Predicting the potential geographic distribution of Thrips palmi in Korea, using the CLIMEX model

The CLIMEX model

CLIMEX is a dynamic model that integrates the weekly responses of a population to climate into a series of annual indices. It assumes that the population of an organism increases during favorable seasons, and decreases during unfavorable seasons. In CLIMEX, the available temperature, moisture, and day length data are combined into a weekly population Growth Index (GI_W) for target species. Annual temperature (TI) and moisture (MI) indices integrate the target species' response to general temperature and moisture conditions, respectively. Responses to extreme conditions are incorporated into the model as a series of stress indices (SI) that estimate the threat posed to that species by extreme or prolonged cold, hot, dry, or wet weather. In addition, restrictions on completion of the life cycle due to prolonged periods of an inappropriate heat summation or day length system can be estimated where data are available. Finally, the growth and stress indices are integrated into an Ecoclimatic Index (EI), scaled from 0 (no persistence) to 100 (maximal population size in relation to climate alone), to represent the overall suitability of the given geographical location for the propagation and persistence of the species. This indicates that the natural occurrence is only possible when EI goes above zero. The EI of CLIMEX simulations are used to represent suitability and establishment of target species' geographical distribution, not only for insect populations, but also for mollusks, plants, microorganisms and mammals (Cooke 1992; Brasier 1996; Baker 1998; Bennett et al. 1998). As a guide to the potential for establishment of populations in target locations, EI values are categorized as follows: locations where EI > 25 are very favorable for population growth and persistence; 10–25 are favorable; and < 10 have low to marginal suitability (Sutherst & Maywald 1985; Sutherst et al. 1995, 1999, 2000, 2007; Vera et al. 2002).

The potential geographic distribution of T. palmi in Korea was predicted by using the “Compare Locations (one species) model” of CLIMEX (version 3.0) (Sutherst et al. 2007). CLIMEX parameter values describing the response of T. palmi to climate conditions were estimated via interpretation from its occurrence in its native country, using climate data from locations where T. palmi is recognized (Hong et al. 1998, Dentener et al. 2002, Vera et al. 2002; Cannon et al. 2007; Sutherst et al. 2007; Strauss 2010; EPPO 2013). Information about the biological growth factors of T. palmi was obtained from the scientific literature (Dentener et al. 2002; Park et al. 2010). The EI values were imported to a map of Korea to create thematic maps, prepared with 12.5 km resolution from the CLIMEX results. We interpolated the raster surface from the EI of each year, using an inverse distance weighted (IDW) technique (Spatial F Analyst) in the ArcGIS software (Sutherst et al. 2007; Ormsby et al. 2010).

Climate data

Two meteorological databases corresponding to 70 locations in South Korea were created for CLIMEX distribution modeling. CLIMEX needs monthly long-term average maximum and minimum temperatures, rainfall and relative humidity at 09:00 and 15:00 h (Sutherst et al. 2007). An initial meteorological database was generated based on past climate data from 2000 to 2010. A second database was created, based on RCP 8.5 climate change scenario from 2015 to 2100 in Korea from the Korea Meteorological Administration (KMA). The RCPs consist of greenhouse gas concentrations and emissions pathways, which are developed for climate modeling on impacts and potential policy responses to climate change (Moss et al. 2008, 2010; Van Vuuren et al. 2011). The RCP 8.5 is a scenario of comparatively high greenhouse gas emissions, combining assumptions about high population and relatively slow income growth with modest rates of technological change and energy intensity improvements. Thus, the RCP 8.5 corresponds to a high greenhouse gas emissions pathway, and hence, also to the upper bound of the RCPs (Riahi et al. 2011).

Both meteorological databases were arranged at 5-year intervals with 12.5 km resolutions for 70 locations over the entire terrestrial area of South Korea (Table 1; Fig. 1). All climatic data are manipulated by the MetManager program (Hearne Scientific Software), which is directly accessible to the CLIMEX simulation program (Sutherst et al. 2007).

Present distribution of T. palmi in Korea

A map of the present geographic distribution of T. palmi in Korea was generated using data from the literature presented by Hong et al. (1998) and 12 years (from 2000 to 2012) of field observation data (Lee, GW, unpublished personal data). The data from the literature contained results of a two-year comprehensive study (1995–1996) on the geographic distribution of T. palmi conducted soon after the first identification of this species in Korea.

Parameters in CLIMEX were calibrated according to the current distribution in the native range of this species from the CABI database (CABI/EPPO 1998). The current distribution of this species in Korea was used to test whether the model parameters were well calibrated.

Parameter fitting

The values of the CLIMEX model parameters, which reflect a species' climatic requirements, are inferred from information on the species' known geographic distribution, relative abundance, and seasonal phenology. Thus, CLIMEX model parameters should be acquired not only from previous published data analyses, but manually adjusted until the modeled geographic distribution matches the observed distribution as closely as possible (Vera et al. 2002; Sutherst et al. 2007). The parameters for the CLIMEX model of T. palmi were acquired from Dentener et al. (2002). However, CLIMEX simulation results using parameters presented by Dentener et al. (2002) failed to match either the present geographic distribution of T. palmi population in Korea or in the worldwide distribution (Hong et al. 1998; Cannon et al. 2007; EPPO 2013). Thus, in this study, the results of a recent temperature-dependent study in Korea, field observations in the winter season and current worldwide geographic distribution of T. palmi were included and analyzed to adjust the parameters in an iterative manner until the CLIMEX model closely matched the known distribution in Korea and worldwide (McDonald et al. 2000; Cannon et al. 2007; Sutherst et al. 2007; Park et al. 2010; EPPO 2013).

