Hostname: page-component-78c5997874-v9fdk Total loading time: 0 Render date: 2024-11-05T06:57:24.216Z Has data issue: false hasContentIssue false

Predicting the geographical potential distribution of species Opisina arenosella Walker in China under different climate scenarios based on the MaxEnt model

Published online by Cambridge University Press:  02 October 2024

Zhiling Wang
Affiliation:
College of Life Science, China West Normal University, Nanchong, China
Zhihang Zhuo
Affiliation:
College of Life Science, China West Normal University, Nanchong, China
Yaqin Peng
Affiliation:
College of Life Science, China West Normal University, Nanchong, China
Danping Xu*
Affiliation:
College of Life Science, China West Normal University, Nanchong, China
*
Corresponding author: Danping Xu; Email: [email protected]
Rights & Permissions [Opens in a new window]

Abstract

As global warming increases with the frequency of extreme weather, the distribution of species is inevitably affected. Among them, highly damaging invasive species are of particular concern. Being able to effectively predict the geographic distribution of invasive species and future distribution trends is a key entry point for their control. Opisina arenosella Walker is an invasive species, and its ability to live on the backs of foliage and generate canals to hide adds to the difficulty of control. In this paper, the current and future distributions of O. arenosella under three typical emission scenarios in 2050 and 2090 are projected based on the MaxEnt model combining 19 bioclimatic variables. Filter through the variables to find the four key environment variables: BIO 1, BIO 6, BIO 11 and BIO 4. The results show that O. arenosella is distributed only in the eight provinces of Tibet, Yunnan, Fujian, Guangxi, Taiwan, Guangdong, Hong Kong and Hainan in the southeastern region. Its high suitability area is concentrated in Taiwan and Hainan. In the long run, highly suitable areas will continue to increase in size, while moderately suitable areas and poorly suitable areas will decrease to varying degrees. This paper aims to provide theoretical references for the control of O. arenosella.

Type
Research Paper
Copyright
Copyright © The Author(s), 2024. Published by Cambridge University Press

Introduction

Opisina arenosella Walker (Lepidoptera: Oecophoridae) is also known as the coconut black-headed caterpillar (BHC). BHC is an essential quarantine leaf-feeding pest native to Sri Lanka and India in South Asia (Lu et al., Reference Lu, Lyu, Tang, Wu, Wyckhuys, Le, Chongchitmate, Qiu and Zhang2023). BHC was first detected and recorded in 1909 in the coastal cities of southern India, and since 1920 it has been reported as an economically relevant pest in Sri Lanka, Hindustan, Bengal Republic, Indonesia, Thailand and Malaysia (Perera et al., Reference Perera, Hassell and Godfray2009). By August 2013, BHC damage was first detected in China in Wanning City, Hainan Province. Subsequently, BHC damage to palms was also reported in Fangchenggang City, Guangxi Province and Foshan City, Guangdong Province (Tang, Reference Tang2017). According to current research, BHC can harm about 34 species of palms (Li et al., Reference Li, Qingzhao, Xiongfei, Guangchang, Bo, Fan and Wei2021) such as coconut (Cocos nucifera), royal palm (Roystonea regia (HBK.) O.F. Cook), Chinese fan palm (Livistona chinensis), date palm (Phoenix dactylifera L.), Oredoxa oleracea Kurth., Metroxylon sagu Rottboell., Caryota urens L., Hyphaene thebacia L., Elaeis guineensis Jacq. and others (Jin et al., Reference Jin, Lin, Han, Ma, Wen and Peng2019). In Hainan Province, China, it mainly includes 21 species of plants that are affected by this pest, such as R. regia, Phoenix sylvestris, Wodyetia bifurcata, L. chinensi, Butia capitat, Hyophore lagenicauli, Washingtonia robusta, Latania lontaroide, Prithchardia pacifica, Bismarckia nobilis, Corypha umbaculifera, Chrysalidocarpus lutescen, Arenga catechu, Archontophoenix alexandrae, Latania verschaffeltii, Arenga pinnat, Licuala grandis, Borassus flabellife, Saccharum sinensis and Musa paradisiaca. However, it is the palms that it mainly harms (Tang, Reference Tang2017). Regarding the BHC, the damage area is mainly concentrated on the leaves of the plant. The females lay their eggs on the underside of the leaf blades and the eggs successfully hatch into larvae, which hide on the underside. The larvae feed on the thin-walled tissues of the leaves and leave behind leaf debris and excreta, and use their excreta and spider threads on the underside of the coconut leaves to construct galleries (Kumar, Reference Kumar2002; Perera et al., Reference Perera, Hassell and Godfray2009). This causes mechanical damage to the leaves, while shading by leaf litter and excreta causes a reduction in photosynthetic area (Nasser and Abdurahiman, Reference Nasser, Abdurahiman, Upadhyay, Mukerji and Chamola2001). Not only that, the insect will damage the leaves 2–3 days later, and the leaves will automatically roll down and fold closed, providing a hiding place for the insect (Li et al., Reference Li, Qingzhao, Xiongfei, Guangchang, Bo, Fan and Wei2021). Also, because the BHC feeds only in the canopy, often at heights of 10 m or more, it is difficult to detect in the early stages of the infestation (Lu et al., Reference Lu, Lyu, Tang, Wu, Wyckhuys, Le, Chongchitmate, Qiu and Zhang2023). It is worth mentioning that only the upper epidermis is left after feeding by the larvae, which, together with the fact that it is a group feeding, results in a large damaged area that externally looks burned. This not only affects the appearance but also leads to a reduction in fruit yield, even early fruit drop and slow growth (Lever, Reference Lever1969; Mao and Kunimi, Reference Mao and Kunimi1994). For coconuts in particular, coconut yields are reduced by 45.4% in coconuts infested with the insect, and 13.8–21% of coconut leaves are damaged. Studies have shown that, in cases of severe damage, it can take up to 4 years to recover normal yields (Mohan et al., Reference Mohan, Nampoothiri, Radhakrishnan Nair and Rajan2010). With regard to the spread of the BHC, in addition to its own superior flight ability (at night), the trade transportation of the host plant (palm) and the development of the coconut chain are the main reasons for it (Jingjing et al., Reference Jingjing, Siming, Jiaqiang, Juan and Te2023). With regard to the control of the BHC, tall palms make it difficult to spray pesticides, as the leaves are not easy to spray comprehensively, and it also increases labour costs. Also, due to the pest's characteristic of hiding behind the leaves to feed, most of the pesticides fall on the front side of the leaves and fall by gravity. This makes spraying difficult and at the same time, greatly reduces the contact rate between the BHC and the pesticide (Jin et al., Reference Jin, Lin, Han, Ma, Wen and Peng2019). Nowadays, all kinds of science and technology are developed interactively, especially the combination of the distribution of various types of pests with modelling software and geographic software. It can predict the future distribution trend and geographic direction of pests, which plays a good role in pointing out control work.

