Effects of Urban Matrix on Plant Diversity of Urban Remnant Mountains in Karst Area: A Case Study of Guiyang, Guizhou Province, China.


 With the rapid development of urbanization, large numbers of cone-shaped or tower-shaped isolated peaks and peak forest were surrounded by a human-dominated urban matrix in the process of urban expansion in Karst areas,forming urban remnant hills (URHs), which are important resources of indigenous plant diversity from the urban environment in these areas.Although the studies on urban plant diversity have been extensive, studies on the plant diversity of the remnant habitats, especially those based on URHs,is very weak.In this study, a total of 143 sample plots were set up on 15 hills based on the combination of slope direction and elevation.Alpha diversity was calculated at four levels, i.e.,tree,shrub,herb and whole plant.Four urban matrix characteristic indices,namely,percent total impervious surface area (PTIA),land use,vegetation coverage (VC) and fragmentation indices (FIs),were measured for each hill.Twenty scales, ranging from 100 m to 2000 m and with increasing steps of 100 m,were used as spatial scales.Pearson correlation analysis and linear regression model were used to study the relationship between URHs plant diversity indices and urban matrix indices.The results showed the following:(1)There was a significant positive correlation between the overall plant diversity of URHs of different sizes and PTIA in the surrounding urban matrix within 1600 m.The overall plant diversity of URHs of different scales was significantly correlated with single or multiple Land-use types.There was no significant correlation between VC,FI and URHs overall plant diversity.(2)There was a significant linear relationship between plant diversity at all levels of different scales URHs and the four urban matrix indices.In addition to the small URHs herbaceous diversity,plant diversity at all levels of URHs was jointly affected by double or multiple factors at the 1800 m spatial scale. plant diversity of URHs of different scales in hilly Karst cities are extremely complex under the combined action of multiple factors and multiple scales.


