Next Article in Journal
Exploring Nonlinear Intra-Annual Growth Dynamics in Fagus sylvatica L. Trees at the Italian ICP-Forests Level II Network
Next Article in Special Issue
Review of Japanese Pine Bast Scale, Matsucoccus matsumurae (Kuwana) (Coccomorpha: Matsucoccidae), Occurring on Japanese Black Pine (Pinus thunbergii Parl.) and Japanese Red Pine (P. densiflora Siebold & Zucc.) from Korea
Previous Article in Journal
Development of a Tree Growth Difference Equation and Its Application in Forecasting the Biomass Carbon Stocks of Chinese Forests in 2050
Previous Article in Special Issue
Current and Future Distribution of Ricania shantungensis (Hemiptera: Ricaniidae) in Korea: Application of Spatial Analysis to Select Relevant Environmental Variables for MaxEnt and CLIMEX Modeling
 
 
Font Type:
Arial Georgia Verdana
Font Size:
Aa Aa Aa
Line Spacing:
Column Width:
Background:
Article

Occurrence Prediction of the Citrus Flatid Planthopper (Metcalfa pruinosa (Say, 1830)) in South Korea Using a Random Forest Model

1
Department of Biology, Kyung Hee University, Seoul 02447, Korea
2
Division of Life Sciences, Incheon National University, Incheon 22012, Korea
3
Department of Biological Science and Biotechnology, Hannam University, Daejeon 34054, Korea
4
Division of Entomology, Seoul National University, Seoul 08826, Korea
5
Department of Applied Biology, Kyungpook National University, Sangju 37224, Korea
6
Department of Life and Nanopharmaceutical Sciences, Kyung Hee University, Seoul 02447, Korea
*
Author to whom correspondence should be addressed.
Submission received: 6 June 2019 / Revised: 28 June 2019 / Accepted: 9 July 2019 / Published: 12 July 2019
(This article belongs to the Special Issue Impacts, Monitoring and Management of Forest Pests and Diseases)

Abstract

:
Invasive species cause a severe impact on existing ecosystems. The citrus flatid planthopper (CFP; Metcalfa pruinosa (Say, 1830)) is an invasive species in many countries. Predicting potential occurrence areas of the species related to environmental conditions is important for effective forest ecosystem management. In this study, we evaluated the occurrence patterns of the CFP and predicted its potential occurrence areas in South Korea using a random forest model for a hazard rating of forests considering meteorological and landscape variables. We obtained the occurrence data of the CFP in South Korea from literature and government documents and extracted seven environmental variables (altitude, slope, distance to road (geographical), annual mean temperature, minimum temperature in January, maximum temperature in July, and annual precipitation (meteorological)) and the proportion of land cover types across seven categories (urban, agriculture, forest, grassland, wetland, barren, and water) at each occurrence site from digital maps using a Geographic Information System. The CFP occurrence areas were mostly located at low altitudes, near roads and urbanized areas. Our prediction model also supported these results. The CFP has a high potential to be distributed over the whole of South Korea, excluding high mountainous areas. Finally, factors related to human activities, such as roads and urbanization, strongly influence the occurrence and dispersal of the CFP. Therefore, we propose that these factors should be considered carefully in monitoring and surveillance programs for the CFP and other invasive species.

1. Introduction

The occurrence of invasive species has increased worldwide due to global warming and international trade [1,2,3]. At the beginning of the invasion process, the community containing the invasive species does not have appropriate species interactions, such as prey–predator, host–parasite, and competition between species to stabilize the community dynamics [4,5,6]. Therefore, they can cause a severe impact on the existing ecosystem, resulting in notorious pests in many countries [7], for example, the gypsy moth (Lymantria dispar L.), the Asian long horn beetle (Anoplophora glabripennis Motschulsky), the emerald ash borer (Agrilus planipennis Fairmaire), found in the USA [8], and the pine wood nematode (Bursaphelenchus xylophilus Steiner and Buhrer), which is found in many countries, including Korea and Japan [9,10].
The citrus flatid planthopper (CFP; Metcalfa pruinosa Say) (Hemiptera, Auchenorrhyncha: Flatidae) is one invasive species which causes several problems in many countries. This species, originally from North America [11], has spread widely throughout America [12]. In Europe, it was first reported in Italy in 1979 [13] and is now found throughout Europe [14,15,16,17,18,19,20]. Meanwhile, in Asia, the presence of this species was first reported in South Korea in 2005 [21] and later in 2009 [22]. Recently, its distribution area has expanded to a nationwide scale [23].
The CFP damages its host plants by feeding on sap and ejecting honeydew, which induces a sooty mold disease [22,24]. The CFP is a polyphagous species feeding on a wide range of host plants in Korea, including woody trees, such as Robinia pseudoacacia L., Castanea crenata Siebold et Zucc., Styrax japonicus Siebold et Zucc., Diospyros lotus L., and Acer palmatum Thunb., and shrubs and herbaceous plants, such as Hibiscus syriacus L., Rhamnus davurica Pall., and Lycium chinense Mill. [22,23].
The occurrence and dispersal of the CFP are highly related to human activities such as roads, highways, and movement of vehicles on a local scale [25,26,27]. However, the occurrence and dispersal of invasive species are influenced by various other factors, including meteorological and geographical conditions, landscape, and host plants [28,29,30]. To effectively manage invasive pests, the evaluation of the occurrence patterns related to environmental conditions and the estimation of the potential occurrence area of the species are essential to provide information to efficiently identify current or future hazardous conditions [28,29,31].
Therefore, various hazard rating systems have been developed for different forest pests. In the early stage of hazard rating, qualitative and statistical methods such as discriminant analysis are used [32,33,34]. Further, recent machine-learning algorithms, such as self-organizing maps, multi-layer perceptron, classification and regression trees, and random forest models, can been used [28,29,35,36].
Despite the importance of hazard ratings for controlling the CFP, there are limited studies on hazard ratings using the CLIMEX model [37] which is used for the prediction of species distribution [38,39,40,41] and for risk assessment of invasive species [42,43]. CLIMEX is a climate-specific model to predict potential distribution and relative abundance of species by considering climatic conditions [44]. However, it does not consider other factors, such as road and landscape conditions. Meanwhile, MaxEnt [45] as a species distribution model is also frequently used to predict the potential distribution of invasive species based on presence-only species records [46]. MaxEnt cannot be applied to abundance data, only to presence data. Recently, studies to predict the distribution of (potential) invasive species, including CFP, using CLIMEX and MaxEnt were reviewed by evaluating their effectiveness [39].
The CFP is one of the quarantine pests listed by the Animal and Plant Quarantine Agency, Republic of Korea (https://www.qia.go.kr/) and its distribution area has been increasing since the first occurrence in Korea. Therefore, it has a high potential as a key insect pest on various plants in agriculture and forests [22]. In this study, we aimed to evaluate the occurrence patterns of the CFP and to predict its potential occurrence area for a hazard rating of forests, considering meteorological and landscape variables in Korea.

