Predicting Polygonum capitatum distribution in China across climate scenarios using MaxEnt modeling

Abstract

Climate change affects the geographical distribution of species. Predicting the future potential areas suitable for certain species is of great significance for understanding their distribution characteristics and exerting their value. Based on the data of 276 effective distribution points of Polygonum capitatum and 20 ecological factors, the maximum entropy (MaxEnt) model and the ArcGIS software were employed to predict the areas suitable for P. capitatum growth, and the main environmental factors affecting the geographical distribution of this species were explored. Under the current climatic conditions, the areas highly suitable for P. capitatum are mainly distributed in southwestern China, with a small number of sites in coastal areas and most sites in Guizhou Province. Under different climate scenarios, the suitable areas were reduced to varying degrees. The dominant environmental variables affecting the distribution of P. capitatum were precipitation in the driest month, annual precipitation, and elevation, with a cumulative contribution rate of 84.1%. Against the background of a changing climate, the areas suitable for P. capitatum in China will be widely distributed in the southwestern region, with Guizhou Province and Yunnan Province as the main distribution areas; some sites will also be distributed throughout the southwest of Tibet Autonomous Region, the south of Sichuan Province, the north of Guangxi Autonomous Region, and the coastal area of Fujian Province. Optimal conditions for P. capitatum include a dry month precipitation range of 13.4 to 207.3 mm, elevations from 460.3 to 7214.3 m, and annual precipitation between 810 and 1575 mm. Given these insights, we recommend enhanced conservation efforts in current prime habitats and exploring potential cultivation in newly identified suitable regions to ensure the species’ preservation and sustainable use.

Introduction

Global warming has become a focal point of international concern, with significant implications for the geographical distribution of species¹. The repercussions of climate change are far-reaching, affecting not only biodiversity but also human and environmental health^2,3. The twentieth century witnessed a surge in greenhouse gas emissions due to land use change and fossil fuel combustion, leading to a discernible warming trend⁴, Projected forecasts under medium emission scenarios suggest that southern China will face an extended period of high temperatures and increased frequency of heavy rainfall events by the mid-twenty-first century⁵. The Intergovernmental Panel on Climate Change (IPCC) has outlined six emission scenarios, predicting an atmospheric CO₂ concentration increase between 540 and 970 ppm in the coming century, potentially raising the global average temperature by 1.4 to 5.8 °C⁶. These climatic shifts are expected to intensify extreme weather events, such as hurricanes, floods, and droughts, which could disrupt ecosystem balances and lead to species extinctions or range expansions, thereby impacting the food chain and biodiversity⁷.

The study of potential future distribution changes for areas suitable for the growth of certain species is crucial for developing targeted conservation measures. Environmental niche models (ENMs), such as the Maximum Entropy model (MaxEnt), are reliable tools for predicting suitable habitats and have been extensively used for this purpose^8–15. MaxEnt, in particular, has demonstrated superior performance in predicting potential suitable areas for species, even with limited data^12,16–21. It operates on the principle of maximum entropy to simulate a species' geographical distribution and its response to climate change, constructing a spatial distribution model that is both readable and accurate^12,18–20.

Polygonum capitatum Buch.-Ham. ex D. Don, a traditional Miao-Nationality herbal medicine, is predominantly found in the mountains and fields of southwestern China. It has been used for its anti-inflammatory, antibacterial, and antioxidant properties²². The potential impact of climate change on P. capitatum is of particular interest, as it could affect the availability and quality of this medicinal resource. While recent studies have explored the chemical composition and pharmacological properties of P. capitatum^23–26, there is a knowledge gap regarding the impact of climate change on its suitable distribution areas. This study aims to bridge this gap by integrating climatic factors, topographic variables, and field distribution data of P. capitatum within China. Using MaxEnt and ArcGIS software, we predict the areas suitable for P. capitatum growth and explore the main environmental factors affecting its geographical distribution under current and future climate scenarios.

In the face of climate change, the habitats of many species, including Onosma species²⁷, are undergoing rapid reshaping, leading to significant range shifts and potential extinctions¹³. For endemic species with limited distribution, such as the Onosma species in Iran, climate change poses an acute threat, as they are particularly sensitive to alterations in their specific environmental requirements³. Our study contributes to the understanding of how P. capitatum might respond to these changes, providing valuable insights for its conservation and sustainable use.

