Accumulation of Urban Insect Pests in China: 50 Years’ Observations on Camphor Tree ( Cinnamomum camphora )

: Since China experienced a rapid and unprecedented process of urbanization and climate change from 1978 onwards, pest outbreaks were frequently reported on urban forests, which reﬂects a signiﬁcant imbalance between natural regulation and human control. Based on information extracted from all journal articles and reports about insect pests on camphor tree ( Cinnamomum camphora ) in urban China, we characterized historical patterns and trends in pest outbreaks over large areas. Our results suggested that (1) most distribution areas of C. camphora in urban China had pest records (14 provinces) over the last 50 years, especially at the south-eastern coastal areas; (2) pests on camphor tree in urban China showed an accelerated growth since the 1990s; and (3) pests on camphor tree in urban China were characterized by native and leaf-feeding species. Urbanization seems to positively correlate with urban pest outbreaks. Changes of urban pest outbreaks could largely be described by synchronic changes of socio-economic indicators, of which CO 2 emissions as metric tons per capita is the most signiﬁcant predictor, followed by GDP and human population. Thus, managers and city planners should allocate resources to socio-economic-related pest outbreaks for a sustainable ecosystem. (4) the possible history of tree cultivate; (5) plant organ / parts and symptoms; and (6) extent and duration of insect pest disturbances and of


Introduction
The rapid change in pest population densities in forests, known as forest pest outbreaks, and their dynamics represent critical systematic evidence for ecosystem stability [1]. For example, pest outbreaks occur when trees are less phenotypically plastic in their ability to cope with conditional changes compared to insect pests [2]. In addition, pest outbreaks can reflect increased urbanization, uncertainties in socio-economic requirements, climatic change impacts, and insect pest sensitivities. Numerous studies have focused on pest outbreaks at community or local scales under climate impacts [3][4][5], which can contribute to forest dynamics modeling, theoretical validating, and calibrating [6]. Results can also be valuable reference information for informed prediction of potential future changes under global climate change, development of adequate management strategies [6], and characterization of life in changing environments, such as biodiversity [7] and adaptive evolution mechanisms [8].
Urban forests are essential green infrastructure, which are recognized for generating a range of benefits and providing valuable ecosystem services (e.g., controlling pollution). They are highly dynamic, inherently vulnerable, and commonly subject to high levels of stress and disturbances [9][10][11]. As a result, managing urban forests for sustainable urban development is becoming increasingly complex  Figure S2, Table S2 in Supplementary Material for photos and detailed descriptions of C. camphora-hosted insect pest damages). Panel (a) represents patterns of the current state of forest insect disturbance reviewed from the literature on remote sensing. Different insect types: bark beetle (olive triangles), broadleaved defoliator (green triangles), coniferous defoliator (orange triangles) and other fluid feeders (red triangles). Panel (b) shows C. camphora is a widely distributed urban tree species in the world (see Appendix 2, Table S2 in Supplementary Material), i.e., the most popular native tree species in tropical and subtropical China (green circles), and has been introduced by other countries (cyan circles). Photo in the lower right corner is C. camphora in the urban street (Changsha, Hunan province, China). Panel (d) indicates native C. camphora has experienced rapid growth of pest outbreaks (see Appendix 4,Table S4 in Supplementary Material for included reports), pink = prevalent area, low frequency; red = severely damaged area, high frequency.

Data Collection
We assembled a database of urban trees and insect pests coupled population time series in China from sources including Google scholar, Scopus, Web of Science, China Knowledge Resource Integrated Database (http://www.cnki.net), and bibliographic databases. Adhering to the PRISMA guidelines [18], we conducted a systematic literature search using these data sources (initial database query 8 June 2017), for all published journal articles, newspaper reports, conference proceeding papers, and dissertations that were on insect pest disturbance of C. camphora trees in urban areas with no limitation on publication dates. We identified specific keyword strings to search relevant studies or reports, including "Chinese/China", "urban/city", "pest/insect", "outbreak", "camphor Figure 1. Geographical distribution of (a) studies and insect types according to Senf et al. [17], (b) C. camphora throughout the world, (c) C. camphora in China, and (d) reported outbreaks of urban insect pests of C. camphora in China over past half-century (see Appendix 2, Figure S2, Table S2 in Supplementary Material for photos and detailed descriptions of C. camphora-hosted insect pest damages). Panel (a) represents patterns of the current state of forest insect disturbance reviewed from the literature on remote sensing. Different insect types: bark beetle (olive triangles), broadleaved defoliator (green triangles), coniferous defoliator (orange triangles) and other fluid feeders (red triangles). Panel (b) shows C. camphora is a widely distributed urban tree species in the world (see Appendix 2, Table S2 in Supplementary Material), i.e., the most popular native tree species in tropical and subtropical China (green circles), and has been introduced by other countries (cyan circles). Photo in the lower right corner is C. camphora in the urban street (Changsha, Hunan province, China). Panel (d) indicates native C. camphora has experienced rapid growth of pest outbreaks (see Appendix 4, Table S4 in Supplementary Material for included reports), pink = prevalent area, low frequency; red = severely damaged area, high frequency.