2. Materials and Methods

2.1. Data Collection

We obtained occurrence data of the CFP (Metcalfa pruinosa) from 105 presence (observed) sites in South Korea from the literature [23,27,47] and government documents from websites of the Gyeonggi-do Agricultural Research Services (https://nongup.gg.go.kr) and the Ministry of Agriculture, Food, and Rural Affairs (http://www.mafra.go.kr) (Figure 1). Our CFP occurrence data included only presence (i.e., observed) sites and we assumed that CFP was in fact absent at the unobserved sites at that moment because the field surveys were extensively conducted to monitor the occurrence of CFP. To evaluate the environmental conditions of the CFP occurrence sites, we selected absence sites with the same number of presence sites randomly with distances over 1.5 km from the presence sites and over 5 km from other selected absence sites to avoid similar environmental conditions. Finally, our dataset comprised 210 sites (105 sites each for presence and absence).
At each selected site (for both presence and absence), we extracted seven environmental variables (i.e., altitude, slope, and distance to road as geographical variables and annual mean temperature, minimum temperature in January, maximum temperature in July, and annual precipitation as meteorological variables) and the proportion of land cover types across seven categories (urban, agriculture, forest, grassland, wetland, barren, and water) from digital maps using a Geographic Information System (ArcGIS 10.1, ESRI). Altitude and slope were extracted from the digital elevation model data obtained from the National Geographic Information Institute (http://www.ngii.go.kr). Distance to road indicated a distance from the observation site to a nearest road and it was measured from a road map by combining road data from the National Spatial Data Infrastructure Portal (http://www.nsdi.go.kr) and the Intelligent Transport Systems of Standard Node Link (http://nodelink.its.go.kr). Meteorological variables were extracted from the digital map of Climatological Normals from the Climate Information Portal of the Korea Meteorological Administration (http://www.climate.go.kr) and Applied SERS A1B scenario data from the Digital Agro-climate Map Database for Impact Assessment of Climate Change on Agriculture (http://www.agdcm.kr) of the Rural Development Administration, Korea. The Environmental Geographic Information Service (https://egis.me.go.kr) operated by the Ministry of Environment, Korea, provided the digital land cover map, which was used to extract land cover area. We used raster digital maps with a resolution of 30 m × 30 m cell size.

2.2. Analysis of Occurrence Patterns

The number of occurrence regions, such as districts (i.e., “si”, “gun”, or “gu” level in Korea) and the area damaged by the CFP were evaluated for different years (Figure 2). The Mann–Whitney U test was used to evaluate the statistically significant differences between the presence and absence sites for environment and land cover conditions. The Mann–Whitney U test was performed with a “stats” package in R [48].

2.3. Prediction of Potential Occurrence

To estimate the occurrence risk of the CFP in the forests on a nationwide scale in Korea, we predicted the potential occurrence area (presence: 1; absence: 0) of the CFP based on the environmental conditions at the occurrence sites using a random forest (RF) model. The RF is an ensemble machine-learning model and it is a non-parametric method of predicting and assessing the relationships among a large number of potential predictor variables and a response variable [49]. In the RF model, we used five environmental variables (altitude, slope, and distance to road (geographical) and minimum temperature in January and maximum temperature in July). We excluded annual mean temperature to avoid collinearity among variables (Figure 3) and annual precipitation, which was not significantly different (p < 0.05) between the presence and absence sites. Although altitude and minimum temperature had high correlation coefficients, we included them as predictors because altitude strongly influences the distribution and abundance of insects [50] and the minimum temperature in January greatly affects the mortality of insects during the winter [6]. We calculated a mean decrease in node impurity by the residual sum of squares to evaluate the relative importance of each variable to determine the presence or absence status of CFP occurrence in the RF model.
We standardized all environmental variables a priori the modeling procedure. We divided our dataset in two parts with a ratio of 8:2 with respect to training and testing the models. We repeated the procedure for the RF model 10 times and averaged the calculated values. The RF models were performed using the package “randomForest” [51] in R. Partial dependence plot function (partialPlot) in “randomForest” and generalized additive model (GAM) in the “mgcv” package [52] were used to present the marginal effect of environmental variables in our model. The model performance was evaluated by accuracy and area under the receiver operating characteristic (AUROC) curve value using the “ROCR” package [53] in R.