Introduction
With the rapid growth of global urbanization and urban populations, many cities have been experiencing large-scale Land-use conversion (Ramalho, et al., 2014). Urban areas continue to expand to (semi-) natural areas outside their boundaries, and a large number of (semi-) natural areas are rapidly lost, split (Hayriye, et al., 2009), degraded or dispersed and embedded in a heterogeneous arti cial built-up environment to form islands or island-like isolated "remnant habitats" (Fernández, et al., 2019;Han, et al., 2019). In most areas, urban expansion has destroyed (semi-) natural habitats (i.e., completely converted to the arti cial environment), and few cities can retain these (semi-) natural habitats in the process of urbanization (Müller, et al., 2015). The process of habitat loss and fragmentation caused by the gradual transformation of intact habitat patches from natural land cover to impervious surfaces in urbanized areas usually has a large and negative effect on plant community species diversity (Fahrig, 2003; Kowarik & von der Lippe., 2018). Therefore, urbanization is considered to be the main cause of urban biodiversity loss and a serious threat to urban biodiversity conservation (McKinney, 2006). Urban remnant habitats are valuable ecological resources and important potential conservation core areas of species diversity in cities (Mona, et al., 2016;Alvey, 2006;Frank, et al., 2020). They can maintain urban native biodiversity and various ecosystem processes and provide ecological well-being for urban residents (Han, et al., 2019;Guirado, et al., 2006). Due to the complexity of the urban arti cial environment, research on the maintenance mechanism of species diversity in urban remnant habitats has become a popular and di cult issue in global urban ecology (Ramalho, et al., 2014) and has attracted extensive attention in the eld of urban ecology (Ramalho, 2012;Schlesinger, et al., 2008). However, the scope and mechanism of the effects of urban matrix environment composition and con guration on the plant diversity of remnant habitats and at different levels are still unclear.
The particularity of urban remnant habitats was that they are completely exposed to the heterogeneity of urban arti cial built-up environments, and the matrix and the surrounding environment, in terms of both material exchange and spatial relationships, are relatively complex. In particular, the mutual in uence between different landscape types and functions and changes in landscape structure and function of the in uence of remnant habitat are complex and diverse (Patarkalashvil, et al., 2017;Shi, et al., 2020). Most previous studies have used the "gradient method" to explore how urbanization factors such as distance from urban centres and population density affect biodiversity in urban substrates (Kinzig, et al., 2005;Malkinson, et al., 2018). Several studies have also shown that the urban species richness had an obvious urban gradient effect (Breuste, et al., 2008;McDonnell & Hahs, 2008).
However, other scholars regarded that urban biodiversity patterns (e.g., birds and plants) were in uenced by the cultural and economic status of urban residents rather than by population density, distance from urban centres, or time after disturbance (Evan, et al., 2020;Luck, et al., 2009;Hope, et al., 2003). The urbanization gradient can be used as a proxy for causal mechanisms such as disturbance state, pollutant load or predation pressure, but it does not directly affect biodiversity (Kinzig, et al., 2005). To a greater extent, it could depends on the differences in adjacent patch characteristics, landscape fragmentation, Land-use composition, and structural changes in remnant habitats with different urbanization gradients. Banksia remnant habitats in Perth, Australia, the results showed that the impact of rapid urban fragmentation on residual vegetation was great and complex, and these effects may take several decades to show a signi cant relationship. Fahrig et al. (2003) showed that multiple mechanisms may lead to positive or negative landscape scale responses to habitat fragmentation and may change with environmental change. Studies on the effects of urbanization on plant diversity in Africa, Europe and Asia have con rmed that patch size and adjacent Land-use type were important factors determining plant species composition and richness, but the direct or indirect effects of disturbance mechanisms and surrounding environmental changes may be different (Frank, et al., 2020;Guirado et al., 2006;. Additionally, other scholars have begun to explore the potential relationship between urbanization and species diversity on a spatial scale. Fernández et al. (2019), using a multitemporal and spatial scale approach, assessed the effects of urban matrix vegetation patterns on the primary productivity of natural heritage habitats in 10 cities in Santiago (Chile), the results showed that the primary productivity of all remnant natural habitats decreased, and this decrease was spatially related to the change in vegetation cover in the surrounding urban matrix within 900 m and was more strongly related to the composition of the matrix. Peng et al. (2019) found that, during the urbanization process, landscape units with a radius of 600-700 m and different land uses had the most appropriate spatial scale range for the conservation of native plant diversity in the study on the optimal landscape pattern for the conservation of native plant diversity in Shunyi District, Beijing. Planchuelo et al. (2020) analysed urban matrix characteristics (e.g., impervious surface area, average oor area) within a 500-m radius around endangered plant species in Berlin, Germany, and revealed the negative impact of urbanization dynamics in Berlin (increase in the impervious surface) on the survival of endangered plant species. Although studies have revealed the potential relationships between some urbanization indicators/predictors and urban biodiversity patterns at different spatial scales, there is no uni ed and optimal methodological indicator to measure and evaluate the impact of urbanization development on urban biodiversity in all regions . Therefore, the selection of appropriate urban matrix drivers and spatial scales could make it easier to collect data and accurately predict or re ect the effects and mechanisms of urbanization on plant diversity in remnant habitats.
The Karst area in southern China, centred on the Guizhou Plateau, is the most typical, complex and abundant Karst area in the world, as well as the largest and most concentrated ecologically fragile area . Due to the special geomorphological form of solitary peaks and peak forest, in the process of urban expansion in this area, a large number of Karst hills of different scales were left as islands or island-like (hemi-) natural remnant habitats in the arti cial built-up heterogeneous urban environment, formed the Karst "urban remnant hills" (URHs) habitat and the landscape mosaic of "city amomg the hills, hills in the city". The spatial form of the Karst hilly city embedded with the hills not only creates a unique city style but also forms the precious remnant ecological resources of the urban hills in the Karst hilly city (Ren, et al., 2018). Thus, this area provides an ideal research place for the ecological study of remnant habitat in an arti cial urban environment.
Guiyang, located in the middle of Guizhou Province, is a typical Karst hilly city with a large number of URHs in the built-up area of the central urban area. In this study,15 URHs in the urban built-up area of Guiyang were selected as the research object, and four indicators of urban matrix characteristics, namely PTIA, Land use, VC and FI, were adopted to explore the following questions