Data Collection
We assembled a database of urban trees and insect pests coupled population time series in China from sources including Google scholar, Scopus, Web of Science, China Knowledge Resource Integrated Database (http://www.cnki.net), and bibliographic databases. Adhering to the PRISMA guidelines [18], we conducted a systematic literature search using these data sources (initial database query 8 June 2017), for all published journal articles, newspaper reports, conference proceeding papers, and dissertations that were on insect pest disturbance of C. camphora trees in urban areas with no limitation on publication dates. We identified specific keyword strings to search relevant studies or reports, including "Chinese/China", "urban/city", "pest/insect", "outbreak", "camphor tree/camphor laurel/C. camphora", "leaf/stem/root", "habit/life history", "record", and "damage/injury/death/wilt/defoliate/tilt/broken/uproot/overthrown". We also scanned the reference section of the studies or reports found in this research for other relevant articles. All titles and abstracts were screened and assessed for relevance using the following entrance criteria: (1) reported at least one insect/pest species that damage at least one plant organ/tissue of leaves, stem/bark/trunk, seeds/flowers, and/or roots; (2) acquired data from field investigations or reports; (3) had similar methods and specific information on site, tree, and insect/pest properties; (4) is a complete article from which all data and results were made available for analysis; and (5) would be possible to calculate the synthetic statistics and address meta-analytic questions.
In addition, articles with the following characteristics were excluded from our analysis: (1) repeated publications or duplicated reports; (2) studies without clear pest records (e.g., physiological studies, morphological studies, or field experiment); (3) studies with poor data quality (data could not be analyzed) and/or quantity (incomplete data), such as studies with population records of insect/pests but without the corresponding records about the host trees; (4) studies with low information availability, from which the diversity of insect pests on C. camphora and tempo-spatial information for pest outbreaks could not be calculated for statistical analysis (attempts have been made to contact the corresponding author(s) to obtain the missing data whenever possible); (5) reports on other tree species of Cinnamomum, such as Cinnamomum porrectum (Roxb.) Kosterm., Cinnamomum subavenium Miq., and Cinnamomum micranthum (Hayata) were excluded; (6) studies in which any specific compounds (or management methods) had been applied to facilitate or inhibit pest outbreaks; and (7) articles that were just summaries of previous works.
The literature search resulted in a total of 72 unique publications that met the strict criteria for our data compilation (see Appendix 4, Table S4 in Supplementary Material for 48 journal articles and 24 newspaper reports).

