Introduction
Bemisia tabaci (Hemiptera: Aleyrodidae) is best described as being a cryptic species complex composed of numerous morphologically indistinguishable species (De Barro et al., Reference De Barro, Liu, Boykin and Dinsdale2011). One member of the complex, Middle East-Asia Minor 1 (silverleaf whitefly; formerly known as biotype B, hereon B. tabaci MEAM1) has become particularly invasive. The IUCN/SSC Invasive Species Specialist Group (http://www.issg.org) and the Global Invasive Species Database (http://www.issg.org/database/welcome/) list it as a top 100 invader. B. tabaci MEAM1 is believed to have originated in the Middle-East/Asia Minor region and has since spread to all continents except to Antarctica through trade in ornamentals (Cheek and Macdonald, Reference Cheek and Macdonald1994; Dalton, Reference Dalton2006). To date, B. tabaci MEAM1 has been recorded in damaging numbers from Australia and the following Pacific Island countries: American Samoa, Cook Islands, Fiji, French Polynesia, Guam, Marshall Islands, New Caledonia, New Zealand, Niue, Northern Mariana Islands and Tonga. In Africa, it has been recorded from Egypt, Iran, Israel, Jordan, Mauritius, Reunion, South Africa, Tunisia, Turkey and Yemen, and in Asia from China, India, Japan, South Korea, Philippines and Taiwan. In the Americas, apart from the USA, it occurs in Antigua & Barbuda Argentina, Barbados, Belize, Brazil, Colombia, Costa Rica, Cuba, Dominican Republic, Ecuador, El Salvador, Grenada, Guadeloupe, Guatemala, Honduras, Martinique, Mexico, Nicaragua, Panama, Puerto Rico, St Kitts, Trinidad & Tobago and Venezuela. In Europe, there have been records in field crops from Cyprus, Greece, Italy, Portugal and Spain and in glasshouses in Austria, Canada, Denmark, France, Germany, the Netherlands, Norway, Poland and the United Kingdom (De Barro, Reference De Barro1995; De Barro et al., Reference De Barro, Liebregts and Carver1998, Reference De Barro, Liu, Boykin and Dinsdale2011; Smith, Reference Smith1999). Damage to agricultural, horticultural and ornamental crops occurs through feeding and the transmission of plant viruses, in particular the begomovirus group of Geminiviridae. The rise in the importance of these viruses as crop pathogens has mirrored the spread of the invasive members of the B. tabaci complex as its broadly polyphagous nature has enabled it to recruit begomoviruses from non-crop hosts and move them into crop species (Jones, Reference Jones2003).
The control of B. tabaci MEAM1 in crops has been hampered by the development of pesticide resistance: amongst the whiteflies, it is perhaps excelled only by B. tabaci MED (formerly known as B. tabaci biotypes Q, J and L), which develops resistance to pyriproxyfen and neonicotinoid insecticides more readily (Cahill et al., Reference Cahill, Byrne, Gorman, Denholm and Devonshire1995; Ma et al., Reference Ma, Gorman, Devine, Luo and Denholm2007; Luo et al., Reference Luo, Jones, Devine, Zhang, Denholm and Gorman2010). In Australia, B. tabaci MEAM1 is being controlled successfully using integrated pest management, including the presence of Eretmocerus hayati (De Barro, Reference De Barro2012).
