Anthropogenic Drivers of Small Island Effects in Urban Remnant Woody Plants

1. Introduction

As urbanization accelerates, natural habitats within and surrounding cities are increasingly being replaced by artificial green spaces or becoming fragmented, leading to a significantly decline in biodiversity [1]. While a few cities with unique geomorphological features still retain some remnants of natural vegetation, these exist in an isolated, island-like distribution within an urban landscape [2]. These urban remnant vegetation patches provide essential habitat for urban wildlife, and offer various multiple ecosystem services [3,4,5]. Additionally, remnant vegetation preserves intact zonal vegetation characteristics, adapting to local climatic conditions, and functions as a vital seed bank for ecological succession, as well as a benchmark for the restoration of near-natural habitats [6]. However, widespread habitat fragmentation and replacement have significantly damaged these patches worldwide, leaving few intact remnants [7,8]. The fragmentation and replacement of these habitats not only disrupt ecological balance but also diminish the resilience of urban ecosystems to environmental changes [9]. This disruption of ecological balance and reduced resilience of urban ecosystems to environmental changes underscore the urgent need for conservation efforts [10]. The conservation and restoration of urban remnant vegetation patches have become a priority in recent years, with scholars focusing on their protection and the impact of habitat fragmentation on vegetation species diversity [11,12]. Since species-area relationships differ among plant life forms [13], by understand these factors in relation to the different life forms of vegetation, we can develop effective strategies to preserve biodiversity in urban environments and mitigate the negative consequences of urbanization.

Island biogeography theory (IBT) is a cornerstone of conservation biology, offering valuable insights into the effects of habitat fragmentation [14]. The concept of IBT has been proven invaluable for understanding patterns of biodiversity in urban ecosystems. Urban green spaces, often isolated by impervious surfaces and built structures, can be conceptualized as islands within a sea of urban matrix [15,16]. The characteristics of this matrix, such as its permeability and the presence of corridors, can significantly influence the movement of organisms between these urban “islands”, consequently, their biodiversity [17]. As one of the focal points of research on island biogeography, the Small Island Effect (SIE) describes a phenomenon when below a certain threshold area, species richness is no longer directly proportional to island area and may even decrease [18]. SIE is particularly relevant for small population conservation, as small populations are more susceptible to stochastic events, and SIE highlights the vulnerability of island ecosystems to species extinction [19]. In urban and landscape ecology, SIE can provide a theoretical foundation for understanding species distribution patterns in fragmented landscapes, aiding in the assessment of habitat quality at various scales [20]. Area thresholds based on SIE emergence allow for more scientific planning of greenfield patches. With the intensification of climate change and human activities, grasping the patterns of SIE existence may have a positive effect on the conservation of urban biodiversity and the maintenance of ecosystem stability [21]. Therefore, a deep understanding of SIE is crucial for predicting future trends in biodiversity [22].

Several hypotheses have been proposed to explain SIE, such as (1) Habitat diversity: smaller islands may experience fewer species deficits due to higher habitat diversity per unit area [23,24]; (2) Extinction: catastrophic events on small islands can lead to total extinctions, preventing species richness from reaching equilibrium levels [25]; (3) Nutrient replenishment: exogenous resources can increase island productivity, affecting species richness. Smaller islands may benefit more from nutrient replenishment per unit area [26]; (4) Disturbance: small islands may be in different stages of disturbance recovery, supporting a wider range of species [27]. While these hypotheses have been explored in various contexts, the existence of SIE in urban remnant vegetation and its underlying factors remain largely understudied.