Study area
This study was carried out in the central urban area of Guiyang (106°07′-107°17′E,26°11′-26°55′N) (Fig. 1), which is located in the hinterland of the Karst region in central Guizhou, the middle of the of the Yunnan-Guizhou Plateau, and the watershed zone between the Yangtze River and the Pearl River. The landform belongs to the Karst hilly basin area dominated by Karst solitary peaks and peak forest .The zonal vegetation in the area is subtropical humid evergreen broad-leaved forests with abundant plant resources in history . In the 1980s,the total built-up area in the central urban are of Guiyang was only 22.19 km², due to the topography of the closed Karst basin, as well as the restriction of backward economic development and a low construction level. By 2018, the total built-up area in the central urban area of Guiyang was 368.39 km2, an increase of 16.6 times in 30 years. According to statistics, the permanent resident population in 2018 was 4.8819 million. The remnant hills in the built-up area are mostly scattered in the heterogeneous urban matrix environment in the form of small ,fragmented and isolated. In fact, most of the URHs at the beganing of being surrounded into built-up area still retained a large patch area, as the densi cation inside the city gradually approaching to the URHs, they are occupied and divided into smaller patches or completely devoured, this process has not yet stopped (Fig. 2).
Furthermore, there is little or no protection and management of the URHs, and the number and area of the patches are decreasing rapidly. The ecological environment is fragile and has obvious spatial heterogeneity (Mu, et al., 2020); thus, it is necessary to protect these special URH habitats .

Selection and sampling design of URHs
The URHs were divided into 3 scales: large ( ≧ 10 hm 2 ), medium (3-10 hm 2 ) and small ( ≦ 3 hm 2 ). Five URHs of each type were randomly selected, and a total of 15 URHs were taken as research objects (Table 1). Sample plots were set according to the four-direction method (east, south, west and north), extending from the top to the foot of the each hill. Large and medium-sized URHs included 3 plots in each direction, while small URHs included 2 plots in each direction, each with an area of 30×30 m 2 , and a total of 143 survey plots were included. Five investigation quadrats were set up in each plot according to the following criteria: tree (10×10 m), shrub (3×3 m) and grass (1×1 m). The interval between each quadrat was no less than 3-5 m, and the quadrats were nested. A eld investigation was conducted from July to October 2019 and from July to October 2020. The species name, number, height, DBH and canopy width of the trees,, the species name, height, canopy diameter and the number of trees of shrubs (including small trees), and the herbaceous species name, plant number and coverage in each quadrat were recorded. Based on the survey results, a database of plant-related information and community characteristics was constructed in Excel. Considering that the response of plants at different levels to disturbance may be different in urban continuous disturbance, therefore, four species diversity indexes (Shannon-Wiener index (H'), Margalef species richness index (R), Simpson index (D) and Pielou index (JH) were calculated for the whole plant diversity and the plant diversity at different levels namely trees, shrubs and herbages (Ma,1995). Classi cation of native species refered to the China Plant Science Data Center (https://www.plantplus.cn/cn) and Flora of Guizhou .

Classi cation of landscape types
The 2018 Pleiades high-resolution satellite image (0.5 m spatial resolution) of the study area was obtained as the basic data source for buffer analysis and urban matrix analysis. According to the Classi cation of Land Use Status (GB/T21010-2017) and based on the ArcGIS software platform, landscape types were divided into two categories: construction land and non-construction land. Construction land was further divided into residential land (R), public management and public service land (A), commercial service facilities land (B), industrial land (M), logistics and storage land (W), transportation facilities land (S), public facilities land (U) and green land (G). The nonconstruction land was further divided into cultivated land (E21), woodland (E22), grassland (E31), water area (E1), natural mountain (EG) and unused land (E32). The spatial attribute database was established for the research area.

Urban matrix characteristic indexes
In this study, four urban matrix characteristics (PTIA, Land use, VC, and FI) were selected to explain the potential effects of urban matrix composition and structural complexity on URH plant community species diversity (Yan, et

Urban matrix characteristic indexes
Since multiple (potential) ecological processes affected by the urban matrix landscape structure occur at different spatial scales, their scope of in uence and relative importance may vary with the size of the landscape (Smith, et al., 2011). In this study area, in order to quantify the potential effects and spatial scale effects of the above matrix indicators on URHs plant diversity at all levels, based on Arcgis10.  (Fig. 4).