Species distribution modelling methods employed by Herrera Campo et al. (Reference Herrera Campo, Hyman and Bellotti2011) resulted in some very poor projected distributions. Herrera Campo et al. (Reference Herrera Campo, Hyman and Bellotti2011) modelled the potential distribution of B. tabaci sensu lato based solely on distribution data, using an ensemble method training five models (climate space, environmental distance, GARP, MaxEnt and support vector machines) with four sets of covariates. The results were highly erratic, ranging from excessively conservative to excessively liberal, and often heavily biased. Even the models that were very liberal had poor sensitivity. The environmental distance trained with all covariates had the highest sensitivity score of 0.941, meaning that it was unable to account for 6% of records, and yet indicated (implausibly) that Central Australia, the Sahara Desert and the entire wet tropics were all modelled as highly suitable. Furthermore, six of the 16 model results indicated that Australia was completely unsuitable for B. tabaci, whilst the remaining ten models indicated suitability in the central desert regions of Australia where cropping could only be undertaken where cropping does not currently occur, but could be undertaken if irrigation was possible. The results of an elaborate weighting scheme in Herrera Campo et al. (Reference Herrera Campo, Hyman and Bellotti2011) indicate climate suitability in areas that are too dry (e.g., Sahara Desert, Namibian Desert and Central Australia) and excessively cold (Southern Argentina and mid-west USA). Clearly, ensemble methods do not, as is often claimed (e.g., Araujo and New Reference Araujo and New2006), necessarily correct for poor performing models (Hannemann et al., Reference Hannemann, Willis and Macias-Fauria2015).
Whilst the present known distribution of B. tabaci sensu lato is broad, the known distribution of B. tabaci MEAM1 is more confined. For example, in the USA, it has a northern limit at Tulare in the San Joaquin Valley in California, and greater numbers at Bakersfield at the south end of the valley (Gruenhagen et al., Reference Gruenhagen, Perring, Bezark, Daoud and Leigh1993). It is abundant at Brawley in the Imperial Valley (Zalom et al., Reference Zalom, Natwick and Toscano1985), and also occurs in smaller numbers at Palm Springs and Phoenix (Arizona), but is absent from New Mexico and Los Angeles (California). In the Texas, it is abundant around Brownsville and the Lower Rio Grande Valley and occurs as far north as Del Rio, but cannot survive in San Antonio or along the northern rim of the Gulf of Mexico. It occurs in southern Florida, especially on crops around Immokales near Fort Myers; further north it is absent from areas north of Orlando, except for a pocket at Charleston (South Carolina) (McKenzie et al., Reference McKenzie, Anderson and Villarreal2004).
In this paper, we develop a climate suitability model to estimate the potential distribution of B. tabaci MEAM1 (De Barro et al., Reference De Barro, Liu, Boykin and Dinsdale2011). We compare the fitted model to the known distribution of another invasive member of the complex, B. tabaci MED, which has been noted invading Canada, China, Guatemala, Japan, Mexico, the Netherlands, New Zealand, South Korea, Uruguay and the United States (McKenzie et al., Reference McKenzie, Hodges, Osborne, Byrne and Shatters2009; Qiu et al., Reference Qiu, Chen, Liu, Peng, Li, Ahmed, Mathur, Du and Ren2009).
Methods
Known distribution and pest status
Global distribution data for B. tabaci MEAM1 and MED were carefully collated from the literature and databases (De Barro, Reference De Barro1995; Van Giessen et al., Reference Van Giessen, Mollema and Elsey1995; Tsai and Wang, Reference Tsai and Wang1996; Liu and Stansly, Reference Liu and Stansly1998; Muniz and Nombela, Reference Muniz and Nombela2001; Jones et al., Reference Jones, Markham, Anderson and Morales2005; Musa and Ren, Reference Musa and Ren2005), IUCN/SSC Invasive Species Specialist Group (http://www.issg.org), Global Invasive Species Database (http://www.issg.org/database/welcome/), samples provided by numerous contributors and where necessary, were geocoded using atlas resources and Google Earth.
The Australian distribution of B. tabaci was determined by a series of surveys across Australia between July 1996 and June 1997. Surveys involved searching available known hosts in a particular location for a maximum of 20 min. The pest status of B. tabaci was determined on a location-specific basis through grower surveys and consultation with key grower representatives. B. tabaci MEAM1 was considered to hold pest status in a particular location if growers were taking management decisions to minimize numbers. Identifications were carried out using random-amplified polymorphic DNA-polymerase chain reaction (De Barro and Driver, Reference De Barro and Driver1997).