Qingdao is a hilly coastal city with rich flora, it has undergone rapid urbanization over the past few decades, yet remnants of natural mountainous areas remain within the urban landscape. Its unique geography and swift urbanization offer an ideal setting to study the SIE, particularly in remnant natural vegetation. Understanding these patterns is crucial for guiding biodiversity conservation efforts and enhancing the ecological services provided by urban green spaces. To address these goals, we selected 17 remnant vegetation patches in the urban area of Qingdao, and surveyed the species diversity pattern of remnant vegetation, and by using six species-area relationship regression models tested whether the SIE exists for remnant vegetation with different life form (i.e., annual herb, perennial herb, shrub and tree). For those life forms where SIE was detected, we also determined the threshold area values. Finally, stepwise and generalized linear regression mode was used to analyze the driving factors behind the generation of SIE with different life forms. These findings are expected to offer valuable insights for optimizing urban green space planning and preserving biodiversity in rapidly urbanizing landscapes.

2. Materials and Methods

2.1. Study Area

The study area is located in Qingdao (119°30′-121°00′ E, 35°35′-37°09′ N), a major sub-provincial city in eastern Shandong province, China, with a total land area of 24900 ha, and about 10.37 million people in 2024. Qingdao is one of the richest areas in plant species at the same latitude, was ranked as the “Forest city” in 2015, with a forest cover rate of 44.7% within built-up area [28]. Qingdao is a hilly city, Mt. Laoshan extends to the main urban area, which has led to a special pattern of vegetation habitat fragmentation in the main urban area under the effect of rapid urbanization.

Our study is based on a randomised sampling, 17 remnant vegetation patches located in Qingdao were surveyed, with mean area of 58.97 ha, ranging from 1.61 ha to 588.00 ha, and altitudes between 62 m to 384 m above sea level. These patches are isolated by surrounding urban buildings, scattering across the city, and face varying degrees of fragmentation. Situated on the edge of Qingdao, Mt. Laoshan serves as a key biodiversity reservoir for the region, acting as a "mainland" source of species for the isolated urban patches. Given its role as a species pool, we hypothesize that the distance and isolation from Mt. Laoshan significantly influence the species composition and richness within the remnant patches, providing a valuable opportunity to explore patterns in species diversity through the lens of island biogeography theory (Figure 1).

2.2. Field Surveys

We conducted plant species surveys in April-June and September-November of 2022 and 2023 in 17 remaining urban vegetation patches. Then, classified them as annual herbs, perennial herbs, trees, and shrubs according to life form (classification is based on the vegetation of China, 1980). During the survey, in order to ensure the robustness of the research results, this study applies the sample line survey method, taking into account the north-south slope direction, and sets at least two sample lines for each remnant vegetation patch, and the length of the sample lines is proportional to the patch size. Each sample line of each patch was surveyed by at least two skillful people before and after the survey to ensure a comprehensive survey of plant species. In addition, vegetation types such as cropland and economic forests with high intensity of anthropogenic disturbance were removed from the survey. The final dataset was formed to be used for subsequent research and analysis.

2.3. Data Analysis

We compared the following six continuous piecewise regression models to test for the existence of a SIE (Eq. 1–6; table. 1, [29]. In this study, the Akaike Information Criterion (AIC) was used to evaluate the fit of the model. If two models with ΔAICc less than 2 are considered to have similar performance. Continuous segmented regression models were selected for Models 1-4, and Models 5 and 6 were control models. If models 1-4 are selected as the best models, it suggests the presence of SIE [30]. Where an island effect was identified, the average of the t-values obtained from the optimal segmented regression model was subsequently calculated as the area threshold for that data set.

The effects of each variable on changes in species richness were assessed separately using multiple linear regression and variance decomposition analyses and through an iterative approach. Area and each environmental variable were first log-transformed to enhance normality and variance alignment of the data. The five smallest islands were used to construct the first dataset, and a binary linear regression model was constructed with species richness as the dependent variable and area separately from each environmental variable. Subsequently, multiple linear regression, variance decomposition, and iterative analyses were performed using the “modEvA” and “vegan” packages to calculate the effects of each variable and area on species richness, respectively, after which a larger island was added to the analyses and built. a new dataset, and repeated the construction of the quadratic equation and calculation of the degree of influence until the largest island was added to the dataset. We used generalized linear regression (GLM, family=Poisson) to analyze trends in the effects of each environmental variable on species richness during the iteration process, which were eventually visualized using the “ggplot2” package. In this process, the drivers of SIE are the various environmental variables that have a decreasing effect on species richness [31]. The aforementioned analysis process was conducted in R version 4.2.3.

