Climate Change Impacts on the Potential Distribution and Range Shift of Dendroctonus ponderosae (Coleoptera: Scolytidae)

: Dendroctonus ponderosae Hopkins (Coleoptera: Scolytidae) is one of the most important bark beetles in North America and causes considerable economic and ecological losses during outbreaks. The distribution of this pest species is likely to be altered by climate change, which may threaten currently una ﬀ ected areas. In this study, we used CLIMEX to project the potential global distribution of D. ponderosae according to both historical climate data (1987–2016) and future climate warming estimates (2021–2100) to evaluate the impact of climate change on this species. Regions with suitable climate for D. ponderosae are distributed in all continents except Antarctica under both historical and future climate conditions, and these are predicted to change continuously with climate change. Overall, climate suitability will increase in middle- and high-latitude regions and decrease in low-latitude regions, and regions most sensitive to climate change are located in the mid-latitude zone. Moreover, the shift directions and ranges of climate-suitable regions under future conditions will di ﬀ er among continents, and the shift distances in the north–south direction are larger than these in the east–west direction for Africa, Asia, Europe, South America, and Oceania, indicating that shift direction is possibly mainly a ﬀ ected by temperature. These projected distributions may provide theoretical guidance for early-warning intervention and risk assessment.


Introduction
The increasing emission of greenhouse gases is already causing global climate change, particularly global warming. Based on representative concentration pathway (RCP) 4.5, globally averaged combined land and ocean surface temperatures warmed by 0.85 • C (0.65-1.06 • C) between 1880 and 2012, and global mean surface temperature by the end of the 21st century (2081-2100) is likely to be 1.1-2.6 • C higher than the 1986-2005 period [1]. As poikilotherms, insects are highly likely to be affected by global climate change, resulting in disruption of their life history and distribution boundaries, which could increase the risk of invasion of new regions, threatening unexpected losses across the world [2,3].
The mountain pine beetle, Dendroctonus ponderosae Hopkins (Coleoptera: Curculionoidea: Scolytidae), a native bark beetle of western North America, is currently widely distributed in three major countries of North America, including the USA, Canada, and Mexico, located in tropical and temperate zones [4,5]. The epic outbreaks of this pest species have caused the mortality of hundreds of millions of trees over large areas, leading to timber loss, increase to fire, habitat loss, alterations to carbon cycling processes The Spatial Analyst Module of ArcMap 10.1, developed by the US Environment Systems Research Institute (ESRI) (RedLands, CA, USA), was used to analyze the projected results from CLIMEX. Inverse distance-weighted interpolation (IDW), mean center, and thematic mapping functions were used to create maps of potential global distributions under historical and future climate conditions. ArcGIS was also used to calculate the areas on each continent assigned to each of the four EI categories for pest species [30].

Climate Data
We followed the method of Zou et al. (2019) [31] to process and obtain historical and future climate data. We used the historical climate data  reported in Zou et al. (2019) [31] as our historical climate data, and for future climate data we selected the same scenario and Global Climate Models (GCMs) as described in this previous study, but divided climate data into different time periods (2021-2100, 2021-2030, 2031-2040, 2041-2050, 2051-2060, 2061-2070, 2071-2080, 2081-2090, and 2091-2100) in order to capture dynamic changes in the potential distribution of D. ponderosae under different climate scenarios. Specific details of the climate data can be found in Zou et al. (2019) [31].

Known Distribution of D. ponderosae and Host Plants
We obtained the current global distribution of D. ponderosae mainly from existing literature regarding this pest [30,[32][33][34][35], the European and Mediterranean Plant Protection Organization (EPPO) [36], and the Global Biodiversity Information Facility (GBIF) database [37]. The known distribution map covers three countries (Canada, the USA, and Mexico). In Canada, D. ponderosae is mainly distributed in the southern provinces including British Columbia, Alberta, and Saskatchewan. In the USA, the species is mainly located in the western states including Washington, Oregon, California, Idaho, Nevada, Utah, and Wyoming. The distribution of D. ponderosae in Mexico covers the entire country.
Various Pinus spp. are host plants of D. ponderosae, and all pine species in western North America are believed to be suitable hosts for this beetle, including lodgepole pine (Pinus contorta Dougl. ex Loud var. latifolia Engelm), ponderosa pine (P. ponderosa Dougl. ex Laws.), western white pine (P. monticola Douglas ex D. Don), eastern white pine (P. strobus L.), and some exotic pines [4,5,12,35]. Therefore, we treated all species of Pinus as possible host plants, for which the distribution is worldwide, including most areas in North America, Europe, Asia, southern areas of South America and Africa, and Oceania. The actual distribution records were obtained from the Global Biodiversity Information Facility (GBIF) database [38] and shown with green dots in Figure 1.

