Introduction

Hyphantria cunea (Drury) (Lepidoptera: Erebidae), the fall webworm moth, originates from North America and is a key invasive species under close surveillance in China. H. cunea, a polyphagous pest, targets a wide array of hosts, affecting over 300 plant species spanning 108 genera and 49 families, including mulberry, elm, willow, and locust tree species (Sun et al. 2021; Seguna et al. 2023). Coniferous species such as Metasequoia glyptostroboides Hu & W.C. Chen, Taxodium distichum (L.) Rich., and Taxodium ascendens Brongn are also damaged to varying degrees by H. cunea (Chen 2022). H. cunea demonstrates resilience, surviving extreme weather ranging from − 16 °C to a scorching 40 °C (Jia et al. 2022). Its peak emergence periods occur in late May and late July. It thrives in sunny, humid environments and is often found in areas with strong sunlight and strong scents such as fishy smells. Photoperiod and temperature play pivotal roles in the development of H. cunea (Wang et al. 2016a, b). Over the period from 2011 to 2022, provincial areas of infestation rose from 8 to 14, county-level infected areas increased from 338 to 611, with a cumulative impact of 655 across county-level administrative regions. By 2022, the infestation had expanded to 30° 50' N (Jiashan County, and Jiaxing City, Zhejiang Province) and 108° 43' E (Fengxi New Town, Xixian New District, and Xi’an City, Shanxi Province). This expansion indicates a four-degree southward and westward extension compared to its 2011 range (Liu and Li 2022). With its wide adaptability, tolerance of extreme temperatures, and capability to survive, even in the absence of food (Ni et al. 2009), its introduction into China in 1979 has significantly impacted agroforestry ecosystems, profoundly affecting China's ecological environment and socioeconomic development. Therefore, it is of vital importance to investigate the spatial and temporal distribution patterns of H. cunea and the underlying influencing factors.

Numerous researchers have dedicated efforts to analyzing the suitability of H. cunea habitats. Xia et al. (2022) employed the MaxEnt model of climate and altitude as influential factors to evaluate and predict potential habitable regions for H. cunea in Henan Province. Li et al. (2018), using the MaxEnt model and combining it with 105 known distribution points of H. cunea and nine environmental factors, predicted suitable areas the distribution range in Sichuan Province. Ji (2019) utilized the random forest model to forecast H. cunea presence in the North China Plain, showing that altitude, vegetation cover, and average temperatures during the wettest and warmest quarters as primary influencers of habitat suitability. Zhao et al. (2018) employed geostatistical analysis to study the spatial and temporal distribution of its population in Beijing from 2008 to 2014. A binary logistic risk model was developed for the spread of H. cunea, evaluating spread risks by integrating township quarantine data, geographical and meteorological conditions, and human-made spread factors. The model identified seven crucial biological temperature variables significantly influencing occurrence and propagation (Ye et al. 2021).

Li et al. (2023) analyzed diffusion dynamics and disaster conditions, examined the diffusion trend in Northeast China, and noted that natural diffusion and human-driven factors spread infestations. They highlighted the coastal area as a conducive environment for infestation occurrence and the infestation-causing factors. Their study emphasized the distinct relative diffusion rates of the combined effects of the environment and prevention measures, reflecting invasive organism characteristics. H. cunea infestation is anticipated to further spread to various protective forest areas (Li et al. 2023). Ji (2019) adopted the random forest model, considering 19 climatic and five environmental factors like altitude, slope, aspect, vegetation cover, and effective photosynthetic radiation. The model simulated the potential habitat distribution of H. cunea from 2011 to 2030, assessing environmental factor importance and predicting potential habitat distribution under different climate scenarios in the 2050s. Lu et al. (2016) used 574 sets of distribution data and 34 environmental layers to predict suitable areas for H. cunea using the GARP niche model.

Additionally, in the field of H. cunea recognition algorithms, Xue et al. (2020) utilized the Keras deep learning framework to extract image data features and proposed an artificial neural network recognition model, IHCDM, with a detection accuracy exceeding 99%. Han (2021) focused on representing three-dimensional information of H. cunea adults, establishing an extraction method based on stereoscopic vision principles and enabling new approaches for identification using machine vision and deep learning.