Table 1. Detection model for small island effects.

model	Equation	Description
1	S=c+(log A≤T)×z1×log A+(log A>T)×((z1×T+z2×(log A-T))	Regression fragments and slope iteration fragments
2	S=c+(log A≤T）(z1×log A+(z2-z1）×T）+(log A＞T)×z2×log A	Regression fragments and slope iteration fragments
3	S=c+(log A>T）×z×(log A-T)	Zero-slope regression fragments and slope iteration fragmentss
4	S=c+(log A≤T)×z1×T+(log A>T)×z1×log A	Regression fragment and direct inheritance fragment
5	S=c+z×log A	Linear regression
6	S=c	Zero-slope regression

*Note: In these equations, S stands for species richness, A for area, while c (intercept), zi (slope) and T (break point) are fitted parameters. Z1 and Z2 denote slopes of first and second non-horizontal segmentation regression lines. In this study, the minimum residual sum-of-squares method was employed for the estimation of the thresholds, and the iterative method was utilized for the nonlinear estimation, with an increase of 0.001 for each iteration. The threshold T corresponding to the minimum residual sum-of-squares was selected as the minimum area threshold.

Table 1. Detection model for small island effects.

model	Equation	Description
1	S=c+(log A≤T)×z1×log A+(log A>T)×((z1×T+z2×(log A-T))	Regression fragments and slope iteration fragments
2	S=c+(log A≤T）(z1×log A+(z2-z1）×T）+(log A＞T)×z2×log A	Regression fragments and slope iteration fragments
3	S=c+(log A>T）×z×(log A-T)	Zero-slope regression fragments and slope iteration fragmentss
4	S=c+(log A≤T)×z1×T+(log A>T)×z1×log A	Regression fragment and direct inheritance fragment
5	S=c+z×log A	Linear regression
6	S=c	Zero-slope regression

*Note: In these equations, S stands for species richness, A for area, while c (intercept), zi (slope) and T (break point) are fitted parameters. Z1 and Z2 denote slopes of first and second non-horizontal segmentation regression lines. In this study, the minimum residual sum-of-squares method was employed for the estimation of the thresholds, and the iterative method was utilized for the nonlinear estimation, with an increase of 0.001 for each iteration. The threshold T corresponding to the minimum residual sum-of-squares was selected as the minimum area threshold.

A total of three key hypotheses were chosen to explain the variables in this study: habitat diversity, nutrient replenishment, and disturbance [32]. Perimeter-area ratio (PAR) and landscape shape index (LSI) represent multiple hypotheses in this study. Specifically, under equal area, patches with longer edges and more complex shapes create a greater diversity of habitats, increasing the opportunity for material exchange with the surrounding environment. However, these patches are also more susceptible to disturbance due to their greater exposure. Variables such as elevation and the proportion of sealed surface within patches （Sealed_patch） represent the habitat diversity hypothesis, reflecting the influence of physical and spatial patch characteristics on species diversity. The nutrient replenishment hypothesis is also represented by the distance to the source of species (DIS), highlighting the potential for nutrient flow and species dispersal from the biodiversity pool (add suitable reference). Finally, the disturbance hypothesis includes variables related to anthropogenic influences, such as sealed surfaces around the patches (Sealed₅₀), anthropogenic disturbance (AT), night light intensity (Light), and GDP, which measure the degree of human impact on the remnant vegetation (add suitable reference). These variables provide a comprehensive framework for assessing the ecological drivers affecting species diversity and the small island effect (SIE) in urban remnant vegetation patches (please see Table 2).