Materials and methods

Species occurrence records and sample collection

The species distribution data required for the MaxEnt model refer to the latitude and longitude information of the target species distribution point. The data used in the study were obtained from two sources: (1) Searching the term “Polygonum capitatum” on the website of the China Digital Herbarium (CVH, http://www.cvh.ac.cn/) and extracting the distribution point data for specimens with available latitude and longitude information; (2) searching the relevant literature using the keyword “P. capitatum” in the China National Knowledge Infrastructure (CNKI, https://kns.cnki.net) and extracting the longitude and latitude information of the sampling points after perusal of the literature. If the collected point information only included the name of the administrative village, the XX village of XX town, XX county, XX province was used to convert the latitude and longitude data through the https://map.yanue.net/. In compiling the species distribution data for P. capitatum, we employed a multifaceted approach to ensure a comprehensive and reliable dataset²⁸. Finally, a total of 317 samples of P. capitatum were collected, of which 206 were from CVH and 111 from CNKI.

The Chinese map used in this study was derived from the national basic geographic information system (https://nfgis.nsdi.gov.cn) with map approval number GS (2023) 2762. To eliminate the overfitting limit caused by the close distance between the distribution points of P. capitatum, we used the ArcGIS 10.8 software to establish a fishing net with a spatial resolution of 2.5 arc minutes. When there are multiple distribution points in a fishing net, only those closest to the center of the grid are retained, and the remaining ones are deleted²¹. Finally, 276 distribution points of P. capitatum were used for MaxEnt model analysis (Fig. 1). This dual-pronged approach not only enriched our dataset but also fortified the reliability of our distribution maps, providing a solid foundation for the MaxEnt model analysis.

Fig. 1.

Distribution records of P. capitatum in China.

Environmental variables

The selection of environmental variables is a critical step in the modeling process, as it directly influences the accuracy and reliability of the predictions²⁹. In line with²⁷, we carefully considered a suite of 19 bioclimatic factors and elevation data (Table 1), which are pivotal in determining the species' distribution. Elevation and climate data were obtained from the World Climate Database (http://www.worldclim.org/), including the current and five future periods (2021–2040, 2041–2060, 2061–2080, and 2081–2100). The spatial resolution was 2.5 arc-min grid cells. The future bioclimatic data were selected from the forecast data of the BCC-CSM2-MR model developed by the China National Climate Center of the Sixth International Coupled Model Comparison Program (CMIP6)³⁰. Two scenarios in the new emission scenario sharing socio-economic pathways (SSPs), including the green development path (SSP126) and the high-speed development path (SSP585), were selected for simulation analysis. The SSP126 represents low greenhouse gas emissions, whereas the SSP585 represents high greenhouse gas emission levels³¹. Since there may be different degrees of correlation between the required environmental variables, to avoid over-fitting caused by using too many environmental variables, which would reduce the accuracy of model prediction, we used the AcrGIS 10.8 software to process the environmental factor data. The spatial resolution of the data was 2.5', variables with a high correlation (|r|> 0.8) were removed, and the environmental layer was converted into ASCII format available for the MaxEnt 3.4.1 software. The knife-cut method is generally used to obtain the percentage contribution of each environmental variable¹², and parameters with a small contribution are eliminated.

Table 1.

Environmental variables considered for modeling the potentially suitable habitat of P. capitatum.

Variables	Description	Units
Bio1	Annual mean temperature	℃
Bio2	Mean diurnal range	℃
Bio3	Isothermality	1
Bio4	Temperature seasonality	1
Bio5	Max temperature of warmest month	℃
Bio6	Min temperature of coldest month	℃
Bio7	Temperature annual range	℃
Bio8	Mean temperature of wettest quarter	℃
Bio9	Mean temperature of driest quarter	℃
Bio10	Mean temperature of warmest quarter	℃
Bio11	Mean temperature of coldest quarter	℃
Bio12	Annual precipitation	mm
Bio13	Precipitation of wettest month	mm
Bio14	Precipitation of driest month	mm
Bio15	Precipitation seasonality	1
Bio16	Precipitation of wettest quarter	mm
Bio17	Precipitation of driest quarter	mm
Bio18	Precipitation of warmest quarter	mm
Bio19	Precipitation of coldest quarter	mm
Elevation	Elevation	m

Model simulation and evaluation

After importing 276 P. capitatum sample distribution points and nine environmental factor datasets into the MaxEnt 3.4.1 model software, 75% of the samples was used for training and 25% for model verification. The model was repeated 10 times, and the remaining parameters were default settings. The results were output in ASCII format, and the relative contribution of each environment variable was evaluated using the knife-cut method¹². Subsequently, the output results were imported into the ArcGIs 10.8 software, and the current and future potential areas suitable for P. capitatum were drawn by using the reclassification tool³² and divided according to Jenks' natural breaks.