Data Filtering and Compilation
Data from all qualifying studies were compiled into datasets. The following information was extracted from each article: (1) first author name, observation sites, geographic coordinates, sampling site description, recording time, and areas of sampled fields; (2) plant taxa and pest species; (3) tree growth characteristics including stand age, stand density, monoculture/mixture, tree height/size/architecture, etc.; (4) information description about the possible history of tree cultivate; (5) description about damaging plant organ/parts and symptoms; and (6) extent and duration of insect pest disturbances and method of detection.
Various attributes determine the distribution and population density of insect pests [19]. Thus to facilitate the diversity and comparison analysis, we classified the complete extracted data sets into several major categories using the variable information of the respective publications. For extracted information, we synthesized pest information based on three categories: (1) tempo-spatial properties; (2) the taxonomic information and habitat characteristics; and (3) symptoms, causes, and types of feeders (see Appendix 5, Tables S5-S7 in Supplementary Material for detailed metadata descriptions and verification criteria).
(1) Tempo-spatial properties Temporal properties included period, year, and month. Spatial properties included province, city/county, and location. One pest occurrence in our list referred to one pest event with a specific location and date [19].
(2) The taxonomic information and habitat characteristics Insect pests were classified into species, genus, family and order. We rectified the Chinese name according to the Chinese Animal Scientific Database (http://www.zoology.csdb.cn/) and the National digital-museum of animal specimens (http://museum.ioz.ac.cn/). We also recorded Latin names and taxonomic information in these national databases. We excluded pest records whose Chinese names were unspecific or cannot be found. We then checked the Latin name by related global databases (e.g., http://www.biolib.cz/en/main/). We applied Pivot Tables in Excel to sort the records and find mismatch taxonomic information based on the similarities in Latin names (see Appendix 6, Table S8 in Supplementary Material for the rectified results from the original records).
Insect pests were classified into native species or non-native species according to locations where the species was first identified.
(3) Types of feeders, symptoms, and causes. We group pest species into three types of feeders according to Cummins (1973)'s trophic functional groups [20]: leaf feeders, stem-feeders, and other-feeders, which differed significantly in structure and function [20,21] (see Appendix 5, Table S7 in Supplementary Material for our cluster). Leaf feeders, also known as foliage feeders or defoliators, are defoliating insects or leaf-eating insects feeding on the foliage of trees, including guilds such as leaf rollers, leaf miners, leaf strip feeder, leaf pit feeders, and leaf sap feeders, skeletonizers, and whole leaf feeders. Stem feeders are stem-eating insects that feed on the stem of trees, including bark beetles, wood borers, sap-sucking insects, and meristem feeders. Other feeders are insects that feed on the fruit, seed, or root of trees and those feeding on mixed plant organs.
For symptoms, we corrected the original description of infestation based on the handbook of the major forest pests in southeast Europe from the Food and Agriculture Organization (http: //www.fao.org/3/a-i4084e.pdf) and analytical approaches for diagnosing plant problems from the US Forest Service and CAL Fire (http://caforestpestcouncil.org/wp-content/uploads/2008/06/Insect-and-Disease-Training-Manual.pdf). Then we categorized symptoms as follows: leaves (12 symptoms), stems (8 symptoms), seeds (5 symptoms), seedlings and cuttings (4 symptoms), flowers (4 symptoms), fruit (6 symptoms), and root system (5 symptoms) (see Appendix 5, Tables S6 and S7 in Supplementary Material for recognizing infestation and damaging symptoms and signs).
In addition, we recorded mean annual temperature, mean annual precipitation, and annual relative humidity of pest-infested areas according to the China Meteorological Administration (http: //data.cma.cn/). We recorded the range of temperature and precipitation, or humidity if the climate information for pest occurrence was vague. We also recorded tree age and categorized the host trees into young trees, mature trees, and old trees according to the United States Department of Agriculture (https://www.fs.fed.us/psw/publications/bingham/bingham3.pdf).

Pest cumulative Characteristics
Natural and anthropogenic factors have resulted in altered environmental conditions that influence changes in abundance and diversity of insect pests [22], which then possibly affect the intensity and frequency of pest outbreaks on host plants [23]. The C. camphora-hosted insect pest outbreak dataset that was compiled are multi-proxy datasets of pest complexes on camphor trees. In order to explain the dataset records to understand the long-term patterns of community change as well as the effects of the diversity of insect pests on the stability of forest functioning, we used cumulative amount and frequency of insect pest detection over time as measures of intensity and frequency of pest outbreaks on camphor trees as used in Senf et al. [17] and Aukema et al. [19]. The cumulative amount and frequency of insect pest detection are usually analyzed by computing total number of detected establishments of insect pests for certain periods or a specific year on the basis of absolute numbers of detections involved in an outbreak across the study sites at any given time. We used linear and non-linear regressions to explore the cumulative detections of insect pest outbreaks and frequency of insect pest detection over time (cumulative frequency or yearly detection) in different categories (through 20 determinative variables), including total occurrence, taxonomic ranks (species, genera, family, and order), habitat characteristics (native and non-native), and feeding preferences (leaf, stem, and others): (1) We evaluated the distribution patterns of pest outbreak in ArcGIS v 9.3 (ESRI, Redlands, CA, USA). The charts were plotted in software OriginPro v 8.0 (OriginLab Co., Northampton, MA, USA); and (2) in order to assess the long-term patterns of pest complex on camphor trees, we regressed pest outbreak frequency (as response variables) against all calendar years (nonlinear terms for cumulative frequency and linear for frequency). We tested for correlations and covariances between the calendar year and variance in intensity and frequency of pest outbreaks on camphor trees using Pearson's correlation coefficients (r). We calculated p-values (t-tests) to determine the significance (at a level of 0.05) of these relationships.

