China was selected as the research area to analyse the distribution of Ch. utilis. Field investigations were carried out from 2020 to 2023 to study the natural population of Ch. utilis, resulting in the collection of 582 distribution records. Additionally, a review of published literature retrieved 19 records on the natural population distribution of Ch. utilis in Yunnan and Guizhou Provinces (Zhang et al. 2019, Lou 2021b, Wang et al. 2022). The Chinese Virtual Herbarium (CVH, www.cvh.ac.cn/, accessed on 19 September 2022) provided 121 records. Ch. utilis was introduced and cultivated in 2000 as part of the farmland-to-forest project. To mitigate the impact of this project, records from 2000 onwards sourced from the CVH were excluded. Following the removal of duplicate records, only one distribution point was retained for each grid (1 km × 1 km), resulting in 62 valid samples (Fig. 1).

Figure 1.  

Distribution of Ch. utilis in China.

In this study, a total of 19 climate variables, three terrain variables and four soil variables were utilised to develop the MaxEnt model (Table 1). The climate data were sourced from the World Climate Database (WorldClim v.2.1, www.worldclim.org/, accessed on 22 September 2022), encompassing current (1970s - 2000s), future 2050s (2040s - 2060s) and future 2090s (2080s - 2100s) scenarios, with a spatial resolution of 30″, approximately 1 km × 1 km. Future climate data were based on the MIROC6 (Model for Interdisciplinary Research On Climate version 6) model in CMIP6 (Coupled Model Intercomparison Project Phase 6), incorporating SSP1_2.6 (SSP1, sustainable development path), SSP2_4.5 (SSP2, medium development path) and SSP5_8.5 (SSP5, conventional development path) scenarios (Riahi et al. 2017, Zhang et al. 2022, Li et al. 2022). The global elevation model (DEM) data, with a resolution of approximately 1 km × 1 km, was also obtained from the World Climate Database, while altitude, aspect and slope data were derived from the DEM. Furthermore, four soil variables were extracted from the Harmonised World Soil Database (www.fao.org/soils-portal/soil-survey/soil-maps-and-databases/harmonized-world-soil-database-v12/en/, accessed on 25 March 2024). It was assumed that the terrain and soil variables remained constant when predicting future suitable habitats.

In order to prevent overfitting, a comprehensive assessment of 26 climate variables was conducted. The screening process consisted of two main steps: (1) Utilising SPSS 23.0 software to examine the correlation amongst the 26 climate variables, with a threshold of 0.80 set for determination; and (2) Running MaxEnt 3.4.1 software with species distribution data and the 26 climate variables to determine the initial percentage contribution of each variable to the model. Following this, variables with a correlation coefficient above 0.80 and a lower contribution rate were excluded (Fang et al. 2022). Ultimately, eleven climate variable factors were selected for the subsequent prediction of Ch. utilis distribution (Table 1).

The prediction of the potential distribution area of Ch. utilis was conducted using the MaxEnt model. The MaxEnt model derived constraint conditions based on the distribution data of species and environmental factors. It assumes that the probability distribution of species emergence is closest to its actual distribution when the entropy is maximised under these constraints (Phillips et al. 2006). In this study, distribution data for 62 Ch. utilis and 26 environmental variables were input into MaxEnt 3.4.1 software. A total of 25% of the distribution points were utilised as test data, while 75% were used as training data. The simulation was repeated 10 times with other parameters remaining constant.

The prediction accuracy was evaluated using the receiver operating characteristic (ROC) curve and the area under the ROC (AUC). AUC values range from 0 to 1, with higher values indicating better prediction accuracy. Specific criteria were defined as follows: 0.5-0.6 for failure, 0.6-0.7 for poor, 0.7-0.8 for fair, 0.8-0.9 for good and 0.9-1 for excellent. (Araújo et al. 2005, Pan et al. 2020).

The output results of the MaxEnt model were visualised using ArcGIS 10.5 software to analyse habitat suitability. Habitat suitability was assessed by categorising habitat into four levels using the natural-breaks classification method: unsuitable habitat (0-0.05), marginally suitable habitat (0.05-0.22), highly suitable habitat (0.22-0.51) and most suitable habitat (0.51-1). The suitable habitat for each province was determined by overlaying administrative divisions with suitable areas. Finally, a transfer matrix was utilised to analyse the relationship between the current suitable habitat and the future suitable habitat.