Impact of Land-use type on overall plant diversity and plant diversity indices at all levels in URHs
The overall plant diversity of large URHs was positively correlated with residential land R. The overall plant diversity of medium URHs was negatively correlated with Land-use types A, G, E21, and E1. There was a signi cant negative correlation between the overall plant diversity of small URHs and the E32 Land-use types (Fig. 6).
The responses of plant diversity indices of different levels of URHs in different scales to Land-use types were different. Large URHs tree diversity indices were negatively correlated with R, E1, and E31; shrub diversity indices was negatively correlated with E1; and herbs diversity indices were positively correlated with G and E1 and negatively correlated with E21 and E22. The tree diversity indices of medium URHs were negatively correlated with M, and shrub diversity indices were positively correlated with E22 and negatively correlated with A and E1, but herbs diversity was not signi cantly correlated with Land-use type. There were a signi cant positive correlation between the diversity indices of tree and shrub and E32 in small URHs and a signi cant negative correlation between herb diversity indices and EG and E32 (Fig. 8).

Effects of VC on the species diversity of URH plant communities
The results showed that there was no signi cant correlation between overall plant diversity indices in all URHs and VC in any scale buffer (Fig. 7).
There were signi cant differences in the spatial response of plant diversity indices at different levels to buffer VC in URHs of different sizes. Among them, the herbs diversity indices of large URHs were signi cantly negatively correlated with VC in the 100 m buffer zone. There was a signi cant positive correlation between the diversity indices of shrubs of medium URH and VC in the range of 300 m. There was a signi cant negative correlation between the diversity indices tree and shrub of small URHs and VC at 700 m and 1800 m, respectively. There was no signi cant correlation between the diversity indices of trees and shrubs of large URH and VC in any scale, as well as trees and herbs of medium URH and herbs or small URH (Fig. 7).

Species composition and diversity of URH plant communities
Previous studies have shown that patch size was a key determinant of species diversity but may differ due to direct or indirect impacts of human activities on urban environments (Guirado, et al., 2006). The results of this study showed that although patch area was one of the factors that explained the plant diversity at different levels URHs, but the relationship was weak. There was no signi cant difference in the plant community composition and structure of URHs in different scale. Medium URHs could maintain higher overall plant diversity, large URHs could maintain high tree and herbaceous diversity, while small URHs showed higher shrub diversity. The urban arti cial environment is characterized by continuous and repeated interference by human factors (Zarzycki, et al., 2002). In the human-dominated urban matrix environment, the maintenance of species diversity and ecological processes in the remnant habitat are largely determined by the type, degree, and frequency of interference of human activity. In particular, the strong human disturbance caused by the transformation of the landscape basement greatly changed the relative in uence of patch size and even topographic factors on the species diversity of the remnant habitats (Matlack, 1993). Therefore, considering patterns of plant diversity in patch characteristics alone is not enough to explain the potential effects of other disturbances.

PTIA and plant diversity at various levels of URH
The PTIA is an important indicator to measure the urban environment and urbanization degree. Some scholars have con rmed that PTIA had a negative impact on the plant diversity in urban area (Yan, et  complicating factors that explain this result may be follows: (1) URHs were mostly surrounded by transportation networks and residential areas, which are more conducive to plant propagation or pollen spread to some extent. (2) There is a phenomenon of park utilization in URHs of different scales, and the arti cial introduction and planting of garden ornamental plants (though most of them are probably alien species) provide a good growth environment and conditions. (3) The phenomenon of reclamation and cultivation by the surrounding residents exists in most of the URHs. To provide good illumination and growth conditions for crops, not only were the upper trees and shrubs cut down, but herbicide was also used to remove the herbaceous plants, which increased the probability of the introduction or invasion of alien species. These factors suggest that rational planning management may be bene cial to maintaining a high level of plant diversity (Muratet, et al., 2008). However, excessive human disturbance would change the disturbance mechanism between the habitat environment and adjacent substrates, leading to differences in the responses of plant diversity at different levels to the environment. their con gurations can signi cantly shape and regulate the composition and diversity of plant communities (Walz, 2015). The results of this study showed that Land-use composition in1600 m width buffer zone signi cantly affected the plant diversity at all levels in URHs, but the correlations were signi cantly different. Similar to the results of Sect. 3.2, residential and industrial Land-use types may have positive impacts on plant diversity, while non-construction Land-use types such as woodland, cultivated land, and grassland may have negative impacts; moreover, they may be in uenced by multiple Land-use types at the same time. This result aslo proved that the composition and structure of land cover and land use on a small scale space are highly complex and dynamic (Luck & Wu, 2002). The complex urban matrix mosaic was formed by the superposition and interaction between the Land-use types of Karst hilly cities and the topography of Karst basin. The high connectivity of impervious surface land uses might lead to the superposition effect of a single land use type on URHS plant community, or lead to different land use between the mechanism of action of offset each other, which might lead to differences in the shown that there is a quadratic relationship between urban matrix vegetation cover and neighbourhood age, but there is not a signi cant positive correlation until approximately 45 years (Grove, et al., 2006;Troy, et al., 2007). This result suggests that improving substrate quality by increasing vegetation cover in a short period may not be a useful strategy to maintain or protect the species diversity and ecosystem integrity of URH plant communities.