Niche model evaluation is generally evaluated by ROC (receiver-operating characteristic curve). This approach has been widely used in the evaluation of the potential distribution prediction model of species and is currently a highly recognized diagnostic test evaluation index³³. The area under the ROC curve (AUC) is one of the indicators to judge the prediction accuracy of the model. Because the AUC value is not affected by the threshold, it has a high reliability. In theory, when the AUC value is 0.5–0.7, the model effect is poor; at values from 0.7–0.9, the model effect is general, and at values > 0.9, the model effect is good³⁴.

The contribution degree represents the contribution of a climate factor to the distribution of a species. The importance value of permutation is the value of the environmental variable on the random permutation training point set, expressed as a percentage. The greater the contribution rate and replacement importance value of the climate variable, the more important the role of this climate variable in the potential geographical distribution of a species. The knife-cut test uses and excludes a variable in turn to establish a new model and measures the importance of bioclimatic variables by comparing the differences in regularization training gain, test gain, and AUC values among models³⁵. Via a comprehensive analysis of the contribution of environmental factors, the importance of replacement, and the test results of the knife-cut method output by the MaxEnt 3.4.1 model, the dominant environmental factors affecting the distribution of P. capitatum were determined, and the response curve of the dominant environmental factors was used to analyze the suitable range of P. capitatum in terms of various environmental factors.

Division of potential suitable habitat grades

The division of suitable areas was performed to predict the past, present, and future potential geographical distribution of this species. The Conversion Tools-ASCII to Raster in the ArcGIS 10.8 software converts the ASCII format files in the simulation results obtained after MaxEnt operation into raster data and divides them. In this study, ArcGIS10.8 was employed to visualize the results of the MaxEnt model, and the average value of the output results of the model repeated 10 times was imported into the ArcGIS10.8 software. The reclassification tool was used to draw the suitable area division map of P. capitatum based on the main environmental factors, according to the logistic value. We used the maximum test sensitivity plus specificity logistic threshold to define the habitats suitable for the distribution of P. capitatum³⁶. According to the natural discontinuity point method, the potential suitable areas were divided into four levels: non-suitable areas (0–0.1594), low-suitable areas (0.1594–0.3), medium-suitable areas (0.30–0.50), and high-suitable areas (0.50–1). According to the response curve of each environmental factor in the operation results of the MaxEnt 3.4.1 model, the ecological demand characteristics of the suitable area and the suitable range of main environmental factors were obtained.

Results

Accuracy of the MaxEnt model

After inputting the longitude and latitude data of the distribution points of P. capitatum and the selected nine environmental factor data into MaxEnt and repeating the operation for 10 times, the average AUC value obtained was 0.95 (Fig. 2). Since this value is greater than 0.9, it is significantly higher than the random prediction value, that is, the prediction result is accurate³⁷. This indicates that the potential distribution area of P. capitatum predicted by the MaxEnt model has a good accuracy and high fitting degree, without over-fitting.

Fig. 2.

Receiver operating characteristic (ROC) curve. The values shown are the average of 10 replications.

Environmental variable analysis

The dominant environmental variables affecting the distribution of suitable areas were determined by the predicted percentage contribution (Table 2) and the overlap test results (Fig. 3). The 9 environmental factors screened were used to determine the dominant environmental factors (Table 2) by the knife-cut method. Precipitation of driest month (bio14, 43.7%), elevation (Elevation, 20.7%), annual precipitation (bio12, 19.7%), and annual temperature range (bio7, 12.3%) had the highest contribution rates, with a cumulative contribution rate of 96.4%. The top four environmental factors were annual temperature range (bio7, 72%), elevation (Elevation, 8.2%), precipitation of driest month (bio14, 7.6%), and annual precipitation (bio12, 7.3%), with a cumulative contribution rate of 95.1%.

Table 2.