Socio-Economic Indicators
The relationship between pest outbreak and calendar year can be complex and synthetic, including indirect drivers [9], such as higher urbanization intensities, unreasonable urban planning, and management, which results in a harsh growing environment and makes it difficult for trees to survive and reach maturity. Therefore, to examine the possible relationship between socioeconomic characteristics and emerging urban pest outbreaks, we regressed a synchronization term between random economic (or population-driven) processes against calendar years. We used degree of urbanization, fossil fuels' share of energy consumption, composition of gross domestic product (GDP), changes of nitrous oxide (N 2 O) emissions, and carbon dioxide (CO 2 ) emissions as proxies for socio-economic indicators. We extracted 22 indicators with clearly increasing or decreasing trends according to economic or population-driven topics in the World Development Indicators (https://www.eui.eu/Research/Library/ResearchGuides/Economics/Statistics/DataPortal/WDI), which were (1) three indicators for degree of urbanization (percentage of urban population of total population, percentage of population living in metropolitan areas that had a population of more than one million people, and the percentage of urban population living in the largest metropolitan area-Chongqing); (2) three indicators for fossil fuels' share of energy consumption (percentage of fossil fuels energy consumption out of total energy consumption, fossil fuels' share of total net electricity production, and coal share of total electricity production); (3) eight indicators for GDP (contribution of forest resources, services, industry, agriculture, trade, imports, exports, and external balance to overall GDP); (4) one indicator for changes of nitrous oxide (N 2 O) emissions (the percentage change in N 2 O from 1990 levels); and (5) seven indicators for CO 2 emissions (CO 2 emissions as metric tons per capita and from solid fuel, manufacturing industries and construction, residential buildings and commercial and public services, electricity and heat production, liquid fuel, and gaseous fuel consumption).
Pearson's correlation coefficients between pest outbreaks and the composite indices of socio-economic factors were calculated. Multiple linear regression analysis was carried out using all the composite socio-economic indices as independent variables, and pest outbreaks during 1964-2017 as dependent variables. A step-wise procedure was used in the decision to include significant independent variables. We here focused on overall accuracy as a measure of the coefficient of determination (R 2 ) to estimate model performances (IBM SPSS Statistics v 21.0, IBM Corp., Armonk, NY, USA). We chose regressive slopes and descriptive variable domains to test for the presence of possible indirect causes.
Insect pests on C. camphora in urban China were characterized by native leaf feeders. Of the 772 pest records, 73% (n = 563) were native species (see Appendix 8, Figure S4, Table S10 in Supplementary Material for all recorded insect pests categorized by native vs. non-native). Leaf-feeding insects account for 73% of all species, followed by stem-feeding insects (105 species, 450 occurrences) and others-feeding insects (19 species, 81 occurrences) (see Appendix 9, Figure S5, Table S11 in Supplementary Material for feeding compositions).

Spatial Aggregation
Pests were found in Jiangxi Province for the first time, which is located in the southern band of the middle and lower Yangtze River. Severely damaged areas were most numerous in coastal regions, which were characterized by a higher intensity and frequency of pest outbreaks (see Appendix 7, Figure S3, Table S9 in Supplementary Material for spatial aggregations). Figure 2 demonstrated that both intensity and frequency of pest outbreaks on C. camphora were monotonically increased since the 1960s. The cumulative frequency was described well in a power function, while frequency of insect pest detection over time was fitted in a linear model (Figure 2 and Appendix 10, Table S12 in Supplementary Material), with higher cumulative frequency observed in native leaf-feeding species. Specifically, frequency of pest outbreaks on C. camphora in urban China was observed at a yearly increase rate of approximately 12 occurrence, 8 species, 7 genera, 4 families, and 1 order (Figure 2a, that is a, y = ax + b, a=0.766, 0.694, 0.438, 0.075, separately), while 9 native species, 6 non-native species (Figure 2b, that is a, y = ax + b, a = 1.231, 0.851, 0.562, separately), or 9 leaf feeders, 2 stem feeders, and 1 other feeder (Figure 2c, that is a, y = ax + b, a = 0.872, 0.19, 0.079 separately) per decade. Insect pests on C. camphora in urban China were characterized by native leaf feeders. Of the 772 pest records, 73% (n = 563) were native species (see Appendix 8, Figure S4, Table S10 in Supplementary Material for all recorded insect pests categorized by native vs. non-native). Leaffeeding insects account for 73% of all species, followed by stem-feeding insects (105 species, 450 occurrences) and others-feeding insects (19 species, 81 occurrences) (see Appendix 9, Figure S5, Table S11 in Supplementary Material for feeding compositions).

