There is mounting scientific evidence that climate change will exacerbate agricultural losses from insect pests in a number of ways (Porter et al. 1991, Gregory et al. 2009, Ziska et al. 2011). First, numerous crop pests have expanded their range northward since the 1960s due, in part, to rising annual temperatures (Bebber et al. 2013). Second, warmer winter temperatures at higher latitudes are expected to reduce winter die-offs thereby resulting in higher densities of pest insects emerging to feed and reproduce in the spring (Bale et al. 2002). Third, warming induced shifts in phenology may lead to earlier spring emergence and subsequent dispersal of overwintering insects (Bell et al. 2015). Finally, climate change is predicted to increase number of generations produced per year in multivoltine insects which in turn can greatly exacerbate the economic damage caused by these agricultural pests (Tobin et al. 2008, Ziter et al. 2012).

The brown marmorated stink bug, Halyomorpha halys (Stål; Hemiptera: Pentatomidae),an extremely polyphagous insect (Hoebeke and Carter 2003), has recently emerged as a harmful invasive pest in North America and Europe. Originally from East Asia (Lee et al. 2013), H. halys invaded North America in 1996 (Hoebeke and Carter 2003) and Europe around 2004 (Haye et al. 2014). H. halys is capable of being transported very long distances via shipping sea freights, ground transport vehicles, and air transportation (Hoebeke and Carter 2003, Gariepy et al. 2014, Gariepy et al. 2015). In addition, both adult and nymphal stages have a strong dispersal capacity (Lee 2015, Lee and Leskey 2015). Consequently, H. halys has rapidly spread across the United States with detections in 43 states (NIPMC 2017). In Canada, the pest has established breeding populations in southern Ontario including Niagara's apple orchards (Frasier 2016); detections have also been made in the provinces of Quebec, Alberta, and British Colombia (Gariepy et al. 2014, NIPMC 2017). After its initial 2007 detection in Switzerland, H. halys has been spreading across Europe (Haye et al. 2014). Breeding populations are present in Switzerland (Haye et al. 2014), France (Callot and Brua 2013), and Italy (Cesari et al. 2015); individuals have also been detected in Liechtenstein (Arnold 2009), Germany (Heckmann 2012), Hungry (Vétek et al. 2014), Greece (Milonas and Partsinevelos 2014), Serbia (Šeat 2015), Romania (Macavei et al. 2015), Spain (Dioli et al. 2016), and Bulgaria (Simov 2016). In Europe, H. halys appears to be moving eastward given the recent detections in Georgia and Russia (Gapon 2016).

H. halys occurs in a wide range of landscapes, including urban areas, where the insect is considered a nuisance pest (Lee 2015). During winter, H. halys enters a facultative reproductive diapause with adults congregating in protected areas including human-made structures. In its native range of East Asia, H. halys is classified as an outbreak pest of fruit trees (Lee et al. 2013). However, in the United States, H. halys has quickly become a major agricultural pest, especially in fruit tree orchards (Leskey et al. 2012a,b, 2015). This phytophagous pentatomomid has >100 reported host plants including ornamental shrubs, hardwood trees, field crops, as well as a wide range of fruit and vegetable crops (Rice et al. 2014). In 2010, the mid-Atlantic apple industry suffered an estimated ~US$37 million in losses from H. halys feeding damage (USAA 2010). Feeding damage has also been reported in the Pacific Northwest and Midwestern United States (NIPMC 2017). In Europe, H. halys has established in Emilia Romagna, Italy, one of the most important tree fruit and wine producing regions in Europe (Maistrello et al. 2016). In just a few short years after first discovery, H. halys has become a significant fruit pest in Italy, with damages estimated upwards of 40% being reported in fruit orchards where this insect is abundant (Maistrello et al. 2017).

The length and favorability of an insect's growing season impacts agricultural losses by influencing host-crop synchrony and pest densities (Gregory et al. 2009, Ziska et al. 2011). For instance, favorable abiotic conditions, including mild winters and early onset of spring, can lead to massive insect pest outbreaks in some years (Roff 1983, Porter et al. 1991). In the mid-Atlantic region of the United States, H. halys’ growing season aligns well with those of apples and peaches, but pest densities and subsequent economic losses vary both spatially and annually (Leskey et al. 2012a, 2015). Similarly, abiotic conditions during the growing season influence the number of generations produced per year by multivoltine insects like H. halys which in turn can influence crop losses. In northern China and Japan, H. halys has been reported to have one or two generations (reviewed by Lee et al. 2013), while four to six generations per year have been suggested in the southern Chinese province of Guangdong (Hoffman 1933). In the United States, H. halys is reported to be univoltine in Pennsylvania, New Jersey, and New York (Nielsen and Hamilton 2009a) and bivoltine in West Virginia, Virginia, North Carolina (Bakken et al. 2015, Leskey et al. 2015), and California (Ingels et al. 2016). In Europe, H. halys appears to be univoltine in Switzerland (Haye et al. 2014) and bivoltine in Italy (Maistrello et al. 2016).