Development rate studies
The development and fecundity of B. tabaci MEAM1 and two populations of Australia (B. tabaci EAN and WAN, formerly B. tabaci biotype AN) were tested experimentally. The B. tabaci MEAM1 culture was established from individuals collected from tomato in Bundaberg, Queensland; B. tabaci EAN was established from collections made on Euphorbia cyathophora in Bundaberg, Queensland and WAN from Emilia sonchifolia in Kununurra, Western Australia. Single-mated females of the B. tabaci MEAM1, Eastern and Western accessions of the Australia biotypes were allowed to oviposit for 24 h onto E. cyathophora seedlings kept in cylinder cages at 15, 20, 25 or 30°C. Each treatment was replicated ten times per temperature. After 24 h, the insects were removed. The time taken for development from newly laid egg to emerging adult was noted.
Modelling
The CLIMEX modelling package (Sutherst and Maywald, Reference Sutherst and Maywald1985; Kriticos et al., Reference Kriticos, Maywald, Yonow, Zurcher, Herrmann and Sutherst2015) was used to fit an ecological niche model to the known distribution of B. tabaci records identified as B. tabaci MEAM1 (De Barro et al., Reference De Barro, Liu, Boykin and Dinsdale2011). CLIMEX is a climate-response model that produces indices derived from the responses of a nominated species to each of the components of climate. Rainfall and relative humidity or evaporation are combined into a single soil moisture index (MI), using a hydrological model to represent the availability of moisture. The effects of extreme climatic values are taken into account in a series of stress indices that estimate the threat to B. tabaci posed by prolonged adverse periods of excessively cold, hot, dry or wet conditions. Finally, the growth and stress indices are combined into an ecoclimatic index, scaled from 0 to 100, to represent the overall favourability of the given geographical location for the permanent survival and propagation of populations of B. tabaci. The annual minimum heat sum for development (in degree-days) is used to estimate the expected number of generations in a year.
B. tabaci MEAM1's climatic requirements were inferred from information on its known global geographical distribution, relative abundance and seasonal phenology. Some life cycle data, such as developmental threshold temperatures, were used to fine-tune or interpret the CLIMEX parameter values and to provide more confidence in these values (table 1). However, published reports on development rates and population dynamics of B. tabaci MEAM1 provide a wide range of parameter values for lifecycle processes, largely dependent on the host plant (Zalom et al., Reference Zalom, Natwick and Toscano1985; Zalom and Natwick, Reference Zalom and Natwick1987; Van Giessen et al., Reference Van Giessen, Mollema and Elsey1995; Tsai and Wang, Reference Tsai and Wang1996; Liu and Stansly, Reference Liu and Stansly1998; Muniz and Nombela, Reference Muniz and Nombela2001; Lin and Ren, Reference Lin and Ren2005; Musa and Ren, Reference Musa and Ren2005).
CLIMEX parameter values for B. tabaci Middle East/Asia Minor 1
CLIMEX parameters were fitted using three assumptions about the conditions under which B. tabaci can exist:
(i) rainfed field environment,
(ii) irrigated crop environment (top-up irrigation to produce 25 mm week−1 in summer, 15 mm week−1 in winter) and
(iii) glasshouses enabling unrestricted over-wintering (cold stress removed).
Published data on geographical distribution and abundance do not clearly define the role of glasshouses in maintaining populations reported in some colder regions. This limited the degree of confidence that we could attribute to estimates of over-wintering ability as determined by low temperatures. When estimating geographical variation in the potential numbers of generations, we assumed that B. tabaci MEAM1 had either over-wintered successfully in very cold locations, perhaps in a greenhouse, or was re-introduced in the spring each year. We made no correction for the modifying effects of crop canopies on microclimates because the target area included both crop- and non-crop host plants, with widely varying canopy densities.