3. Results

3.1. Species Composition

Across the 17 remnant vegetation patches in Qingdao city, we recorded 331 species of 225 genera and 81 families, including 24.77% annual herbaceous species, 36.25% perennial herbs, 21.15% shrub species and 17.82% tree species (Figure 2). Across all patches, the proportion of herbaceous species (mean = 56.99 %, range 35.71% to 70.18%) exceeded that of woody plants (mean = 43.01 %, range 29.83% to 64.29 %). Within herbaceous plants, the proportion of annual herbs among patches (mean = 51.37%, range from 44.44% to 66.67%) exceeded that of perennial herbs (mean = 48.63%, range from 33.33% to 59.14%), Pearson correlation analysis showed that there was a significant correlation between annual and tree (p = 0.02), while there was no significant correlation between any of the other taxa (Figure 3). The highest and lowest richness of species recorded in a single patch were 170 and 33, respectively, whereas the mean species richness was 73. Of all species, the top three frequency family were Asteraceae (50 species), Gramineae (28 species) and Rosaceae (24 species), containing about 38.4% of the total species of remnant vegetation species. The top two frequency genus were Artemisia (9 species) and Erigeron (5 species).

3.2. Detection of the Small Island Effect

The fitting results indicate that annual and perennial herbs do not exhibit the small island effect (SIE), as evidenced by Model 5 being the only optimal model (∆AICc > 2 compared to suboptimal models), suggesting stable species richness across areas and no significant SIE. In contrast, for trees and shrubs, Models 1 to 5 show similar explanatory power (∆AICc < 2), indicating the presence of SIE. The area thresholds for the presence of SIE are 6.38 ha for trees and 11.91 ha for shrubs. Additionally, for total plant species, Models 1 to 5 also have similar explanatory power, with an area threshold of 86.44 ha (table 3).

Table 3. Species-area relationship fitting results for 17 remnant vegetation patches vegetation. AICc, corrected Akaike information criterion; c, intercept; K, number of estimable parameters; T, area threshold; z1, first slope; z2, second slope. The number of Models please check Table 1.

Group	Models	Parameters				Model Comparison
		C	Z1	Z2	T	K	AICc	ΔAICc
Total plants	5	-138.068	40.408	0	0	2	149.04	0
	1	-25.023	16.308	31.382	4.926	3	150.654	1.613
	2	-179.611	16.308	47.689	4.926	3	150.654	1.613
	3	66.875	186.37	0	6.216	2	150.998	1.958
	4	-1091.611	186.37	0	6.216	2	150.998	1.958
	6	72.941	0	0	0	1	166.259	17.218
Annual herb	5	-24.248	8.586	NA	NA	2	103.611	0
	1	-24.328	8.603	-0.026	5.126	3	106.6	2.989
	2	-24.195	8.603	8.577	5.126	3	106.6	2.989
	3	19.5	33.434	NA	6.216	2	110.172	6.561
	4	-188.325	33.434	NA	6.216	2	110.172	6.561
	6	20.588	NA	NA	NA	1	116.198	12.587
Perennial herb	5	-56.323	14.717	NA	NA	2	112.715	0
	1	23.465	-3.211	18.839	4.506	3	114.95	2.235
	2	-61.426	-3.211	15.628	4.506	3	114.95	2.235
	3	18.375	66.19	NA	6.216	2	116.54	3.825
	4	-393.063	66.19	NA	6.216	2	116.54	3.825
	6	20.529	NA	NA	NA	1	131.324	18.609
shrub	3	6.804	10.188	—	4.216	2.000	115.033	0.000
	4	-36.147	10.188	—	4.216	2.000	115.033	0.000
	5	-35.959	10.153	—	—	2.000	115.098	0.065
	1	-9.808	4.794	10.478	5.346	3.000	115.852	0.819
	2	-65.823	4.794	15.272	5.346	3.000	115.852	0.819
	6	17.059	—	—	—	1.000	124.510	9.476
tree	3	7.742	6.977	—	4.216	2.000	107.947	0.000
	4	-21.674	6.977	—	4.216	2.000	107.947	0.000
	5	-21.539	6.952	—	—	2.000	107.999	0.051
	1	7.882	0.754	8.868	5.046	3.000	109.413	1.465
	2	-36.867	0.754	9.622	5.046	3.000	109.413	1.465
	6	14.765	—	—	—	1.000	114.729	6.782