Climate is often a good predictor for continental scale invasive risk forecasts because invasive species are often limited more by climate than biotic interactions (Dukes and Mooney 1999, Ziska et al. 2011). Kriticos et al. (2017) modeled the potential global distribution of H. halys using the bioclimatic niche model, CLIMEX. This model highlighted several regions at risk for further spread and establishment under current climatic conditions. Climate change will add to the complications of managing such a species because the geographic distribution of H. halys is primarily driven by climate although the non-climatic factors likely influence population densities (Venugopal et al. 2016). Projected rising temperatures may enhance H. halys development and survival particularly in the northern latitudes where winter minimum temperatures are predicted to rise disproportionally faster (Nakicenovic and Swart 2000, Musolin 2007). For instance, mean January and February temperatures influence H. halys overwintering mortality and Kiritani (2006) estimated that each 1°C rise above 4°C would reduce winter mortality by 13.5% in Japan. However, climate change may also negatively impact H. halys distribution and abundance in more southern latitudes where average daily temperatures are likely to rise above H. halys’ upper temperature threshold (Zhu et al. 2016). To date, several ecological niche models have been used to predict the potential global distribution of H. halys under current climatic conditions (Zhu et al. 2012, Haye et al. 2015, Zhu et al. 2017, Kriticos et al. 2017) and possible mid-century climate scenarios (Zhu et al. 2016). However, none of these models address potential changes in H. halys voltinism and phenology that may occur under a changing global climate.

To address this shortcoming, CLIMEX (Hearne Scientific Software Pty Ltd, South Yarra, Australia), a process-based bioclimatic niche modeling package (Kriticos et al. 2015a), was used to compare the potential global geographic distribution and abundance of the invasive H. halys under recent baseline climate and possible future climate scenarios. CLIMEX was chosen because it has a proven track record of reliably modeling the potential distribution of invasive species (Sutherst and Maywald 2005, Kriticos et al. 2015b, Yonow et al. 2016) and can be used to inform biosecurity planning and pest management in light of a changing global climate (Sutherst et al. 2000, Kriticos and Brunel 2016, Olfert et al. 2016). This approach identifies areas suitable for H. halys and could serve as a potential framework to quantify the impacts of climate change by incorporating multiple future climatic scenarios to examine shifts in suitable areas. CLIMEX model output was also used to assess how climate change may influence H. halys phenology and voltinism across representative locations in its non-native range.

Materials and Methods

The CLIMEX Model

In this study, the CLIMEX bioclimatic model for H. halys developed by Kriticos et al. (2017) was used to evaluate distribution and population dynamics (Table 1). This model's range projections are consistent with the known global distribution of this species and exhibits higher sensitivity compared to previous attempts to estimate the potential distribution of H. halys (Zhu et al. 2012; Haye et al. 2015; Zhu et al. 2017). CLIMEX is a process-based niche modeling package that estimates suitable areas based on a combination of long-term meteorological data and parameter values describing the lspecies' abiotic preferences (Kriticos et al. 2015a). Unlike correlative niche models, CLIMEX model parameters can be calibrated inferentially, utilizing species occurrence data and phenological observations, as well as deductively, drawing upon experimental observations of species behavior under laboratory conditions to inform parameter section. CLIMEX uses weekly and annual Growth Indices, (GI_W) and (GI_A) respectively, to describe the potential for population growth as a function of optimal Soil Moisture (MI) and Temperature (TI) indices. A second set of parameters, Stress Indices (cold, wet, hot, dry), depict what extreme temperature and moisture conditions limit species distribution and survival. As a result, the CLIMEX model better matches observed northern latitudinal occurrences in North America and China by clearly defining H. halys tolerance to cold temperatures (Zhu et al. 2017). In addition, Kriticos and colleagues' (2017) model defines this insect's dry range limit more realistically whereas previous MaxEnt models overestimated suitable habitat in arid regions of North America and the Middle-East (Haye et al. 2015). Nevertheless, there are some important discrepancies between the aforementioned niche models concerning their estimations of H. halys establishment potential in Africa that should be noted. For instance, Zhu and colleagues' (2012) GARP model and Haye and colleagues' (2015) MaxEnt model underestimate suitability in sub-tropic regions of southern Africa given H. haly's presence in the subtropical regions of south eastern China and Taiwan. However, Kriticos and colleagues' (2017) CLIMEX model overestimates suitability in large tropical regions of equatorial Africa including DR Congo, Congo Republic, and Uganda.