Model fitting, verification and validation
The CLIMEX growth parameter values were adjusted to describe seasonal growth and over-wintering of B. tabaci at Brawley, California, where data on B. tabaci show high population levels under irrigation (Coudriet et al., Reference Coudriet, Prabhaker, Kishaba and Meyerdirk1985), at Fresno, northern California, where B. tabaci MEAM1 has been reported in smaller numbers (Gruenhagen et al., Reference Gruenhagen, Perring, Bezark, Daoud and Leigh1993) and at Katherine, Australia where it is only marginally suitable (De Barro, unpublished data). The model was fitted using distribution data for the Middle East, Asia Minor and the USA, and verified using data from Australia and Eastern Asia. Distribution data in Africa, Central America, the Caribbean, Europe and South America were used to validate the model. The model-fitting procedure involves an implicit cross-validation between distribution data, parameters derived from observed phenologies (e.g., Zalom et al., Reference Zalom, Natwick and Toscano1985; Zalom and Natwick, Reference Zalom and Natwick1987) and parameters estimated from laboratory experiments (e.g., Butler et al., Reference Butler, Henneberry and Clayton1983).
Soil moisture index
The lower soil moisture limit (SM0) was set at 0.1, which approximates the permanent wilting point of many crop plants (Kriticos et al., Reference Kriticos, Sutherst, Brown, Adkins and Maywald2003); reasoning that if the soil moisture is insufficient for host plants to grow, then B. tabaci MEAM1 is also unlikely to be able to grow. The optimal range for growth was set from 0.5 to 1.5 to cover the range of soil moisture levels at which most crops are able to grow optimally. The upper soil moisture limit, SM3, was set to 2, accounting for the ability of B. tabaci MEAM1 and its hosts to grow to a limited capacity under fairly wet conditions such as those found in Karnataka in south-western India.
Temperature index
The base temperature for development (DV0) was set to 12°C, which compares favourably with empirical findings from Zalom and Natwick (Reference Zalom and Natwick1987) fitted to population peaks of a base temperature of 10°C. The lower temperature limit for optimal growth (DV1) was set to 28°C and the upper optimum (DV2) was set to 32°C in accordance with results from growth rate studies (Zalom et al., Reference Zalom, Natwick and Toscano1985; Zalom and Natwick, Reference Zalom and Natwick1987; Van Giessen et al., Reference Van Giessen, Mollema and Elsey1995; Tsai and Wang, Reference Tsai and Wang1996; Liu and Stansly, Reference Liu and Stansly1998; Muniz and Nombela, Reference Muniz and Nombela2001; Musa and Ren, Reference Musa and Ren2005). Butler et al. (Reference Butler, Henneberry and Clayton1983) noted that eggs failed to hatch at a constant 36.0°C, and that aestivation occurred with fluctuating temperatures between 27 and 43°C. Accordingly, the upper limit for growth (DV3) was set to 42°C.
The development rates for B. tabaci MEAM1 and the two Australia accessions showed no significant differences in the slopes of the fitted temperature development response functions, and differed only marginally in the development rates for a given temperature (fig. 1; De Barro and Hart, Reference De Barro and Hart2000). These results suggest that the optimal temperature for development is likely to be above 30°C (fig. 1), lending further support to the choice of parameters for the optimum temperature range.
Observed development rates as a function of temperature for three species within the Bemisia tabaci complex: Middle East Asia Minor 1 (B. tabaci MEAM1), Asia II 3 and Australia (B. tabaci EAN and WAN). Taxonomy follows De Barro et al. (Reference De Barro, Liu, Boykin and Dinsdale2011).