The “Compare Locations” simulation option in the CLIMEX model system readily identifies field sites as marginal to optimal and represent the range of conditions experienced by the species. This enables the identification of sites that are marginal for different reasons (i.e., weather stresses; cold, hot, dry and wet). T. palmi is essentially a tropical species and is mainly restricted in its establishment and geographic distribution by temperature parameters especially cold stresses (McDonald et al. 2000; Dentener et al. 2002; Cannon et al. 2007; Sutherst et al. 2007). The temperature at which there is a 50% mortality level of T. palmi is rapidly increasing below 0°C in field conditions (McDonald et al. 2000). Field lethal time data of T. palmi were inverse to death rate. The zero death rate temperature was then projected linearly as 4.4°C for T. palmi (McDonald et al. 2000). Thus, the cold stress parameter was set with an average temperature at 4.4°C for the CLIMEX simulation. Other temperature parameters were set by linear trend estimation of the results of a recent temperature-dependent experiment, which showed a non-linear relationship between temperature and development of T. palmi (Sutherst et al. 2007; Park et al. 2010).

Dry and wet stress threshold parameters, including stress rates, were adjusted in an iterative manner using parameters from Dentener et al. (2002), observed weather data (2000–2012) from KMA, and pre-loaded meteorological data in the CLIMEX program, and compared with recent distribution data (Cannon et al. 2007; Sutherst et al. 2007; EPPO 2013). Cold-wet and cold-dry stress indices were not included in this CLIMEX simulation due to disagreement between simulation results and actual geographic distribution of T. palmi in Korea. Detailed parameters and their literature sources are summarized in Table 2.

Present distribution of T. palmi in Korea

According to the literature, the geographic distribution of T. palmi is confined to the southern coastal areas (the northernmost latitude is 35°54′N), and includes Jeju Island, the southernmost territory in Korea (Hong et al. 1998; opened squares in Fig. 3). A similar geographic distribution pattern was found in field observations conducted during 2000–2012, where the northernmost latitude was 36°16′N (closed squares in Fig. 3). This result indicates that geographic range of T. palmi has expanded approximately 45 km northward since its first introduction in Korea.

Modeled distribution of T. palmi using past meteorological data

Based on the CLIMEX simulation using past meteorological data (1995–2010) with calibrated model parameters, the simulated distribution of T. palmi is consistent with the known range of its natural distributions in South Korea (Figs 3, 4), indicating that the CLIMEX model can explain the present geographic distribution of T. palmi reliably. This result validates the use of CLIMEX to describe the ecological requirements of T. palmi and future potential geographic distribution in Korea.

The predicted potential distribution during 2000–2010 is mainly along the southern (EI value from 8 to 49) and eastern (ranging from 3 to 25) areas. The distribution in the western coastal area expands as simulation years progress: EI values range from 2–22 in 2000 and from 12–31 in 2010 (Figs 3, 4). For example, in 2000 and 2005, Jeonju (No. 43 in Fig. 1, Table 1) and Boryeong (No. 34 in Fig. 1, Table 1), were marginally suitable for the distribution of T. Palmi (i.e. Eis < 10; Fig. 4). Meanwhile, in 2010, EI values indicate that a northern district of geographic distribution of T. palmi changed from marginal to favorable (10 < EI < 25) and most of the southern area turned very favorable for T. palmi population growth and persistence (EI > 25). Although the discrepancy between the simulation and field observations appears to be large in the eastern coastal area, it is probably related to the distribution of host crop plants for T. palmi, and not related to the reliability of the CLIMEX model. A moderated agricultural cultivation practice is performed in the eastern coastal area because high mountains (Baekdu-Daegan) are located along the coastal lines (Fig. 3).

Modeled distribution of T. palmi using RCP8.5 climate change scenario

After the CLIMEX parameters were fitted under the past climate average (Table 2), the RCP 8.5 scenario was chosen to reflect the range of possible future climatic conditions from 2020 to 2100. The potential distribution areas estimated to be climatically suitable for T. palmi are illustrated in Figure 5. The EI values in each graphic panel represent means of 10-year EI values at each location.

Between the years 2020 and 2030, the potential geographic distributions of T. palmi remain very similar to the present distribution, with a similar range of EI values (see Figs 3, 4). All the western and southern coastal areas are favorable to very favorable for growth and persistence of T. palmi. However, the distribution in most of the eastern area is greatly impeded and locations with EI values greater than 15 are found in only two locations (Nos. 39 and 48). After 2030, the potential distribution expands rapidly to the entire central inland region with a slight retreating and expanding pattern. In 2080 all the EI values calculated appear to be greater than 25, indicating that the entire Korean Peninsula is potentially suitable for permanent establishment of T. palmi.