Shaanxi Province is located in the northwestern part of China (105°29′E-111°15′E and 31°42′N-39°35′N). The area is 2.06 × 10¹¹ m², where the Loess Plateau accounts for 40% of the whole land area of Shaanxi Province. The topography is low in the middle and high in the south and north. The northern part is the Loess Plateau, most of which belongs to the warm temperate monsoon climate, and the northern part of Shaanxi along the Great Wall belongs to the middle temperate monsoon climate; the precipitation varies from 0.40 to 0.60 m. The central part is the Guanzhong Plain, which belongs to the warm temperate monsoon climate, and the rainfall ranges from 0.50 to 0.70 m. The southern part is the Qinling Mountains, which belong to the north subtropical monsoon climate, and the precipitation varies from 0.70 to 0.90 m. Shaanxi Province spans from subtropical, warm temperate, and temperate zones from south to north, with significant climate differences.

The marginal land is highly variable due to its various definitions (Richards et al., 2014; Jiang et al., 2018). Generally, the land unsuitable for food production, ambiguous lower quality land, and economically marginal land (Shortall, 2013) are considered marginal land. According to the method of Yin et al. (2018), this study selected seven types of land (including shrubland (SHL), sparse forest land (SFL), dense grassland (DG), moderately dense grassland (MDG), sparsely dense grassland (SDG), saline land (SAL), and bare land (BL)) as marginal land based on the land use data in 2018 (Xu et al., 2018). Eventually, the total area of marginal land in Shaanxi Province is 1.07 × 10¹¹ m², accounting for 51.9% of the total area of Shaanxi Province.

The Maxent model, based on statistics and machine learning, is a popular software for predicting species distribution (Phillips et al., 2006), and it is available for free (https://biodiversityinformatics.amnh.org/open_source/maxent/). A model was built to forecast species distribution based on environmental layers for a group of grid cells and a collection of observed species sample locations (Phillips et al., 2004). The model’s appropriateness was represented by the function of its environment variables in each grid cell. The higher the value of the function of a particular grid cell, the more suitable the conditions are for the species. The model was a probability distribution of all grid cells, and the selected distribution is the one with maximum entropy (Merow et al., 2013).

The percent contribution and the jackknife test are common indicators to evaluate the importance of environmental predictors in the model (Zhang J.-M. et al., 2021). The percent contribution is only heuristically defined as a function of the particular path of the optimal solution of the Maxent model (Phillips, 2017). The jackknife test creates a model in three ways: (a) removing one variable at a time and relying on the remaining variables; (b) each variable is used individually; and (c) using all variables (Charrua et al., 2020). The response curves depict the Maxent model, which was generated by using only the provided variables (Phillips, 2017), which showed the quantitative relationship between environmental variables and predicted sustainability (Yi et al., 2016; Zhang et al., 2019).

Because the results of the same species distribution predicted by this model somewhat varied each time (Cullen, 2017), the average results of 10 model runs were used in this study. The mean standard deviation of the model after 10 replicates was 0.08. The AUC values of all models performed well in the training data, ranging from 0.86 to 0.94 with a mean of 0.90. While in the test data, the AUC values ranged from 0.65 to 0.94 with a mean of 0.79. The area under the receiver operating characteristic curves with error bars and the average AUC across models (Figure 1) show an excellent model. The mean results of the model after 10 replicates calculations are used to ensure the reliability of the model results. Therefore, the model successfully predicted the distribution of Jerusalem artichoke in theory.

Percentage contributions of the variables in the Maxent model are shown in Figure 2A with a descending order of SLP, DPT, Bio5, Bio1, Bio16, SR7, Bio15, Bio17, and Bio3. Figure 2B illustrates the jackknife test of regularized training gain of variable importance, which shows the variables SLP, Bio5, Bio1, Bio16, DPT, Bio15, SR7, Bio17, and Bio3 in a descending order. SLP and DPT were the most potent predictors of Jerusalem artichoke distribution since SLP and DPT variables contained more information than other factors. The training gain was dramatically reduced when the SLP and DPT variables were excluded. When Bio3 and Bio17 were employed in isolation, they had a minor influence on calculating the spread of Jerusalem artichoke (Figure 2B).