Table 1.

CLIMEX parameter values for Halyomorpha halys

Phenology and Voltinism

Twelve representative locations with established non-native H. halys populations (EDDMapS 2017, NIPMC 2017) were selected to assess H. halys sensitivity to different possible future climates. Representative locations in the invaded range included ten locations in North America (eight in the United States and two in Canada) and two locations in Europe (Table 2). The latitude, longitude, and elevation of each representative 10′ grid cell location are shown in Supplementary Table 1. Rural and urban irrigation practices in Walla Walla, Portland, and Sacramento, United States were simulated by assuming that 2.5 mm of water was applied to farms and urban gardens daily to better match observed H. halys’ densities at these drier sites (Hansen 2015, Leskey et al. 2015, Ingels et al. 2016, Kriticos et al. 2017). Essentially, in any week which had <21 mm of precipitation, the model would make up the deficit as an irrigation supplement. To examine the impact of climate change on H. halys phenology, shifts in the weekly growth index (GI_w) were predicted under the two GCMs for the years 2050 and 2100. The number of weeks where GI_w > 0 (0 = no growth and 1 = optimal growth) provided an estimation of the total growing season length for H. halys. Therefore, GI_w is a decent indicator of population growth and this “modeled seasonal suitability” was found to match accurately observed trends in insect seasonal phenology based on trap counts (De Villiers et al. 2013, Kriticos et al. 2015b, Yonow et al. 2016). In addition, the simulated number of generations predicted based on the CLIMEX model, using a value of 595 degree-days above 12°C, was utilized to compare recent historical and possible future climates.

Table 2.

Ecoclimatic Index values (% of total continental areas) for Halylomorpha halys based on recent historical climate (1975H), 2050 and 2100 with two general circulation models (CSIRO Mk3.0 and MIROC-H) under the SRES A2 for the World, North America, and Europe

Results

Potential Distribution Under Possible Future Climate Scenarios

Overall, the modeled potential global distribution under recent historical climate agrees well with the known distribution of this species in both its native and invaded ranges (see Kriticos et al. 2017 for model validation). In the Southern Hemisphere, model output indicates that portions of eastern South America, central and southern Africa, and Australia's eastern coast and New Zealand's North Island are highly suitable for H. halys establishment (Fig. 1). Modeled climatic suitably estimates in the Southern Hemisphere are consistent with the known geographic distribution of H. halys with the exception of equatorial Africa where the CLIMEX model projects an excessive capability of H. halys to cope with wet tropical climates. There is also potential for further range expansions in both North America (Fig. 2) and Europe (Fig. 3). Compared to recent historical climate conditions, the suitable H. halys range was shown to shift poleward and to contract from its known southern range limits in North America and Asia beginning in 2050 and continuing into 2100 (Fig. 4). In the Northern Hemisphere, China and the United States become less favorable for H. halys while Europe, Eastern Russia, and Canada become more susceptible to H. halys establishment and spread under both GCM scenarios by the end of the 21st century (Fig. 4). In East Asia, southern Russia constitutes a steep cold stress gradient that limits northward range expansion under future climate scenarios (Fig. 4). In contrast, potential suitable ranges in the Southern Hemisphere contract in South America, Africa, and Australia under both GCM scenarios (Fig. 4). On a global scale, the percentage of highly suitable range declined under future climate scenarios, but more so under the CSIRO Mk3.0 scenarios (Table 2).

Fig. 1.

The modeled global climatic suitability (CLIMEX Ecoclimatic Index) for Halyomorpha halys under recent historical climate. Blue circles designate recorded occurrences of H. halys.

Fig. 2.

Modeled climatic suitability (CLIMEX Ecoclimatic Index) for Halyomorpha halys under recent historical climate in North America. Blue circles designate recorded occurrences of H. halys.