Table 3. Species-area relationship fitting results for 17 remnant vegetation patches vegetation. AICc, corrected Akaike information criterion; c, intercept; K, number of estimable parameters; T, area threshold; z1, first slope; z2, second slope. The number of Models please check Table 1.

Group	Models	Parameters				Model Comparison
		C	Z1	Z2	T	K	AICc	ΔAICc
Total plants	5	-138.068	40.408	0	0	2	149.04	0
	1	-25.023	16.308	31.382	4.926	3	150.654	1.613
	2	-179.611	16.308	47.689	4.926	3	150.654	1.613
	3	66.875	186.37	0	6.216	2	150.998	1.958
	4	-1091.611	186.37	0	6.216	2	150.998	1.958
	6	72.941	0	0	0	1	166.259	17.218
Annual herb	5	-24.248	8.586	NA	NA	2	103.611	0
	1	-24.328	8.603	-0.026	5.126	3	106.6	2.989
	2	-24.195	8.603	8.577	5.126	3	106.6	2.989
	3	19.5	33.434	NA	6.216	2	110.172	6.561
	4	-188.325	33.434	NA	6.216	2	110.172	6.561
	6	20.588	NA	NA	NA	1	116.198	12.587
Perennial herb	5	-56.323	14.717	NA	NA	2	112.715	0
	1	23.465	-3.211	18.839	4.506	3	114.95	2.235
	2	-61.426	-3.211	15.628	4.506	3	114.95	2.235
	3	18.375	66.19	NA	6.216	2	116.54	3.825
	4	-393.063	66.19	NA	6.216	2	116.54	3.825
	6	20.529	NA	NA	NA	1	131.324	18.609
shrub	3	6.804	10.188	—	4.216	2.000	115.033	0.000
	4	-36.147	10.188	—	4.216	2.000	115.033	0.000
	5	-35.959	10.153	—	—	2.000	115.098	0.065
	1	-9.808	4.794	10.478	5.346	3.000	115.852	0.819
	2	-65.823	4.794	15.272	5.346	3.000	115.852	0.819
	6	17.059	—	—	—	1.000	124.510	9.476
tree	3	7.742	6.977	—	4.216	2.000	107.947	0.000
	4	-21.674	6.977	—	4.216	2.000	107.947	0.000
	5	-21.539	6.952	—	—	2.000	107.999	0.051
	1	7.882	0.754	8.868	5.046	3.000	109.413	1.465
	2	-36.867	0.754	9.622	5.046	3.000	109.413	1.465
	6	14.765	—	—	—	1.000	114.729	6.782

3.3. Model Analysis

In our analysis of the drivers behind the SIE across different plant life forms, we found that the drivers of the SIE vary among plant life forms, and their influence on species diversity is closely linked to patch size. For total plants, Sealed₅₀, showing a consistent downward trend, which suggests that smaller patches are more vulnerable to the negative effects of surrounding impervious surfaces. In contrast, Elevation, PAR, and Sealed_patch showed a continuous upward trend, indicating that species diversity increases with patch size as these variables increase. For DIS, GDP, LSI, initially observed a sharp decline followed by a relatively slower increase, meaning these factors exert a strong influence on species diversity in smaller patches but stabilize as the area grows. Notably, AT displayed a significant downward trend followed by a marked upward shift, indicating a more dynamic relationship with the vegetation patches. Meanwhile, Light exhibited a gradual upward trend followed by a slow decline, reflecting a more moderate but steady impact across patches of varying sizes.