The description and understanding of a species are inextricably linked to its distribution and drivers, which are the basis of both ecology and geography (Barker and MacIsaac, Reference Barker and MacIsaac2022). Species distribution models (SDMs) are critical modelling tools for studying the direction of species' geographic distributions, linking and combining species occurrence data with environmental predictors and utilizing them in a variety of ways (Guisan and Zimmermann, Reference Guisan and Zimmermann2000). The results derived from SDM can be used to study and excavate the target species in depth in terms of response to the environment, the most critical environmental factors, the predicted probability of occurrence of the species in time and space and the predicted presence or absence of the species. From the above aspects, hypotheses related to determining ecological conditions for the distribution of target species can be formulated (Bradie and Leung, Reference Bradie and Leung2017). With the rapid development of SDM, many more methods have been derived, including CART (Breiman et al., Reference Breiman, Friedman, Olshen and Stone1984), MARS (Friedman, Reference Friedman1991), GARP (Stockwell, Reference Stockwell1999), GLMS (Guisan et al., Reference Guisan, Edwards and Hastie2002) and GAMS (Guisan et al., Reference Guisan, Edwards and Hastie2002). In addition to this, it is worth mentioning that the most noteworthy algorithms for machine learning include random forest (Breiman, Reference Breiman2001), artificial neural network (Olden et al., Reference Olden, Lawler and Poff2008), gradient boosting (De'Ath, Reference De'Ath2007), support vector machine (Guo et al., Reference Guo, Kelly and Graham2005) and maximum entropy (MaxEnt) (Phillips et al., Reference Phillips, Dudík and Schapire2004). In this context, MaxEnt has been widely used for species prediction and is one of the most popular methods in climate modelling studies (Elith et al., Reference Elith, Graham, Anderson, Dudík, Ferrier, Guisan, Hijmans, Huettmann, Leathwick and Lehmann2006). It can produce relatively robust results with only a small number of presence/absence (or pseudo-absence) samples (Guisan et al., Reference Guisan, Zimmermann, Elith, Graham, Phillips and Peterson2007; Elith et al., Reference Elith, Phillips, Hastie, Dudík, Chee and Yates2011). MaxEnt distinguishes between the presence and absence of environmental conditions for the classification of the target species through maximum entropy (Elith et al., Reference Elith, Phillips, Hastie, Dudík, Chee and Yates2011), and then finds the probability distribution of maximum entropy under a set of constraints based on the occurrence data of the target species (Phillips et al., Reference Phillips, Anderson and Schapire2006). As a purely modelling algorithm, it can be integrated with predictive variables such as climate and remote-sensing variables. These data are used to predict the relative occurrence rate of a target species in a designated landscape (Rhoden et al., Reference Rhoden, Peterman and Taylor2017).

Therefore, in this study, MaxEnt modelling was used to combine the occurrence data of the target species, BHC, with 19 bioclimatic variables to predict the habitat areas in China. This work was able to obtain the most important environmental variables affecting its distribution, which in turn allowed us to predict the geographic distribution of the pest under three climate scenarios for the next two periods. This provides some ideas for the control of BHC in the future, both in terms of climate and geographic distribution.

