The distribution records of E. annuus and E. philadelphicus were accessed and downloaded from the Center for Agriculture and Bioscience International (CABI, 2022a; CABI, 2022b), the Global Biodiversity Information Facility (GBIF, 2022a; GBIF, 2022b), the China Digital Herbarium (https://www.cvh.ac.cn/, accessed April 2023), and related literature (Zhang et al., 2011b; Fan et al., 2020; Sennikov et al., 2020). We imported all the distribution points collected into ArcGIS, referred to relevant literature, and removed unreasonable data and distribution points with missing environmental data. Spatial filtering was applied using the spThin function in RStudio 4.1.3 to minimize spatial autocorrelation. Following this filtering, the remaining distribution points were separated by a minimum distance of 9 km, indicating their suitability for model building. After filtering the raw data, a total of 2,279 valid distribution records for E. philadelphicus and 793 distribution points for E. annuus were retained.

We obtained elevation (ele) data and historical (1970–2000) and future (2040–2060) (the 2050s) data for 19 bioclimatic variables (bio1–bio19) in Worldclim version 2.1 (https://www.worldclim.org, accessed May 2023), with a spatial resolution of 5 arc_min. For future environmental data, we selected four different Shared Socio-economic Pathways in 2050 under the BCC-CSM2-MR model: SSP126, SSP245, SSP370, and SSP585. SSP126 means lower forcing scenarios (radiative forcing reaches 2.6 W/m² in 2100; the temperature may be lower than 2.0°C relative to the pre-industrial revolution). SSP245 and SSP370 mean medium forcing scenarios (radiative forcing becomes 4.5 and 7.0 W/m² in 2100). SSP585 means the highest forcing scenarios (radiative forcing reaches 8.5 W/m² in 2100) (Lan et al., 2022). Soil salinity data (Esp and Ece) were provided by the FAO World Soil Database version 1.2 (http://www.fao.org/soils-portal/soil-survey/soil-maps-and-databases/har-monized-world-soil-database-v12/en/, accessed May 2023). The collinearity of environmental variables may influence parameter optimization during the calibration process of the MaxEnt model (Dormann et al., 2013). The variable screening process was as follows: (1) we imported 22 environmental variables and the current distribution into MaxEnt to remove zero-contribution rate variables; (2) the sampling function of ArcGIS was used to extract the environmental data of all distribution points; and (3) to reduce the collinearity between highly correlated variables, correlation analysis and principal component analysis (PCA) were processed by SPSS version 25.

To optimize the model and prevent overfitting, it is important to set relevant parameters (Zhao et al., 2022). The regularization multiplier (RM) and feature combinations (FCs) are regarded as two critical parameters for optimal settings in model calibration (Muscarella et al., 2014), which can be selected in the basic settings. The feature combinations represent different transformations of the covariable (Lu et al., 2020) including linear features (L), quadratic features (Q), product features (P), hinge features (H), and threshold features (T) (Muscarella et al., 2014). RM was set from 0.5 to 4, with a 0.5 interval, and six different FC combinations, namely, L, LQ, LQP, LQH, LQHP, and LQHPT. We utilized the R package (ENMevals) to select the FC and RM values as the optimization parameters for subsequent prediction when the optimized minimum information criterion (Delta AICc) is 0 (Warren & Seifert, 2011; Ye et al., 2021). The results showed that RM was 0.5 for both E. philadelphicus and E. annuus, and the FCs were LQPTH.