Spatial Aggregation
Pests were found in Jiangxi Province for the first time, which is located in the southern band of the middle and lower Yangtze River. Severely damaged areas were most numerous in coastal regions, which were characterized by a higher intensity and frequency of pest outbreaks (see Appendix 7, Figure S3, Table S9 in Supplementary Material for spatial aggregations). Figure 2 demonstrated that both intensity and frequency of pest outbreaks on C. camphora were monotonically increased since the 1960s. The cumulative frequency was described well in a power function, while frequency of insect pest detection over time was fitted in a linear model (Figure 2 and Appendix 10, Table S12 in Supplementary Material), with higher cumulative frequency observed in native leaf-feeding species. Specifically, frequency of pest outbreaks on C. camphora in urban China was observed at a yearly increase rate of approximately 12 occurrence, 8 species, 7 genera, 4 families, and 1 order (Figure 2a, that is a, y = ax + b, a=0.766, 0.694, 0.438, 0.075, separately), while 9 native species, 6 non-native species (Figure 2b, that is a, y = ax + b, a = 1.231, 0.851, 0.562, separately), or 9 leaf feeders, 2 stem feeders, and 1 other feeder (Figure 2c, that is a, y = ax + b, a = 0.872, 0.19, 0.079 separately) per decade. Figure 2. Temporal patterns of cumulative frequency and yearly pest outbreaks (1964-2017) as classified by four traditional taxonomic ranks (species, genera, family, and order) (a), pooled as whole occurrence or binned by two types of habitat characteristics (native and non-native) (b), or three groups of feeding preferences (leaf, stem, and others) (c). Statistical parameters (r, Slope, and P) indicate how general temporal models fit the data (see Appendix 10, Table S12 in Supplementary Material for non-linear simulation and linear regression results). Cumulative power models (bold solid lines in left panels): occurrence (black), species (olive), genera (dark green), family (green), order (light green), native (blue), non-native (cyan), leaf feeders (red), stem feeders (orange), other feeders Figure 2. Temporal patterns of cumulative frequency and yearly pest outbreaks (1964-2017) as classified by four traditional taxonomic ranks (species, genera, family, and order) (a), pooled as whole occurrence or binned by two types of habitat characteristics (native and non-native) (b), or three groups of feeding preferences (leaf, stem, and others) (c). Statistical parameters (r, Slope, and P) indicate how general temporal models fit the data (see Appendix 10, Table S12 in Supplementary Material for non-linear simulation and linear regression results). Cumulative power models (bold solid lines in left panels): occurrence (black), species (olive), genera (dark green), family (green), order (light green), native (blue), non-native (cyan), leaf feeders (red), stem feeders (orange), other feeders (light magenta). Linear growth models (fitted red lines with a shaded area of 95% confident intervals in right panels): gray circles represent the frequency of detection for each recorded year. Unrepresentative data from early reports or province-wide surveys were excluded (open triangles). Pearson correlation was used to evaluate the correlation coefficient between variables.