Cold stress
Two cold stress mechanisms limit the distribution of B. tabaci MEAM1 to locations where it was thought to be able to overwinter outside of glasshouses. The threshold temperature model simulates frost sensitivity: when minimum temperatures fall below TTCS (0°C), stress accumulates moderately quickly (0.01 week−1). The degree-day cold stress mechanism provides a limitation under cool conditions where daily maximum temperatures are insufficient for host photosynthesis and B. tabaci MEAM1 feeding to offset metabolic base rate respiration losses. In the fitted model, if B. tabaci MEAM1 does not receive a minimum of 15°C days above DV0 (12°C) each week, stress accumulates slowly (−0.00019 week−1). The overwintering limits were explored using Google Earth, to identify the presence of glasshouses in cropping regions in cold areas (e.g., the high elevation cropping area in western Azerbaijan, Iran, and the Shandong and Beijing provinces of northeastern China). In northwestern China (Burqin and the Emin Valley), cropping is seasonal and depends on irrigation. There is little evidence of glasshouses in this region, and it may be that reports of B. tabaci MEAM1 presence in these cold, dry areas are from seasonal migrations or commodity contamination from regions with more favourable climates. Tuscon, Arizona, is modelled as being barely able to support a year-round population of B. tabaci MEAM1. This is consistent with Dennehy et al. (Reference Dennehy, Degain, Harpold, Zaborac, Morin, Fabrick, Nichols, Brown, Byrne and Li2010) noting difficulty in collecting B. tabaci (Biotype Q, MED) regularly in this region, though intense management of pests using pesticides could have contributed to low population numbers.
Dry stress
The dry stress threshold (SMDS) was set at 0.1, reasoning that if the host plants were unable to grow due to drought-stress, then B. tabaci MEAM1 populations would likely decline also. The dry stress accumulation rate was fitted to barely enable persistence in the locations in the Middle East where it has been reported (Israel, Jordan and western Syria).
B. tabaci MEAM1 has been reported in extremely xeric locations such as Karameh in southern Iran and the cropping areas in El Centro, Southern California and Maricopa, Arizona in the USA. These locations depend on irrigation for agriculture, and hence the fitted dry stress under the natural rainfall scenario indicates that these locations are unsuitable for year-round persistence.
Wet stress
The wet stress threshold SMWS (2.0) and stress accumulation rate HWS (0.002 week−1) were fitted to preclude B. tabaci MEAM1 from Goa in western India, while still allowing it to persist in Karnataka where it has been reported frequently.
Hot-wet stress
An equipment failure provided an opportunity to observe that under laboratory conditions, B. tabaci MEAM1 populations thrived under hot dry conditions, but declined at similar temperatures with high humidity (De Barro, pers. obs.). The hot-wet stress parameters were adjusted to make Katherine in the Northern Territory of Australia barely suitable.
Generation time
In the literature, the criteria used to define the generation time have not been standardized, making comparison of different datasets difficult (Zalom et al., Reference Zalom, Natwick and Toscano1985; Zalom and Natwick, Reference Zalom and Natwick1987; Van Giessen et al., Reference Van Giessen, Mollema and Elsey1995; Tsai and Wang, Reference Tsai and Wang1996; Liu and Stansly, Reference Liu and Stansly1998; Muniz and Nombela, Reference Muniz and Nombela2001; Musa and Ren, Reference Musa and Ren2005). Further, as generations quickly overlap each season, and there is significant variation in timing of emergence of the first and last individuals from each generation, estimates of the number of generations is necessarily an approximation. To account for this variation, we investigated the sensitivity of CLIMEX projections to changes in the thermal sum (number of degree-days, °C days) required to complete a generation. We examined the developmental threshold temperature (10, 12 and 14°C) and the thermal sum (330, 365 and 400°C days) required to complete one generation. The generation time was taken as the sum of the pre-oviposition period and the development times of each lifecycle stage, to produce 50% emergence of adults. The fitted value for PDD of 365°C days above DV0 (12°C) compares favourably with the value of 369.5°C days above 10°C noted by Zalom and Natwick (Reference Zalom and Natwick1987).