The MaxEnt model was utilised to predict potentially suitable habitats for Ch. utilis. The evaluation of the ROC curve results showed that the average AUC value of the training data was 0.997, indicating high reliability in the prediction outcomes. (Fig. 2).

Figure 2.  

Receiver operating characteristic curve of Ch. utilis for MaxEnt model.

The precipitation in the driest month (Bio14) made the highest percentage contribution to the prediction model at 45.70%. This was followed by altitude (Alt) and isothermality (Bio03), which contributed 33.60% and 17.70%, respectively, resulting in a cumulative contribution of 97.00%. The contributions of the other factors were minimal (Table 2).

The curve of the distribution probability and environmental factor response, as shown in Fig. 3, indicates a pattern of slow increase, rapid increase, rapid decrease and slow decrease with increasing environmental factor values. The probability of occurrence exceeded 0.50. The environmental characteristics within the distribution area of Ch. utilis included a range of precipitation in the driest month (Bio14) of 16.31-25.26 mm, an altitude (Alt) range of 1288.86-2005.11 m and an isothermality (Bio03) range of 23.03-26.91.

Figure 3.  

Response curves of the probability of the main climate factors.

The total suitable area of Ch. utilis in China under the current climate was 10.55 × 10⁴ km² (Table 3 and Fig. 4). Suitable habitats were predominantly found in the south-western region of China, encompassing north-eastern Yunnan Province, south-eastern Sichuan Province, most of Guizhou Province, southern and northern Chongqing Municipality, western Hubei Province and southern Shanxi Province.

Figure 4.  

Potential distribution of Ch. utilis in China under the current climate.

The most suitable habitat area was 0.85 × 10⁴ km², accounting for 8.06% of the total suitable area. These habitats were primarily located in the junctional areas of Yunnan, Guizhou, Sichuan Provinces and Chongqing Municipality, specifically in the Daloushan Mountains and Wumeng Mountains. The highly suitable habitat area was 2.16 × 10⁴ km², making up 20.47% of the total suitable area. These habitats were mainly found in north-eastern Yunnan Provinces, south-eastern Sichuan Provinces, the central and northern parts of Guizhou Provinces, and the southern and northern parts of Chongqing Municipality. The marginally suitable habitat area was 7.54 × 10⁴ km², accounting for 71.47% of the total suitable area. It is predominantly found in south-western Hubei Province, southern Shaanxi Province, north-eastern Chongqing Municipality, north-eastern Yunnan Province, south-eastern Sichuan Province and most of Guizhou Province. A comparison with the natural distribution revealed that the model predicted a much larger range. Despite some deviations, the core distribution area matched the current distribution area.

We predicted the potential distribution of Ch. utilis in China in two future periods (2050s and 2090s) and three greenhouse gas scenarios (SSP1, SSP2 and SSP5).

In the 2050s and 2090s, the total suitable habitat area for Ch. utilis decreased by 3.79% and 1.04%, respectively, under the SSP1 scenario, reaching 10.15 × 10⁴ km² and 10.44 × 10⁴ km^2, respectively (Table 3 and Fig. 5). Conversely, the total area increased by 0.38% and 6.16% under the SSP2 scenario, totalling 10.59 × 10⁴ km² and 11.20 × 10⁴ km², respectively, for the same periods. Similarly, under the SSP5 scenario, the total area expanded by 1.14% and 10.52%, reaching 10.67 × 10⁴ km² and 11.66 × 10⁴ km² for the 2050s and 2090s, respectively. The most suitable habitats for Ch. utilis generally decreased, except in the 2050s-SSP1 scenario. While the highly suitable habitats decreased under the SSP1 scenario, they increased under the SSP2 and SSP5 scenarios. Conversely, marginally suitable habitat areas increased, except in the 2050s-SSP1 scenario. Generally, in the future, the suitable habitat area will decrease in the southern region (including Yunnan, Guizhou, Hunan and Hubei Provinces; Chongqing Municipality; and Guangxi Zhuang Autonomous Region), but increase in the northern region (such as Gansu, Sichuan and Shaanxi Provinces).