Causes and Controller Attribution
We chose temporal trends (Figure 3), synchronic covariant coefficients (Figure 4), and multi-regressive model simulations (Table 1) to test for the presence of possible indirect causes for pest outbreaks.   The linear relationship between the random economic or population-driven indicators and calendar years was highly significant (all p < 0.001 expect p for CO2 emissions from solid fuel consumption and CO2 emissions from liquid fuel consumption, which was 0.01 and 0.03 respectively). Most indicators were more significantly positively related to the calendar year than insect pest detection over the same time (r for 13 indicators > 0.6). Regression slope of pest outbreaks and socioeconomic indicators over time indicated that there was comparatively higher convergence between pest outbreaks and some socio-economic indicators (r for 14 indicators > 0.0 and p < 0.05).
The correlation analysis indicated that pest outbreak dynamics response to the urban population (% of total population), population in urban agglomeration of more than one million (% of total population), fossil fuel energy consumption (% of total energy consumption), GDP compositions (including forest rents, services, agriculture, trade, and imports and exports of goods and services), changes of N2O emissions, and CO2 emissions (Figure 4). Overall, cumulative amount of pest outbreaks showed a greater response to socio-economic indicators than yearly pest outbreaks. The outbreak of leaf feeder was significantly correlated with socio-economic indicators (p < 0.05), although other types of feeders were nearly unaffected.  - Model: (Y i ) = (a i ) + (Σ(b ij ×X j )), i = 1~20 and j = 1~22. (Y i ) donate the 20 outcome variables of the regression equation, i.e., the observed cumulative and yearly pest outbreaks as indicated in Figure 2: Y 1~Y8 = yearly cumulative amount (Y 1~Y4 ) and yearly(Y 5~Y8 ) frequency of pest outbreaks by taxonomic species, genera, family, and order, respectively, Figure 2a; Y 9~Y14 = yearly cumulative amount (Y 9~Y11 ) and yearly (Y 12~Y14 ) frequency of pest outbreaks pooled as whole occurrence or binned by two types of habitat characteristics (native and non-native), respectively, Figure 2b; Y 15~Y20 = yearly cumulative amount (Y 15~Y17 ) and yearly (Y 18~Y20 ) frequency of pest outbreaks binned by three groups of feeding preferences (leaf, stem, and others), respectively, Figure 2c. While the (X j ) refer the included (Step-wise method, p < 0.05) seven predictors in the model selected from the 22 socio-economic indicators shown in Figure 3: X 2 = population in urban agglomeration of more than one million (% of total population), Figure 3Ab; X 6 = electricity production from coal sources (% of total electricity production), Figure 3Bc; X 8 , X 10 , X 13 , X 14 = percentage of total GDP as derived from services value added (X 8 , Figure 3Cb), agriculture value-added (X 10 , Figure 3Cd), exports of goods and services (X 13 , Figure 3Cg), and external balance on goods and services (X 14 , Figure 3Ch), respectively; X 16 = CO 2 emissions as metric tons per capita (Figure 3Da).
The linear relationship between the random economic or population-driven indicators and calendar years was highly significant (all p < 0.001 expect p for CO 2 emissions from solid fuel consumption and CO 2 emissions from liquid fuel consumption, which was 0.01 and 0.03 respectively). Most indicators were more significantly positively related to the calendar year than insect pest detection over the same time (r for 13 indicators > 0.6). Regression slope of pest outbreaks and socio-economic indicators over time indicated that there was comparatively higher convergence between pest outbreaks and some socio-economic indicators (r for 14 indicators > 0.0 and p < 0.05).
The correlation analysis indicated that pest outbreak dynamics response to the urban population (% of total population), population in urban agglomeration of more than one million (% of total population), fossil fuel energy consumption (% of total energy consumption), GDP compositions (including forest rents, services, agriculture, trade, and imports and exports of goods and services), changes of N 2 O emissions, and CO 2 emissions (Figure 4). Overall, cumulative amount of pest outbreaks showed a greater response to socio-economic indicators than yearly pest outbreaks. The outbreak of leaf feeder was significantly correlated with socio-economic indicators (p < 0.05), although other types of feeders were nearly unaffected.
Multi-regression analysis (Table 1) indicates that most of historical trends of the pest outbreak metrics could be explained by socio-economic indicators (p for 17/20 models < 0.05). Changes of cumulative amount of pest outbreaks could largely be described by synchronic changes of socio-economic indicators (R 2 for 10/10 models > 0.93), of which CO 2 emissions as metric tons per capita is the most significant predictor, followed by the contribution of external balance on goods and services in total GDP. Synchronic changes of socio-economic indicators only explain 26%~41.2% of the changes for yearly pest outbreaks (R 2 for 7/10 models = 0.26~0.412), and 70% of the models were only significantly simulated from the contribution of exports of goods and services in total GDP (p for 5/7 models < 0.05). The cumulative amount of outbreaks for pest species = 65.21 + 39.11 × CO 2 emissions as metric tons per capita + 4.27 × contribution of external balance on goods and services in total GDP (R 2 = 0.969, p < 0.001).