Fig. 3.

Modeled climatic suitability (CLIMEX Ecoclimatic Index) for Halyomorpha halys under recent historical climate in Europe. Blue circles designate recorded occurrences of H. halys.

In North America, potential H. halys range expanded northward into Canada as cold limitations were overcome while substantial range contraction occurred in the southeastern United States under future climate scenarios (Fig. 5). These range contractions were more pronounced under the MIROC-H model, but overall projected changes in climatic suitability (EI values) were similar between the two GCMs (Table 2). Projected range losses in the southern latitudes (EI < 1) and reduced climatic suitability in the northern latitudes were associated with increased heat stress (HS) (Table 3). Accumulated heat stress > 100 indicated that stress was sufficient to eliminate the majority of potential H. halys range in Texas, Oklahoma, Kansas, Missouri, Tennessee, Kentucky, Arkansas, Louisiana, South Carolina, Mississippi, and Florida by 2100 under both GCMs (Fig. 5). By 2050, much of inland California would no longer be suitable for H. halys, including Sacramento, due to heat stress (Fig. 5) (Table 3). By 2100, the two GCM outputs indicated that Walla Walla, United States would also no longer be able to support year round H. halys populations as a result of increased heat stress levels (Table 3). However, positive annual growth indices at Sacramento and Walla Walla (Supplementary Table 2) and high rates of weekly spring and fall growth rates (Supplementary Fig. 1) under future climate scenarios indicate the possibility of transient H. halys populations at these increasingly hot sites. In Canada, large portions of Ontario, Quebec, New Brunswick, and Nova Scotia become suitable or highly suitable for H. halys establishment by 2100 under both GCMs (Fig. 5). Under future climate scenarios, suitable to highly suitable areas increased substantially across Europe (Table 2) and considerable range expansion was observed in northern Europe as cold stress limits were overcome (Fig 6). EI values rose more under the MIROC-H model compared to the CSIRO Mk3.0 model (Table 2). Under the CSIRO Mk3.0 model, suitable areas increased more than highly suitable areas compared to recent historical conditions (Table 2). The MIROC-H model indicated that the percentage of highly suitable areas, including large proportions of the United Kingdom, France, Belgium, the Netherlands, Germany, Poland, and Denmark, would undergo a 30-fold increase by 2100 compared to projections under the recent historical climate data (Table 2). In southern Europe, EI values declined in portions of Spain, Italy, Serbia, and Greece under both GCMs by 2100 due to rising heat stress levels (Fig. 6).

Fig. 4.

The modeled global climatic suitability (CLIMEX Ecoclimatic Index) for Halyomorpha halys based on CSIRO Mk3.0 and MIROC-H general circulation model projections under the A2 SRES emission scenario for 2050 and 2100.

Fig. 5.

The modeled climatic suitability (CLIMEX Ecoclimatic Index) for Halyomorpha halys based on CSIRO Mk3.0 and MIROC-H general circulation model projections in North America under the A2 SRES emission scenario for 2050 and 2100.

Table 3.

CLIMEX indices for Halyomorpha halys based on recent historical climate (1975H) and general circulation model CSIRO Mk3.0 and MICRO-H at 12 representative locations in North America and Europe

Fig. 6.

The modeled climatic suitability (CLIMEX Ecoclimatic Index) for Halyomorpha halys based on CSIRO Mk3.0 and MIROC-H general circulation model projections in Europe under the A2 SRES emission scenario for 2050 and 2100.