3. Results

3.1. Occurrence Patterns of CFP

The number of occurrence regions (districts as “si”, “gun”, or “gu” in Korea) increased as a function of year after the first occurrence. It was less than five districts in 2009, but increased greatly to over 20 districts in 2010. From 2010 to 2015, the rate of increase was relatively low. However, in 2016, the number of occurrence regions abruptly increased from 43 to 60 districts, showing an exponential growth pattern. The damaged area displayed similar patterns regarding the number of occurrence regions from 2014 to 2017 (r = 0.95, p < 0.05) when the data were available.
Geographical and meteorological conditions were significantly different between the presence and absence sites (Mann–Whitney U test, p < 0.05) (Figure 4). The CFP was mostly observed in areas with low altitude (103.2 m ± 8.9 standard error (SE)), low slope (5.0° ± 0.6 SE), and short distance to the road (24.2 m ± 4.0 SE). Annual mean temperature and maximum temperature in July were higher at the presence sites than at the absence sites (Mann–Whitney U test, p < 0.01), while the minimum temperature in January was lower at the presence sites (Mann–Whitney U test, p < 0.01). Annual precipitation was not significantly different between the presence and absence sites.
The ratios of all land cover types, except wetland, were significantly different between the presence and absence sites (Mann–Whitney U test, p < 0.01) (Figure 5). The presence sites were mostly observed in forests (33.8% ± 2.2 SE), agricultural land (31.6% ± 2.3 SE), and urban areas (19.5% ± 1.5 SE). The proportion of presence in forests was lower than that of absence, whereas the proportion of urban and agricultural areas of presence sites were higher than those of the absence sites.

3.2. Prediction of Potential Occurrence Area

The RF model predicted the potential occurrence areas of the CFP based on the geographical and meteorological conditions. The model showed high predictability: 0.78 (±0.01 SE) accuracy and 0.80 (±0.01 SE) AUROC curve value (Figure 6a). The model displayed that the occurrence probability was mostly high in lowland and urban areas (Figure 1 and Figure 6b).
The value of mean decrease in node impurity presented the distance to road as the most important variable for the prediction of the CFP occurrence sites in the RF model, followed by altitude and slope (Figure 7). This was reflected in the partial dependence plot of occurrence probability responding to the changes in the dependent variables (Figure 8). The occurrence probability of the CFP mostly changed as a function of distance to road and altitude, in particular, at low values of distance to road (mostly <50 m) and altitude (mostly <200 m), which displayed high occurrence probability. The probability was not highly influenced at high altitude and long distance to road.

4. Discussion

4.1. Occurrence Patterns of CFP

In South Korea, CFPs were first observed as adults only in Hanrim of Gimhae-si in the southern part of South Korea in 2005. The nymphs and adults were observed in Jinyoung of Gimhae-si, which is the town next to Hanrim, in 2008 [21,22]. Meanwhile, the CFP was discovered at several locations in the southern to central regions of the Korean Peninsula, including Seoul and Suwon in 2009 [22], and its occurrence and damage area increased greatly from 2010 (Figure 2). Currently, the CFP is widely distributed in South Korea. Thus, considering the distribution area of the CFP over time, the dispersal of the CFP was limited, with a small dispersal area, at the early stage of the invasion process from the first occurrence in 2005 until 2009. This phenomenon is commonly observed in the dispersal of invasive species [4,54]. During the early stage of invasion, the species might colonize and increase its population size, then disperse to the neighboring areas [8].
Regarding the increase in occurrence area, we have to note that there was a big difference between 2009 and 2010 (Figure 2). We infer that the actual distribution of the CFP in 2009 was wider than that displayed by the recorded data. The observed number of locations in 2009 might be smaller than that of the actual ones because the CFP was not well recognized as an invasive species and a pest insect by entomologists and policy decision-makers in South Korea until 2009. Meanwhile, the distribution area of the CFP increased exponentially after 2015, showing a stepwise dispersal pattern in the number of occurrence areas. This pattern was also reported in the dispersal of invasive species such as pine wilt disease in both Korea [9] and Japan [10]. This may be related to the increase in exponential population growth. However, the mechanism of this dispersal pattern was not clear. Therefore, further study is needed to understand the dispersal patterns of invasive species to effectively manage them.

4.2. Invasion Route and Factors Promoting Dispersals

Two possible routes were proposed for the CFP invasion into South Korea: From North America, which is the origin of the CFP, or from European countries [22], which would imply that the CFP arrived in South Korea through the imports of horticultural nursery stock. Similarly, the possible means of the CFP invasion were mostly by passive transport of the CFP adults or eggs with infested plants of many countries, including the Czech Republic and Romania [55,56]. Guillemaud, et al. [57] reported two main factors promoting biological invasion: Human activities and the biological adaptation of species. Human activities, including international trade and movement of vehicles among local regions, are highly responsible for recent biological invasions by moving invasive species and causing long distance dispersal [4,5,9,14,25,27,58,59,60,61,62]. Human activities also promote the dispersal of invasive species by modifying and disturbing natural habitats and providing the possibility of the introduction of new species into ecosystems [63].
Our results also revealed that human activities are the main influencing factors for the occurrence and dispersal of the CFP in South Korea. The CFP was observed mostly in areas close to the road with low altitude and slopes (Figure 4). This was also supported by the difference in occurrence patterns in different land use types. The occurrence of the CFP was observed mainly in urban and agricultural areas (Figure 5), which are highly influenced by human activities. Meanwhile, the proportion of presence sites in forest areas was smaller than that of absence sites. However, the occurrence proportion in forest areas was still the highest (over 33.8%) among land use types. This indicates that the occurrence areas in the forest are mostly in lowland areas which are close to the road and urban and agricultural areas.
Meanwhile, flight activities of the CFP adults contribute to local dispersal [14,56]. The annual dispersal distance of the CFP was estimated at 0.2–0.5 km in Austria [64]. The dispersal distance in natural conditions can be variable depending on various environmental conditions, such as temperature, precipitation, landscape, geology, and distribution of host plants [4,5,8,14,58,59,60], although human activities play a substantial role in their dispersal. Therefore, an integrated approach for the evaluation of the dispersal of invasive species is recommended with the consideration of environmental factors in several different categories, if data are available.