Percent contributions of environmental variables for P. capitatum suitable sites.

variable	Percent contribution (%)	Permutation importance (%)
bio14	43.7	7.6
elevation	20.7	8.2
bio12	19.7	7.3
bio7	12.3	72
bio11	1.7	1.5
bio2	0.8	0.5
bio15	0.7	1.1
bio5	0.3	0.5
bio18	0.2	1.3

Fig. 3.

Jackknife training gain of each environmental variable for P. capitatum.

In the knife-cut test, the greater the test gain obtained when only a single environment variable is used, the more information the environmental variable contains that other environmental factors do not have. As shown in Fig. 3, when only one variable was used, the three factors with the highest values in the test gain and the regularized training gain were precipitation of driest month (bio14), annual precipitation (bio12), and annual temperature difference (bio7). By excluding a single variable, annual precipitation (bio12), precipitation of driest month (bio14), and elevation further decreased the model regularization training benefit, indicating that the growth of P. capitatum was more correlated with these environmental variables. The dominant environmental variables affecting the growth and distribution of P. capitatum were precipitation of driest month, annual precipitation, and elevation. In addition, the annual range of temperature, the average daily range, the highest temperature in the hottest month, the precipitation in the coldest season, the average temperature in the coldest season, and the seasonality of precipitation also affected the distribution of P. capitatum.

To better demonstrate the influence of environmental factors on the potential distribution of P. capitatum, we selected the above three dominant environmental factors to draw the response curve of the occurrence probability of P. capitatum. When bio14 is higher than 0 mm, the probability of P. capitatum occurrence decreases sharply. When bio14 is higher than 13.4 mm, the probability of its occurrence starts to increase sharply. At a level of 207.3 mm, the probability of occurrence is highest (Fig. 4a). At an elevation above 460.3 m, the possibility of the occurrence of P. capitatum increases sharply. At an elevation of 730.2 m, the possibility of P. capitatum occurrence peaked and started to decline, stabilizing at an elevation from 6523.8–7214.3 m (Fig. 4b). When bio12 is higher than 810 mm, the probability of P. capitatum occurrence increased. At a level of 1,165 mm, the probability of occurrence was highest and started to decrease gradually (Fig. 4c). Based on these findings, the suitable habitat conditions for the occurrence of P. capitatum are precipitation of driest month (bio14) from 13.4 to 207.3 mm, elevation from 460.3 to 7214.3 m, and annual precipitation (bio12) from 810 to 1,575 mm.

Fig. 4.

Response curves of the occurrence probability of P. capitatum to precipitation of the direst month (Bio14) (a), response curves of existence probability of P. capitatum to elevation (b), and response curves of existence probability of P. capitatum to annual precipitation (Bio12) (c) in China.

Distribution under the current climate

Using the ArcGIS 10.8, we visualized the results predicted by MaxEnt 3.4.1 and calculated the area proportion of each suitable grade (Table 3). The current area suitable for P. capitatum in China is 115.09 × 10⁴ km², accounting for 11.99% of China's terrestrial area. The highly suitable area is 34.43 × 10⁴ km², mainly distributed in southwest China and in some coastal areas (Fig. 5), including most of Guizhou Province, northwestern and southeastern Yunnan Province, southeastern Sichuan Province, northern Guangxi Autonomous Region, southern Hunan Province, northern Guangdong Province, and southern Fujian Province. The highly suitable area is concentrated in Guizhou Province, whereas there are some scattered highly suitable areas in Guangdong Province, Hunan Province, Guangxi Autonomous Region, and Fujian Province. The medium suitable area amounts to 40.80 × 10⁴ km², mainly distributed in the southwest of Hubei Province, northeast and southeast of Chongqing Municipality, southern Sichuan Province, most of Yunnan Province, central Guangxi Autonomous Region, western and southern Hunan Province, most of Fujian Province, and northern Guangdong Province. The lowly suitable area is 115.09 × 10⁴ km², mainly distributed in southern and eastern Sichuan Province, central and southwestern Yunnan Province, southern Tibet Autonomous Region, northern Chongqing Municipality, southern Shaanxi Province, central Guangxi Autonomous Region, western and southern Hunan Province, northern Hubei Province, and local areas of Jiangxi Province, Guangdong Province, Fujian Province, Anhui Province, Taiwan Province, Zhejiang Province, and Jiangsu Province.

Table 3.

Areas suitable for the growth of P. capitatum under different climate change scenarios (10⁴ km²).