Known Distribution of D. ponderosae and Host Plants
We obtained the current global distribution of D. ponderosae mainly from existing literature regarding this pest [30,[32][33][34][35], the European and Mediterranean Plant Protection Organization (EPPO) [36], and the Global Biodiversity Information Facility (GBIF) database [37]. The known distribution map covers three countries (Canada, the USA, and Mexico). In Canada, D. ponderosae is mainly distributed in the southern provinces including British Columbia, Alberta, and Saskatchewan. In the USA, the species is mainly located in the western states including Washington, Oregon, California, Idaho, Nevada, Utah, and Wyoming. The distribution of D. ponderosae in Mexico covers the entire country.
Various Pinus spp. are host plants of D. ponderosae, and all pine species in western North America are believed to be suitable hosts for this beetle, including lodgepole pine (Pinus contorta Dougl. ex Loud var. latifolia Engelm), ponderosa pine (P. ponderosa Dougl. ex Laws.), western white pine (P. monticola Douglas ex D. Don), eastern white pine (P. strobus L.), and some exotic pines [4,5,12,35]. Therefore, we treated all species of Pinus as possible host plants, for which the distribution is worldwide, including most areas in North America, Europe, Asia, southern areas of South America and Africa, and Oceania. The actual distribution records were obtained from the Global Biodiversity Information Facility (GBIF) database [38] and shown with green dots in Figure 1.

Parameter Fitting
Parameters for D. ponderosae CLIMEX indices are grouped into stress-related parameters and growth-related parameters. Generally, the parameters are determined by referring to the biological characteristics of the pests and the templates of the software. The combined use of manual adjustment of parameters and semiautomatic parameter fitting procedure may avoid the error of subjectivity [28].
Semi-automatic parameter fitting involved a number of steps. First, we created the reference file according to the known distribution of D. ponderosae. We then set an initial set of parameter values and accurate ranges that could be fitted by the semi-automatic parameter fitting procedure (see Appendix A, Table A1). Moreover, default GA operating characteristics were employed, with a

Parameter Fitting
Parameters for D. ponderosae CLIMEX indices are grouped into stress-related parameters and growth-related parameters. Generally, the parameters are determined by referring to the biological characteristics of the pests and the templates of the software. The combined use of manual adjustment of parameters and semi-automatic parameter fitting procedure may avoid the error of subjectivity [28].
Semi-automatic parameter fitting involved a number of steps. First, we created the reference file according to the known distribution of D. ponderosae. We then set an initial set of parameter values and accurate ranges that could be fitted by the semi-automatic parameter fitting procedure (see Appendix A, Table A1). Moreover, default GA operating characteristics were employed, with a parameter set population of 100, a crossover rate of 0.75, and a mutation rate of 0.1, and the model was applied for 250 generations. Eventually, the software derived the best fitting parameters with a fitness value of 55.55, as listed in Table A1 in Appendix A.
Manual adjustment of stress-related parameters was based on the results of semi-automatic parameter fitting, and we adopted all stress-related parameters except cold stress accumulation rate (THCS), heat stress threshold (TTHS), and dry stress accumulation rate (HDS). We adjusted THCS to −0.1 week −1 to fit the northern boundary of the known distribution, and adjusted THHS to 42 • C to fit the tropical desert area known to be suitable for D. ponderosae in northern Mexico, with reference to the reports of Patterson (1930) and Sanfranyik et al. (1999) [39,40]. We then adjusted HDS to −0.005 week −1 with reference to typical semi-arid conditions.
Manual adjustment of temperature indices was performed by setting the minimum development temperature threshold (DV0) to 5 • C, in line with results from Reid et al. (1970) who stated that the lowest temperature for the development of D. ponderosae is 5 • C [41]. The lower optimum temperature (DV1) and the upper optimum temperature (DV2) were set to 18 and 25 • C, respectively. Because Bentz et al. (1991) found that the optimum developmental temperatures for all life stages of D. ponderosae are between 23 and 25 • C, and the optimum temperature for egg hatching is between 20 and 25.6 • C [42], we adjusted the lower temperature to 18 • C to fit the distribution in areas bordering southern Canada and Mexico. According to experimental results, D. ponderosae becomes negatively phototactic at temperatures above 35 • C [43], and flight ability is severely restricted above 38 • C [44], hence we set the maximum developmental temperature threshold (DV3) to 38 • C. The annual heat sum for population restriction (PDD) was set at 833 degree-days because some studies reported that D. ponderosae requires more than 833 degree-days to complete the life cycle within a single year (i.e., to be univoltine) [32,[45][46][47].
Manual adjustment of moisture indices was performed according to Creeden et al. (2014) since we know that outbreaks of D. ponderosae are highly correlated with a dry climate [48]. Additionally, D. ponderosae is distributed in tropical desert areas (e.g., California and New Mexico), hence moisturerelated indices were set low; the lower soil moisture threshold (SM0) was 0.05, the lower optimal soil moisture threshold (SM1) was 0.1, and the upper optimal soil moisture threshold (SM2) was 0.2. Furthermore, the upper soil moisture threshold (SM3) was set to 1.5, allowing D. ponderosae to colonize the southwestern area of British Columbia. The final parameters are listed in Table A2 in the Appendix A.