4.3. Host Plants

The high values of CFP occurrence in urban, agricultural, and forest areas reflect the diversity of both their habitats and host plants. The CFP is polyphagous; it has a variety of host plants and can inhabit very different environmental conditions. Therefore, the CFP can damage various types of plants, including ornamental or agricultural plants and herbaceous or woody plants [22,23,56]. It has a high probability of dispersing to new areas with commercial trade and transportation. Meanwhile, in this study, we did not include the distribution of host plants in the analysis and in the model for the potential occurrence prediction of the CFP because of its polyphagous nature and the difficulty in characterizing the distribution of host plants on the digital map. From another aspect, polyphagous species have a wide distribution of host plants. Therefore, our results would not be influenced by the exclusion of host plants distribution.

4.4. Hazard Rating with the CFP

The hazard rating of forests regarding susceptibility to insect pests and diseases provides information to efficiently identify current and future hazardous conditions in the concerned forests [28]. Therefore, it is very important in forest ecosystem management [29]. In this study, we evaluated the hazard rate of forests on a nationwide scale in Korea through the prediction of the potential occurrence of the CFP, showing that most areas in the South Korean territory, excluding high mountain areas, have a high occurrence probability of the CFP (Figure 5). Distance to road displayed the highest contribution regarding predicting the occurrence of CFP, followed by altitude and slope (Figure 7). The marginal effect of each variable to the model output displayed the influence of each variable to the hazard rating (Figure 8). The occurrence probability of the CFP showed a wide variation according to the distance to road, altitude, and slope. The occurrence probability was high at low values of these variables, while it remained at low values above certain values of each variable (300 m, 300 m, and 20° for the distance to road, altitude, and slope, respectively). The importance of roads in the invasion of species was supported by previous studies [25,26,27], showing that roads that allow human activities contribute highly to the early invasion and dispersal of species, because sites near roads are more likely to receive invasive species. These results indicate that these variables greatly affect the dispersal and occurrence of the CFP. Therefore, we propose that these factors should be considered carefully in the monitoring and surveillance programs for the CFP and other invasive species.
Meanwhile, minimum and maximum temperatures in January and July induced little variation in the occurrence probability. However, both temperature and occurrence probability displayed a positive relationship. This indicates that most areas in South Korea have a suitable temperature for the CFP. Although temperature is a fundamental factor for the development and distribution of species, it did not display an influential impact on the model output. This might be due to the relatively small differences among study sites where the South Korean territory was not big enough to display temperature differences. Therefore, if this was applied in a much larger area, temperature factors could have been influential on the distribution of the species. In this regard, the CLIMEX model has been frequently used for the prediction of the potential distribution area based on temperature conditions. Strauss [26] predicted the CFP’s potential geographical distribution and identified areas at risk in Austria using CLIMEX®. The potential distribution of CFP and its dispersal were evaluated in South Korea using CLIMEX according to the climate change scenario [38]. Similar to our study, their results showed that CFP has spread in all regions of South Korea under the current climate. Both CLIMEX and MaxEnt are frequently used for the prediction of potential distribution of various species with different objectives, such as conservation of endangered species, control of invasive species, etc. However, there are some limitations in their applications. CLIMEX is based on climate-oriented data, whereas MaxEnt is not applicable to abundance data, only presence/absence data. Meanwhile, the RF model used in this study can be applied to presence/absence data as well as abundance data with various types of predictors, although presence/absence data were used in this study.
Although our models predicted the occurrence probability and potential distribution areas based on various environmental factors, the actual occurrence probability and hazard rating could be influenced by the adaptation ability of invasive species, which was not considered in our study. Species have a great ability to adapt to new environments, including changes in temperature and habitats, and this promotes the colonization of invasive species and dispersal to new regions. Therefore, further study is needed on the influence of biological adaptation of invasive species on dispersal and settlement.

5. Conclusions

The CFP occurrence areas were mostly located at low altitudes and near roads and urbanized areas. The RF model predicted the potential distribution areas of CFP with five environmental variables, showing that the CFP has a great potential to be distributed over the whole of South Korea, excluding high mountainous areas. Finally, factors related to human activities, such as roads and urbanization, strongly influenced the occurrence and dispersal of the CFP. Therefore, these variables should be considered carefully in monitoring and surveillance programs for the CFP and other invasive species to effectively control the CFP.

Author Contributions

Conceptualization, D.-S.L. and Y.-S.P.; data curation, D.-S.L., Y.-S.B., B.-K.B., S.L., and J.K.P.; methodology, D.-S.L. and Y.-S.P.; writing—original draft preparation, D.-S.L. and Y.-S.P.; writing—review and editing, D.-S.L., Y.-S.B., B.-K.B., S.L., J.K.P., and Y.-S.P.

Funding

This study was supported by the ′R&D Program for Forest Science Technology (FTIS 2017042A00-1823-CA01) provided by Korea Forest Service (Korea Forestry Promotion Institute) and by the National Research Foundation of Korea (NRF) funded by the Korean government (MSIP) (grant number NRF-2016R1A2B4011801).

Acknowledgments

The authors are grateful to survey members involved in the project supported by the ′R&D Program for Forest Science Technology provided by Korea Forest Service for their participation in the field survey.

