1、Climate change impacts on the global potentialgeographical distribution of the agricultural invasivepest, Bactrocera dorsalis (Hendel)(Diptera: Tephritidae)Yujia Qin1,2& Cong Wang3& Zihua Zhao1& Xubin Pan3& Zhihong Li1Received: 21 August 2018 /Accepted: 6 May 2019# Springer Nature B.V. 2019AbstractC
2、limatechange is a majorfactordriving shifts in the distribution of invasive pests. The orientalfruit fly, Bactrocera dorsalis, native to mainland Asia, has spread throughout Southeast Asiaand sub-Saharan Africa. Recently, the species has extended its Asian range northward intoregions previously thou
3、ght unsuitable which presents a major new risk to temperate zoneagriculture and has invaded Italy. Thus, it is necessary to study how climate change mayimpact on the global distribution of B. dorsalis. MaxEnt models were used to map suitablehabitat for this species under current and future climate c
4、onditions averaged from four globalclimate models under two representative emission pathways in 2050 and 2070. The resultshighlighted that a total of 30.84% of the worlds land mass is currently climatically suitableincluding parts of the western coast and southeast of the USA, most of Latin America,
5、 parts ofMediterranean coastal European regions, northern and coastal Australia, and the north islandof New Zealand. Under future climate conditions, the risk area of B. dorsalis in the northernhemisphere was projected to expand northward, while in the southern hemisphere, it would besouthward, espe
6、cially by 2070 under RCP85 with very high greenhouse gas emissions. Futuremanagement of this pest should consider the impacts of the global climate change on itspotential geographical distribution.1 IntroductionInvasive species can have far-reaching ecological and economic impacts worldwide (Davis20
7、09;Hulme2009;Macketal.2000). Climate change may directly or indirectly influenceClimatic Changehttps:/doi.org/10.1007/s10584-019-02460-31 and 2 contributed equally to this work.Electronic supplementary material The online version of this article (https:/doi.org/10.1007/s10584-019-02460-3) contains s
8、upplementary material, which is available to authorized users.* Zhihong LExtended author information available on the last page of the articlebiological invasions by altering the likelihood of introduction, probability of establishment,geographical range size, environmental impacts, economic costs,
9、and/or ease of management(Hulme 2017). Agricultural insect pests are a tremendous burden because of the high lossesthey can inflict on crops. Many studies investigating the impact of climate change on pestdistributions show an expansion to higher latitudes of the geographical range of pests (Chenet
10、al. 2017; Hughes 2000;Wangetal.2017).Bactrocera dorsalis (Hendel) (Diptera: Tephritidae), the oriental fruit fly, is a notoriousglobal important invasive pest attacking more than 250 vegetables and fruits (Clarke et al.2005). Bactrocera dorsalis synonyms have also included B. invadens Drew Tsuruta &
11、 White,B. papaya Drew & Hancock, and B. philippinensis Drew & Hancock (FAO 2014;Schutzeet al. 2015). Bactrocera dorsalis is endemic to the Indo-Asian region, it was first recorded inTaiwan (1912) and Hainan (1934) in China (Wang 1996), and it recently has invaded intoCentral China where it was previ
12、ously considered climatically unsuitable to the fly due tooverwintering cold stress (Han et al. 2011; Stephens et al. 2007). This invasion began in theearly 2000s and northward expansion continues (Wang et al. 2009;Yuanetal.2008). OutsideAsia, it was first reported in Hawaii (1945) and there have be
13、en numerous detections inCalifornia in the USA (Papadopoulos et al. 2013), it is established in several South Pacificcountries (Vargas et al. 2015), and it has invaded and been eradicated twice in Australia(Cantrell et al. 2002). In 2003, B. dorsalis first arrived Africa in Kenya; within 15 years, t
14、hisfly has spread across all of the sub-Saharan Africa expect for small parts of South Africa(Manrakhan et al. 2015) and costs about $2 billion USD to the export markets in Africa (Ekesiet al. 2016). Recently this species was found in Italy (Nugnes et al. 2018) and is undereradication (2018).The pot
15、ential geographical distribution of this pest has been confounded by its confusedtaxonomic history as B. dorsalis using CLIMEX (Stephens et al. 2007), B. invadens usingMaxEnt and GARP (De Meyer et al. 2010) or four species separately using ENFA (Hill andTerblanche 2014). De Villiers et al. (2016) bu
16、ilt a new CLIMEX model with year-roundabundance data with a special focus on Africa under “four in one” scenario, where synony-mized species B. invadens, B. papaya, and B. philippinensis have been reclassified asB. dorsalis. However, the global risk map they created did not include those central pro
17、vinces(Jiangsu, Anhui, and Hubei etc.) where the species is already established (Qin et al. 2018).In this study, we used the maximum entropy (MaxEnt) modeling to assess the current(19502000) habitat suitability of B. dorsalis under “four in one” scenario with presence-onlydata and the projected futu