Parameters for Host Plants
Pinus spp. are among the most widely distributed plant species in the world, owing to their wide niche range. It is reported that pines can tolerate low temperatures of −60 • C and high temperatures of 50 • C, and they lack a strict demand for soil, while most species prefer a well-drained soil-flourishing in sandy conditions [49]. Additionally, the leaves of pines have the structural characteristics of xerophytes, allowing them to adapt to long periods of drought and cold. Kishchenko (2004) [50] investigated the effects of climatic factors on the growth of pines and found that air temperature and the average above-zero temperature for shoot growth varies greatly across pine species, hence it is essential to set wide parameter ranges when considering trends in whole Pinus spp. The templates for, for semi-arid zones and for temperate zones, were selected to be the initial references for parameters, but the temperature indexes and moisture indexes were adjusted to match their wide niche range and the known distribution. Besides, it is known that the cold resistance of content in Chinese red pine (P. tabuliformis) is high, thus we set the cold stress parameters referring to the template of P. tabuliformis and extended the range to coordinate some species in colder regions such as Eurasian pine (P. sylvestris) [51]. Furthermore, most pines are photophilic and require sufficient daylength to grow, hence we introduced the light index (LI) parameter and set them to 10 and 15, respectively, referring to the parameters for P. tabuliformis. The final parameters for pines are shown in Table A2 in Appendix A and the predicted results were in good consistency with the known distribution.

Classification of Ecoclimatic Index (EI) Values
The EI values give an overall measure of the suitability of a pest to given locations, and they are generally classified into groups to describe the favorability more specifically, in which the favorability actually levels the continuous model's predictions of how environmental characteristics are appropriate in study areas, then determines which actually favor the species presence [52]. The classification standard should be defined in accordance with actual occurrence severities in different regions for actual species. For D. ponderosae, it was reported to have seriously devastated the pine forests in British Columbia and Alberta of Canada, and in western American areas [53,54], where the EI values range from 20 to 48, thus it is reasonable to set the cutoff value to 20 for very favorable regions. As the records for occurrence areas with low level are rare for classification, we referred to Sutherst (2003) [55], who indicated that EI <10 shows that a location is marginal for a given species, while EI is equal to 0 indicating that the location is not favorable for the long-term survival, then we set the cutoff values dividing favorable regions and marginal regions to 10, and the cutoff dividing marginal regions and unfavorable regions to 0, respectively. Finally, EI values were grouped into four classes: unfavorable (EI = 0), marginal (0 < EI ≤ 10), favorable (10 < EI ≤ 20), and very favorable (EI ≥ 20).