The distribution records and environmental variables obtained for the two species were separately imported into MaxEnt 3.4.4. In the modeling, 25% of the data were used to test the model and 10 repetitions were set for each type of subsample. The maximum number of iterations was 5,000, and the 10 percentile training presence was added. We selected environmental parameters using Jackknife analysis. This analysis is commonly used to analyze the influence of various environmental variables on the predictive outcomes or to determine the key environmental factors affecting species distribution (Peterson & Cohoon, 1999; Phillips et al., 2004). By using the Jackknife analysis, the primary environmental variables that affect the distribution of a species can be analyzed. The principle behind this method is to sequentially omit each environmental variable during the computation, build a model using the remaining variables, and then analyze the correlation between the omitted environmental variable and the omission error. If the absence of an environmental variable leads to a significant increase in omission error, it indicates that this environmental parameter significantly affects the predictive outcomes of the model (Peterson et al., 2007). The RM was set to 0.5, and the FC was set to LQPTH. The results from MaxEnt were imported into ArcGIS and converted to Raster. The map displaying suitable habitats was classified into four categories: negligible risk, low risk, medium risk, and high risk. Jenks Natural Break Classification (NBC) was used in this study, and this method minimizes the variation within each range, so the areas are as close as possible in value to each other (Gao & Shi, 2021; Zhang et al., 2022b). According to the result plots from MaxEnt, the classification for E. philadelphicus was 0–0.12, 0.12–0.31, 0.31–0.69, and 0.69–1, and that for E. annuus was 0–0.08, 0.08–0.31, 0.31–0.69, and 0.69–1 with Fixed cumulative value 5 Cloglog threshold. In this study, the area under the receiver operating characteristic (ROC) curve, also known as the area under the curve (AUC), was chosen as the evaluation metric. The AUC value ranges from 0 to 1, which can be interpreted as follows: 0 to 0.5 (prediction no better than random); 0.5 to 0.7 (average prediction effect); 0.7 to 0.9 (good prediction effect); and 0.9 to 1.0 (excellent prediction effect) (Franklin & Miller, 2010). The data from different climatic conditions were reclassified and assigned to obtain future changes in the suitable habitats for the four climatic scenarios compared to historical conditions using the raster calculator. The raster to point tool was used and the mean distribution centroid was calculated for each climate condition.

The study examined the shifts of the global potential suitable habitats of E. annuus under historical climate condition and different future climate scenarios in 2050 ( Figure 1 ; Table 1 ). Results revealed that the potential suitable habitats of E. annuus were mainly concentrated in central and coastal North America, Europe, eastern Asia, southeastern parts of South America, southern Australia, and southern Africa ( Figures S2, S3 ). The analysis further predicted the expansion of potential suitable habitats under all future scenarios, with the largest expansion under the SSP245 climate scenario, mainly in Asia, Africa, and America. Under future climate conditions, the native range in North America shows an expansion trend, especially under the SSP245 climate scenario, where there is a large-scale expansion of suitable habitats. In this climate scenario, apart from North America, the introduced ranges in Europe and Asia also experience significant expansion. Additionally, Africa and South America, where the species is currently not distributed, show notable expansion under the SSP245 climate scenario.

The analysis of potential suitable habitat shifts under historical condition and different future climate scenarios in 2050 ( Figure 2 ; Table 1 ) revealed that the main distribution of E. philadelphicus was predominantly inhabited in subtropical areas at 30°–40° N latitude: encompassing native North America, eastern Asia, Europe, eastern Australia, small parts of South America, and Africa ( Figure S6 ). Although an overall expansion trend was noted for E. philadelphicus, the increase in suitable habitats was not apparent across the four future climate scenarios ( Figure 2 ; Figure S7 ). The results suggest that the native range in North America has remained largely unchanged. In contrast, the introduced range in Europe is projected to experience a slight contraction in suitable habitats under future climatic conditions. In the eastern coastal regions of Oceania, where the species is currently absent, there is a slight expansion of suitable habitats. In East Asia, where the species has already invaded, there is no significant change in suitable habitats projected under future climatic conditions.

The study depicted the shifts of the potential suitable habitats of E. annuus in China under historical condition and different future climate scenarios in 2050, which are shown in Figure 3 ; Table 2 ; it was found that within China, most of the central, eastern, and southern regions were suitable for its growth ( Figures S4, S5 ; Table S1 ). Specifically, under the SSP245 climate scenario, the suitable habitats of E. annuus exhibited notable expansion and contraction compared to historical conditions. The regions experiencing expansion primarily included Heilongjiang, Inner Mongolia, and Yunnan, and contraction areas were Tibet, Sichuan, Qinghai, Gansu, Henan, Hebei, and Shandong.