Figure 5.  

Potential distribution of Ch. utilis under future climate scenarios. Panels (a, b, c) indicate the potential distributions under the three scenarios in the 2050s; panels (d, e, f) indicate the potential distributions under the three scenarios in the 2090s.

To highlight the changes in suitable area between the current and future scenarios, we used a transition matrix to analyse internal changes (in the current and 2090s) without considering the total area of unsuitable areas.

Under the SSP1 scenario, the area of grade change from the present to the 2090s was 11.27 × 10⁴ km² (Table 4). Specifically, the marginally suitable habitat of 1.04 × 10⁴ km² transitioned into unsuitable habitat (0.83 × 10⁴ km²) and highly suitable habitat (0.21 × 10⁴ km²). The highly suitable habitat area of 0.57 × 10⁴ km² transformed into marginal habitat (0.50 × 10⁴ km²) and most suitable habitat (0.07 × 10⁴ km²), while the most suitable habitat area of 0.20 × 10⁴ km² transitioned into highly suitable habitat. Additionally, an area of 1.00 × 10⁴ km² became a more suitable habitat, with 0.72 × 10⁴ km² of this area becoming a new suitable habitat. Conversely, the habitat quality decreased by an area of 1.53 × 10⁴ km², with 0.83 × 10⁴ km² of suitable habitat being lost.

Under the SSP2 scenario, the area of grade change from the present to the 2090s was 11.81 × 10⁴ km² (Table 5). The marginally suitable habitat with an area of 1.14 × 10⁴ km² became unsuitable habitat (0.61 × 10⁴ km²) and highly suitable habitat (0.53 × 10⁴ km²), the highly suitable habitat with an area of 0.46 × 10⁴ km² became marginally suitable habitat (0.34 × 10⁴ km²) and most suitable habitat (0.12 × 10⁴ km²) and the most suitable habitat with an area of 0.14 × 10⁴ km² became the highly suitable habitat. An area of 1.91 × 10⁴ km² became a more suitable habitat, of which 1.26 × 10⁴ km² of the area became a new suitable habitat. Conversely, the habitat quality decreased by 1.09 × 10⁴ km² and 0.61 × 10⁴ km² of suitable habitat was lost.

Under the SSP5 scenario, the area of grade change from the present to the 2090s was 12.17 × 10⁴ km² (Table 6). The marginally suitable habitat with an area of 0.89 × 10⁴ km² became unsuitable habitat (0.51 × 10⁴ km²), the highly suitable habitat (0.38 × 10⁴ km²), the highly suitable habitat with an area of 0.39 × 10⁴ km² became marginally suitable habitat (0.30 × 10⁴ km²) and the most suitable habitat (0.09 × 10⁴ km²) and the most suitable habitat with an area of 0.15 × 10⁴ km² became the highly suitable habitat. An area of 2.09 × 10⁴ km² became a more suitable habitat, of which 1.62 × 10⁴ km² became a new suitable habitat. Conversely, the habitat quality decreased by 1.26 × 10⁴ km² and 0.51 × 10⁴ km² of suitable habitat was lost.

Amongst the 11 environmental factors considered in the modelling process, precipitation had the highest total contribution rate at 46.00%, followed by terrain at 33.70%, temperature at 19.10% and soil at 1.20% (Table 2). This indicates that precipitation and terrain were the primary factors influencing the distribution of Ch. utilis. The results from the Jackknife test and single response curves revealed that the precipitation of the driest month (Bio14) was the most significant factor, with an optimal growth range between 16.31-25.26 mm, followed by altitude (Alt, 1288.86-2005.11 m) and isothermality (Bio03, 23.03-26.91) (Fig. 3). Previous studies have indicated that precipitation is a crucial environmental factor influencing the distribution and growth of bamboo (Zhou 1991), which is in line with the findings of this study (Lou 1991, Zhou 1991, Li et al. 2019). Specifically, precipitation during the driest month emerged as the primary environmental factor shaping the distribution of Ch. utilis, showing a high contribution rate and significant permutation importance in the MaxEnt model. It has the same habitat preference as Chimonobambusa hejiangensis (Wang et al. 2019). Furthermore, because of its strong adaptability to soil, the contribution rate related to soil was small (Liu et al. 2022). Currently, the natural Ch. utilis is mainly naturally distributed in south-western China within an altitude range of 1400 m to 2200 m (Ding et al. 2011). The model predicts that the optimal growth range altitude for the bamboo species is between 1288.86-2005.11 m, suggesting that it may be found in low-altitude areas. The differences in distribution at high altitudes may be attributed to the limited sample size used in model construction (62) and the incomplete coverage of the species' entire environmental range in the sample data, impacting prediction results (Chen et al. 2012). As the sample size increases, prediction accuracy follows a non-linear pattern and eventually stabilises (Ji et al. 2019). To improve predictive accuracy, it is advised to incorporate training data that spans the species' entire environmental range when utilising the MaxEnt model for habitat predictions.