Period	Not suitable	Total suitable area	Low suitability area	Medium suitability area	High suitability area
Current	844.91	115.09	40.80	39.86	34.43
(2021–2040)-SSP126	879.27	80.73	33.11	33.19	14.43
(2041–2060)-SSP126	890.89	69.11	30.56	32.53	6.02
(2061–2080)-SSP126	882.13	77.87	32.13	31.28	14.47
(2081–2100)-SSP126	879.16	80.84	35.23	34.04	11.56
(2021–2040)-SSP585	863.29	96.71	45.94	33.82	16.95
(2041–2060)-SSP585	875.98	84.02	36.68	34.64	12.70
(2061–2080)-SSP585	907.11	52.89	40.49	12.37	0.04
(2081–2100)-SSP585	922.81	37.19	30.82	6.33	0.03

Fig. 5.

Predictions of the potentially suitable area of P. capitatum under current climate conditions.

Potential distribution under future climate conditions

Under the future climate scenarios, the potential area suitable for P. capitatum in China shows a decreasing trend (Table 3). Under the green development path (SSP126) climate scenario, by 2021–2040, 2041–2060, 2061–2080, and 2081–20,100, the proportion of the potential suitable area is expected to decrease to 8.41%, 7.20%, 8.11%, and 8.42%, respectively. Under the high-speed development path (SSP585) climate scenario, the respective decreases are expected to be 10.07%, 8.75%, 5.51%, and 3.87%.

By analyzing the variation trend of high suitable areas in the core distribution area of P. capitatum under different climate scenarios, we observed that under the green development path (SSP126) climate model (Fig. 6a), the distribution of highly suitable areas decreases and then increases, whereas under the rapid development path of the high fossil fuel consumption (SSP585) climate model (Fig. 6e), the distribution of highly suitable areas decrease over time, indicating that P. capitatum is more sensitive to climate change. Under the SSP126 scenario in 2041–2060 (Fig. 6b), the highly suitable areas in southeastern China are greatly reduced and concentrated in Guizhou and Yunnan provinces, and the lowly and moderately suitable areas in southeastern China are slightly reduced and sporadically distributed, concentrated in the southwestern region. Under the SSP126 scenario in 2061–2080 (Fig. 6c), the highly suitable area is reduced again and evolves into a moderately–lowly suitable area. Under this scenario, the highly suitable area is at risk of disappearing, and the medium and lowly suitable areas in the southeast are gradually disappearing. Under the SSP126 scenario in 2081–2100 (Fig. 6d), the highly suitable area expands towards the northeast and decreases in Yunnan Province, and the moderately and lowly suitable areas increase in a few regions around Yunnan Province and Guizhou Province and the southeastern coastal areas. Under the SSP585 scenario in 2041–2600, the highly suitable area in the southeast almost disappears, that in Yunnan Province is reduced and the moderately suitable area slightly decreases toward the northeast and evolves into a lowly suitable area (Fig. 6f). Under the SSP585 scenario in 2061–2080, the highly suitable areas in eastern Guizhou Province and western Yunnan Province are further reduced, and the junction with Guangxi Autonomous Region is increased. The moderately and lowly suitable areas in the same regions are slightly increased in the northwest direction and decreased in the northeast direction (Fig. 6g). Under the SSP585 scenario in 2081–2100, the highly suitable areas almost disappear, and the moderately and lowly suitable areas have also decreased in size (Fig. 6h). Under this scenario, the growth of P. capitatum is expected to be greatly affected, most likely because of unstable habitat conditions and higher human activities. The suitable climatic conditions of the original highly suitable area would no longer exist, greatly reducing the growth and distribution of this species, with a risk of extinction. This calls for effective measures to protect these plant resources.

Fig. 6.

Potential distribution of P. capitatum under different climate scenarios.