The shifts of the potential suitable habitats of E. philadelphicus in China under historical climate condition and different future climate scenarios in 2050 are shown in Figure 4 ; Table 2 , and it was found that within China, the range of suitable habitats was similar to the E. annuus but slightly smaller ( Figures S8, S9 ). The overall potential suitable areas for E. philadelphicus are projected to expand mainly in regions such as Tibet, Shanxi, Shandong, Hebei, and Inner Mongolia, while the contraction area was less obvious compared with the historical climate scenario under climate change.

For the prediction of potential suitable habitats of E. annuus, five environmental variables were screened for E. annuus, the temperature seasonality (bio4), the precipitation of the wettest quarter (bio16), the precipitation of the driest month (bio14), the mean temperature of the wettest quarter (bio8), and the precipitation seasonality (bio15), contributing 29.9%, 27.4%, 23.4%, 16.8%, and 2.4%, respectively ( Table 3 ). Analysis using the Jackknife method indicated that when considering “With only variable”, the precipitation of the driest month (bio14) had the highest gain, indicating that this variable contained the most effective information and had the most predictive ability; in the “Without variable” condition, ignoring the temperature seasonality (bio4) minimized the gain, indicating that this variable contained the most information that was not present in the other variables ( Figure S1A ).

The model prediction for E. philadelphicus included five selected variables: the precipitation of the driest month (bio14), the temperature annual range (bio7), the precipitation of the wettest month (bio13), the topsoil sodicity (ESP), and the topsoil salinity (ECE) contributing 46.6%, 27%, 19%, 3.8%, and 3.5%, respectively ( Table 3 ). Jackknife analysis revealed that under the condition “With only variable”, the precipitation of the driest month (bio14) had the highest gain. Excluding the temperature annual range (bio7) resulted in minimal gain when analyzing the gain in the “Without variable” condition ( Figure S1B ).

After optimization of FC and RM parameters, as shown in Figure 5 , for E. annus, the AUC value was 0.944, indicating an excellent level of model performance; for E. philadelphicus, the AUC value was 0.8844, suggesting a good model performance.

In this study, we aimed to predict the shifts in distribution centroids of the two invasive alien weeds and compare the overlap and different areas of potential suitable habitats of E. annuus and E. philadelphicus to reveal the invasion trend of the two species within China.

Figure 6A illustrates the trend of centroid transfer of E. annuus and E. philadelphicus in China from the current climate to the four scenarios (SSP126, SSP245, SSP370, and SSP585) in 2050. The distribution centroids for E. annuus showed a northward spreading trend shifting from Hubei to Hebei Province under SSP245 while the distribution centroids of E. philadelphicus were concentrated in Hubei Province. There was a sizable overlap between suitable habitats of E. annuus and E. philadelphicus in southern China. The potential suitable habitats for only E. annuus was larger than that of E. philadelphicus and mainly in the northern end of the overlapping areas ( Figure 6 ). Additionally, the potential suitable habitats for only E. philadelphicus was mainly in Yunnan Province under historical climate condition, but overlapped with suitable habitats for E. annuus under future climate scenarios ( Figure 6 ).

We identified topsoil salinity (Ece) and topsoil sodicity (Esp) as key factors influencing suitable areas for E. philadelphicus, which was consistent with observations that this species lacks resistance to salt and sodium. The results indicate that E. philadelphicus exhibits a preference for humid conditions, and according to the results, the precipitation of the driest month (bio14) was the variable with the highest contribution rate for this species. Conversely, E. annuus demonstrated higher adaptability and the ability to survive in poor soils, even growing in stone walls. Compared with previous studies, this study added soil variables into consideration, which greatly affect the distribution of weeds, but the collected distribution points may not cover all habitats where it is distributed, so there was still limitation in the prediction results.