FI and plant diversity at different levels of URHs
Although the impacts of landscape fragmentation on the plant communities in remnant habitats are diverse and complex, it doesn't necessarily to have a direct impact (Ramalho, et al., 2014). In this study, we found that buffer FI had no signi cant effect on the plant diversity of large and medium URHs. However, there was a negative effect  2006) that "different taxa/species respond to urban spatial forms at different spatial scales", and also indicated that the overall plant diversity of URHs, the in uencing factors and the spatial scale range of plant diversity at all levels were complex and diverse.
According to the results of this study and the theory of "extinction debt" and previous studies, it is known that the response of remnant habitat plant communities to urbanization migth be signi cantly negative correlation after decades or longer.The results of this study con rmed the previous hypothesis that the response of plants at different levels to urban matrix disturbance was different when the URHs were continuously disturbed by human beings. Therefore, future studies should consider the impact of dynamic changes of landscape patterns in urban matrix (historical evolution of land use) on plant diversity in remnant habitats. Therefore, future studies should consider the impact of dynamic changes of landscape patterns in urban matrix (historical evolution of land use) on plant diversity of remnant habitats, and re ne the effects of intensity and frequency of human activity interference (engineering construction, engineering facilities, slope hardening and greening, cultivation, cemeteries, etc.) within remnant habitats on plant communities and different levels of plants.It is suggested that in the process of urban development and construction in the future, multiple in uencing factors and mechanisms should be considered for different remnant habitats in the time dimension and different spatial scales.

Conclusions
The ecological processes and community structure changes of URHs in the arti cial built-up environment of Karst hilly cities are more complex due to the combined effects of multiple factors on different scales. The complexity of the in uencing factors increase the challenge of URHs protection and management.Based on this study, it is concluded that (1) There was a signi cant positive correlation between the overall plant diversity of URHs of different sizes and PTIA in the surrounding urban matrix within 1600 m.The overall plant diversity of URHs of different scales was signi cantly correlated with single or multiple Land-use types, but the impact effects were signi cantly different. There was no signi cant correlation between VC, FI and URHs overall plant diversity. (2) There was a signi cant linear relationship between plant diversity at all levels of different scales URHs and the four urban matrix indices. In addition, the herbaceous plant diversity of small size remnant hills showed a signi cant linear relationship with only Land use, but in other URHs, the plant diversity at all levels was affected by double or multiple factors at the range of 1800 m. Plant species diversity and community structure of URHs of different scales in hilly Karst cities are extremely complex and diverse under the combined action of multiple factors and multiple scales. Therefore, in the process of urban green space ecology system planning, urban spatial planning and urban renewal, the multiple in uence factors and their mechanism to different scales URHs at different scales should be considered to provide a scienti c reference and to help protect remnant habitat plant community species diversity.  Area of study Note: The designations employed and the presentation of the material on this map do not imply the expression of any opinion whatsoever on the part of Research Square concerning the legal status of any country, territory, city or area or of its authorities, or concerning the delimitation of its frontiers or boundaries. This map has been provided by the authors.  Analysis of the species composition of the URH plant community in the built-up area of Guiyang city