Materials and methods

Sources and processing of BHC occurrence data

In this study, data on the occurrence of BHC in China and other countries were selected. These data were mainly obtained by searching websites on the Internet, published related literature, and newspapers and books. The website is powered by the European and Mediterranean Plant Protection Organization (EPPO, https://gd.eppo.int) and Global Biodiversity Information Network (GBIF, http://www.gbif.org/) databases for BHC occurrence data. The rest of the data were obtained through CNKI and Web of Science in the relevant literature. Of these avenues, data for which precise latitude and longitude were available were recorded, and those that provided only geographic location names were obtained through Google Earth (http://ditu.google.cn), using county location descriptions for latitude and longitude. Then, all the points are filtered, duplicates and errors are removed and they are carefully checked. Finally, the geographic coordinates (latitude and longitude) were converted into an Excel spreadsheet according to the requirements of the MaxEnt model and then saved in ‘CSV’ format. In order to avoid overfitting, ENMTools version 1.0.4 of the R platform was used to spatially filter the data according to longitude, retaining only one point in each grid cell (5 × 5 km). Finally, a total of 106 BHC distribution points were obtained.

Acquisition and processing of environment variable data

MaxEnt version 3.4.4 and ArcGIS version 10.4.1 were used in this study. The input MaxEnt data required environmental variable data in addition to the geographic coordinates (latitude and longitude) of the point of occurrence. Environmental data were obtained from WorldClim (http://www.Worldclim.org), containing 19 bioclimatic variables and 48 monthly climatic variables, and the data were in ASC format for use in ArcGIS software. Current climate data (2020s) and future climate data (2050s and 2090s) were obtained from the World Climate Data website (https://www.worldclimatedata.org/), and the spatial resolution of the above data is 2.5 arc-minutes (about 4.5 km2). The Sixth International Coupled Model Comparison Program (CMIP6) model proposes several shared socio-economic pathways (SSPs) designed based on different socio-economic assumptions, and in this study, three SSP emission scenarios based on the simulation of the CanESM 5 model were selected (2.6 W m−2 (SSP1-2.6), 4.5 W m−2 (SSP2-4.5) and 8.5 W m−2 (SSP5-8.5)) under the climate data.

Regarding the way these data were handled, the first step was to take these environmental variables and use the Jackknife test in MaxEnt to determine the extent to which each environmental variable contributes to the construction of the model and to exclude variables with a contribution of 0. The first step was to compare the environmental variables with their Pearson correlation coefficients. To avoid overfitting and improve model accuracy, the remaining variables were compared for their Pearson correlation coefficients and the two environmental variables with absolute values greater than 0.8 were removed. This is because it indicates a strong linear relationship between them. Finally, the remaining environmental variables were used to model predictions of current and future fitness ranges.

MaxEnt modelling methodology and accuracy assessment

The BHC geolocation information converted to CSV format was entered into the MaxEnt software along with the environmental variable data after harmonizing the projection and resolution. Seventy-five per cent of the distribution records were used as a randomised training dataset to build the predictive model, and 25% of the remaining distribution records were used as a test dataset. The number of iterations was then chosen as 1000, the number of background points was chosen as 10,000 and the model was repeated ten times for averaging using the cross-validation method. Finally, this work plots the receiver operating characteristic curve (ROC) and area under curve (AUC) to evaluate the model fit. The ROC curve is a graphical tool used to evaluate the performance of a binary classification model. The AUC takes values ranging from 0 to 1, with values of 0–0.6 indicating extremely poor predictive performance; values between 0.7 and 0.8 indicate fair predictive performance; values between 0.8 and 0.9 indicate good predictive performance; and values between 0.9 and 1.0 indicate excellent predictive performance. Overall, the closer the value is to 1, the better the model fits (Wang et al., Reference Wang, Xu, Liao, Xu and Zhuo2023).

Criteria for delineating BHC potential habitat areas

The MaxEnt model result file was imported into the ArcGIS software and reclassified by the software's reclassification function so as to derive the possible geographic distribution areas of BHC in China. The distribution values were classified according to the method of assessing probability in the IPCC report, and the suitability areas were categorised into four classes based on the suitability index P: high suitability areas (P ≥ 0.66), medium suitability areas (0.33 ≤ P < 0.66), low suitability areas (0.05 ≤ P < 0.33) and unsuitable areas (P < 0.05) (Wang et al., Reference Wang, Xu, Liao, Xu and Zhuo2023). The number of rasters in each area was then calculated to obtain the area of suitable habitat in each category for different time periods and climate scenarios. The comparative data and calculated differences were then graphed for more visual analysis.

Results

MaxEnt modelling and accuracy testing

In this work, the MaxEnt model successfully predicted the current and future suitable habitats for BHC. The model ran for a total of ten iterations, achieving an AUC of 0.993. The training AUC and test AUC were 0.9936 and 0.9931, respectively, with a standard deviation of 0.006. The value then performs excellently according to the criterion of the AUC mentioned in 2.3 (fig. 1). This indicates that the prediction results of the model have a very satisfactory level of reliability. It can scientifically simulate and reflect the geographical distribution of BHC in China, thus providing a basis for further research.

Figure 1. Receiver operating characteristic curve and AUC result of MaxEnt modelling.