We found that climate change has a similar effect on the distribution of E. annuus in its native range of North America and its introduced range in Europe and Asia. Overall, there is a noticeable expansion trend, but the general area of suitable habitat remains unchanged. On the other hand, for E. philadelphicus, we observed that there was no significant change in its distribution, both in its native range and in its introduced range. These results support the hypothesis of ecological niche conservatism in the case of E. philadelphicus. According to the ecological niche conservatism hypothesis, species tend to maintain their original ecological niche and adaptability as much as possible when facing climate change, resulting in a relatively slow response to environmental changes (Bates and Bertelsmeier, 2021). Therefore, when a species exhibits a higher degree of ecological niche conservatism, its range may remain relatively stable under climate change without significant changes. However, these results may also be influenced by other factors such as habitat suitability, competition, and human disturbance. Further research is needed to explore the mechanisms of species distribution changes and the role of ecological niche conservatism in the context of climate change.

The prediction results indicate that both weeds have extensive suitable habitats in China. E. annuus, being more adaptable, is expected to have a wider range of highly suitable areas including Fujian, Guizhou, and Taiwan. The southeast area will be the most suitable area for both weeds, indicating high invasive risk; E. annuus even shows a tendency to spread in the northeast area. Thus far, the current distributions of these two weeds were far from saturated in their suitable habitats, and appropriate quarantine control measures should be taken. Since both weeds are highly prolific in terms of seed production and rely on wind-borne spread, it is advisable to eradicate them before flowering (Zhang, 2011a; Pacanoski, 2017).

Both of these species originated from North America, and it is worth noting that North America and China have similar climatic zones south of 40° N. Thus, most North American species introduced to China adapt easily to their new habitat and establish themselves effectively in a relatively short period of time (Peel et al., 2007). Given this, it is imperative to enhance the prevention and control of quarantine at ports to prevent further introduction of these two invasive alien weeds.

The MaxEnt models were optimized using the R package “Enmeval”, resulting in an AUC value of 0.944 for E. annuus and 0.884 for E. philadelphicus. The predicted range of suitable habitats by the MaxEnt model aligned well with its historical distribution range, indicating that the models showed a good predictive ability. In this study, modeling primarily relied on known distribution records and climate data, but due to the diversity and complexity of climate change, the distribution data and climatic variables may still have some limitations. Additional factors, such as slope, slope orientation, soil capacitance, and organic matter content (Zhang et al., 2022a), were not taken into consideration. Furthermore, the invasion patterns of exotic organisms will be affected not only by climate but also by natural enemies, interspecific interactions, human interference, and other factors; thus, in the future, model prediction should encompass a broader range of factors to optimize the model evaluation and enhance the accuracy of prediction results (Jia et al., 2020).

When comparing the shifts in potential suitable habitats of two invasive alien weeds under different climate scenarios separately, it was observed that both species are expected to expand globally. The SSP.2-4.5 climate scenario, with 3°C of warming, is projected to result in the largest expansion of E. annuus. This may be due to the increasing temperature favoring its physiology. However, if temperatures rise even higher, they may inhibit its growth, but the specific mechanism remains to be studied. The expansion of E. philadelphicus showed minimal change under four climate scenarios. Within China, the potential suitable areas of both invasive alien Erigeron weeds expanded the most under the SSP585 climate scenario. These results indicate that the change of suitable habitats in response to climate change is not a simple linear relationship. Therefore, it is crucial to strengthen the monitoring and prevention in provinces with suitable habitats.

The current climate and future scenarios indicate that the distribution centroid of E. philadelphicus is predicted to primarily reside in Hubei and is unlikely to move much under the climate change scenarios we examined. In contrast, the distribution centroid of E. annuus transitioned to the north from Hubei to Hebei under SSP245 scenarios, indicating higher adaptability to climate change. Thus, careful monitoring of potential range expansion of E. annuus in China is warranted.

By comparing the ecological niche overlap of the two weeds, it was found that the potential suitable habitats of E. annuus is projected to have a wider global distribution of potential suitable habitats, primarily in North America and Asia. Within China, it is predicted that E. annuus may extend to the northeastern region, and will invade parts of Yunnan Province with climate change that were suitable only for E. philadelphicus under historical climate conditions. Hence, it can be speculated that the invasive potential of E. annuus is relatively stronger. However, it remains to be studied whether the two weeds interfere with or complement each other during the invasion process. Both weeds have been established in China and caused losses, due to the diverse geography, climatic conditions, and farmland in China, and therefore reasonable and effective methods should be taken for their prevention and control to minimize losses.