Under the current climate, the total suitable habitat area for Ch. utilis in China is 10.55×10⁴ km² (Table 3), with the most suitable habitats mainly concentrated in the junction area of Guizhou, Sichuan Province and Chongqing Municipality (Fig. 4). These regions are located in the transition area between the Sichuan Basin and the Yunnan-Guizhou Plateau, characterised by a sub-tropical monsoon climate with hot and rainy summers and mild and humid winters. This climate aligns with Ch. utilis's preference for low temperatures and high precipitation (Wu and Han 1982, Lou 1991). Research indicates that the most suitable habitats are likely to harbour rich genetic diversity and are core regions for germplasm resource distribution (Wu et al. 2018). Therefore, it is recommended to prioritise investigations, resource collection and protection efforts in these regions. The highly suitable habitat for Ch. utilis is mainly found in north-eastern Yunnan Guizhou, south-eastern Sichuan Guizhou, most of Guizhou Province and the southern and northern parts of Chongqing Municipality. These regions provide suitable conditions for the large-scale introduction and domestication of Ch. utilis (such as Dafang County and Qingzhen City in Guizhou Province and Chaotian and Zhaohua areas of Guangyuan City at the junction of Sichuan, Shaanxi and Gansu Provinces) (Lou 2021a). Moreover, it is recommended to conduct resource surveys in these areas to investigate the possible existence of wild Ch. utilis populations. There were also many areas of marginally suitable habitats in western Hubei Province, south-western Shaanxi Province, south-eastern Sichuan Province, south-western Chongqing Municipality, most parts of Guizhou Province and south-western Taiwan Province, where the actual local conditions should be considered for the introduction of Ch. utilis.

From the current to the 2050s, the area of Ch. utilis varied from -3.79% to 1.14%, with a noticeable expansion only observed under the SSP5 scenario. Moving forward to the 2090s, the suitable habitat area for Ch. utilis varied from -1.04% to 10.52%, showing an expansion trend under both the SSP2 and SSP5 scenarios (Table 3). In the face of future climate change, the boundaries of Ch. utilis' suitable habitat underwent irregular changes, with a noticeable northward expansion trend, although the overall area change remained minimal. The suitable habitats in Yunnan, Hubei and Guizhou Provinces, Chongqing Municipality and Guangxi Zhuang Autonomous Region were significantly impacted by climate change in future scenarios, leading to a decrease. These changes are attributed to rising temperatures and falling precipitation in south-western China（Pan et al. 2020, Zhang et al. 2020）. Measures should be implemented to rescue and protect the existing Ch. utilis resources in these areas to prevent the loss of valuable germplasm resources. Additionally, it is essential to make efforts to introduce, cultivate and promote the existence of Ch. utilis in the above regions. Numerous studies have indicated a tendency for species' suitable habitats to shift towards higher latitudes and decrease in size due to climate change (Li et al. 2021, Li et al. 2022). In contrast, our research findings suggest a rising trend in the area of suitable bamboo habitats (Gui et al. 2024). Despite the emergence of new suitable habitats, the overall quality of the habitats is declining (Tables 4, 5, 6), potentially leading to a decrease or even disappearance of the most suitable habitat area if current trends persist (Rinawati et al. 2013, Jian et al. 2022).