Field Observational Database
Except for large-scale monitoring like remote sensing, a high-quality database based on field observation in a large tempo-spatial scale is especially important to characterize the historical patterns and trends for pest accumulations. Existing databases provide limited information for pest species. For example, Senf et al. [17] reviewed the worldwide literature on the remote sensing of insect outbreaks (Figure 1a), resulting in 27 insect species and attributes that include feeding preference, habitat characteristics, host species, and spatial properties; Aukema et al. [19] compiled a list for nonindigenous forest insects with taxonomic information and feeding preference information established in the United States between 1860 to 2006, which resulted in more than 450 insects; Liebhold et al. [24] assembled 62 species of damaging invasive forest insects established in the continental USA, with feeding preference, date of detection, and distribution information recorded. Besides, there is no database based on field studies concentrated in urban areas.
We present the first dataset of insect pests on native C. camphora in urban China over the past half-century based on all relevant pest attributes, which included taxonomic ranks, types of feeders, habitat characteristics, symptoms, signs, damage, and causes (see Appendix 11, Table S14 in Supplementary Material for the dataset). Compiled pest attributes and pest records in a large tempo-spatial scale could help us to collect and integrate data temporally and spatially, to analyze the causes and progress of the pest accumulation and to predict future infestation patterns.

Habitat Characteristics
Both the native species and non-native species showed a decreasing trend from southeast coastal areas to northwest inland areas, which reflected the similarity in their need for environmental conditions. Studies had shown that water potential remained positively associated with pest outbreaks of both native species and non-native species in urban areas [25].
In general, native species have maintained higher fitness than non-native species on C. camphora in China over the past half-century, with a faster accumulation process than non-native species. Related studies also indicated there are more obstacles to the assemblage of non-native species, such as their association with invasion pathways and variation in pest life history [26]. In addition, a lot of studies provided evidence that warming is significantly related to native species in urban areas [27,28], while the environmental effects on non-native species need to be further studied. Studies had further revealed the complexity and specificity in the establishment processes for non-native species, such as the more significant impacts of connectivity of host plant patches, the number of international tourist arrivals, exotic vegetation cover, and regional GDP on invasive species than on native species [29,30]. However, phenotypic plasticity and resource availability can allow non-native pest species to succeed across heterogeneous environments, such as urban forest ecosystems. Since 2010, urban forests in China may experience more frequent fluctuations in allocation, growth, and reproduction.

Types of Feeders
Regional pest outbreaks were characterized by co-occurrence patterns of insect pests in different types of feeders, especially the pest aggregation of leaf feeders, stem feeders, and other feeders, or an aggregation of leaf feeders and stem feeders. The pest accumulation and detection in all of the types of feeders also showed a decreasing trend from eastern coastal areas to western interior regions. Related studies proved there exists similarity among insects in different feeding-types in response to conditional changes. For instance, the positive impacts of temperature and tree water potential were observed both in leaf feeders and stem feeders in cities [31]. In addition, there was an association between urban land-use and species richness and abundance both in leaf feeders and stem feeders [32].
Leaf feeders played a more significant role in the overall insect pest outbreaks of C. camphora in China, followed by stem feeders, then other feeders. Studies had indicated that warming and CO 2 had strong positive effects on foliar consumption for insect pests in urban areas [27,33]. Insects also responded differently to experimentally induced water deficit in plants classified by sap-feeding insects and chewing insects [34]. However, the response of stem feeders and other feeders to conditional changes in urban areas remains to be studied.