Parameter Verification
The final parameters were verified to obtain reasonable prediction results. Firstly, the modeled potential global distribution under historical climate condition matched the known distributions very well, all occurrence records of D. ponderosae worldwide were within the suitable range in the model. Then, the condition of temperature and humidity in the predicted distribution were similar to the climate of western North America, the native range of the species. In addition, the environmental condition of climate-unsuitable regions presented significant disadvantages to D. ponderosae. The high temperature in southern South America and southern Africa, and the low temperature in northern North America and northern Asia both make it hard for D. ponderosae to survive. The reason for the unsuitable regions in the coastal zone is the excessive wet pressure, for D. ponderosae prefers relatively dry conditions [48]. Furthermore, the accumulated temperatures in the climate-suitable regions all meet the development demand of D. ponderosae, which is reported to complete life cycle in one to two years [56].

Analysis of Results
Since D. ponderosae feed only on pines, it is essential to consider the distribution of pines when predicting that of the insect. Therefore, we applied the method of Berzitis et al. (2014) [57] and used EI values for pines (EI P ) to calculate an index of host plant availability (θ), which was then multiplied by the original EI value for D. ponderosae (EI D ) to obtain a new measure of suitability for D. ponderosae (EI D ), incorporating both climate and host plant availability. The θ value was calculated based on the sigmoidal function.
in which q indicates how steeply the curve rises, and h determines the value of EI P at which h equals 0.5. The value of q and h were set at 7 and 12, respectively, to ensure a low value of θ when an area is unsuitable for pine growth (EI D < 10) and a high value (up to 1) of θ when an area is favorable or highly favorable for pine growth (EI D ≥ 10; Figure 2). Using this approach, we acquired EI values for D. ponderosae (EI D ) that superimposed the distribution of host plants under corresponding historical and future climate conditions. After superimposition of EI values onto host plants, EI values for D. ponderosae were imported into ArcGIS software, and the distribution maps of insects under historical (1987-2016) and future (each decade from 2021 to 2100) climatic conditions were obtained using the inverse distance weighted (IDW) function (Figures 3 and 4), and we analyzed the area proportions of different climate-suitable regions in each decade ( Figure 5). Additionally, we compared differences between each decade using historical data and drew maps ( Figure 6) from which we could assess the impacts of climate change on the potential distribution. We then applied the methods proposed by Ge et al. (2019) [58] and obtained curves showing EI differences varying by latitude ( Figure 6).
To capture the location shifts of habitats for D. ponderosae more accurately, we divided each continent into separate polygons based on 0 ≤ EI < 10 (unsuitable regions) and 10 ≤ EI ≤ 100 (suitable regions). We then determined the central point in each polygon using the 'Mean Center' function of ArcGIS software, the shift of which represents the corresponding habitat shift. In addition to shifts paths, we measured the relative shift distances in two cardinal directions (east-west and north-south) between 'future' points and historical central points, as shown in Figure 7. After superimposition of EI values onto host plants, EI values for D. ponderosae were imported into ArcGIS software, and the distribution maps of insects under historical (1987-2016) and future (each decade from 2021 to 2100) climatic conditions were obtained using the inverse distance weighted (IDW) function (Figures 3 and 4), and we analyzed the area proportions of different climatesuitable regions in each decade ( Figure 5). Additionally, we compared differences between each decade using historical data and drew maps ( Figure 6) from which we could assess the impacts of climate change on the potential distribution. We then applied the methods proposed by Ge et al. (2019) [58] and obtained curves showing EI differences varying by latitude ( Figure 6).
To capture the location shifts of habitats for D. ponderosae more accurately, we divided each continent into separate polygons based on 0 ≤ EI < 10 (unsuitable regions) and 10 ≤ EI ≤ 100 (suitable regions). We then determined the central point in each polygon using the 'Mean Center' function of ArcGIS software, the shift of which represents the corresponding habitat shift. In addition to shifts paths, we measured the relative shift distances in two cardinal directions (east-west and north-south) between 'future' points and historical central points, as shown in Figure 7.