Key environmental factors affecting the geographic distribution of BHC

In MaxEnt's modelling, for key environmental factors, the mean percentage contribution (PC) is usually used to describe their importance. According to the screening method mentioned in section ‘Acquisition and processing of environment variable data’, screening was performed using the Jackknife test and Pearson correlation coefficient. Table 1 presents the PC values of each environmental factor after screening. Among them are the values of four environmental variables, BIO 1 (annual mean temperature), BIO 6 (minimum temperature of the coldest month (°C)), BIO 11 (mean temperature of the coldest quarter (°C)) and BIO 4 (temperature seasonality (standard deviation × 100)), which are four environmental variables with values of 88, 4.9, 3.7 and 3.4%, respectively. It can be seen that they are all closely related to temperature. It is interesting to note that BIO 1 has the highest PC value, 88% for this one alone. This is a huge difference from the other three environmental factors. Figure 2 shows the Jackknife test plot for the MaxEnt model. When used independently, the gain values provided by BIO 1, BIO 11 and BIO 6 all show high values, and only BIO 4 has a large gap in value. It can be concluded that the geographic distribution of BHC in China is largely influenced by these four temperature-related factors, especially BIO 1.

Table 1. Per cent contribution (%) and permutation importance (%) of environmental variables in predicting the occurrence of O. arenosella in the MaxEnt model.

Figure 2. Importance of environmental variables to O. arenosella by jackknife test.

Analysis of response curves for environmental factors

Based on the identification of key environmental factors, this work further investigates them. Response curves were used for each key environmental factor to evaluate its relative impact on MaxEnt predictions. Figure 3 shows the response plots of the MaxEnt model with respect to the key environmental factors. These curves give a good picture of how the different key variables affect the course of the probability of BHC existence while maintaining the average sample values of all other variables. The values of the response curves in the figure are averaged over ten repeated runs of MaxEnt. The response curve of BIO 1 (annual mean temperature) can be seen in fig. 3A. The curve starts to show an upward trend when the temperature rises to about 17.0°C. Above 20.0°C the curve rises sharply until it reaches about 25.0°C, where it remains steady and its output value stabilises at about 0.96. The output value is more informative when it is greater than 0.6. That is to say, when the temperature is greater than 23.6°C and less than 25.0°C, the interval is suitable for the survival of BHC. Moreover, its survival probability is positively correlated with the increase in temperature in this interval, and the survival of the insect is most suitable when the annual mean temperature is 25.0°C. For BIO 4 (temperature seasonality (standard deviation × 100)) (fig. 3B), after the temperature seasonality value of 300, the output value starts to drop off a cliff. Until the temperature seasonality value stabilises around 750, the output value will be 0. This indicates that the temperature seasonality value between 300 and 400 is the most suitable for the survival of the insect, but with the rise of this value, there is a negative correlation with the output value. However, minimum temperature of the coldest month (fig. 3C) is different. As can be seen from the figure, the whole curve shows an increasing trend, with a positive correlation between the higher temperature and the higher output value. The temperature starts to curve up from 5.0°C and reaches a plateau around 17.5°C. When the temperature rises to 13.6°C, the output value is 0.6. It shows that minimum temperature of the coldest month is suitable for the survival of this worm in the interval of 13.6–17.5°C. Similarly, BIO 11 (mean temperature of the coldest quarter) (fig. 3D) showed a positive correlation between the increase in temperature and output values. The interval between 10.0 and 21.8°C was suitable for the survival of BHC.

Figure 3. Response curves of the environmental variables that contributed most to the MaxEnt models. (A) Annual mean temperature (Bio 1); (B) temperature seasonality (standard deviation × 100) (Bio 4); (C) coldest month minimum temperature (Bio 6); (D) mean temperature of coldest quarter (°C) (Bio 11).

Projections of BHC suitable habitat (current/future)

Figures 4 and 5 show the contemporary and future distribution of suitable habitat at BHC, respectively. The colouring patterns are mainly in red, orange, yellow and white to indicate highly suitable areas, moderately suitable areas, poorly suitable areas and unsuitable areas, respectively. In general, at any given time, the highly suitable area is mainly found in the island provinces of Hainan and Taiwan. Almost the entire island of Hainan Province is covered by highly suitable areas, especially the sea-rimmed areas, except for a small amount of moderately suitable and poorly suitable areas, of which the poorly suitable area is the smallest.

Figure 4. Current suitable climatic distribution of O. arenosella in China. The probability of O. arenosella is shown by the colour scale in the area. Red indicates a highly suitable area with a probability of higher than 0.66, orange indicates a moderately suitable area with a probability of 0.33–0.66, yellow indicates a poorly suitable area with a probability ranging from 0.05 to 0.33, and white represents unsuitable areas.

Figure 5. Potential distribution of suitable areas of O. arenosella under different climate change scenarios in China. The probability of O. arenosella is shown by the colour scale in the area. Red indicates a highly suitable area with a probability of higher than 0.66, orange indicates a moderately suitable area with a probability of 0.33–0.66, yellow indicates a poorly suitable area with a probability ranging from 0.05 to 0.33, and white represents unsuitable areas.