Figure 3 depicts the response curves of the nine significant environmental variables on the distribution of Jerusalem artichoke. It also indicates that the Jerusalem artichoke–favored areas in Shaanxi Province are places where the annual mean temperature (Bio1) is 16.5–18.0°C, the isothermality (Bio3) is 265–275, the max temperature of the warmest month (Bio5) is 30–31°C, the precipitation seasonality (Bio15) is 50–60%, the precipitation of wettest quarter (Bio16) is 0.01–0.02 m, the precipitation of the driest quarter (Bio17) is 0–0.005 m, the soil depth (DPT) is 1.50–1.60 m, the average slope (SLP) is 0–5°, and the July solar radiation (SR7) is 1.66–1.67 × 10⁷ W/m².

The prediction of the Maxent model is shown in Figure 4A, which indicates that the suitable areas for growing Jerusalem artichoke in the Guanzhong Plain (Xianyang, Baoji, Weinan, and Hanzhong) and the Southern Shaanxi (mainly Ankang and Shangluo). Superimposing Figure 4A and the marginal land of Shaanxi Province (Figure 4B) in ArcGIS 10.2 yielded the suitable marginal land for growing Jerusalem artichoke (Figure 4C), which revealed that moderately dense grassland was the dominating land type used for growing Jerusalem artichoke. The classification result of marginal land is shown in Figure 4D, which indicates that the areas suitable for Jerusalem artichoke are mainly in northern and southern parts of Shaanxi Province.

The results of the statistics in Figure 4C are shown in Table 2. The results show that the suitable area of shrubland, sparse forest land, dense grassland, moderately dense grassland, sparsely dense grassland, saline land, and bare land are 1.06 × 10¹⁰ m², 9.94 × 10⁹ m², 1.46 × 10¹⁰ m², 4.00 × 10¹⁰ m², 1.26 × 10¹⁰ m², 1.37 × 10⁸ m², and 1.51 × 10⁸ m², respectively. Accounting for 99.7% of the total appropriate area in the southern (Hanzhong, Ankang, and Shangluo) and northern (Yan’an and Yulin) regions of Shaanxi Province, shrubland, sparse forest land, and dense, sparsely dense, and moderately dense grasslands are best for cultivating Jerusalem artichokes.

The results of the statistics in Figure 4D indicated that incompetent, moderate potential, good potential, and high potential areas for growing Jerusalem artichoke are approximately 1.89 × 10¹⁰ m², 6.29 × 10¹⁰ m², 2.19 × 10¹⁰ m², and 3.19 × 10¹⁰ m², respectively (Table 3). The total suitable area (moderate potential, good potential, and high potential) for growing Jerusalem artichoke in Shaanxi Province is approximately 8.81 × 10¹⁰ m², with the southern (Hanzhong, Ankang, and Shangluo) and northern (Yan’an and Yulin) regions accounting for 42.8% of the total area of Shaanxi Province.

Studies have shown that soil, topography, and solar radiation are also important factors affecting the growth of Jerusalem artichoke (Zorić et al., 2016). Unlike previous studies that were limited to the effects of climatic factors on species distribution (Zhang et al., 2020; Tarnian et al., 2021), this study considered the effects of soil depth, topography, land use, and solar radiation on the distribution of Jerusalem artichoke. This multifactorial pattern explained and predicted species distribution better than a temperature-based pattern, and it was more in line with plant growth needs.