Phenology

Weekly growth indices for H. halys were plotted for 6 out of the 12 representative locations selected based on available H. halys phenology records and the presence of vulnerable crops (Fig. 7). Weekly growth indices for the other six representative locations are shown in Supplementary Fig. 1. Overall, the number of weeks where the growth index was positive (GI_w) increased, resulting in longer growing seasons under both GCMs (Table 3). Rising spring temperatures resulted in early spring emergence while warming fall temperatures led to later fall activity (Fig. 7). However, these increased opportunities for growth were often negated by weeks where the average maximum temperature exceeded H. halys’ limiting high temperature threshold (DV3 = 33°C) (Table 1). Overall, excessively warm weeks in June through August reduced weekly growth rates at all representative locations with the exception of Zurich-Obefleden, Switzerland (Fig. 7A–E). At Zurich-Obefleden, Switzerland, observed recent historical weekly growth rates doubled in July-August (CSIRO Mk3.0) and September (MIROC-H) by 2100, respectively (Fig. 7F). By 2100, the projected timing of peak and minimal growth periods varied greatly between the GCMs. For many locations, model output using MIROC-H resulted in lower EI values compared to CSIRO Mk3.0 (Table 3) due to decreased mid-summer growth rates for 2100, including Allentown, United States (Fig. 7A), Asheville, United States (Fig. 7B), and Emilia-Romogna-Modena, Italy (Fig. 7E). When an irrigation scenario was applied, Portland, United States exhibited 5 additional weeks of positive H. halys growth (Fig. 7D), which in turn, doubled the climatic suitable (EI) rating at this site under the recent historical climate (Table 3). Sacramento and Walla Walla, United States also exhibited enhanced growth and climatic suitability when an irrigation scenario was applied (Table 3) (Supplementary Fig. 1E and F). Under future climate scenarios, weekly growth indices at Portland, Walla Walla, and Sacramento were enhanced by the inclusion of an irrigation scenario (Table 3), which also neutralized any potential dry stress accumulation at these sites (Supplementary Table 2).

Fig. 7.

Weekly growth index (GI_w) for Halyomorpha halys based on recent historical climate and general circulation model outputs for (A) Allentown, United States, (B) Asheville, United States, (C) Niagara-Fonthill, Canada, (D) Portland, United States under top-up irrigation scenario, (E) Emilia-Romagna-Modena, Italy, and (F) Zurich-Obefelden, Switzerland.

Discussion

In the context of this study, the global potential range of H. halys expanded polewards and contracted from its southern temperature range limits under future climate scenarios. The future climate scenarios explored here suggest that the invasion threat posed to Europe and Canada will greatly intensify. Overall, H. halys exhibited longer growing seasons as well as enhanced population growth in the spring, and more possible generations per year. However, summer growth was greatly reduced in all but one representative location. Therefore, the effect of climate change on H. halys is complex with both positive and negative impacts varying both spatially and temporally. This study highlights some areas of emerging or increasing threat of invasion by H. halys under possible future climates. It also identifies areas where H. halys abundance and subsequent pest outbreaks may decline. Nevertheless, these range suitability and relative abundance projections are only potential distributions in which pest outbreaks may occur based on climatic factors and not predicted future distributions.

In the United States, prime horticultural areas in the Northeast, Pacific Northwest, Upper Midwest, and portions of California remained suitable or became more suitable for H. halys in terms of future climate scenarios. Although soil moisture was not a limiting factor for H. halys persistence across the majority of sites examined in this study, the inclusion of a moderate irrigation scenario did enhance H. halys growth potential and climatic suitability at representative locations in California and the Pacific Northwest under both baseline and future climate scenarios. Given that many horticultural production regions are heavily irrigated, current ecological niche models may be underestimating the threat H. halys poses to horticultural crops grown in more arid regions (Zhu et al. 2012, Kriticos et al. 2017). Therefore, U.S. pest managers in these regions should continue to monitor changes in H. halys densities and be prepared for the possibility of pest outbreaks in the future. At present, H. halys has very limited distribution in Canada and no agricultural losses have been reported. However, future climate scenarios indicate that key fruit and vegetable growing regions in Ontario, Quebec, and British Columbia will become increasingly suitable for H. halys (Statistics Canada 2012). Therefore, it would seem prudent to consider the feasibility of coordinated containment strategies to prevent the future spread of this pest into Canada's prime horticultural regions.

In terms of the future climate scenarios explored here, substantial northward range expansions are indicated across Europe. The large decrease in climatic suitability in southern Spain by the end of the century is encouraging, although much of the country's northern boundaries remain climatically suitable. By 2100, Modena, Italy, which is located within a key fruit growing region in Europe, became marginal to completely unsuitable for H. halys due to increases in summer temperatures above its upper temperature threshold. However, the majority of Emilia-Romagna region remained suitable for H. halys particularly in areas east of Modena where orchards and vineyards are abundant (Maistrello et al. 2016). Since H. halys is capable of travelling >5 km per day (Lee and Leskey 2015), it is plausible that the pest will disperse from sites like Modena during the summer and seek temporary refuge in cooler land types especially forests and mountains (Bakken et al. 2015). Thus, Italy's horticultural regions will likely remain at risk from H. halys in the future, but the timing and intensity of infestations will likely vary depending on land type. Given that future climate scenarios indicate that the risks for Europe are likely to increase in conjunction with the present risk of further spread (Kriticos et al. 2017), European countries may want to consider which biosecurity measures are needed to reduce further spread of H. halys across Europe. Preventing its entry into high-risk countries may be the most cost-effective response.