For shrubs, apart from LSI and Sealed_patch consistently declined, all other variables exhibited a trend of initial decline followed by an increase, though the patterns varied across different factors. Specifically, DIS, GDP, Sealed₅₀ and Light followed a trend of slowly decreasing and then a gradual increase. Elevation and AT showed a slowly decline followed by a sharp increase, while PAR displayed a sharp decline followed by a slow rise. These patterns indicating that these variables initially exert a negative impact on shrub diversity especially for smaller patches.

In the case of trees, aside from DIS and GDP showed a continuous decline, all other variables also exhibited an initial decline followed by an increase, though with varying patterns. Light and PAR experienced a sharp decrease followed by an increase. Sealed₅₀, LSI, and Sealed_patch showed a trend of decreasing slowly and then increasing, and Elevation and AT showed a trend of decreasing slowly and then increasing sharply, reflecting their variable but growing importance as patch size increase (Figure 4).

4. Discussion

Rapid urbanization in Qingdao has led to habitat fragmentation of remnant vegetation patches. The coastal and mountainous character of Qingdao has led to a special diversity pattern of remnant vegetation. In our study, we examined 17 remnant vegetation patches within the urban area, which only account for just 0.32% of the total city area. Despite their small size, these patches support an impressive diversity, with 331 species recorded. Our findings highlight that the dominant plant families in these remnant vegetation patches are Asteraceae, Rosaceae and Gramineae. Among these, Asteraceae and Gramineae showed the greatest diversity in terms of both genera and species. The most frequently observed genera included Artemisia, Erigeron and Rosa. These species are globally widespread and possess stronger resilience and high reproduction rates, reflecting the adaptability of these families, particularly Asteraceae and Gramineae, in maintaining biodiversity even within small, isolated urban green spaces. The result that is consistent with the results of studies in Yunnan [33], Harbin [34], Beijing [35], Shanghai [36]. Additionally, Asteraceae and Gramineae are usually introduced into cities due to their ornamental, economic, and medicinal values, which has may also contribute to their spread and colonization [37,38]. These findings emphasize the importance of conserving remnant patches, which serve as valuable reservoirs of biodiversity despite their limited spatial footprint.

MacArthur and Wilson (1967) hypothesized that the Small Island Effect (SIE) occurs when extinction events outnumber colonization events on small islands, resulting in a steeper decline in species richness as island size decreases. Our research in Qingdao’s urban remnant patches has provided empirical evidence to support this hypothesis, particularly for woody plants such as shrubs and trees. The variability in the expression of the SIE among different life forms of plants is a critical finding. While woody plants exhibit a clear response to the SIE, herbaceous plants do not, likely due to their superior dispersal capabilities and smaller size, which allow them to maintain higher population densities and lower extinction probabilities across patch sizes [39]. This distinction has significant implications for conservation strategies, suggesting that a one-size-fits-all approach may not be effective and that the specific ecological traits of plant communities must be considered.

The area thresholds of SIE identified for tree and shrubs—6.38 ha and 11.91 ha, respectively—underscore the necessity for larger remnant patches to sustain woody plants diversity. Additionally smaller patches being more prone to SIE, experience a more rapid decline in diversity, thus rendering urgent and vital the rescue conservation efforts targeted at these threatened small patches that fall below these thresholds. Interestingly, the total plant species does not exhibit SIE until a much larger threshold of 86.44 ha is reached, which may be attributed to herbaceous plants mitigate the impacts of the SIE. Their presence in smaller patches compensates for the decline in woody plant diversity, helping to maintain higher overall species richness despite isolation effects. As Schrader (2020) noted that species groups with high dispersal abilities or low extinction probabilities tend to show smaller breakpoint areas, as extinction rates only surpass colonization rates on very small islands [13]. To mitigate the effects of SIE, it is essential to focus on maintaining and enhancing the size and quality of patches, particularly for woody plants, which are more vulnerable to the effects of habitat fragmentation.