Conflicts of Interest

The authors declare no conflict of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or in the decision to publish the results.

References

  1. Hughes, L. Biological consequences of global warming: Is the signal already apparent? TREE 2000, 15, 56–61. [Google Scholar] [CrossRef]
  2. Walther, G.R.; Post, E.; Convey, P.; Menzel, A.; Parmesan, C.; Beebee, T.J.C.; Fromentin, J.M.; Hoegh-Guldberg, O.; Bairlein, F. Ecological responses to recent climate change. Nature 2002, 416, 389–395. [Google Scholar] [CrossRef] [PubMed]
  3. Hulme, P.E. Trade, transport and trouble: Managing invasive species pathways in an era of globalization. J. Appl. Ecol. 2009, 46, 10–18. [Google Scholar] [CrossRef]
  4. Choi, W.I.; Song, H.J.; Kim, D.S.; Lee, D.-S.; Lee, C.-Y.; Nam, Y.; Kim, J.-B.; Park, Y.-S. Dispersal patterns of pine wilt disease in the early stage of its invasion in South Korea. Forests 2017, 8, 411. [Google Scholar] [CrossRef]
  5. Lee, D.-S.; Nam, Y.; Choi, W.I.; Park, Y.-S. Environmental factors influencing on the occurrence of pine wilt disease in Korea. Korean J. Ecol. Environ. 2017, 50, 374–380. [Google Scholar] [CrossRef]
  6. Choi, W.I.; Jeon, M.J.; Park, Y.S. Structural dynamics in the host-parasitoid system of the pine needle gall midge (Thecodiplosis japonensis) during invasion. PeerJ 2017, 5, e3610. [Google Scholar] [CrossRef]
  7. USDA. Major Forest Insect and Disease Conditions in the United States: 2015; U.S. Department of Agriculture: Washington, DC, USA, 2017; p. 45.
  8. Liebhold, A.M.; Tobin, P.C. Population ecology of insect invasions and their management. Annu. Rev. Entomol. 2008, 53, 387–408. [Google Scholar] [CrossRef]
  9. Choi, W.I.; Park, Y.-S. Dispersal patterns of exotic forest pests in South Korea. Insect Sci. 2012, 19, 535–548. [Google Scholar] [CrossRef]
  10. Togashi, K.; Shigesada, N. Spread of the pinewood nematode vectored by the Japanese pine sawyer: Modeling and analytical approaches. Popul. Ecol. 2006, 48, 271–283. [Google Scholar] [CrossRef]
  11. Wilson, S.W.; McPherson, J.E. Life histories of Anormenis septentrionalis, Metcalfa pruinosa and Ormenoides venusta with description of immature stages. Ann. Entomol. Soc. Am. 1981, 74, 299–311. [Google Scholar] [CrossRef]
  12. Metcalf, Z.P.; Bruner, S.C. Cuban Flatidae with new species from adjacent regions. Ann. Entomol. Soc. Am. 1948, 41, 63–118. [Google Scholar] [CrossRef]
  13. Zangheri, S.; Donadini, P. Appearance in the Venice district of a Nearctic bug: Metcalfa pruinosa Say (Homoptera, Flatidae). Redia 1980, 63, 301–305. [Google Scholar]
  14. Preda, C.; Skolka, M. Range expansion of Metcalfa pruinosa (Homoptera: Fulgoroidea) in Southeastern Europe. Ecol. Balk. 2011, 3, 79–87. [Google Scholar]
  15. Balakhnina, I.V.; Pastarnak, I.N.; Gnezdilov, V.M. Monitoring and control of Metcalfa pruinosa (Say) (Hemiptera, Auchenorrhyncha: Flatidae) in Krasnodar Territory. Entomol. Rev. 2014, 94, 1067–1072. [Google Scholar] [CrossRef]
  16. Malumphy, C.; Baker, R.; Cheek, S. Citrus Planthopper, Metcalfa pruinosa; Plant Pest Notice No. 19; Central Science Laboratory: Aberdeen, UK, 1994.
  17. Drosopoulos, A.; Broumas, T.; Kapothanassi, V. Metcalfa pruinosa (Hemiptera, Auchenorrhyncha: Flatidae) an undesirable new species in the insect fauna of Greece. Ann. Benaki Phytopathol. Inst. 2004, 20, 49–51. [Google Scholar]
  18. Nickel, H. Arrival of the citrus flatid planthopper Metcalfa pruinosa (Say, 1830) in Germany and northern Switzerland. Entomo Helv. 2016, 9, 129–136. [Google Scholar]
  19. Della Giustina, W. Metcalfa pruinosa (Say 1830), new for French fauna (Hom.: Flatidae). Bulletin de la Société Entomologique de France 1986, 91, 89–92. [Google Scholar]
  20. European and Mediterranean Plant Protection Organization (EPPO). EPPO Global Database; European and Mediterranean Plant Protection Organization: Paris, France, 2018. [Google Scholar]
  21. Lee, H.S.; Wilson, S.W. First report of the Nearctic flatid planthopper Metcalfa pruinosa (Say) in the Republic of Korea (Hemiptera: Fulgoroidea). Entomol. News 2010, 121, 506–513. [Google Scholar] [CrossRef]
  22. Kim, Y.; Kim, M.; Hong, K.-J.; Lee, S. Outbreak of an exotic flatid, Metcalfa pruinosa (Say) (Hemiptera: Flatidae), in the capital region of Korea. J. Asia-Pac. Entomol. 2011, 14, 473–478. [Google Scholar] [CrossRef]
  23. Kim, D.-E.; Kil, J. Occurrence and host plant of Metcalfa pruinosa (Say) (Hemiptera: Flatidae) in Korea. J. Environ. Sci. Int. 2014, 23, 1385–1394. [Google Scholar] [CrossRef]
  24. Wilson, S.W.; Lucchi, A. Feeding activity of the flatid planthopper Metcalfa pruinosa (Hemiptera: Fulgoroidea). J. Kans. Entomol. Soc. 2007, 80, 175–178. [Google Scholar] [CrossRef]
  25. Pantaleoni, R.A. The ways in which Metcalfa pruinosa (Say, 1830) (Auchenorrhyncha Flatidae) invades a new area. Bollettino dell’Istituto di Entomologia Guido Grandi della Università degli Studi di Bologna 1989, 43, 1–7. [Google Scholar]
  26. Strauss, G. Pest risk analysis of Metcalfa pruinosa in Austria. J. Pest Sci. 2010, 83, 381–390. [Google Scholar] [CrossRef]
  27. Kil, J.H.; Lee, D.H.; Hwang, S.M.; Lee, C.W.; Kim, Y.H.; Kim, D.E.; Kim, H.M.; Yang, H.S.; Lee, J.C. Detailed Studies on Invasive Alien Species and Their Management (VI); National Institute of Environmental Research: Incheon, Korea, 2011.
  28. Park, Y.-S.; Chung, Y.-J. Hazard rating of pine trees from a forest insect pest using artificial neural networks. For. Ecol. Manag. 2006, 222, 222–233. [Google Scholar] [CrossRef]
  29. Park, Y.-S.; Chung, Y.-J.; Moon, Y.-S. Hazard ratings of pine forests to a pine wilt disease at two spatial scales (individual trees and stands) using self-organizing map and random forest. Ecol. Inform. 2013, 13, 40–46. [Google Scholar] [CrossRef]
  30. Futai, K. Pine wilt in Japan: From first incidence to the present. In Pine Wilt Disease; Zhao, B.G., Futai, K., Sutherland, J.R., Takeuchi, Y., Eds.; Springer: Tokyo, Japan, 2008; pp. 5–12. [Google Scholar]