Changes in weekly growth indices (GI_w) for Allentown and Asheville, United States under recent historical conditions closely conformed to results of phenological studies conducted at these locations (Nielsen and Hamilton 2009a, Bakken et al. 2015). Unfortunately, the literature did not provide much data on H. halys phenology in European countries and Canada, but the simulated weekly growth indices for Zurich-Obefleden, Switzerland were consistent with a phenology study conducted by Haye et al. (2014) in nearby Zurich-Wollishofen, Switzerland. In this study, future climate scenarios resulted in longer growing seasons. However, both GCM projections suggest that rising mid-summer temperatures could restrict H. halys population growth. These results are consistent with Olfert et al. (2016) who also found mid-summer declines in weekly growth rates of the parasitoid, Peristenus digoneutis (Loan; Hemeoptera: Braconidae), across Canada under similar future climate scenarios. Dramatic shifts in H. halys phenology could have important management implications in terms of chemical control. In the mid-Atlantic region of the United States, H. halys is reported to feed on tree fruit over the entire growing season with the majority of feeding damage occurring late in the growing season (August–early September) by first-generation adults (Nielsen and Hamilton 2009b). In response, growers in this region have increased the frequency of costly insecticide applications (Leskey et al. 2012a) which have inadvertently decreased natural occurring predator and parasitoid densities. In turn, secondary pest outbreaks of the woolly apple aphid and San Jose scale have been reported in several mid-Atlantic apples orchards that employ insecticides to control H. halys (Leskey et al. 2012b). In the absence of their natural enemies, these once minor pests are now able to grow to economically damaging levels (Hardin et al. 1995) indicating that current chemical control strategies are not financially and ecologically sustainable in the long run. Given that the modeled decreases in mid-summer H. halys growth across this region coincides with late season tree fruit development, climate change may reduce mid- to late season fruit damage, thereby lessening the need for late season insecticide applications. However, the possible future climates explored here also indicated earlier spring activity and enhanced spring growth potential. Earlier spring emergence coupled with reduced over-wintering mortality (Kiritani 2006) could increase the risk of early season feeding damage in some horticultural crops.

Although model projections for voltinism were fairly consistent with numbers reported in the literature, these estimates should be approached with some caution. The number of generations predicted by Kriticos and colleagues' (2017) CLIMEX model was based off of a degree day model with a relaxed minimal developmental threshold (DV0) of 12°C. Since estimated temperature thresholds for complete development are 14.17°C for U.S. populations (Nielsen et al. 2008) and 12.97°C for Swiss populations (Haye et al. 2014), Kriticos and colleagues' (2017) CLIMEX model parameters may be overestimating the number of annual generations possible. For instance, an additional generation was projected in Sacramento, United States, under recent historical climate than has actually been observed (Ingels et al. 2016). In the northeastern United States, the completion of a partial to full second generation in Pennsylvania and New York, as projected in this study's model output, is plausible based on previous H. halys voltinism model estimations conducted by (Nielsen et al. 2016), but have yet to be confirmed at these locations. Unfortunately, captures in pheromone traps commonly used to monitor H. halys phenology do not necessarily provide clear estimates pertaining to the number of generations per season as generations often overlap (Leskey et al. 2015). Regardless, projected changes in H. halys voltinism under different possible future climates were consistent with similar studies examining the impact of climate change on insect voltinism (Tobin et al. 2008, Ziter et al. 2012, Olfert et al. 2016). Rising global temperatures are expected to increase the number of annual generations in multivoltine insects, including stinkbugs (Kiritani 2006), and this study suggests that H. halys may become multivoltine throughout much of its non-native range over the next 30 to 80 yr.

Overall, the projected changes in global H. halys distribution and relative abundance were similar between mid-century GCM scenarios. In contrast, CSIRO Mk3.0 and MIROC-H model outputs for 2100 differed greatly, which highlights the uncertainties associated with current climate modeling projections. However, it should be noted that variance the between the two GCMs (under a single high emission scenario) does not represent the full range of uncertainties associated with climate change modeling. The use of additional GCMs would cover a wider range of climatic sensitives in relation to climate change (Kriticos and Brunnel 2016). Future studies should also consider employing more moderate greenhouse gas emission scenarios given the dynamic nature of global policy in regards to climate change.