Spatial autocorrelation is related to the possible mutual influence or relationship among the observed data of certain variables within a given study area. Spatial positive correlation refers to the more significant correlation with the aggregation of spatial distribution (Meng et al. 2005). This study aimed to delineate the temporal and spatial trends of H. cunea across China, providing a foundation for prevention and control efforts nationwide against H. cunea.

Materials and methods

Study area and data

We took the red area in Fig. 1 as our research object.

Fig. 1
figure 1

Map of the study area

Full size image

The influence of climate, topography, human disturbances, and vegetation on the survival and spread of H. cunea were determined. Data on climate factors came from the National Earth System Science Data Center (http://www.geodata.cn), terrain data from earth data (https://urs.earthdata.nasa.gov), anthropogenic factors data from ORNL LandScan Viewer–Oak Ridge National Laboratory (https://landscan.ornl.gov). Vegetation data came from the Data Center for Resources and Environmental Sciences, Chinese Academy of Sciences (https://www.resdc.cn/).

Work framework

Basic data required for the experiment were prepared, including information on the distribution of H. cunea in 2021 and 2022, and data on environmental factors, administration division, and boundaries (Fig. 2). Extraction of the H. cunea distribution data and statistical processing of the environmental variable data were completed in the data processing phase. In the data analysis stage, spatial correlation analysis selected the influence factors to be studied. GeoDetector analysis analyzed the influence of a single factor and the interaction of two factors on the spatial distribution of H. cunea. Visualization of variable data (climate, topography, human disturbance, vegetation) was performed to observe distribution characteristics. The effects of each factor on the distribution of H. cunea and the spatiotemporal distribution trends at a national scale were analyzed in conjunction with the experimental results. This study may provide the necessary references for the prevention and control of H. cunea infestations at a national scale.

Fig. 2
figure 2

Framework for spatiotemporal trend analysis of the Hyphantria cunea

Full size image

Environmental variables and their selection criteria

Based on the ranking of variables conducive to the habitat of H. cunea (Liu and Li 2022; Xue et al. 2023), 16 primary environmental variables were selected that impact H. cunea distribution. These variables were categorized into four groups according to their characteristics: climatic, topographic, human disturbance, and vegetation factors (Table 1). Within this set of factors, climatic elements encompass various metrics such as average, minimum, and average temperatures during the coldest and driest months, the highest and average temperatures in the hottest month, average temperatures during the wettest month, and temperature range, as well as precipitation-related factors including total precipitation, precipitation of the driest and wettest months, and precipitation in the coldest and hottest months. Terrain factors include altitude, human disturbance factors encompass population density, and vegetation factors relate to density of vegetation.

Table 1 List of environmental variable data
Full size table

Spatial autocorrelation statistics

This study used the global Moran index and the generalized G statistic to determine the clustering of H. cunea distribution to depict patterns of aggregation.

When Moran’s index is positive, it signifies a positive spatial correlation. A higher value implies a more pronounced spatial correlation. Conversely, a negative value indicates a negative spatial correlation, with smaller values indicating greater spatial disparity. Specifically, when Moran’s index is 0, it indicates a random spatial distribution.

The generalized G statistic serves as a global indicator of spatial autocorrelation. It is commonly used to gauge the extent of clustering, whether high or low, within the data in a given study area. Additionally, it aids in identifying the presence of hotspots and coldspots within the same area. The relatively small generalized G statistic indicates that the low value is spatially correlated with the value below the average level. The Z value is the significance test of the statistic. For positive representation, if the observed generalized G index is greater than the expected, the data is clustered in the high-value area. In contrast, for the negative representation, if the observed generalized G index is less than the expected, the data is clustered in the low-value area.

Introduction of geographic detector

The detector method is a statistical method to reveal the driving force behind it by detecting spatial heterogeneity (Wang et al. 2010 and 2016b). In this study, the detector’s factor analysis was utilized to evaluate the individual contribution of each influencing factor to the spatial variability of H. cunea. In addition, interaction detection was utilized to appraise the nature and strength of the interplay between these two factors in influencing H. cunea spatial distribution. The formula is:

$$q=1-\frac{{\sum }_{h=1}^{P}{N}_{h}{\sigma }_{h}^{2}}{N{\sigma }^{2}}$$
(1)

where, h = 1,…, P is the stratification of factor X; \({N}_{h}\) and \(N\) are the number of h layer samples and the total number of samples, respectively. \({\sigma }_{h}^{2}\) represents the variance of layer h, while \({\sigma }^{2}\) denotes the variance of the entire area. q signifies the impact of the factor on the spatial differentiation of H. cunea, with its value ranging from 0 to 1. A higher q value indicates a substantial influence of the factor on the spatial differentiation of H. cunea.

