Platycodon grandiflorus, the sole species from the genus Platycodon and a member of the Campanulaceae family, has been known for its medicinal, culinary, and ornamental uses for approximately 2000 years. Its distribution is primarily in eastern Asia, including China, Korea, Japan, and Russia. With increasing market demand and the depletion of wild resources, understanding the distribution and factors affecting its habitat suitability is crucial for its conservation and sustainable utilization. This study aims to predict the suitable habitat for P. grandiflorus in China considering current and future climate scenarios. The MaxEnt model, with an AUC value of 0.846, demonstrated good predictive ability for the current and future distribution of P. grandiflorus. It also identified the central, eastern, and southern regions as suitable habitats, with the critical environmental factors being precipitation, temperature, and elevation. Future scenarios under both SSP126 and SSP585 projections indicate an increase in suitable habitats, particularly in northeastern and central China, albeit with a shift in the distribution center towards the northeast by 2041-2060 and 2081-2100 under different scenarios. As a result, P. grandiflorus’ distribution is significantly influenced by environmental factors, with precipitation and temperature being pivotal. In summary, this study predicts an expansion of suitable habitats under future climate scenarios, suggesting that climate change may facilitate the growth and distribution of P. grandiflorus in new areas. The northward shift in the distribution center underlines the impact of global warming on plant distribution. These findings are crucial for the conservation, effective utilization, and strategic planning for the cultivation of P. grandiflorus in the face of climate change.

Distribution point collection

According to the Global Biodiversity Information Facility (GBIF, http://www.gbif.org/) as well as Chinese Virtual Herbarium (CVH), accessed on 7 January 2024, the distribution points of P. grandiflorus in China were screened and recorded. Repeated locations were removed and the remaining locations were screened manually so that only one location appeared in every 10 km × 10 km grid cell. 405 known distribution locations of P. grandiflorus were documented through the use of ArcGIS 10.8 (Esri, Redlands, CA, USA) (Fig. 1). The data points were saved in CSV format for subsequent analysis by ArcGIS.

Environmental parameters

Fifty-two environmental parameters which could influence the distribution of P. grandiflorus were selected. This study used the latest WorldClim version 2.1 (https://worldclim.org/, accessed on 13 January 2024) to obtain current, future climate projection data as well as elevation data at a spatial resolution of 2.5 min. Climate type data used include January to December precipitation and average temperatures, as well as 19 bioclimatic factors. The Shared Socioeconomic Pathways (SSPs) are based on five narratives of socioeconomic development (Riahi et al. 2017). For this study, we selected the low forcing scenario SSP126 and the high forcing scenario SSP585 to predict the impacts on P. grandiflorus. SSP126 represents a low-material, low-resource, and low-energy green development pathway, while the SSP585 scenario depicts a future socioeconomic pathway with high emissions and high carbon use (O’Neill et al. 2016). In this study, we adopted BCC-CSM2-MR climate model which was reported to accurately simulates temperature and precipitation in China (Wu et al. 2019). The BCC-CSM2-MR model was developed by the National Climate Centre (Beijing, China) and participated in the International Coupled Model Comparison Program (ICMCP) and enhanced its climate simulation capability in Eastern Asia, especially for the China region. We projected the potential distribution areas of P. grandiflorus for 2041-2060 and 2081-2100 under the SSP126 and SSP585 scenarios. In addition to the above-mentioned variables, we considered two topographic factors and six soil factors. The topographic factors included slope and aspect. Soil factors included topsoil clay fraction, topsoil sand fraction, topsoil organic carbon, the acidity and alkalinity of the soil, the total nutrient fixing capacity of the soil, available water storage capacity of the soil unit. We obtained the topographical data from the Geospatial Data Cloud (http://www.gscloud.cn/, downloaded on 15 January 2024)，the soil data through the Harmonized World Soil Database (https://www.fao.org/soils-portal/data-hub/soil-maps-and-databases/harmon- ized-world-soil-database-v12/en/, downloaded on 15 January 2024). These environment variables were converted to ASCII format by the ArcGIS Conversion Tools for downstream analysis. High correlations and covariances between the extracted environmental variables can easily lead to overfitting of the model and affect the prediction results, so not all variables were needed (Hu and Hua 2014). Therefore, for every environmental factor, its contribution to the model predictions was first evaluated using the jackknife test (a module in MaxEnt (version 3.4.1)), and the environmental factors that contributed less (<1 %) were eliminated. The correlation between the remaining variables was then calculated in SPSS 20.0 adopting the Pearson correlation coefficient method. We considered two variables with |r|≥0.8 to be significantly correlated, and excluded one of the variables with relatively low biological significance to minimize model overfitting (Xu et al. 2019). After screening, only 17 of the initial 52 environment variables were included for evaluation (Table 1).