In addition to the uncertainties related to future global greenhouse gas emission patterns, CLIMEX itself makes several assumptions that introduce further uncertainty to model projections. First, CLIMEX utilizes a baseline climate data set that is centered on 1975 (Kriticos et al. 2012). Given the increase in global temperatures since 1975 (Rahmstorf et al. 2012) and the ongoing observed shifts in insect pest distributions (Bebber et al. 2013), up-to-date global climatologies for ecological niche and species distribution modeling packages are greatly needed. Second, CLIMEX does not account for potential changes in the frequency and intensity of extreme short-term weather events predicted under future climate change scenarios (Kriticos et al. 2012, Walsh et al. 2014), which could have important management implications for H. halys. Third, the CLIMEX model parameters used in this study assume that the existing invasive populations have the same adaptive potential as their native counterparts in East Asia. However, North American populations are likely the result of a single introduction event comprising of individuals originating from the Beijing region in China (Gariepy et al. 2014, Xu et al. 2014). In contrast, there have been multiple introductions in Europe resulting in dominant haplotypes differing between countries (Gariepy et al. 2015). Given the differences in genetic diversity across native and invaded ranges, it is reasonable to assume invasive populations may respond differently to climate change compared to their more genetically diverse East Asian counterparts. Finally, CLIMEX's projected changes in H. halys distribution and abundance are based solely on its response to climate. Non-climatic factors, such as land use type, host plant distribution, and biotic interactions, have been shown to influence H. halys distribution and densities at regional spatial scales (Haye et al. 2014, Rice et al. 2014, Venugopal et al. 2016). In the mid-Atlantic United States, Bakken et al. (2015) found greater H. halys abundance at high altitudes, a surrogate for decreasing temperature. Thus, mountainous terrains, like the Appalachian Mountains, may act as a temperature refuge for H. halys during the increasingly hot summer months that are likely to occur under future climate change. The relationship between host plants and H. halys development is an important component of this pest's population dynamics. Despite its broad host range, H. halys is only reported to complete its life cycle on a few select host plants (Bergmann et al. 2016). For example, H. halys survival on apples (Malus domestica Borkh) is low (Funayama 2002) while peach (Prunus persica L.) is able to support all H. halys life stages (Rice et al. 2014). Similarly, Acebes-Doria et al. (2016) found that H. halys fitness was enhanced by feeding on multiple host plants and higher pest densities are often associated with orchards neighboring unmanaged woodland (Bakken et al. 2015, Venugopal et al. 2016). Crop phenology is also likely driving H. halys movement between host plants as H. halys prefers to feed on reproductive structures of many economically important crop species (Rice et al. 2014). Climate change may influence these seasonal feeding patterns by shifting the availably of preferred host plant life stages.

In the instance of H. halys, global climate change directly influenced this lspecies' potential distribution and abundance. Concurrently, climate change may also influence pest-crop outcomes (Bale et al. 2002), thereby indirectly influencing pest dynamics. To better address future economic risks associated with H. halys under global climate change, host plant responses to a changing climate also need to be considered (Gregory et al. 2009). Earlier spring emergence is likely to occur across much of this pest's range in the future, which may result in a crop-pest mismatch if H. halys responds more strongly to rising temperatures than their preferred host plants. For instance, Funayama (2013) found that higher maximum spring temperatures in northern Japan were associated with lower H. halys densities as the insect emerged weeks before suitable host plants were available. It is plausible that such phenological desynchronization could ultimately lead to decreased crop damage if H. halys emerges before preferred crops are available.

In conclusion, the results of this study provide an indication of the possible change in potential future distribution and abundance of H. halys in light of climate change. Regardless of CLIMEX's limitations and uncertainties related to future greenhouse gas emission scenarios, it would seem prudent to assess the feasibility of a coordinated containment strategy in Europe and Canada as both current (Zhu et al. 2012, 2017; Haye et al. 2015; Kriticos et al. 2017) and future climate scenarios (Zhu et al. 2016) indicate suitable climatic conditions for further spread. The most sensible response in light of uncertainties associated with climate change may be to educate the general public about the threats caused by this species and organize an efficient early warning system through citizen science (Maistrello et al. 2016) in regions currently under high risk of H. halys establishment as well as areas that are projected to become suitable for this species in the future.