Furthermore, the preliminary Maxent model (Supplementary Appendix A) revealed that not all environmental variables substantially impact Jerusalem artichoke growth. A previous research has shown that modeling with strongly correlated variables might give unexpected results (Sillero and Barbosa, 2020). Moreover, compared with the multifactor-integrated assessment method, the Maxent model can obtain the contribution of environmental factors for species and respond to the impact of each environmental factor on the species. To improve the calculation efficiency of the model, this study screened 15 environmental variables that have a significant effect on the distribution of Jerusalem artichoke in combination with the percent contribution of the preliminary Maxent model, and then selected nine important environmental factors by Pearson’s correlation analysis, excluding covariance among environmental variables. Then to make the model more resilient, the mean results from running the Maxent model 10 times are used, and each model has good accuracy.

Meanwhile, the distribution of Jerusalem artichoke in this study is consistent with previous research (Xu et al., 2013). The total area suitable for the growth of Jerusalem artichoke in this study is approximately 8.81 × 10¹⁰ m², which is less than 11.0 × 10¹⁰ m² given in Liu’s study (Liu, 2011) owing to differences in research methods, environmental factors selection, classification criteria, and land use data. The results of the model, as stated previously, are credible.

Because of differences in evaluation methods, the jackknife and percent contribution results had minor differences in terms of variable orders. Still, the variables that significantly affected the Jerusalem artichoke were consistent, confirming that SLP and DPT were the main environmental variables affecting the distribution of Jerusalem artichoke. They were also compatible with the percent contribution of the preliminary model (Supplementary Appendix A). These findings revealed that topography and soil conditions were the most relevant variables for the Jerusalem artichoke species being modeled.

Furthermore, Jerusalem artichoke is a long day plant that requires a short photoperiod to flower, grows best at 18–26°C (Zhuang et al., 2011), needs abundant rainfall (>0.15 m) (Swanton et al., 1992), and it can survive in both flat and mountainous areas (Wang et al., 2019). The results of this study are compatible with these conditions. Moreover, the findings demonstrated that seasonality significantly impacted Jerusalem artichoke growth, consistent with earlier studies (Zorić et al., 2016; Krivorotova and Sereikaite, 2018).

According to studies, Jerusalem artichoke has a huge potential for energy plant development in the arid and semi-arid areas in the western part of China (Liu et al., 2011). Simultaneously, the development and utilization of Jerusalem artichoke are of great practical significance to assist rural communities in overcoming poverty (Nie et al., 2022). According to the investigation, most marginal land areas are underutilized, and large-scale rational planting has yet to be realized. Jerusalem artichoke only grows sporadically on roadsides and wastelands in some regions. Thus, this study selected the Jerusalem artichoke as an energy plant and researched its potential on marginal lands that can guide the rational planting of Jerusalem artichoke, which is in line with the practical requirements.

Furthermore, the area suited for Jerusalem artichoke differed from Figure 4A, owing to the perfect environmental conditions for cultivating Jerusalem artichoke in Xi’an, Baoji, Xianyang, and Weinan. As shown in Figure 4D, the marginal land ideal for the growth of Jerusalem artichoke in Guanzhong only accounted for 5.40% of Shaanxi Province, indicating that industrialization and urbanization significantly impact the distribution of Jerusalem artichoke species. Furthermore, when the Guanzhong Plain (Xi’an, Baoji, Xianyang, and Weinan) was excluded, the results in Figure 4D showed that the area suitable for planting Jerusalem artichoke in Northern Shaanxi was greater than that in Southern Shaanxi. But, the good potential and high potential locations more ideal for producing Jerusalem artichoke in Northern Shaanxi were less than those in Southern Shaanxi. The differences between Southern and Northern Shaanxi in terms of nine environmental variables are shown in Figure 5, and this result shows significant differences between Southern and Northern Shaanxi at Bio1, Bio15, Bio16, Bio17, and SR7. Thus, these environmental variables make Southern Shaanxi more suitable for planting Jerusalem artichoke than Northern Shaanxi in terms of quality. It can be seen that temperature, precipitation, and solar radiation are important factors affecting the classification of Jerusalem artichoke suitability. Thus, this study’s findings have greater practical implications and can guide smart Jerusalem artichoke planning.