In terms of the contemporary area of suitable distribution areas, table 2 shows us the area of different suitable areas and the percentage of each province. Table 2 shows that not many provinces are suitable for BHC distribution, mainly Tibet, Yunnan, Fujian, Guangxi, Guangdong, Hong Kong, Taiwan and Hainan, of which only a small amount of moderately suitable area, poorly suitable area and poorly suitable area exist in Tibet, Yunnan and Guangdong, while only a small amount of moderately suitable area, poorly suitable area and poorly suitable area exist in Fujian, Guangxi and Hong Kong. Among them, only a few moderately suitable and poorly suitable areas exist in Tibet, Yunnan and Guangdong, while only a few poorly suitable areas exist in Fujian, Guangxi and Hong Kong. The highly suitable areas coincide with those presented in fig. 4, with highly suitable areas existing only in Hainan and Taiwan Provinces. The percentages of highly suitable areas in Hainan and Taiwan Provinces were 59.6 and 5.8%, respectively, of the province, with the highest percentage in Taiwan Province. In terms of the future size of the suitable distribution area, table 3 shows the distribution of the area for two future periods and three climate scenarios, as well as the increase or decrease in comparison with contemporary data. Notably, the highly suitable area of BHC showed an increase under all three climate scenarios in the 2050s (SSP1-2.6, SSP2-4.5 and SSP5-8.5), while the other two suitable areas showed an overall decrease in area. Interestingly, the overall trend was consistent in the 2090s compared to the 2050s, although the area did not increase or decrease as consistently across climate scenarios as it did in the 2050s. Of these, the highly suitable area has the largest increase in area under the 2090s SSP5-8.5 scenario, with an increase of 24.03%. The largest decrease in area was in moderately suitable area under the 2090s SSP2-4.5 scenario, with a decrease of 49.40%. Overall, the highly suitable area will continue to increase in size in the future, while the moderately suitable area and poorly suitable area will decrease to varying degrees.

Table 2. Analysis of main suitable distributions of O. arenosella in China.

a Indicates the total area of the corresponding province.

Table 3. Prediction of the suitable areas for O. arenosella under current and future climatic conditions.

Discussion

In today's scientific research, SDM has been applied in a variety of scientific fields, such as ecology, biology and geography (Elith and Leathwick, Reference Elith and Leathwick2009). It provides a key approach to the control of harmful invasive species (Bowen and Stevens, Reference Bowen and Stevens2020) and the conservation of rare and endangered species (Escalera et al., Reference Escalera, Hernandez, Soto-Rojas and Suazo-Ortuño2018), and can also be used to predict potential impacts on target species by assessing changes in climate (Nneji et al., Reference Nneji, Salako, Oladipo, Ayoola, Onadeko, Adedeji, Omotoso, Ugwumba and Adeola2019). Most SDM methods can be used to form statistical relationships between species occurrence data and environmental variables and for prediction. However, in this study, the MaxEnt model was selected as a prediction tool for the invasive species BHC among many SDMs. MaxEnt is an integration of environmental variables as a simulation of the general environment, and the presence and location of the target species can be predicted using very little species occurrence data. It includes the density probability of presence location and further derives the proportion of area present in different provinces and survival suitability (Elith et al., Reference Elith, Phillips, Hastie, Dudík, Chee and Yates2011). Because of these superiorities of MaxEnt as a predictive model, the model was chosen to predict its geographic distribution for BHC. However, all models have certain shortcomings based on their strengths, and MaxEnt is no exception, with certain recognised performance limitations. For example, in terms of background samples, although a small sample size is sufficient to support the completion of the modelling, the representativeness of the sample selection affects the prediction results. But the model's currently known limitations can be mitigated relatively well. For example, screening for environmental factors effectively avoids the question of whether the background sample is representative. For example, for missing species occurrence data, MaxEnt can be modelled with easily accessible presence-only datasets (Elith et al., Reference Elith, Phillips, Hastie, Dudík, Chee and Yates2011).

The predicted results for BHC show that temperature is an important factor influencing the distribution of this pest. The key four environmental factors (BIO 1, BIO 6, BIO 11 and BIO 4) were all related to temperature, with BIO 1 (Annual Mean Temperature) being the most influential factor. There is an effect of climate change on the population dynamics of BHC in the field (Ramkumar et al., Reference Ramkumar, Muralimohan, Kiranmayi and Srinivasa2006; Shameer et al., Reference Shameer, Nasser, Mohan and Hardy2018). It has been shown that lower temperatures can have a negative effect on population growth. At lower temperatures, the spawning period of BHCs is prolonged, which reduces population growth (Muralimohan et al., Reference Muralimohan, Srinivasa and Pn2013). This happens to also corroborate the response curve for the most influential BIO 1 environmental factor (PC = 88%), i.e. temperature is positively correlated with the probability of presence when the average temperature of the year is in the interval 23.6–25.0°C.

In summary, there are not many highly suitable areas for BHC in China, nor is there a trend of rapid expansion. The suitable areas are mainly located in the cities along the southeast coast, with the highly suitable areas existing mainly in Hainan Province and Taiwan Province. It is noteworthy that the whole province of Hainan and the southern part of Taiwan Province have a tropical monsoon climate. This coincides with the distribution of high suitability areas of BHC in these two provinces. Regarding areas with a tropical monsoon climate, the temperature is mainly characterised by high temperatures throughout the year, which again confirms the previous point. For invasive pests, early detection is an important method for effective containment, eradication and control of target species. Therefore, the prediction work in this study provides directions for prevention in places where invasive pest species have not yet been detected. Places that may have suitable areas for BHC in the future should be detected in time to strengthen the precaution. According to the habit of this pest, it usually provides a protective barrier by making channels on the surface under the leaves using excreta. Therefore, the use of insecticides to control this pest is not very desirable, either in terms of cost or feasibility. As early as 1924, attempts at biological control were made in India (Jayanth and Nagarkatti, Reference Jayanth and Nagarkatti1984). In the future, rigorous testing of areas where this is likely to occur and the timely placement of natural enemies will be a more effective program.