  31. Mason, G.N.; Lorio, P.L., Jr.; Belanger, R.P.; Nettleton, W.A. Rating the Susceptibility of Stands to Southern Pine Beetle Attack. Integrated Pest Management Handbook; Agriculture Handbook No. 645; USDA, Forest Service: Washington, DC, USA, 1985.
  32. Valentine, H.T.; Houston, D.R. A discriminant function for identifying mixed-oak stand susceptibility to gypsy moth Lymantria dispar defoliation. For. Sci. 1979, 25, 468–474. [Google Scholar]
  33. Hicks, R.R., Jr.; Howard, J.E.; Coster, J.E.; Watterston, K.G. Rating forest stand susceptibility to southern pine beetle in east Texas using probability functions. For. Ecol. Manag. 1980, 2, 269–283. [Google Scholar] [CrossRef]
  34. Lorio, P.L., Jr.; Sommers, R.A. Gulf coastal plain (Lousiana). In Site, Stand, and Host Characteristics of Southern Pine Beetle Infestations; USDA Forest Service, Tech. Bull., 1612. Comb. For. Pest Res. Devlop. Prog.; Coster, J.E., Searcy, J.L., Eds.; USDA Forest Service: Pineville, LA, USA, 1980. [Google Scholar]
  35. Nam, Y.; Koh, S.-H.; Jeon, S.-J.; Youn, H.-J.; Park, Y.-S.; Choi, W.I. Hazard rating of coastal pine forests for a black pine bast scale using self-organizing map (SOM) and random forest approaches. Ecol. Inform. 2015, 29, 206–213. [Google Scholar] [CrossRef]
  36. Baker, F.A.; Verbyla, D.L.; Hodges, C.S., Jr.; Ross, E.W. Classification and regression tree analysis for assessing hazard of pine mortality caused by Heterobasidion annosum. Plant Dis. 1993, 77, 136–139. [Google Scholar] [CrossRef]
  37. Sutherst, R.W.; Maywald, G.F.; Bottomley, W.; Bourne, A. CLIMEX for Windows, Version 2 User’s Guide; CRC for Tropical Pest Management: Brisbane, Australia, 2004. [Google Scholar]
  38. Byeon, D.-H.; Jung, J.-M.; Lohumi, S.; Cho, B.-K.; Jung, S.; Lee, W.-H. Predictive analysis of Metcalfa pruinosa (Hemiptera: Flatidae) distribution in South Korea using CLIMEX software. J. Asia-Pac. Biodivers. 2017, 10, 379–384. [Google Scholar] [CrossRef]
  39. Byeon, D.-H.; Jung, S.; Lee, W.-H. Review of CLIMEX and MaxEnt for studying species distribution in South Korea. J. Asia-Pac. Biodivers. 2018, 11, 325–333. [Google Scholar] [CrossRef]
  40. Vanhanen, H.; Veteli, T.O.; Paivinen, S.; Kellomaki, S.; Niemela, P. Climate change and range shifts in two insect defoliators: Gypsy moth and nun moth—A model study. Silva Fenn. 2007, 41, 621. [Google Scholar] [CrossRef]
  41. Sridhar, V.; Verghese, A.; Vinesh, L.S.; Jayashankar, M.; Jayanthi, P.K. CLIMEX simulated predictions of Oriental fruit fly, Bactrocera dorsalis (Hendel) (Diptera: Tephritidae) geographical distribution under climate change situations in India. Curr. Sci. 2014, 106, 1702–1710. [Google Scholar]
  42. Logan, J.A.; Régnière, J.; Powell, J.A. Assessing the impacts of global warming on forest pest dynamics. Front. Ecol. Environ. 2003, 1, 130–137. [Google Scholar] [CrossRef]
  43. Kumar, S.; Neven, L.G.; Zhu, H.; Zhang, R. Assessing the global risk of establishment of Cydia pomonella (Lepidoptera: Tortricidae) using CLIMEX and MaxEnt niche models. J. Econ. Entomol. 2015, 108, 1708–1719. [Google Scholar] [CrossRef]
  44. Canberra Australia: Commonwealth Scientific and Industrial Research Organisation (CSIRO). CLIMEX Version 4: Exploring the Effects of Climate on Plants, Animals and Diseases; CSIRO: Canberra, Australia, 2015.
  45. Phillips, S.J.; Anderson, R.P.; Schapire, R.E. Maximum entropy modeling of species geographic distributions. Ecol. Model. 2006, 190, 231–259. [Google Scholar] [CrossRef] [Green Version]
  46. Elith, J.; Phillips, S.J.; Hastie, T.; Dudík, M.; Chee, Y.E.; Yates, C.J. A statistical explanation of MaxEnt for ecologists. Divers. Distrib. 2011, 17, 43–57. [Google Scholar] [CrossRef]
  47. Korea Forestry Promotion Institute (KOFPI). Development of Prediction Models for Dispersal and Population Change of Forest Exotic Insects (2nd Year Report); Korea Forestry Promotion Institute: Seoul, Korea, 2018. [Google Scholar]
  48. R Core Team. R: A Language and Environment for statistical Computing; R Foundation for Statistical Computing: Vienna, Austria, 2017. Available online: https://www.R-project.org/ (accessed on 23 June 2018).
  49. Breiman, L. Random forests. Mach. Learn. 2001, 45, 5–32. [Google Scholar] [CrossRef]
  50. Hodkinson, I.D. Terrestrial insects along elevation gradients: Species and community responses to altitude. Biol. Rev. 2005, 80, 489–513. [Google Scholar] [CrossRef]
  51. Liaw, A.; Wiener, M. Classification and Regression by randomForest. R News 2002, 2/3, 18–22. [Google Scholar]
  52. Wood, S.N. Generalized Additive Models: An Introduction with R; Chapman and Hall/CRC: Boca Raton, FL, USA, 2017. [Google Scholar]
  53. Sing, T.; Sander, O.; Beerenwinkel, N.; Lengauer, T. ROCR: Visualizing classifier performance in R. Bioinformatics 2005, 21, 3940–3941. [Google Scholar] [CrossRef] [PubMed]
  54. Lee, S.D.; Park, S.; Park, Y.-S.; Chung, Y.-J.; Lee, B.-Y.; Chon, T.-S. Range expansion of forest pest populations by using the lattice model. Ecol. Model. 2007, 203, 157–166. [Google Scholar] [CrossRef]
  55. Lauterer, P. Citrus flatid planthopper—Metcalfa pruinosa (Hemiptera: Flatidae), a new pest of ornamental horticulture in the Czech Republic. Plant Prot. Sci. 2002, 38, 145–148. [Google Scholar] [CrossRef]
  56. Grozea, I.; Gogan, A.; Virteiu, A.M.; Grozea, A.; Stef, R.; Molnar, L.; Carabet, A.; Dinnesen, S. Metcalfa pruinosa Say (insecta: Homoptera: Flatidae): A new pest in Romania. Afr. J. Agric. Res. 2011, 6, 5870–5877. [Google Scholar] [CrossRef]
  57. Guillemaud, T.; Ciosi, M.; Lombaert, É.; Estoup, A. Biological invasions in agricultural settings: Insights from evolutionary biology and population genetics. Comptes Rendus Biol. 2011, 334, 237–246. [Google Scholar] [CrossRef] [PubMed] [Green Version]
  58. Auffret, A.G.; Cousins, S.A.O. Humans as long-distance dispersers of rural plant communities. PLoS ONE 2013, 8, e62763. [Google Scholar] [CrossRef] [PubMed]