Potential Distribution of D. ponderosae under Historical Climate Conditions
The projected potential distribution of D. ponderosae under historical climate conditions (1987-2016) and the known global occurrence records of D. ponderosae are shown in Figure 3. The predicted potential distribution covers all currently known distribution areas and coincides with the severity of occurrence, confirming that our model fits the actual occurrence records with a high degree of accuracy.
The percentages of very favorable regions, favorable regions, and marginal regions accounting for the global land area are projected to be 27.96%, 16.80%, and 28.64%, respectively. The land between 67° N and 55° S is predicted to be climatically suitable for D. ponderosae, except for northern

Potential Distribution of D. ponderosae under Historical Climate Conditions
The projected potential distribution of D. ponderosae under historical climate conditions  and the known global occurrence records of D. ponderosae are shown in Figure 3. The predicted potential distribution covers all currently known distribution areas and coincides with the severity of occurrence, confirming that our model fits the actual occurrence records with a high degree of accuracy.
The percentages of very favorable regions, favorable regions, and marginal regions accounting for the global land area are projected to be 27.96%, 16.80%, and 28.64%, respectively. The land between 67 • N and 55 • S is predicted to be climatically suitable for D. ponderosae, except for northern Africa, central and western Oceania, and western China, which are marginal or unfavorable for D. ponderosae under historical climate conditions.

Potential Distribution of D. ponderosae under Future Climate Conditions
The projected potential future distributions of D. ponderosae in each decade (2021-2100) are shown in Figure 4, overlaid onto hosts distributions. The area proportions of different climatesuitable regions in each decade are shown in Figure 5. In general, the potential future distributions of D. ponderosae expand considerably following global warming, with extensive northward expansion in habitats in the northern hemisphere, while tropical and subtropical habitats tend to shrink. Overall, the potential future distribution is predicted to continue increasing from 2021 to 2100, mainly due to the areas increase in favorable and marginal regions over the decades, which are predicted to increase 1.87% and 2.14%, respectively. However, the area covered by favorable regions will likely decrease by 1.18% in total.
Changes in climatic suitability are shown in Figure 6. EI differences in mid-latitude regions (30-60° S and 30-60° N) are positive and peak at ~60° N, suggesting that climate change will have positive impacts on D. ponderosae, and the impacts will be greatest around 60° N. Additionally, differences in EI between each decade in historical data tend to increase over time, suggesting that the impact of climate change on D. ponderosae suitability may also increase over time. Differences in EI for each grid between different decades and historical climate conditions (Figure 6b-i) show an apparent deepening trend with the progression of time, and most regions in the northern hemisphere become redder, suggesting that habitats become more climate-suitable for D. ponderosae, while most regions in the southern hemisphere become bluer, indicating a decline in climate suitability for D. ponderosae in these regions. In Africa, the suitability of southern regions decreases in a southerly direction, and northern areas in Algeria and Morocco become more suitable in future, leaving the large area inbetween mostly unchanged. In Asia, most habitats tend to become more suitable for D. ponderosae, especially in Russia and western China, where changes are greatest. By contrast, the suitability of habitats in northeastern Kazakhstan, India, and coastal states of southern Asia tend to decrease. In Europe, habitats broadly become more suitable, especially in the north. In North America, changes in Canada, Alaska, and western areas are positive, while negative changes are predicted for central

Potential Distribution of D. ponderosae under Future Climate Conditions
The projected potential future distributions of D. ponderosae in each decade (2021-2100) are shown in Figure 4, overlaid onto hosts distributions. The area proportions of different climate-suitable regions in each decade are shown in Figure 5. In general, the potential future distributions of D. ponderosae expand considerably following global warming, with extensive northward expansion in habitats in the northern hemisphere, while tropical and subtropical habitats tend to shrink. Overall, the potential future distribution is predicted to continue increasing from 2021 to 2100, mainly due to the areas increase in favorable and marginal regions over the decades, which are predicted to increase 1.87% and 2.14%, respectively. However, the area covered by favorable regions will likely decrease by 1.18% in total.
Changes in climatic suitability are shown in Figure 6. EI differences in mid-latitude regions (30-60 • S and 30-60 • N) are positive and peak at~60 • N, suggesting that climate change will have positive impacts on D. ponderosae, and the impacts will be greatest around 60 • N. Additionally, differences in EI between each decade in historical data tend to increase over time, suggesting that the impact of climate change on D. ponderosae suitability may also increase over time. Differences in EI for each grid between different decades and historical climate conditions (Figure 6b-i) show an apparent deepening trend with the progression of time, and most regions in the northern hemisphere become redder, suggesting that habitats become more climate-suitable for D. ponderosae, while most regions in the southern hemisphere become bluer, indicating a decline in climate suitability for D. ponderosae in these regions. In Africa, the suitability of southern regions decreases in a southerly direction, and northern areas in Algeria and Morocco become more suitable in future, leaving the large area in-between mostly unchanged. In Asia, most habitats tend to become more suitable for D. ponderosae, especially in Russia and western China, where changes are greatest. By contrast, the suitability of habitats in northeastern Kazakhstan, India, and coastal states of southern Asia tend to decrease. In Europe, habitats broadly become more suitable, especially in the north. In North America, changes in Canada, Alaska, and western areas are positive, while negative changes are predicted for central and southern regions. In South America, most areas tend to become more suitable, except in the east and west near the sea. In Oceania, positive changes occur in southeastern habitats, and negative changes occur in northeastern and southwestern habitats, while other areas remain unchanged. Overall, the climate suitability of temperate habitats tends to increase, while that of tropical habitats tends to decrease. and southern regions. In South America, most areas tend to become more suitable, except in the east and west near the sea. In Oceania, positive changes occur in southeastern habitats, and negative changes occur in northeastern and southwestern habitats, while other areas remain unchanged. Overall, the climate suitability of temperate habitats tends to increase, while that of tropical habitats