Based on the current distribution of the species, we compared and analyzed the spatial pattern changes of potential suitable areas under eight climate scenarios (Table 4 and Fig. 7). Under the SSP126 climate model, the gain rate was 3.87%–7.60%, and the expansion range was mainly in Linzi City of Tibet Autonomous Region, Lijiang City of Yunnan Province, and Xichang City of Sichuan Province. The loss rate was 42.26%–61.88%, and most areas were lost in Chengdu City of Sichuan Province, Chongqing Municipality, Enshi Prefecture of Hubei Province, and Huaihua City of Hunan Province. The stable rate was 34.25%–50.14%, and most of these areas were located in Guizhou Province, most areas of Yunnan Province, Yibin City of Sichuan Province, and Hechi City of Guangxi Autonomous Region. Under the SSP585 climate model, the gain rate was 0.58%–4.00%, and area expansion was mainly observed in Linzi City of Tibet Autonomous Region, Xichang City of Sichuan Province, Xi 'an City of Shaanxi Province, and Tianshui City of Gansu Province, among others. The loss rate was 41.64%–78.01%, and most areas were lost in Chengdu City of Sichuan Province, Chongqing Municipality, Zunyi City of Guizhou Province, Enshi Prefecture of Hubei Province, Pu 'er City of Yunnan Province, Fuzhou City of Fujian Province. The stable rate was 17.99%–49.78%, and stable areas were mainly distributed in the western part of Guizhou Province, the eastern and northwestern parts of Yunnan Province, and the southern part of Sichuan Province.

Table 4.

Changes in the predicted distribution coverages of P. capitatum under different climate scenarios.

Period	Area (104 km2)			Change (%)
	Gained	Lost	Stable	Gain rate	Loss rate	Stable rate
(2021–2040)-SSP126	5.40	61.87	51.82	4.53	51.95	43.51
(2041–2060)-SSP126	4.20	67.16	37.17	3.87	61.88	34.25
(2061–2080)-SSP126	12.29	68.35	81.09	7.60	42.26	50.14
(2081–2100)-SSP126	11.18	73.10	74.40	7.05	46.07	46.89
(2021–2040)-SSP585	14.57	70.70	84.53	8.58	41.64	49.78
(2041–2060)-SSP585	4.95	62.68	55.45	4.02	50.93	45.05
(2061–2080)-SSP585	5.05	75.06	24.18	4.84	71.97	23.18
(2081–2100)-SSP585	3.91	76.25	17.59	4.00	78.01	17.99

Fig. 7.

Dynamic change map of the predicted areas suitable for the growth of P. capitatum under different climate scenarios.

As shown in Fig. 8, the distribution center of P. capitatum under the current climatic conditions is located in Guizhou Province (106° 44′ E, 26° 52′ N). Under the climatic conditions of SSP126, the distribution center is located in the northern part of Yunnan Province (104° 03′ E, 26° 28′ N). 103° 13′ E, 26° 22′ N), whereas from 2061 to 2080, it is located at the junction of eastern Yunnan Province and western Guizhou Province (104° 34′ E, 26° 15′ N). From 2081 to 2100, the distribution center is expected to be located in Guizhou Province (104° 36′ E, 26° 38′ N). In general, the distribution center of P. capitatum moves to high elevations and in the southwestern direction under the SSP126 scenario. Under the climatic conditions of SSP585, the distribution center of P. capitatum migrates to the west and towards higher elevations. In 2021–2040, the distribution center of P. capitatum is located at the junction of northern Guizhou Province and northern Yunnan Province (104° 37′ E, 27° 21′ N), whereas in 2041–2060, it is expected to be located in northwest Guizhou Province (104° 34′ E, 26° 58′ N). In 2061–2080, the distribution center would be located in the northeastern part of Yunnan Province (103° 21′ E, 26° 43′ N), whereas in 2081–2100, it is expected to be located in the southern part of Sichuan Province (101° 51′ E, 26° 47′ N).

Fig. 8.

Trends of the distribution transfer of P. capitatum under different climate scenarios.

Discussion

MaxEnt model performance

The MaxEnt model’s predictive accuracy is notably superior compared to other environmental niche models³⁸. Generally, there is a random subset of data modeling, and the predictive ability of the model is evaluated based on the AUC value³⁹. The most significant advantage of the MaxEnt model lies in its ability to utilize solely the observed distribution points of the target species for habitat prediction and evaluation. This capability ensures that, even with limited sample sizes, incomplete datasets, or small data volumes, MaxEnt can still generate reliable predictions²¹. In addition, the logistic output improves the model, and the large difference in output values better corresponds to the large difference in applicability. The background sampling of the target group has better prediction performance than the random background sampling. Interestingly, even though random background sampling significantly reduces computational time, it does not compromise the model's performance⁴⁰. Fine-tuning model parameters to suit specific species characteristics can further augment the MaxEnt model's accuracy, leading to more precise predictions⁴¹. In this study, the AUC was 0.95, indicating a high prediction reliability.