Accelerated Urban Pest Outbreaks in China Since the 1990s
Urbanization over the last century has been accompanied by accelerated tree growth [35,36], increased pest fitness and abundance [37], and enhanced pest suitability and plasticity [38]. Related studies put emphasis on improper plantations when discussing the causes for pest damage (see Table 1 for detailed causes). However, the random economic or population-driven process can explain the pest damage in some cases: Researchers have realized the acceleration of insect pest outbreaks. For example, an accumulation of non-native pest insects (8 species per 10 years) was detected on Eucalyptus in California [9]. An accumulation of nonindigenous forest insects (25 species per 10 years) was also observed around the United States [19]. In our research, pest outbreaks of C. camphora in urban China showed an increasing trend with a well-fitted power function over time. Especially, pest outbreaks of C. camphora in urban China accelerated since the 1990s. Research also indicated that global urbanization has increased rapidly with a nearly tripled number of mega-cities and a doubled population living in medium-sized cities since the 1990s [39]. Global emissions of all major greenhouse gases have increased considerably since 1990, such as net emissions of CO 2 that increased by 42% according to the United States Environmental Protection Agency (https://www.epa.gov/climate-indicators/climate-changeindicators-global-greenhouse-gas-emissions). In addition, increasing drought under global warming since the 1980s has been observed [40]. However, there was a significant reduction in the pest abundance since 1990s in some studies [41,42] about croplands, forests, and grassland, which may also indicate that pest outbreaks may not always be explained by the direct impacts of environmental factors.
Insect pests are widely distributed in growing regions of C. camphora in urban China. In particular, the coastal areas had been seriously disturbed with a higher cumulative number and frequency of pest detection, which agreed with the research that showed a highly concentrated pest distribution in the north-eastern region of USA [24]. It was also proved that the increase of pest records was most likely due to economic development and urban expansion since the 1970s [43]. The study further showed that industrialization and urbanization in China were spatially concentrated in eastern coastal cities [44]. On the other hand, relatively high temperatures and suitable humidity of eastern areas in China [45] would motivate pest growth and reproduction. Especially, extreme climatic events like droughts were observed at a significantly increasing trend since the 1960s [46,47].
Particularly, Jiangxi had the highest frequency and diversity of insect pests on C. camphora when compared to other provinces. The changeable climate in Jiangxi is characterized by frequent droughts [48] and increasing extreme precipitation events since the 1960s [49]. Especially in 1964, pests on C. camphora were first detected in Jiangxi, and extreme weather phenomena represented by drought and heavy rain occurred sequentially before the pest outbreak [50]. Similarly, there was pest detection in urban areas in Montes Claros, Brazil, with the tropical savanna climate [51]. In addition, the increased pest damage was also observed on heavily watered trees in urban areas [21]. It may highlight the sensibility, suitability, and plasticity of insect pests when coping with conditional changes [52].

Conclusions
It's essential to estimate the ecosystem stability for urban trees to manage urban sustainability. In this study, we mapped the tempo-spatial changes of vulnerability for C. camphora in urban China between 1964 and 2017 based on pest records from field observation. As a result, a total of 772 pest occurrences (7 orders, 58 families, 225 genera, and 233 species) were recorded, with native leaf feeders dominating the ecosystems in terms of species richness, abundance, and accumulation rate. Pest outbreaks were highly focused in the southeastern coastal areas, with the largest variation and highest frequency over the last 50 years. We also revealed the most likely indirect causes (random economic or population-driven process) for temporal patterns of intensity and frequency of pest outbreaks on camphor trees. Our results can provide detailed information for pest identification in the future and reference information for urban planning and urban forestry, integrative pest management, and data monitoring, as well as for forest dynamics modeling, theoretical validating, and calibrating.
Supplementary Materials: The following are available online at http://www.mdpi.com/2071-1050/12/4/1582/s1, Figure S1: Tempo-spatial distribution of urbanization in China, Figure S2: Photos showing for recorded pest outbreaks in native C. camphora among groups of leaf feeding (a-k), stem feeding (l-p) and others (q-r), Figure  S3: Regional patterns and provincial distributions of total (a) occurrence, (b) species, (c) genera, (d) family, and (e) order for C. camphora hosted pest outbreaks over past half Century, Figure S4: Tempo-spatial distribution of pest outbreaks on the recording of overall (a) and provincial (b-o) insect pest disturbances by types of insect pest habitat characteristics, Figure S5: Tempo-spatial distribution of pest outbreaks on the recording of overall (a) and provincial (b-o) insect pest disturbances by types of feeding types, Table S1: Panel captions for Figure S1, Table S2: Panel captions for Figure S2, Table S3: Worldwide distribution of C. camphora shown in Figure 1b, Table  S4: Selected studies for data extraction, Table S5: Meta data for our pest database, Table S6: Recognizing common signs and symptoms, Table S7: Description of types of feeders clusters, Table S8: List of insect pests found on native urban tree species C. camphora in China in the period 1964-2017, Table S9: Summary of reported outbreaks (Frequency) of insect pests of C. camphora in native geographic range in Chinese urban area, Table S10: Summary of reported native vs. non-native insect pests of C. camphora in native geographic range in China among different province, Table S11: Summary of reported insect pests of C. camphora in native geographic range in China among different province by three feeding types, Table S12: Non-linear simulation and linear regression parameters for Figure 2, Table S13: Linear regression parameters for Figure 3, Table S14: the dataset.