Simulations
The model was run with the CliMond database of long-term average climate data on a 30′ grid (Kriticos et al., Reference Kriticos, Webber, Leriche, Ota, Bathols, Macadam and Scott2012). It consisted of a 30-year average of monthly minimum and maximum temperatures, relative humidity at 09:00 and 15:00 h and total precipitation. This climatology (CM30_1995H_V2) was centred on 1995, spanning the period 1980–2010. CLIMEX interpolates the monthly average temperatures to weekly values in order to describe the daily temperature cycle for each week. The initial modelling was undertaken without irrigation, and locations records that could not be fitted with reasonable dry stress and MI values were scrutinized using Google Earth to identify telltale signs of irrigation (e.g., green circles or rectangles amongst a background of brown or yellow exposed soil and the nearby presence of a river or dam). Simulations were then undertaken with 2.1 and 3.6 mm day−1 top-up irrigation in winter and summer, respectively. The Global Map of Irrigated Areas Version 5 (Siebert et al., Reference Siebert, Doll, Hoogeveen, Faures, Frenken and Feick2005; Portmann et al., Reference Portmann, Siebert and Doll2010) was used to define areas where the irrigation and natural rainfall scenarios were included in a final composite risk map. The GMIA dataset included some dubious artefacts in desert regions, including Central Western Australia, Algeria and Libya. In these regions, the dataset indicated that irrigation was being practiced over small, proportions of the landscape. Searching these areas using Google Earth did not reveal any signs of irrigation. Consequently, we set a threshold of 5 ha per 10′ grid cell in GMIA V5 to indicate that irrigation was being practiced. This is equivalent to approximately 0.05%.
Results
Known distribution and pest status
As befits the reputation of the B biotype as an invasive species, the known distribution of B. tabaci MEAM1 is extensive, being present on every continent except on Antarctica (fig. 2). The model results for the validation data in Africa, Central America, the Caribbean, Europe and South America accord perfectly. The known distribution of B. tabaci MEAM1 overlaps that of MED in places in the Mediterranean, North America, Central America and northern China (fig. 2). In China, both species are found far to the north of their modelled potential distribution, as far north as Beijing. We attribute this to B. tabaci MED and MEAM1 probably overwintering in greenhouses, which are abundant in this region and easily recognized using Google Earth. In the absence of cold stress, the annual growth index (GIA) in this region is adequate for strong population development, which accords with observations that it is a major pest in this region (Ren et al., Reference Ren, Wang, Qiu and Xiao2001). B. tabaci MEAM1's pest status in Australia seems confined to the sub-tropics, with populations further south only appearing to warrant pest control in greenhouse situations.
Known distribution of Bemisia tabaci, Middle East Asia Minor 1 and Mediterranean genotypes sensu De Barro et al. (Reference De Barro, Liu, Boykin and Dinsdale2011): (a) global and (b) Australia indicating the locations where it is known to be present, known to be a pest and not known to occur.
The known Australian geographical distribution is shown in fig. 2, as are the areas where it is known not to occur. In Western Australia, B. tabaci MEAM1 occurs at Carnarvon, but has yet to establish in Kununurra. In the Northern Territory, it has been recorded as far north as Darwin, and while it is present in the Darwin rural area, it has failed to persist further inland in Katherine, despite having initially established there. In Queensland, it is found as far north as Cooktown and extends southwards in a peri-coastal band. In the south-east of the State it extends further inland, spanning the Darling Downs as far west as St George and north to Roma, Emerald, Richmond and Dimbulah. On the east coast of New South Wales, B. tabaci extends as far south as Bateman's Bay and inland to Bourke, the Sydney Basin, Namoi (Narrabri) and Gwyder (Moree) Valleys, Narromine, Tamworth and Warren. In New South Wales, numbers tend to be low except in the Tweed Valley region around Cudgen and the Sydney Basin. In the locations south of Cudgen, infestations tend to be in urban areas in close proximity to dwellings; however, in the Sydney Basin, field infestations are in close proximity to glasshouse infestations. Thus, in areas south of Cudgen, including the Sydney Basin, the protected environments afforded by urbanization and glasshouses almost certainly confound local climatic suitability. B. tabaci MEAM1 has also been recorded from glasshouses in South Australia, New South Wales, Victoria and Western Australia, usually on plant material imported from Queensland and New South Wales. These data indicate that the species is well suited to and distributed in the tropics, but is less well adapted to the cooler non-tropical regions and its presence may well be supported by the provision of over-wintering sites in protected cropping and urban settings.