18、re effects from climate change on the species distribution at aglobal scale to provide the basis for future preparedness.2 Materials and methods2.1 Occurrence dataGlobal B. dorsalis occurrence data was initially extracted from the CABI Crop ProtectionCompendium (CABI CPC, https:/www.cabi.org/cpc), t
19、he Global Biodiversity InformationFacility (GBIF, https:/www.gbif.org/), and was combined with the collection data provided inQin et al. (2018). Detection and eradication data were not included; the distribution recordcouldbeseeninZengetal.(2019). To minimize sampling bias, data were assigned to 9 k
20、m9 km climate data grids. Seven hundred fifty-seven GBIF + CABI + collection records for B.Climatic Changedorsalis were finally selected for the modeling (Fig. 1). The conversion and editing of the datawere conducted in ArcGIS 10.2 (ESRI Inc., Redlands, CA, USA). URL http:/ Climate dataClimate data
21、were downloaded from the WorldClim website (http:/worldclim.org/)(Hijmanset al. 2005). Current climate conditions were average monthly climate data for minimum,mean, and maximum temperature and precipitation for 19502000 (Hijmans et al. 2005).Future climate conditions were downscaled global climate
22、model (GCM) data from CMIP5(IPPC Fifth Assessment) (IPCC 2014). Four GCMs including HadGEM2-ES (HE), IPSL-CM5A-LR (IP), MIROC-ESM-CHEM (MI), and NorESM1-M (NO) estimated for 2050(average for 20412060) and 2070 (average for 20612080) were selected to offer a widerange of temperature and rainfall chan
23、ges. These models were selected to give a wide range ofrainfall and temperature changes, rather than to represent the likelihood of future climatechange (Warszawski et al. 2014). The models were each run under two representativeconcentration pathways (RCPs) which are two greenhouse gas concentration
24、 estimates, astringent mitigation scenario RCP26 and a scenario with very high greenhouse gas emissionsRCP85 (IPCC 2014).Nineteen bioclimatic variables wereobtained ata spatialresolutionat 5 arc-min(9kmattheequator). Because some of these variables are highly correlated, principal component analysis
25、(PCA) and correlation analysis were carried out in IBM SPSS Statistics version 22(https:/ variables with Pearson correlation coefficients having absolute values 0.8 thatare useful and eco-physiologically relevant for modeling.2.3 MaxEnt modelingThe potential geographical distribution of B. dorsalis
26、under current and projected to differentfuture climate scenarios was conducted in MaxEnt (v3.3.3k, http:/biodiversityinformatics.amnh.org/open_source/maxent/) with presence-only data (Phillips et al. 2006). MaxEnt isFig. 1 Global distribution data from collection efforts (Qin et al. 2018)(green poin
27、ts), CABI (pink points), andGBIF (blue points) used to build and evaluate the MaxEnt models of B. dorsalisClimatic Changewidely used to assess habitat suitability for species under current and future climate scenarios(Elith et al. 2006; Merow et al. 2013; Sultana et al. 2017;Wangetal.2017; Wei et al
28、. 2018).Moreover, some studies found that this software performed well regardless of the number orgeographical extent of species records, compared with the performance of MahalanobisTypicalities, random forests methods, and other methods (Bosso et al. 2016; Hernandezet al. 2006; Elith et al. 2011).
29、In this study, models were calibrated using 25% random testpercentage, 5000 maximum iterations, 10 percentile training presence threshold rule, 15replicates under subsample run type and default features followed Young et al. (2011)andWang et al. (2017). Habitat suitability was depicted at four level
30、s: negligible risk, low risk,medium risk, and high risk according to the MaxEnt plots. Results were converted into rasterfiles and risk areas were calculated for each continent including Asia, Africa, North America(Canada and USA), Latin America, Europe, and Oceania in ArcGIS 10.2. Jackknife testing
31、was used to measure variable importance, and receiver operating characteristic (ROC) curvesaveraged over the replicated runs were drawn to evaluate the model performance by the areaunder ROC curve (AUC) values. AUC ranges between 0 and1,models with AUC value of 0.5represents a model with discriminat
32、ion ability no better than random, while a model with AUCvalue higher than 0.75 represents a fair ability to discriminate (Swets 1988; Pearce and Ferrier2000).3Results3.1 Bioclimatic variables selectionPrincipal component analysis and correlation analysis of 19 bioclimatic variables were con-ducted
33、for variable selection. The first five principal components explained 92.072% of thetotal variance. The first and second components mainly attributed to temperature (bio 1, bio3,bio3, bio5, bio6, bio7, bio8, bio10, bio11), the third and fourth attributed to precipitation(bio13, bio14, bio16, bio17),