This study combines 19 bioclimatic variables modelled using the MaxEnt model to explore both the contemporary and future geographic distribution of BHC in China. The bioclimatic variables are mainly composed of two elements: temperature and precipitation. It can well simulate the climate conditions in different situations and is also a more scientific indicator of climate nowadays. Meanwhile, the collection of data on the occurrence of BHC was screened and integrated. On this basis, the MaxEnt model was used to predict the potential distribution area of BHC in China and effectively reflect the changes in the size of the suitable area.

Conclusion

In this study, we successfully simulated the potential geographic distribution of the BHC under three climate change scenarios (SSP1-2.6, SSP2-4.5 and SSP5-8.5) for the current and future periods (2050s and 2090s) based on the MaxEnt model. Under the current climate conditions, the BHC is distributed only in the provinces of Tibet, Yunnan, Fujian, Guangxi, Taiwan, Guangdong, Hong Kong and Hainan. Among them, high suitability zones existed only in Hainan and Taiwan. The total areas of high, medium and low suitability zones were 2.32 × 104, 1.68 × 104 and 15.56 × 104 km2, respectively. The environmental factor that had the greatest impact on the geographic distribution of the bug was BIO 1 (annual mean temperature), followed by BIO 6 (minimum temperature of the coldest month), BIO 11 (mean temperature of the coldest quarter) and BIO 4 (temperature seasonality). All these factors are related to temperature. In other words, temperature has an important effect on the distribution of BHCs, which is also consistent with other related studies. The expected temperature increase is projected to lead to the geographic expansion of this pest's distribution. Also, the potential suitable distribution areas predicted by the BHC were generally consistent with the actual distribution areas. This work aims to provide new control perspectives for the invasive pest, the BHC.

Data

The data supporting the results are available in a public repository from: Zhiling Wang. List of locations used for Opisina arenosella, with longitude, latitude and sources. figshare. Dataset. https://doi.org/10.6084/m9.figshare.25188527.v1. EPPO (2023) Opisina arenosella. EPPO datasheets on pests recommended for regulation. https://gd.eppo.int (accessed 21 May 2023), and GBIF.org (08 February 2024) GBIF Occurrence Download https://www.gbif.org/occurrence/download/0007921-240202131308920.

Author contributions

Conceptualisation, Zhiling Wang and Danping Xu; data curation, Yaqin Peng; formal analysis, Zhiling Wang and Yaqin Peng; investigation, Zhiling Wang and Yaqin Peng; methodology, Zhiling Wang; software, Yaqin Peng and Danping Xu; supervision, Zhihang Zhuo and Danping Xu; writing – original draft, Zhiling Wang; writing – review and editing, Zhihang Zhuo and Danping Xu. All authors have read and agreed to the published version of the manuscript.

Financial support

This work was funded by the Ministry of Sichuan Province Science and Technology Support Program (2022NSFSC0986) and China West Normal University Support Program (20A007, 20E051, 21E040 and 22kA011).

Competing interests

None.

Institutional review board statement

Not applicable.

Informed consent statement

Not applicable.

Footnotes

*

These authors contributed equally to this work and should be regarded as co-first authors.