Central Point Shifts in Potential Distributions under Future Climate Conditions
Shifts in the central points of climate-unsuitable regions (0 ≤ EI < 10, red points) and climate-suitable regions (10 ≤ EI ≤ 100, blue points) over future decades (2021-2100) in each continent are shown in Figure 7, and the differences of directions and gradients of shifts among all continents are obvious. Regarding unsuitable regions (0 ≤ EI < 10), the directions of central point shifts are relatively stable in Asia (Figure 7b1 Figure 7 (a3-f3) shows the relative shift distances of central points in two directions (north-south and east-west) for each of the eight future decades compared with the historical central point . From the curves, we see that the largest range of central point shifts is in Asia (Figure 7b3), followed by North America (Figure 7d3), South America (Figure 7e3), and Africa (Figure 7a3), while the smallest range of shifts occurs in Oceania (Figure 7f3), reflecting differences in sensitivity to climate change in different continents. In addition, the shift distance of suitable regions (10 ≤ EI ≤ 100) in all continents is larger than that of unsuitable regions (0 ≤ EI < 10). Except for North America, shift distances of suitable regions in the north-south direction are consistently larger than those in the east-west direction (the shift distance in the north-south direction in North America is slightly smaller).

Discussion
Our predictions indicated that suitable regions in Canada were mainly distributed in the central and southern regions of the country, which were consistent with previous research by Carroll et al. (2006) and Sanfranyik et al. (2010) [9,29]. However, the distribution range for climate-suitable regions in our study appears to be wider, for we predicted the northern and western border to be further north and west, especially for very favorable regions, possibly due to differences in models and climate data used for simulation. Regarding applied SDMs, both previous studies employed the Sanfranyik model, which only considers the minimum requirements for temperature and precipitation to support D. ponderosae survival, neglecting the adverse effects of high temperatures and excessive precipitation. Additionally, the climate data we used to simulate future conditions were extracted from multi-model ensemble mean climate datasets based on CMIP5 (available at http://pcmdi9.llnl.gov/search/esgf-llnl), while the climate data in the previous studies were based on the CGCM1 general circulation model [59]. Thus, differences in the source of meteorological data may also contribute to differences in the results.
As a cold-intolerant pest [60], cold temperatures in winter are key constrains limiting the distribution of D. ponderosae, and overwintering success is critical for sustained pest outbreaks [61][62][63]. In our study, cold stress (CS) mainly occurred in northern North America (Northwest Territories and Nunavut in Canada), Greenland, and northeastern Asia (Mongolia and Russia). Accompanying climate change, the boundary for CS moves northward, with latitude increasing by 3 • -8 • in different regions, and the area for CS remains contraction constantly, especially in Canada and Russia (see Appendix B, Figure A1). With reduced CS limitations in the future, D. ponderosae will be more likely to cause outbreaks in severe cold regions and expand its range into higher latitudes (see Appendix B, Figure A2). We should therefore pay attention to the potential threat of climate change to ecological security in severe cold regions.
Our investigation of the distribution shift paths of D. ponderosae considered both unsuitable and suitable regions in each continent as homogeneous polygons to generate central points, ignoring differences in EI among the same polygons. The shift direction varied between continents, providing guidance for more accurate outbreak prediction of this pest. Shift distances were largest in Asia and smallest in Oceania, possibly due to the areas covered by these continents, since larger spans in latitude and longitude are more likely to be affected by changes of temperature and moisture. Moreover, the distance in the north-south direction was almost always larger than that in the east-west direction, indicating that temperature plays a major role in the distribution and expansion of D. ponderosae, consistent with the results of Carroll et al. (2006) [29] mentioned above. In further studies, we will consider assigning the corresponding weighting to different regions of the distribution according to EI values to generate heterogeneous polygons that better reflect suitability and to obtain more precise central points to allow complex image processing and calculations, although this approach may yield unexpected errors.