Phenology
In CLIMEX, the stress indices have the strongest impact on the species modelled range, whereas the growth indices have a stronger effect on the modelled suitability patterns within the range boundaries, and also drive directly the modelled phenological patterns (Kriticos et al., Reference Kriticos, Maywald, Yonow, Zurcher, Herrmann and Sutherst2015). The modelled growth indices are presented in fig. 3 for three locations where the phenology is known. Naranjo et al. (Reference Naranjo, Castle, De Barro, Liu, Stansly and Naranjo2010) observes that the impacts of B. tabaci in the San Joaquin Valley (near to Fresno) are muted compared with the impacts observed in the Imperial Valley (Brawley), attributing this to the warmer climate in the southern site, especially during the early part of the season. The modelled temperature and growth indices accord with these observations (fig. 3). At Katherine, in the Northern Territory, the apparent strong potential for growth (fig. 3) is offset by hot-wet stress. Here, it has been observed to establish during the cooler, drier months and undergo local extinction during the warmer, wetter months (Naranjo et al., Reference Naranjo, Castle, De Barro, Liu, Stansly and Naranjo2010).
Seasonal CLIMEX growth (green), temperature (red) and moisture (blue) indices for Bemisia tabaci at Brawley and Fresno, California and Katherine, Northern Territory under an irrigated rainfall scenario (details are given in the main text). Monthly rainfall (bars) and minimum and maximum temperatures are shown in the top panel.
Potential distribution and pest status
The modelled global potential distribution of B. tabaci MEAM1 reflects the belief that it is a species adapted to the dry tropics and sub-tropics. Its modelled climate suitability under natural rainfall conditions throughout much of its putative native range in the Middle East is marginal or poor (fig. 4a), but increases substantially in those areas where irrigation is practiced (fig. 4b). The composite irrigation scenario provides a pest risk map that highlights those areas that may be at risk if suitable hosts are grown (fig. 4c). A difference map between the natural rainfall and the composite irrigation scenarios is presented in fig. 4d. The contrast in suitability under irrigation highlights B. tabaci MEAM1's adaptation to hosts that are ruderals sensu Grime (Reference Grime and Grime1979), and its ability to take advantage of habitats that become suitable for only short periods, by having a short lifecycle and high per capita rate of natural increase and a durable resting stage in order to persist. The modelled potential distribution encompasses the known global distribution, save those locations where its presence is maintained through glasshouses (e.g., northern China, South Korea and northern Iraq). In these areas, the CLIMEX ecoclimatic index is zero, indicating that conditions are unsuitable for year-round persistence due to excessive cold stress during winter (Supplementary Information, fig. S1). However, the GIA is suitable for supporting significant population growth during the warmer months when field crops are grown (Supplementary Information, fig. S2). The ecoclimatic index suggests that substantial areas in the Americas, Africa and eastern Asia are likely to be at risk of further invasion by B. tabaci MEAM1.
Global climate suitability and known distribution of Bemisia tabaci, Middle East Asia Minor 1: (a) natural rainfall, (b) irrigation, (c) composite natural rainfall and irrigation scenario combined using the global map of irrigated areas (Siebert et al., Reference Siebert, Doll, Hoogeveen, Faures, Frenken and Feick2005; Portmann et al., Reference Portmann, Siebert and Doll2010) and (d) comparison between natural rainfall and irrigation scenario. The irrigation scenario is described in the main text.
Discussion
The known distribution of B. tabaci MEAM1 is extensive, confirming its reputation as an extremely important invasive species. Moreover, its modelled potential distribution is even greater, especially in South America, Africa and Asia (fig. 3, bottom panel). The CLIMEX model performed very well, giving perfect model sensitivity in relation to all known qualified distribution records, and no excessive model prevalence. The areas modelled as climatically suitable, but without known distribution records all appear to be biologically plausible.