Critical environmental factors influencing species distribution

The distribution patterns of plant species are restricted by environmental factors, among which climatic factors are widely considered to be the main factors affecting the distribution of plants. Climate change can promote or inhibit the growth and development of plants, thus affecting their spatial distribution⁴². In our analysis, elevation, temperature (specifically, precipitation during the driest month), and precipitation (including annual totals and precipitation during the driest month) emerged as primary environmental drivers affecting the distribution of P. capitatum. Precipitation is crucial for plant survival, influencing growth and physiological traits⁴³. Although rainwater contains sufficient oxygen, which promotes root respiration after rain, thus facilitating plant growth, some plants prefer humid environments, whereas others grow better under dry conditions. With decreasing rainfall, soil moisture decreases, resulting in a considerable decrease in species richness and species diversity⁴⁴. In our study, the significance of precipitation during the driest month and annual precipitation were particularly pronounced, accounting for 43.7% and 19.7% of the environmental influence, respectively. This underscores high water demand for P. capitatum. As a shade-tolerant species, this plant prefers a shady and humid environment. It can be cultivated on slopes and fields, but it grows poorly in day soils and abandoned mines. The annual temperature difference replacement importance value reached 72%, indicating that temperature changes also play an important role in shaping the distribution pattern of P. capitatum.

Impact of climatic changes on suitable habitats for P. capitatum

The ongoing acceleration of global warming and increased frequency of extreme weather phenomena have led to the contraction and northward or upward migration of suitable habitats for various species⁴⁵. The pace at which species distributions adjust to climatic changes often lags due to their inherent physiological constraints⁴⁶. Although some species may be able to cope with climate change via phenotypic effects or natural selection⁴⁷, their ability to adapt to novel environmental conditions is limited⁴⁸. This necessitates a focused examination of the dynamics of species distribution areas—expansions and contractions. The disappearance of species from their native habitats signals their inability to adapt to altered conditions, potentially leading to extinction⁴⁹. In this study, the BCC-CSM2-MR climate system model was used. Based on two different climate scenarios, we could determine that the area suitable for P. capitatum growth is expected to shrink. However, the overall reduction in suitable habitats significantly outweighs the areas of expansion, underscoring the urgent need for effective conservation strategies to mitigate the impacts of climate change on the distribution of P. capitatum. The insidious effects of climate change are anticipated to significantly alter the environmental conditions that define the suitable habitats for P. capitatum. As elucidated in the study of Thymus species in Iran⁵⁰, climate change not only induces shifts in the bioclimatic envelope but also accelerates the rate of these changes, potentially outpacing the adaptive capacity of species. The alterations in temperature and precipitation patterns, which are core components of the species' ecological niche, could lead to a mismatch between the species' distribution and its optimal habitat conditions.

Conclusions

This study applies the MaxEnt model to forecast P. capitatum's distribution under climate scenarios, showing climate change's profound effect on its habitat. Key factors—dry month precipitation, annual rainfall, and elevation—show moisture's critical role for the species, favoring mesic environments. Under SSP126, habitats may shrink then grow, but overall decline; SSP585 predicts continuous loss, emphasizing the need for climate mitigation and adaptive conservation to protect biodiversity.

Adaptation to climate change is manifested in the shifting distribution patterns of P. capitatum. The species' distribution center is anticipated to migrate towards higher elevations and southwestern regions under the SSP126 scenario, indicative of an adaptive response to environmental changes. Conversely, under the more severe SSP585 scenario, a westward and upward shift is projected, reflecting a natural tendency to seek more favorable conditions in the face of climate adversity. The insights garnered from this study underscore the imperative for conservation strategies that are attuned to the ecological needs of P. capitatum and the anticipated impacts of climate change. The implementation of such strategies will be paramount in preserving this valuable medicinal resource.

Author contributions

Conceptualization, J.L., Y.M., Y.L. and X.G.; Methodology, J.L., Y.M., D.Z. and Y.L.; Formal analysis, J.L. and Y.M.; Investigation, J.L., Y.M. and D.Z.; Software and Supervision, D.Z.; Validation, Y.L.; Data curation and Project administration, Y.L. and X.G.; Writing – review & editing, J.L., D.Z., Y.L. and X.G.; Funding acquisition, J.L. and Y.L. All authors contributed to the article and approved the submitted version.

Data availability

The datasets used and/or analysed during the current study available from the corresponding author on reasonable request.

Competing interests

The authors declare no competing interests.