References

Barker, JR and MacIsaac, HJ (2022) Species distribution models: administrative boundary centroid occurrences require careful interpretation. Ecological Modelling 472, 110107. doi: 10.1016/j.ecolmodel.2022.110107.CrossRefGoogle Scholar
Bowen, A and Stevens, M (2020) Temperature, topography, soil characteristics, and NDVI drive habitat preferences of a shade-tolerant invasive grass. Ecology and Evolution 10, 10785. doi: 10.1002/ece3.6735.CrossRefGoogle ScholarPubMed
Bradie, J and Leung, B (2017) A quantitative synthesis of the importance of variables used in MaxEnt species distribution models. Journal of Biogeography 44, 13441361. doi: 10.1111/jbi.12894.CrossRefGoogle Scholar
Breiman, L (2001) Random forests. Machine Learning 45, 532. doi: 10.1023/A:1010933404324.CrossRefGoogle Scholar
Breiman, L, Friedman, J, Olshen, RA and Stone, CJ (1984) Classification and Regression Trees. United states: Springer Nature, 368 pp.Google Scholar
De'Ath, G (2007) Boosted trees for ecological modeling and prediction. Ecology 88, 243251. doi: 10.1890/0012-9658(2007)88[243:BTFEMA]2.0.CO;2.CrossRefGoogle ScholarPubMed
Elith, J and Leathwick, JR (2009) Species distribution models: ecological explanation and prediction across space and time. Annual Review of Ecology, Evolution, and Systematics 40, 677697. doi: 10.1146/annurev.ecolsys.110308.120159.CrossRefGoogle Scholar
Elith, J, Graham, CH, Anderson, RP, Dudík, M, Ferrier, S, Guisan, A, Hijmans, RJ, Huettmann, F, Leathwick, JR and Lehmann, A (2006) Novel methods improve prediction of species’ distributions from occurrence data. Ecography 29, 129151. doi: 10.1111/j.2006.0906-7590.04596.x.CrossRefGoogle Scholar
Elith, J, Phillips, SJ, Hastie, T, Dudík, M, Chee, YE and Yates, CJ (2011) A statistical explanation of MaxEnt for ecologists. Diversity and Distributions 17, 4357. doi: 10.1111/j.1472-4642.2010.00725.x.CrossRefGoogle Scholar
Escalera, L, Hernandez, R, Soto-Rojas, C and Suazo-Ortuño, I (2018) Predicting Ambystoma ordinarium habitat in central Mexico using species distribution models. Herpetologica 74, 117126. doi: 10.1655/Herpetologica-D-16-00078.1.CrossRefGoogle Scholar
Friedman, JH (1991) Multivariate adaptive regression splines. Annals of Statistics 19, 1141. doi: 10.1214/aos/1176347963.Google Scholar
Guisan, A and Zimmermann, NE (2000) Predictive habitat distribution models in ecology. Ecological Modelling 135, 147186. doi: 10.1016/S0304–3800(00)00354-9.CrossRefGoogle Scholar
Guisan, A Jr., Edwards, TC and Hastie, T (2002) Generalized linear and generalized additive models in studies of species distributions: setting the scene. Ecological Modelling 157, 89100. doi: 10.1016/S0304-3800(02)00204-1.CrossRefGoogle Scholar
Guisan, A, Zimmermann, NE, Elith, J, Graham, CH, Phillips, S and Peterson, AT (2007) What matters for predicting the occurrences of trees: techniques, data, or species’ characteristics? Ecological Monographs 77, 615630. doi: 10.1890/06-1060.1.CrossRefGoogle Scholar
Guo, Q, Kelly, M and Graham, CH (2005) Support vector machines for predicting distribution of Sudden Oak Death in California. Ecological Modelling 182, 7590. doi: 10.1016/j.ecolmodel.2004.07.012.CrossRefGoogle Scholar
Jayanth, KP and Nagarkatti, S (1984) Testingbessa remota [Dip.: Tachinidae] against Opisina arenosella [Lep.: Cryptophasidae] and other lepidopterous hosts in India. Entomophaga 29, 415419. doi: 10.1007/BF02372163.CrossRefGoogle Scholar
Jin, T, Lin, Y, Han, S, Ma, G, Wen, H and Peng, Z (2019) Host performance of Trichogramma species on Opisina arenosella, and evaluation of their biological control efficacy. Journal of Asia-Pacific Entomology 22, 990996. doi: 10.1016/j.aspen.2019.06.010.CrossRefGoogle Scholar
Jingjing, D, Siming, F, Jiaqiang, Z, Juan, S and Te, L (2023) Analysis of insect pests identified to species that were intercepted at Chinese ports of entry. Management of Biological Invasions 14, 641657. doi: 10.3391/mbi.2023.14.4.05.CrossRefGoogle Scholar
Kumar, R (2002) Studies on Ecology, Reproductive Biology and Management of Opisina arenosella (Walker) (Lepidoptera: Studies on Ecology, Reproductive Biology and Management of Opisina arenosella (Walker) (Lepidoptera: Oecophoridae) (Doctor Type). University of Agricultural Sciences, Bangalore.Google Scholar
Lever, RJAW (1969) Pests of the Coconut Palm. Finland: FAO Agricultural Study.Google Scholar
Li, L, Qingzhao, L, Xiongfei, Y, Guangchang, M, Bo, L, Fan, Y and Wei, Y (2021) Observation on the ultrastructures of antennal sensilla in Opisina arenosella. Journal of Environmental Entomology 43, 10701078.Google Scholar
Lu, H, Lyu, B, Tang, J, Wu, Q, Wyckhuys, KAG, Le, KH, Chongchitmate, P, Qiu, H and Zhang, Q (2023) Ecology, invasion history and biodiversity-driven management of the coconut black-headed caterpillar Opisina arenosella in Asia. Frontiers in Plant Science 14. doi: 10.3389/fpls.2023.1116221.Google ScholarPubMed
Mao, H and Kunimi, Y (1994) Longevity and fecundity of Brachymeria lasus (WALKER) (Hymenoptera: Chalcididae), a pupal parasitoid of the Oriental Tea Tortrix, Homona magnanima DIAKONOFF (Lepidoptera: Tortricidae) under laboratory conditions. Applied Entomology and Zoology 29, 237243.CrossRefGoogle Scholar
Mohan, C, Nampoothiri, CK, Radhakrishnan Nair, CP and Rajan, P (2010) Leaf-eating caterpillar (Opisina arenosella)-induced yield loss in coconut palm. International Journal of Tropical Insect Science 30, 132137. doi: 10.1017/S174275841000024X.CrossRefGoogle Scholar
Muralimohan, K, Srinivasa, YB and Pn, S (2013) Effect of temperature on developmental biology of Opisina arenosella (Walker). Current Biotica 7, 228232.Google Scholar
Nasser, M and Abdurahiman, UC (2001) Biological control of the coconut caterpillar Opisina arenosella (Lepidoptera: Xylorictidae): achievements and prospects. In Upadhyay, RK, Mukerji, KG and Chamola, BP (eds), Biocontrol Potential and its Exploitation in Sustainable Agriculture: Volume 2: Insect Pests. Boston, MA: Springer US, pp. 285305.Google Scholar
Nneji, LM, Salako, G, Oladipo, S, Ayoola, A, Onadeko, A, Adedeji, B, Omotoso, O, Ugwumba, A and Adeola, A (2019) Species distribution modelling predicts habitat suitability and reduction of suitable habitat under future climatic scenario for Sclerophrys perreti: a critically endangered Nigerian endemic toad. African Journal of Ecology 58, 481491. doi: 10.1111/aje.12713.CrossRefGoogle Scholar
Olden, JD, Lawler, JJ and Poff, NL (2008) Machine learning methods without tears: a primer for ecologists. The Quarterly Review of Biology 83, 171193.CrossRefGoogle ScholarPubMed
Perera, PACR, Hassell, MP and Godfray, HCJ (2009) Population dynamics of the coconut caterpillar, Opisina arenosella Walker (Lepidoptera: Xyloryctidae), in Sri Lanka. Bulletin of Entomological Research 78, 479492. doi: 10.1017/S0007485300013237.CrossRefGoogle Scholar
Phillips, SJ, Dudík, M and Schapire, R (2004) A maximum entropy approach to species distribution modeling. International Conference on Machine Learning, 655663. doi: 10.1145/1015330.1015412.Google Scholar
Phillips, SJ, Anderson, RP and Schapire, RE (2006) Maximum entropy modeling of species geographic distributions. Ecological Modelling 190, 231259. doi: 10.1016/j.ecolmodel.2005.03.026.CrossRefGoogle Scholar
Ramkumar, , Muralimohan, K, Kiranmayi, L and Srinivasa, YB (2006) Discrete generation cycles in the tropical moth Opisina arenosella. Current Science India, 91, 811816.Google Scholar
Rhoden, CM, Peterman, WE and Taylor, CA (2017) Maxent-directed field surveys identify new populations of narrowly endemic habitat specialists. Peerj 5, e3632. doi: 10.7717/peerj.3632.CrossRefGoogle ScholarPubMed
Shameer, KS, Nasser, M, Mohan, C and Hardy, ICW (2018) Direct and indirect influences of intercrops on the coconut defoliator Opisina arenosella. Journal of Pest Science 91, 259275. doi: 10.1007/s10340-017-0904-6.CrossRefGoogle ScholarPubMed
Stockwell, D (1999) The GARP modelling system: problems and solutions to automated spatial prediction. International Journal of Geographical Information Science 13, 143158. doi: 10.1080/136588199241391.CrossRefGoogle Scholar
Tang, Y (2017) Studies on the Suitability of Coconut Weaver Moth to Four Host Plants and its Distribution Pattern on Coconut Trees (Master Type). Hainan University, China.Google Scholar
Wang, Z, Xu, D, Liao, W, Xu, Y and Zhuo, Z (2023) Predicting the current and future distributions of Frankliniella occidentalis (Pergande) based on the MaxEnt species distribution model. Insects 14, 458. doi: 10.3390/insects14050458.CrossRefGoogle ScholarPubMed
Figure 0

