratios exceeding 2.0 (25/46), with the top 50% exceeding 3.0. Four species had relatively few specimens on campus (e.g., very low exposures), but each had Preference Ratios exceeding 13.0 (Table 1): Euonymus maackii (2 host trees; PrefR: 17.031); Excoecaria acerifolia (1; 17.031); Wisteria villosa (2; 17.031); and Sapium sebiferum (4; 13.625). Thus, certain species are clearly preferred as mistletoe hosts regardless of their relative exposure.

Impact of pruning on mistletoes in human-managed landscapes

Intuitively, we found pruned trees in our study had significantly larger DBH, smaller crown widths, and fewer mistletoe clumps than comparable unpruned trees (Fig. 5). That is to say, pruning (e.g., pollarding) both removes mistletoes and decreases the crown width to DBH ratio. This may explain the unexpected results we found for other variables in our study. For example, despite our expectation that exotic tree species might be more likely parasitized by mistletoes (Shackleton 2016; Silva and Fadini 2017), in our study, the nativity variable did not show significance either way. We found 60% of highly exposed species without any mistletoes and 60% of highly exposed species with high mistletoe prevalence rates were not native to the Sichuan Basin. But, of the 590 pollarded specimens, only 206 were native, meaning that non-native specimens accounted for 65.1% of the pollarded trees. Looking at it another way, native species had a greater exposure to mistletoe infection on campus overall (accounting for 61.0% of total specimens), but in terms of pollarding, 16.4% of non-native trees were pollarded (384/2,342), but only 5.6% of native trees were pollarded (206/3,670). Thus, in human-managed landscapes, such as a university campus, the effect of variables like tree nativity are likely confounded by management variables, such as pruning.

Moreover, multiple studies found mistletoes tend to preferentially parasitize larger trees (Maruyama et al. 2012; Gairola et al. 2013; Díaz-Limón et al. 2016; Shackleton 2016; Sreekar et al. 2016). Previous studies from Chengdu similarly found that larger trees within a species (in terms of height and DBH) were more likely to be infected by mistletoes (Ren 2015; Ma et al. 2020). Although we found similar results in that trees with larger crown widths were significantly more likely to have mistletoes (e.g., prevalence) and have more mistletoe clumps per infected tree (e.g., infection intensity) than trees with smaller crown widths, not all traditional tree size measures are equal in predicting mistletoe infection. Our findings showed that DBH may not be a good proxy for tree “size” when predicting likelihood of mistletoe infection in human managed landscapes (such as university campuses). Though trees with larger DBH also tend to have larger canopies, and we did find crown width significantly correlated with DBH on campus (R2 = 0.3768; p < 0.001), pruned trees had significantly smaller crown widths, significantly larger DBH, and significantly less likely to be infected by mistletoe than unpruned trees (Fig. 5). Consequently, when studying mistletoes in human-managed landscapes, measures of tree size that assess the actual size of the crown, such as tree height (Ma et al. 2020) or crown width should be preferred over DBH.

Limitations and recommendations for future research

It is important to note that our “pruning” variable (later excluded from our final model) was limited only to those specimens that displayed signs of significant recent mechanical cutback (e.g., pollarding, a rather extreme form of pruning). However, we know other trees on campus had also undergone recent pruning (including, in some cases, the mechanical removal of mistletoes), but the lack of detailed pruning history records made the inclusion of these pruning events impossible in our study. Nevertheless, the uncertainties related to this variable merit further research.

More research is not only necessary to assess patterns of urban mistletoe distribution, prevalence, and infection intensity, but also of bird-mistletoe interactions and the positive and negative tradeoffs of the ecosystem services and disservices that mistletoes provide. This is especially pertinent for understanding faunal biodiversity dynamics in Chinese urban forests. From the landscape perspective, mistletoes are often quite beautiful, with attractive, usually-pendulous, evergreen foliage, showy flowers, and brightly-colored fruits (Ren 2015; Watson 2001; Fig. 2). But mistletoes are widely viewed as invasive pests, perceived as damaging to trees and detrimental to forest health (Watson 2001; Maruyama et al. 2012; Duan et al. 2021). But, in Australia, mechanical removal of mistletoes resulted in declines of one-quarter to one-third of bird species in manipulative wood plot experiments (Watson and Herring 2012). Nevertheless, studies on mistletoe distribution in Chinese cities often end with recommendations to completely eradicate mistletoes (mechanically and/or chemically) from the urban forest due to their perceived threat to tree vitality (See for example: Duan et al. 2021). But these recommendations are not based on hypothesis-driven manipulative studies, or sufficient scientific data to support these recommendations. Indeed, previous urban forestry studies have noted Chinese cities often lack science-based urban forest policies and data-driven best management practices (Wang et al. 2018). Yet, some mistletoe species could be functionally acting more as epiphytes than parasites and their possible negative effects on trees could be outweighed by the ecosystem services mistletoes provide (Koenig et al. 2018). Since mistletoes provide many ecosystem services that are particularly important in the complex ecosystems of modern urban forests, more studies are necessary in Chinese cities to ensure mistletoe management decisions are adequately supported by data.

Moreover, beyond the variables in our model, other factors that could affect mistletoe distribution, prevalence, and intensity should be considered in future models. For example, mistletoe distribution is closely related to multiple environmental factors, including soil fertility (Watson 2001), soil moisture and ambient humidity (Skrypnik et al. 2020), as well as the ecological behaviors and preferences of mistletoe seed and pollen dispersers (Watson and Rawsthorne 2013). Further, of all 353 mistletoe host trees at the time of our survey, we found only three that hosted both mistletoe species, one Ginkgo biloba, one Prunus cerasifera, and one Robinia pseudoacacia. But, by August 2022, we found an additional P. cerasifera and four Chaenomeles speciosa infected by both mistletoe species, underscoring the dynamic nature of mistletoe parasitism. Yet, all eight trees varied in number of mistletoe clumps (e.g., parasitic intensity) from both mistletoe species. Either one or the other dominated. This suggests the two mistletoe species may be in competition. Future studies should explore this in more detail by identifying mistletoe species at the host tree level and investigate the relative distribution, prevalence, and infection intensity for each mistletoe species separately and over time. This will help reveal what relationship may exist between mistletoe species and whether they differ in host preferences.

Conclusion

After surveying mistletoe distribution, prevalence, and infection intensity on Sichuan University’s Wangjiang Campus, 353 landscape trees hosted mistletoes out of 6,012 total trees. Rather than assume mistletoe infection is necessarily detrimental to landscape-level health, more research is necessary to determine the differential impact at various scales (e.g., individual tree, species, landscape, and city-wide scales). Since bird preferences and community ecology influence which trees are parasitized by mistletoes, more research is necessary to explore these interactions in different urban contexts, especially related to the magnitude of ecosystem services and disservices these relationships provide. More research is also necessary to understand how different types of pruning affect mistletoe distribution, prevalence, and intensity over time, especially in different socio-cultural and environmental contexts.