When using geographic detectors for factor analysis, the independent variable is the type quantity and the dependent variable the numerical quantity. Given that the variables under investigation are numerical in nature, we opted to employ the geographic detector (in the R language version) for our analysis. This version offers the flexibility to choose appropriate classification techniques and intervals for discretizing numerical data into categorical variables.

Results and discussion

Spatial autocorrelations and aggregate analysis

A global spatial autocorrelation analysis was carried out for the distribution of H. cunea for 2021 and 2022.

The global Moran I index results are shown in Table 2. A Moran I index for 2021 was 0.2980, the Z score 84.7492, and the p-value 0.0000, indicating an aggregation pattern showing a positive spatial correlation. With the aggregation of spatial location, the more significant correlation, and the possibility of randomly generating this clustering pattern is less than 1%, i.e., 99% confidence. The Moran I index for 2022 was 0.3588, the Z score 101.7841, with a p-value of 0.0000, which indicates that it was also an aggregation pattern showing a positive spatial correlation. The correlation is more significant with the aggregation of spatial locations, and the possibility of randomly generating this clustering pattern is less than 1%, 99% confidence level. Therefore, the distribution of H. cunea in 2021 and 2022 showed an aggregation pattern.

Table 2 Global Moran index of Hyphantria cunea for 2021 and 2022
Full size table

The value of the generalized G statistic in 2021 was 0.0000, the Z score 57.9142 > 1.9600, the observed value of General G was greater than the expected value, and the p-value was 0.0000 < 0.0100, indicating a high clustering pattern (Table 3). The probability of randomly generating this high clustering pattern is less than 1%, and it was the same for 2022. Therefore, the distribution of H. cunea in 2021 and 2022 showed a high clustering pattern in high value areas.

Table 3 H. cunea general G statistics for 2021 and 2022
Full size table

Proliferation trends

The change in the distribution of H. cunea from 2021 to 2022 provides an idea of its spread. In Fig. 3, the blue line represents the decrease in the spread of H. cunea in 2022 compared to 2021, while the red line represents the increase in 2022 in the same district. The increase in area was mainly concentrated in Shandong and Hebei provinces, while Shandong Province was most widely affected. The largest increase in area was in Gounan County, Lianyungang City, Jiangsu Province, from 1481.81 ha in 2021 to 6660.00 ha in 2022, an increase of 5178.19 ha, far exceeding the increase in Quyang County, Baoding City, Hebei Province, which was second with an increase of 3648.49 ha. H. cunea area reduction was more obvious in counties concentrated in Liaoning Province and Hebei Province. The largest reduction in the infestation was in Funing County, Yancheng City, Jiangsu Province, from 13,740.53 ha in 2021 to 2460.67 ha in 2022, a total reduction of 134,876 ha and the second largest reduction in area was in Lishan District, Huaibei City, Anhui Province, with a reduction of 13,441.38 ha due to a significant effect of prevention and control measures.

Fig. 3
figure 3

Map of cumulative area change of H.cunea

Full size image

The standard deviation ellipse analysis and center of gravity shift analysis of H. cunea distribution in 2021 and 2022 can further explain the spread of the pest. As shown in Fig. 4, the blue dots are the center of gravity in 2021, the yellow dots the center of gravity in 2022; the red ellipses (2021) and the pink ellipses (2022) are both first-level standard deviation ellipses. From the results of the standard deviation ellipse creation, the ellipse with 468 km as the long semi-axis and 251 km as the short semi-axis at 118.11° E, 37.4° N covers about 68% of the area affected by H. cunea in 2021. The rotation angle of the ellipse is 26.07°; the ellipse with 451 km long semi-axis and 236 km short semi-axis encompasses about 68% of the area affected by H. cunea in 2022, with an ellipse rotation angle of 27.48°. It shows that the concentration of H. cunea is roughly distributed in the northeast-southwest. The generated area of the ellipse in 2022 was slightly smaller and its flatness slightly increased compared with that in 2021. From the results of the center of gravity migration, it is seen that the center of gravity of the impact was in Shandong Province, and moving to the northwest. The likely cause for this shift could be attributed to climate changes (Lewis 2006).