  59. Wilson, J.R.; Dormontt, E.E.; Prentis, P.J.; Lowe, A.J.; Richardson, D.M. Something in the way you move: Dispersal pathways affect invasion success. Trends Ecol. Evol. 2009, 24, 136–144. [Google Scholar] [CrossRef] [PubMed]
  60. Lozon, J.D.; MacIsaac, H.J. Biological invasions: Are they dependent on disturbance? Environ. Rev. 1997, 5, 131–144. [Google Scholar] [CrossRef]
  61. Cox, G.W. Alien Species and Evolution: The Evolutionary Ecology of Exotic Plants, Animals, Microbes, and Interacting Native Species; Island Press: Washington, DC, USA, 2004. [Google Scholar]
  62. Strauss, G. Environmental risk assessment for Neodryinus typhlocybae, biological control agent against Metcalfa pruinosa, for Austria. Eur. J. Environ. Sci. 2012, 2, 102–109. [Google Scholar] [CrossRef]
  63. Lake, J.C.; Leishman, M.R. Invasion success of exotic in natural ecosystems: The role of disturbance, plant attributes and freedom from herbivores. Biol. Conserv. 2004, 117, 215–226. [Google Scholar] [CrossRef]
  64. Kahrer, A.; Strauss, G.; Stolz, M.; Moosbeckhofer, R. Beobachtungen zu Faunistik und Biologie der vor kurzem nach Österreich eingeschleppten Bläulingszikade (Metcalfa pruinosa). Beiträge zur Entomofaunistik 2009, 10, 17–30. [Google Scholar]
Figure 1. Occurrence (presence) sites of Metcalfa pruinosa based on the literature [23,27,47].
Figure 1. Occurrence (presence) sites of Metcalfa pruinosa based on the literature [23,27,47].
Forests 10 00583 g001
Figure 2. Change in (a) the number of occurrence regions and (b) area damaged by Metcalfa pruinosa in South Korea. Data were from databases (Gyeonggi-do Agricultural Research Services (https://nongup.gg.go.kr) and the Ministry of Agriculture, Food, and Rural Affairs (http://www.mafra.go.kr)). Damaged area data in Figure 2b were only available from 2014.
Figure 2. Change in (a) the number of occurrence regions and (b) area damaged by Metcalfa pruinosa in South Korea. Data were from databases (Gyeonggi-do Agricultural Research Services (https://nongup.gg.go.kr) and the Ministry of Agriculture, Food, and Rural Affairs (http://www.mafra.go.kr)). Damaged area data in Figure 2b were only available from 2014.
Forests 10 00583 g002
Figure 3. Correlation coefficient between environmental variables measured at 210 presence or absence sites.
Figure 3. Correlation coefficient between environmental variables measured at 210 presence or absence sites.
Forests 10 00583 g003
Figure 4. Differences in environmental conditions between the presence and absence sites of Metcalfa pruinosa. Bars denote averages and standard errors. The p-value (p) was calculated based on the Mann–Whitney U test. (a) Altitude, (b) slope, (c) distance to road, (d) annual precipitation, (e) annual average temperature, (f) maximum temperature in July, and (g) minimum temperature in January.
Figure 4. Differences in environmental conditions between the presence and absence sites of Metcalfa pruinosa. Bars denote averages and standard errors. The p-value (p) was calculated based on the Mann–Whitney U test. (a) Altitude, (b) slope, (c) distance to road, (d) annual precipitation, (e) annual average temperature, (f) maximum temperature in July, and (g) minimum temperature in January.
Forests 10 00583 g004
Figure 5. Differences in land cover types between presence and absence sites of Metcalfa pruinosa. The p-value (p) was calculated based on the Mann–Whitney U test. Urban: Urban area; Agr: Agricultural land; Wet: Wetland; and Bar: Barren.
Figure 5. Differences in land cover types between presence and absence sites of Metcalfa pruinosa. The p-value (p) was calculated based on the Mann–Whitney U test. Urban: Urban area; Agr: Agricultural land; Wet: Wetland; and Bar: Barren.
Forests 10 00583 g005
Figure 6. Prediction of potential distribution area of Metcalfa pruinosa with the random forest (RF) model. (a) Mean receiver operating characteristic (ROC) curve of the random forest (RF) model with 10 replications and (b) predicted potential distribution with different occurrence probability and actual observed sites with green circles.
Figure 6. Prediction of potential distribution area of Metcalfa pruinosa with the random forest (RF) model. (a) Mean receiver operating characteristic (ROC) curve of the random forest (RF) model with 10 replications and (b) predicted potential distribution with different occurrence probability and actual observed sites with green circles.
Forests 10 00583 g006
Figure 7. Relative importance of independent variables in the RF model for projecting presence or absence of Metcalfa pruinosa. Importance was calculated using node impurity in the RF model. Error bars indicate averages and standard errors. Dist. to road: distance to road, Min. temp.: minimum temperature in January, and Max. temp.: maximum temperature in July.
Figure 7. Relative importance of independent variables in the RF model for projecting presence or absence of Metcalfa pruinosa. Importance was calculated using node impurity in the RF model. Error bars indicate averages and standard errors. Dist. to road: distance to road, Min. temp.: minimum temperature in January, and Max. temp.: maximum temperature in July.
Forests 10 00583 g007
Figure 8. Partial dependence of occurrence probability of Metcalfa pruinosa against the changes of each environmental variable in the RF model. (a) Altitude, (b) distance to road, (c) slope, (d) minimum temperature in January, and (e) maximum temperature in July. Solid lines represent averages and gray areas display the 95% confidence interval.
Figure 8. Partial dependence of occurrence probability of Metcalfa pruinosa against the changes of each environmental variable in the RF model. (a) Altitude, (b) distance to road, (c) slope, (d) minimum temperature in January, and (e) maximum temperature in July. Solid lines represent averages and gray areas display the 95% confidence interval.
Forests 10 00583 g008