Environmental factors have been shown to play a significant role in influencing plant colonization and species richness, with several ecological theories proposed to explain the mechanisms behind the SIE. The habitat diversity hypothesis, for instance, suggest that the loss of certain habitat types below a critical area threshold can lead to the failure of dependent plant species to survive, which is supported by our findings for shrubs and trees [23]. As elevation increases in smaller patches, the terrain tends to become more uniform, limiting the diversity of microhabitats that are crucial for different species [40]. Additionally, smaller patches at higher elevations are more exposed to harsh climatic conditions like wind and temperature extremes, which can negatively impact plant survival and further reduce habitat diversity [14]. As the extent of sealed surfaces increases, critical habitat elements such as soil moisture, root space, and nutrient cycling are disrupted [41]. This limits the capacity of the patch to support diverse plant species, particularly those with specific habitat requirements like shrubs and trees. The surrounding sealed environments may also exacerbate the urban heat island effect [42], which can further degrade the internal conditions of the patch and make it less hospitable for certain plants. Furthermore, as the PAR and LSI amplifies the edge-to-interior ratio in smaller patches, leading to more pronounced edge effects and habitat degradation. These effects, such as increased light exposure, temperature variation, and wind exposure, can disrupt stable core habitat conditions that many woody plants depend on. The greater edge effects introduced by a higher PAR and LSI create a more fragmented habitat, reducing core areas that provide essential resources like shelter from environmental stressors and stable conditions for growth [33]. Together, these factors limit the variety of microhabitats available within the patch, leading to a decline in species that require more stable or specialized habitats, thus contributing to SIE.

The nutrient replenishment hypothesis posits that exogenous resources in the vicinity of an island can increase island productivity, thereby affecting the number of species on the island [43]. In both shrubs and trees, PAR, LSI, and DIS showed clear trends, the greater the edge effect from PAR and LSI, the greater the potential for exchanging material with the outside world and the more nutrients received [44]. Under a certain area threshold, the patch itself is small in size and more likely to receive nutrients, so the edge effect has little effect on nutrient supplementation, but as the patch area increases, the nutrient supplementation becomes insignificant, and at this time, the edge effect plays a positive role, and is able to provide nutrient supplementation for the larger island, which plays a positive role for plant survival [26]. Our species composition results show a high degree of overlap with the plant species observed in Laoshan （Laoshan vegetation species reference Laoshan Flora, 2003）. Qingdao City When the patches are closer to the species pool or “mainland” (Mt. Laoshan in our case), are more likely to get replenish through nutrient replenishment and species dispersal. Consequently, enhancing the ecological quality of Laoshan Mountain is likely to have a beneficial impact on the protection of plant species diversity in the remnant vegetation patches within the main urban area.

The disturbance hypothesis, which aligns with our finding, posits that small islands, analogous to fragmented habitat patches, may be in various stages of recovery from disturbance, with transient or random events potentially reducing the expected positive relationship between area and species richness [45]. Our findings indicate that in smaller patches, edge effects associated with PAR and LSI intensify disturbance, leading to a decline in species numbers [46]. This is particularly relevant as it suggests that the configuration of habitat patches can significantly influence biodiversity. The DIS, Sealed₅₀, AT, Light, and GDP indices, which serve as indicators of urbanization intensity, further highlight the multifaceted nature of disturbance, with higher values correlating with increased disturbance around remnant patches [47]. Small patches, which more susceptible to external disturbance, can also exhibit a higher biodiversity, which can act as a conduit for species movement and gene flow. However, this comes at the cost of increased vulnerability to disturbance that can rapidly alter habitat conditions, thus affecting species survival and overall biodiversity [48]. These insights are crucial for biodiversity conservation efforts, especially in the context of urbanization where habitat fragmentation is a significant concern.