The modelled potential distribution of B. tabaci MEAM1 completely overlaps that of MED and the point distribution records for B. tabaci s.l., as well as the known country-level records globally. Further, the observed development rates of B. tabaci MEAM1 and Australia (AN) differ only slightly (De Barro and Hart, Reference De Barro and Hart2000). Such overlaps and similarities reflect the difficulties in distinguishing the different taxonomic entities within the B. tabaci complex (Muniz and Nombela, Reference Muniz and Nombela2001; Qiu et al., Reference Qiu, Chen, Liu, Peng, Li, Ahmed, Mathur, Du and Ren2009; Boykin, Reference Boykin2014). It is possible that the niche differentiation between the various species within the complex are less distinguished by climatic niche, and more by host preference, virus transmission patterns, etc. (Iida et al., Reference Iida, Kitamura and Honda2009; De Barro et al., Reference De Barro, Liu, Boykin and Dinsdale2011; Tajebe et al., Reference Tajebe, Boni, Guastella, Cavalieri, Lund, Rugumamu, Rapisarda and Legg2014). The fundamental climatic niche of the various species in the B. tabaci complex may, in fact, be extremely similar. Inter-specific competition between these morphologically indistinguishable conspecifics has been noted previously, with invaders displacing resident species (McKenzie et al., Reference McKenzie, Anderson and Villarreal2004; Liu et al., Reference Liu, De Barro, Xu, Luan, Zang, Ruan and Wan2007; McKenzie et al., Reference McKenzie, Hodges, Osborne, Byrne and Shatters2009; Legg et al., Reference Legg, Sseruwagi, Boniface, Okao-Okuja, Shirima, Bigirimana, Gashaka, Herrmann, Jeremiah and Obiero2014). A detailed understanding of the relative competitive ability of each of the species in the complex under various environmental conditions is likely to remain elusive. Therefore, it is perhaps prudent to consider all of those areas identified as climatically suitable in the CLIMEX model as being at risk of invasion by B. tabaci MEAM1, at least until there is evidence that another member of the complex can restrict its spread through inter-specific competition.
There is growing concern that Cassava brown streak virus and Uganda cassava brown streak virus will move from East Africa to West Africa, especially to Nigeria, via whitefly introductions or the movement of infected cassava cuttings. Nigeria is the largest cassava producer in the world (FAOSTAT, 2015), and if these two viruses are introduced into the region, where cassava production is already hampered by Cassava Mosaic Disease virus, serious food security and economic implications will ensue. It is imperative to stop the spread of both whiteflies and infected planting material to the region.
The pest risk posed by B. tabaci MEAM1 has several components. The invasion risks posed under natural rainfall conditions are extended by the use of irrigation and the presence of glasshouses, which allow survival under otherwise lethally dry or cold conditions. The set of climate suitability maps, including the composite scenario, provides the basis for assessing emerging invasion risks if the use of glasshouses or irrigation were to be extended. The underlying model can also be used with confidence to project the changes that may be expected under future climate changes.
The potential threats of further invasion by B. tabaci MEAM1 suggest that there may be value in exploring the potential for deployment of Eretmocerus spp. and other agents for classical biological control into areas that are suitable for B. tabaci MEAM1, but presently remain uninvaded. Such pre-emptive regional assessments could speed up the response to new incursions, thereby reducing the production shocks in newly invaded areas. Doubtless, the new molecular diagnostic tools available for distinguishing the species within the B. tabaci complex will be critical in matching control agents to target species. It may also be economically justifiable to develop crop host genotypes that are resistant (Smith et al., Reference Smith, Nagle, MacVean, Vallad, van Santen and Hutton2018).
Supplementary material
The supplementary material for this article can be found at https://doi.org/10.1017/S0007485320000061
Open access