34、 and the fifth to mean diurnal range which is bio2; finally, sixuncorrelated bioclimatic variables were selected from IBM SPSS Statistics for the analysis ofB. dorsalis (Table 1).3.2 Potential geographical distribution under current climate conditionsThe current distribution map shown in Fig. 2 is b
35、ased on the distribution databases andcollection efforts, and climate suitability is shown as different colors on the map. Wecategorized habitat suitability according to the MaxEnt plots of this species (Fig. S1) into fourlevels: negligible risk (0.000.08), low risk (0.080.23), medium risk (0.230.46
36、), and highrisk (0.461.00). The prediction from MaxEnt under current climate conditions indicated thatB. dorsalis could potentially establish throughout much of the tropics and subtropics. Parts ofthe western coast and southeast of the USA, most of Latin America, parts of the Mediterraneancoastal re
37、gions of Europe, northern and coastal Australia, and the north island of New Zealandwere also suitable for the species (Fig. 2). The extent of the land area that is climaticallysuitable for B. dorsalis under current and future climate conditions was quantified for eachcontinent (Table 2). A total of
38、 30.84% of the worlds land mass (excluding Antarctica), or13,271.98104km2, is currently climatically suitable. The results of the analysis indicatedthat, expressed as a percentage of total land area, more than 70% of Latin America, and greaterClimatic Changethan 20% of Asia, Africa, and Oceania, was
39、 projected to be currently climatically suitable.Whereas the potential suitable areas in North America (only including Canada and USA) andEurope were lower (2.62% and 6.93%, respectively) (Table 2).3.3 Climate change impact on the potential geographical distributionFour distribution maps of suitable
40、 habitat for B. dorsalis under a range of possible futureclimate scenarios for 2050 and 2070 are summarized in Fig. 3 which showed mean predictedresults of all four global climate models under RCP26 and RCP85, the lowest and highest ofTable 1 Principal component analysis (PCA) performed on 19 biocli
41、matic variables for B. dorsalisBioclimate variables Principal Components12345Annual Mean Temperature (bio1) 0.533 0.836 0.078 0.016 0.034Mean Diurnal Range (bio2)a0.034 0.024 0.301 0.258 0.841Isothermality (bio3) 0.884 0.206 0.242 0.045 0.209Temperature Seasonality (bio4)a0.968 0.125 0.072 0.049 0.0
42、59Max Temperature of Warmest Month (bio5) 0.034 0.883 0.224 0.105 0.377Min Temperature of Coldest Month (bio6)a0.852 0.463 0.116 0.087 0.175Temperature Annual Range (bio7) 0.839 0.146 0.260 0.153 0.419Mean Temperature of Wettest Quarter (bio8) 0.029 0.934 0.057 0.050 0.220Mean Temperature of Driest
43、Quarter (bio9) 0.769 0.593 0.008 0.042 0.098Mean Temperature of Warmest Quarter (bio10)a0.024 0.985 0.125 0.019 0.032Mean Temperature of Coldest Quarter (bio11) 0.831 0.533 0.019 0.034 0.067Annual Precipitation (bio12) 0.128 0.103 0.535 0.777 0.216Precipitation of Wettest Month (bio13) 0.030 0.002 0
44、.036 0.964 0.135Precipitation of Driest Month (bio14)a0.083 0.051 0.940 0.171 0.080Precipitation Seasonality (bio15) 0.193 0.267 0.783 0.093 0.278Precipitation of Wettest Quarter (bio16)a0.001 0.006 0.045 0.980 0.135Precipitation of Driest Quarter (bio17) 0.108 0.067 0.937 0.190 0.100Precipitation o
45、f Warmest Quarter (bio18) 0.349 0.222 0.329 0.454 0.475Precipitation of Coldest Quarter (bio19) 0.356 0.005 0.374 0.548 0.281aSix uncorrelated variables used in the analysis, values in italics were above 0.8 explaining more varianceFig. 2 Current potential geographical distribution of B. dorsalis ba
46、sed on MaxEnt (v3.3.3k) (averagedfrom19502000). White areas indicated negligible-risk areas, yellow areas indicate low-risk areas, orange areasindicate medium-risk areas, and red areas indicate high-risk areasClimatic ChangeTable2ProjectedriskareagloballyforB.dorsalisunderthecurrentandfutureclimates
47、cenarios,expressedasanarea(104km2)andasapercentageofthetotalareapercontinentCurrent(19502000)2050-RCP262050-RCP852070-RCP262070-RCP85Riskarea/104km2%totalareaRiskarea/104km2%totalareaRiskarea/104km2%totalareaRiskarea/104km2%totalareaRiskarea/104km2%totalareaAsia907.3920.35972.6421.82995.2622.33978.8
48、221.961035.8323.24Africa1408.5846.641546.6751.211581.0252.351572.0752.061639.5154.29NorthAmericaa51.312.62110.785.65144.727.38109.835.60202.9710.35LatinAmerica1407.7673.331437.0274.861441.7175.101439.5174.991461.3176.12Europe70.396.93104.3210.27127.5212.55114.5611.28193.9719.09Oceania247.5127.59285.
49、1331.79292.4132.60280.8031.30308.7634.42Worldb4092.9430.844456.5733.584582.6434.534495.5833.874842.3436.49aOnlyincludingCanadaandUSA.bTheareagivenfortheworldexcludestheAntarcticaClimatic Changefour greenhouse gas concentration pathways. Bactrocera dorsalis was predicted to have agreatly extended ran
50、ge on all continents especially by 2070 under RCP85 (Table 2, Fig. S2). Itshowed a northward trend in Asia, Northern Africa, North America, and Europe and asouthward trend in South America, Southern Africa, Australia, and New Zealand (Fig. 3d).The total suitable land mass was increased by 363.63 104