In summary, the SIE is a complex phenomenon influenced by a multitude of factors, including plant traits, and environmental conditions. In order to better conserve biodiversity, we should maintain habitat diversity and avoid anthropogenic disturbance as much as possible, which can be achieved by means of increasing the area of green space and reducing the destruction of pristine habitats. At the same time, Mt. Laoshan and the forests around the remnant vegetation patches are important to maintain the species diversity of the remnant vegetation patches, so attention should be paid to the protection of the green space environment around the patches. The results of our study are of great significance for the conservation of biodiversity, and provide a scientific reference value for the conservation strategy of different life forms vegetation. Given the limited number of remnant natural patches within Qingdao’s urban areas, our study focused on a sample of 17 patches, which provides a foundational perspective on the Small Island Effect (SIE) in this region. Future research endeavors should expand this analysis to encompass a broader range of urban environments, and allow for a more comprehensive understanding of the drivers behind SIE. Additionally, investigating the differences in SIE between natural remnants and artificially created green spaces could offer further insights into the mechanisms of SIE and its implications for urban ecology and conservation efforts. This expanded scope of research can also explore how the configuration and connectivity of habitat patches influence biodiversity and ecological processes within urban landscapes. By doing so, we can enhance our predictive capacity regarding the impacts of urbanization on native species and work towards more effective biodiversity conservation strategies in the face of habitat fragmentation.

5. Conclusions

This study detected the Small Island Effect (SIE) in Qingdao’s urban remnant vegetation, particularly in woody plants, with significant area thresholds identified for trees (6.38 ha) and shrubs (11.91 ha). Then, trends in the contribution of the environmental variables to changes in species richness over the iterations were analyzed. Our findings revealed the following: (1) In total, 331 species were surveyed in 17 remnant vegetation patches, proofing remnant vegetation patches can maintain a high number of remnant vegetation species in urban areas. (2)The SIE was only detected on woody plants, with the threshold area 6.38 ha for tree and 11.91 ha for shrub, suggesting woody plants are more sensitivity than herbaceous plants to SIE; (3) The drivers of SIE differ between shrubs and trees. total plants largely influenced by Sealed₅₀, DIS, GDP, LSI and AT, shrub more sensitive to threats posed by changes in LSI, Sealed_patch, PAR. tree is significantly correlated to DIS, GDP, Light and PAR. The study supports the habitat diversity, disturbance, and nutrient replenishment hypotheses, emphasizing the need for targeted conservation strategies that consider the specific ecological traits of plant communities. In light of the aforementioned findings, it is recommended that the conservation of different life forms is differentiated in urban planning and construction to enhance the maintenance of native habitats of remnant vegetation patches.

Remnant vegetation is suitable for creating near-natural landscapes because of its high spontaneity, low maintenance costs, and ornamental values. The use of these remnant vegetation in the design of urban artificial green spaces will yield higher ecological and economic benefits. The attributes of the patches themselves have an important influence on the patches, which are expressed in this study as patch area and landscape shape index and perimeter-area ratio. When constructing and managing urban green space, urban planners should consider these properties to enhance the species richness of remnant vegetation and enrich the urban green space, especially for patches located away from natural species pool. The green patches should have the smallest possible perimeter-to-area ratio; in addition, large patches should be protected because they have greater species conservation capacity. However, due to the severe shortage of urban land, it is impractical to design large patches when planning green spaces in urban areas; but it seems feasible and effective to utilize more small patches to construct ecological corridors to promote vegetation dispersal. In addition, external disturbance is highly disruptive to plant species richness, which means that urban planners and managers may need to take additional specific measures to reduce this risk, e.g., by reducing anthropogenic disturbance such as cutting, trampling, building, and cultivation. Our insight into patterns of remnant vegetation richness in Qingdao is of great significance for biodiversity conservation and provides scientific reference value for the conservation strategies of different life forms of vegetation and the construction of urban landscaping.