Predicting the suitable habitat distribution of Polygonatum kingianum under current and future climate scenarios in southwestern Yunnan, China

Climate change affects the distribution patterns and chemical composition of medicinal plants, leading to quality differences. Polygonatum kingianum Collett & Hemsl. (P. kingianum) is a Chinese medicine and food homolog. With the increasing market demand, the sustainable supply and quality assurance of P. kingianum are particularly important. This study used the maximum entropy (MaxEnt) model to predict the potential habitat suitability under current and future (2050s and 2070s) climate change scenarios, including Shared Socioeconomic Pathways (SSP) 245 and 585. In addition, chemical information of P. kingianum from different geographical sources was collected based on fourier transform near infrared spectroscopy (FT-NIR), and the geographical origin of P. kingianum was identified using a combination of two-dimensional correlation spectroscopy (2DCOS) and residual convolutional neural network (ResNet) to evaluate the impact of different geographical sources on the quality of P. kingianum. We explored the correlation between absorbance and environmental factors from the perspective of spectral information. The research results indicated that annual mean temperature (Bio01), temperature seasonality (Bio04), annual precipitation (Bio12), precipitation of driest month (Bio14) and elevation (Ele) are the main environmental factors affecting the distribution of P. kingianum. In the current climate scenario, the suitable habitat was mainly located the southwest of China, including Yunnan, Sichuan, Guizhou and Chongqing. The area of lowly (0.2 ≤ suitability indices < 0.4) suitable habitat was 57.20 × 10⁴ km², accounting for 51 % of the total suitable habitat area. The total land area comprising potential suitable habitat is predicted to expand under future climate scenarios. Under the emission scenario 245 (SSP245), the change rate of highly suitable habitat in 2070 was 30.26 %. The combination of 2DCOS and ResNet model successfully identified samples of P. kingianum from different geographical sources, with an accuracy rate of 100 %. In summary, the research results provided a fast and accurate method for the geographical tracing of P. kingianum, and provided theoretical guidance for the protection and sustainable utilization of wild resources.