Fig. 4
figure 4

Elliptic analysis of the standard deviation of H. cunea as affected by distribution area

Full size image

Visualization of environment variable data

The growth, propagation, geographical distribution, diversity, and abundance of H. cunea are all affected by environmental factors. (Austin 2002; Kreft and Jetz 2007). To better observe the effect of climate, terrain, human disturbance, and vegetation distribution, a distribution map of various environmental variables was developed (Figs. 5, 6, 7).

Fig. 5
figure 5

National temperature distribution map

Full size image
Fig. 6
figure 6

National precipitation distribution map

Full size image
Fig. 7
figure 7

Elevation, population cover density, and vegetation cover density

Full size image

As shown in Fig. 5, the average national temperature in 2021 was influenced by regional terrain. The southeastern region is relatively high, while the northwestern region, Xinjiang, is significantly higher than other regions. Altitude and latitude exerted an influence on the national average temperature in 2021, as well as on average temperatures during the coldest and driest months. At the same latitude, the western region generally experiences lower temperatures compared to the central and eastern regions, while the southern region tends to have higher temperatures than the northern region. Temperatures in the Turpan Basin of Xinjiang and Sichuan is higher than that in the surrounding areas. In 2021, the highest temperature, the hottest monthly mean temperature, and the wettest monthly mean temperature were concentrated in the eastern part of the country and the Turpan Basin of Xinjiang, while the Tibet Autonomous Region, Qinghai Province and the northwestern part of Sichuan Province had relatively low temperatures due to altitude. The variation in national temperatures in 2021 was stable from north to south.

In 2021, the precipitation in southern China was generally higher than that in northern areas of the country. Total precipitation and the wettest monthly precipitation decreased with increasing latitude. Precipitation in the driest and in the coldest months are affected by topography, and in some areas is higher than in the surrounding areas. Due to the influence of monsoons, the hottest months are in the southwest, southeast, and northeast, and the precipitation is significantly higher than in other areas.

The national terrain as a whole presents a three-step format from west to east from high to low. Population is mainly concentrated in the eastern region outside the Qinghai-Tibet Plateau and in other areas with high elevations and unsuitable for plant growth (Guo et al. 2023). The vegetation cover decreases with an increase in population density.

GeoDetector analysis

The GeoDetector is a research method based on spatial heterogeneity that quantitatively detects the main driving factors and the interactions between different drivers that express a spatiotemporal phenomenon (Zhao et al. 2020). It has higher explanatory efficiency compared with other spatial heterogeneity detection tools (Xu et al. 2023). Spatial heterogeneity is defined as the difference in the distribution area of H. cunea between different regions. The detector results showed that all these factors significantly explained the distribution of H. cunea (Table 4). The results were ranked by their q values as: the driest month precipitation (X10) > altitude (X14) > the coldest average monthly temperature (X3) > temperature range (X8) > minimum temperature (X2) > precipitation (X9) > average temperature of the driest month (X4) > the coldest month precipitation (X12) > the hottest month precipitation (X13) > average temperature (X1) > the wettest average monthly temperature (X7) > population coverage density (X15) > highest temperature (X5) > the hottest average monthly temperature (X6) > the wettest monthly precipitation (X11) > vegetation cover density (X16). From these, the precipitation of the driest month has the greatest explanatory power, indicating that the spatial distribution of H. cunea in China is most strongly controlled by this factor and has the strongest consistency with H. cunea in spatial distribution. This may be due to the fact that precipitation can affect insect oviposition (Rahman et al. 2023). The secondary factor was altitude, indicating that altitude is also important affecting the spatial distribution. This observation is consistent with impact factor analysis of H. cunea by Liu et al. (2023).