Figure 1. Receiver operating characteristic curve and AUC result of MaxEnt modelling.

Figure 1

Table 1. Per cent contribution (%) and permutation importance (%) of environmental variables in predicting the occurrence of O. arenosella in the MaxEnt model.

Figure 2

Figure 2. Importance of environmental variables to O. arenosella by jackknife test.

Figure 3

Figure 3. Response curves of the environmental variables that contributed most to the MaxEnt models. (A) Annual mean temperature (Bio 1); (B) temperature seasonality (standard deviation × 100) (Bio 4); (C) coldest month minimum temperature (Bio 6); (D) mean temperature of coldest quarter (°C) (Bio 11).

Figure 4

Figure 4. Current suitable climatic distribution of O. arenosella in China. The probability of O. arenosella is shown by the colour scale in the area. Red indicates a highly suitable area with a probability of higher than 0.66, orange indicates a moderately suitable area with a probability of 0.33–0.66, yellow indicates a poorly suitable area with a probability ranging from 0.05 to 0.33, and white represents unsuitable areas.

Figure 5

Figure 5. Potential distribution of suitable areas of O. arenosella under different climate change scenarios in China. The probability of O. arenosella is shown by the colour scale in the area. Red indicates a highly suitable area with a probability of higher than 0.66, orange indicates a moderately suitable area with a probability of 0.33–0.66, yellow indicates a poorly suitable area with a probability ranging from 0.05 to 0.33, and white represents unsuitable areas.

Figure 6

Table 2. Analysis of main suitable distributions of O. arenosella in China.

Figure 7

Table 3. Prediction of the suitable areas for O. arenosella under current and future climatic conditions.