Our study provides theoretical guidance for early-warning intervention and risk assessment. The project results indicate that under both historical and future climate conditions, climate-suitable areas were not only distributed in North America but spread across all continents (except Antarctica), including northern South America, most areas of Europe, northwest and southeast Asia, southeast Africa, and southeast Oceania (Figure 3), where outbreaks of D. ponderosae have not yet been recorded. This may alert governments and agencies in these regions to the high risk of D. ponderosae invasion, and encourage strengthening quarantine and defense measures to prevent the spread of the pest in the international trade. Despite the fact that the project results declared the decreasing trend of climate suitability in large regions of North America, the risk of a massive outbreak of the pest still exists in the near future, and efforts for controlling and prevention should be further strengthened. Moreover, the predicted results under future climate conditions show that the potential distribution may change with climate change (Figure 4), with favorable regions expanding in range toward higher latitude, especially in the northern hemisphere. Therefore, relevant defense measures and responses to the possible harm caused by D. ponderosae invasion should be taken in advance.
SDMs are based on the assumption that species are at equilibrium with their environments, but it may involve species with unrepresentative records of new conditions and prediction to novel environments when using in non-equilibrium settings (e.g., invasions and climate change). Several problems with SDMs have been identified, including different (combinations of) environmental factors that may limit distributions or biotic interactions that may change substantially in the new context. Besides, genetic variability, phenotypic plasticity, and evolutionary changes may also alter the results [64][65][66]. Nevertheless, the use of correlative models currently remains one of few practical approaches for predicting or hindcasting distributions [17]. The CLIMEX model used in the present study aims to capture the core features of species climatic requirements from minimal observations, and results are known to be relatively accurate and applied widely to explore the distribution of plants, animals, and microorganisms, especially in the absence of sufficient experimental data [28]. However, the model may have some restrictions because it only uses a single set of parameters for all regions and ignores any other non-climatic constraints that may alter the distribution, such as topography, predators, and human activities. For D. ponderosae, its developmental rate in southern populations (i.e., warmer habitats) was slower than that in more northern populations (i.e., cooler habitats), which may be an evolutionary adaptation to ensure seasonality, and it may lead to a significant underestimation in potential range expansion of the northern limits of this beetle because the actual rate of range expansion may be greater than predicted [67]. Furthermore, the insufficient occurrence data and incorrect parameter selection may also lead to some problems of the model [68]. When considering the effects of host trees, we identified all Pinus species as possible hosts of D. ponderosae conservatively, but actually they presented the difference in resistance of being attack to D. ponderosae. It is reported that Great Basin bristlecone and foxtail pines have relatively high levels of constitutive defenses which make them less vulnerable to climate-driven D. ponderosae range expansion relative to other high elevation pines [69,70]. Additionally, the defense of pines in recently invaded areas are predicted to be lower than in areas with longer-term D. ponderosae presence, for lacking targeted defense traits [71]. Therefore, it may be of great value to explore temporal and spatial changes in parameters and insect responses to different hosts to improve the accuracy of projections.

Conclusions
Climate change can significantly affect the potential distribution of D. ponderosae, and the impacts on the suitability of the pest will vary with latitude. With climate change, the suitability will increase in middle and high latitude regions and decrease in the low latitude regions, while the most sensitive regions to climate change located in the mid-latitude zone. Moreover, the shift directions and ranges of climate suitable regions under future condition will differ among continents, which is likely mainly affected by temperature. The projections of potential distribution can provide theoretical guidance for early-warning techniques and risk assessment, thus reducing the economic and ecological loss caused by the pest.