Share and Cite

MDPI and ACS Style

Lee, D.-S.; Bae, Y.-S.; Byun, B.-K.; Lee, S.; Park, J.K.; Park, Y.-S. Occurrence Prediction of the Citrus Flatid Planthopper (Metcalfa pruinosa (Say, 1830)) in South Korea Using a Random Forest Model. Forests 2019, 10, 583. https://0-doi-org.brum.beds.ac.uk/10.3390/f10070583

AMA Style

Lee D-S, Bae Y-S, Byun B-K, Lee S, Park JK, Park Y-S. Occurrence Prediction of the Citrus Flatid Planthopper (Metcalfa pruinosa (Say, 1830)) in South Korea Using a Random Forest Model. Forests. 2019; 10(7):583. https://0-doi-org.brum.beds.ac.uk/10.3390/f10070583

Chicago/Turabian Style

Lee, Dae-Seong, Yang-Seop Bae, Bong-Kyu Byun, Seunghwan Lee, Jong Kyun Park, and Young-Seuk Park. 2019. "Occurrence Prediction of the Citrus Flatid Planthopper (Metcalfa pruinosa (Say, 1830)) in South Korea Using a Random Forest Model" Forests 10, no. 7: 583. https://0-doi-org.brum.beds.ac.uk/10.3390/f10070583

Note that from the first issue of 2016, this journal uses article numbers instead of page numbers. See further details here.

Article Metrics

Back to TopTop