Table 4 Results of factor detector
Full size table

The results from the interaction detector show that, aside from the interaction between average temperature (X1) and vegetation cover density (X16), all other factor interactions exhibited enhancement (Table 4). This encompasses both bi-factor and nonlinear enhancements. None of the factors operated in isolation, emphasizing that the spatial differentiation pattern of H. cunea was not determined by a single factor, but rather resulted from the collective influence of multiple factors. Notably, the most substantial interactions were identified between the average temperature of the coldest month (X3) and altitude (X14), as well as between the driest month’s precipitation (X10) and altitude (X14). The interactions between the average temperature of the driest month (X4) and altitude (X14), the coldest month's average temperature (X3) and the driest month’s precipitation (X10), as well as the coldest month’s average temperature (X3) and the wettest average monthly temperature (X7) are moderate. The interactions between the wettest monthly precipitation (X11) and vegetation cover density (X16), and the highest temperature (X5) and the coldest monthly precipitation (X12) were the smallest.

Conclusion

The significant economic losses across various provinces and cities due to the spread of H. cunea necessitates urgent measures for prevention and control of this infestation. Spatial autocorrelation statistical analysis indicated a high degree of clustering in its distribution in 2021 and 2022. During these years, it was predominantly concentrated in the eastern regions of the country, particularly Shandong, Jiangsu, Anhui, Liaoning, Henan, and Hebei provinces. Standard deviation ellipse analysis and centroid displacement analysis revealed that the centroid of H. cunea in China lay within Shandong Province, with a directional shift towards the northeast. The initial introduction of H. cunea in Shandong Province primarily occurred as incomplete 3rd generations in the Jiaodong Peninsula, but the region has increasingly witnessed complete 3rd generations, leading to a significant expansion across the province. Additionally, compared to western genetic populations, the eastern ones exhibited substantially higher genetic diversity, potentially contributing to increased reproductive generations (Dai et al. 2023; Lu et al. 2023). In other areas analyzed through standard deviation ellipse analysis, such as Guannan County in Lianyungang City, Jiangsu Province, Quyang County in Baoding City, and Guantao County in Handan City, Hebei Province, the affected areas experienced a notable positive expansion exceeding 50,000 acres per county. Timely attention to preventive measures like chemical treatments, the introduction of natural enemies, and manual operations, alongside crucial quarantine or prevention steps, are essential. An observation from the centroid shift suggests a decelerating spread rate in the new epidemic area along the Yangtze River region. In the insect’s native habitat, the United States, latitude influences the moth's genetic makeup (Black et al. 2023). Therefore, in Chinese regions with similar latitudes to the United States, the relatively slower expansion in the leading edge area of the infestation compared to the older infected area could be attributed to lower genetic richness and diversity levels.

With regards to the environment, H. cunea exhibits robust survivability against extreme conditions, tolerating temperatures ranging from  −16 to 40 °C and sustaining growth and development even without food for up to 15 days (Gomi et al. 2007). Visual analysis of climate data divides temperature and precipitation into multiple variables, including annual average temperature, annual lowest temperature, average coldest month temperature, hottest month precipitation, and driest month precipitation. These variables not only affect the reproduction and development of H. cunea but also impact its natural enemies, such as the parasitic wasp, Brachymeria lasus (Walker). Temperature influences the development rate and parasitic success of B. lasus (Tian 2021). The eastern regions, roughly situated at the boundary of precipitation and temperature, boast relatively higher vegetation density, providing suitable temperatures, precipitation, and growth environments for larvae and adult development, thereby contributing to the severity of the infestation in these areas. The geographic detector not only revealled the causal relationship between species distribution and driving factors but also strong adaptability to collinearity among various variables. It has been widely utilized in studies on pest control, such as for the Asian tiger mosquito (Aedes albopictus Skuse), and management has implemented effective control recommendations based on factors that hold relatively strong explanatory power (Chen and Wei 2023; Wei et al. 2023). Therefore, concerning the two most critical controlling factors highlighted by the geographic detector analysis—driest month precipitation and altitude—these factors should also be considered in efforts for H. cunea control. Additionally, the interactions among various factors were examined, and the outcomes indicate an amplifying effect, particularly observed between the average temperature of the coldest month and altitude, as well as precipitation in the driest month and altitude.

In conclusion, this study carried out a spatial autocorrelation analysis of H. cunea’s distribution in China for the years 2021 and 2022. Evaluation using Moran’s I index and generalized G statistics assessed clustering patterns. Additionally, the GeoDetector method examined spatial heterogeneity. The research explored the spatial distribution and influencing factors of H. cunea spread, providing a viable foundation for the formulation of prevention and control measures. However, integrating more localized factors and conducting field surveys remain necessary to enhance the specificity and effectiveness of control efforts.