Prediction of the Potential Distributions of Prunus salicina Lindl., Monilinia fructicola, and Their Overlap in China Using MaxEnt

Prunus salicina Lindl. (P. salicina) is an essential cash crop in China, and brown rot (BR) is one of its most important diseases. In this study, we collected geographic location information on P. salicina and Monilinia fructicola (G. Winter) Honey (M. fructicola), one of the BR pathogenic species, and applied the MaxEnt model to simulate its potential suitable distribution in China. There have been discussions about the dominant environmental variables restricting its geographical distribution and their overlap. The results showed that the mean temperature of the coldest quarter, precipitation of the warmest quarter, precipitation in July, and minimum temperatures in January and November were the main climatic variables affecting the potential distribution of P. salicina, while the coldest quarter, precipitation of the driest month, precipitation of March, precipitation of October, maximum temperatures of February, October, and November, and minimum temperature of January were related to the location of M. fructicola. Southern China had suitable conditions for both P. salicina and M. fructicola. Notably, the overlap area of P. salicina and M. fructicola was primarily located southeast of 91°48′ E 27°38′ N to 126°47′ E 41°45′ N. The potential overlap area predicted by our research provided theoretical evidence for the prevention of BR during plum planting.


Introduction
Prunus salicina Lindl. (P. salicina) is a species of plum in the Rosaceae family. It is also known as Jiaqingzi, Bulin, plum, and Jade Emperor plum. P. salicina is cultivated globally, and it is an important temperate fruit [1,2]. P. salicina is a deciduous fruit tree that originated in China and has been cultivated for over 3000 years [3]. At present, more than 20 provinces and municipalities in China grow Prunus. In 2020, China had 1.95 million hectares of harvested area and produced 6.48 million tons of plums, equivalent to 73.86% of the world's plum planting area and 52.98% of the world's plum output, according to the Statistics Department of the World Food and Agriculture Organization [4]. Among the plethora of pathogenic agents attacking Prunus crops, brown rot (BR) is the most important disease [5]. It can damage stone fruit trees, such as peaches (Prunus persica (L.) Batsch), nectarines (Prunus persica var. nucipersica (Suckow) C. K. Schneid), plums, apricots (Prunus armeniaca L.), and sour cherries (Prunus cerasus L.), along with some fruit trees, including apples (Malus pumila Mill.), and pears (Pyrus communis L.) [6][7][8][9]. BR is caused by Monilinia spp., which belongs to the phylum Ascomycota, family Sclerotiniaceae, and order Helotiales [10][11][12][13][14].
BR is widely distributed worldwide, with the most obvious damage in Australia, Asia, America, and Europe [10,[15][16][17][18]. The disease can not only harm the buds, branches, flowers, and fruits during the growth stage but also induce fruit canker in the storage period, resulting in the decline of fruit yield and quality and increased economic losses [10,11,19]. The control of BR in the production, storage, and transportation of fruits can not only reduce financial loss but also limit the spread of the Monilinia spp. to stop the effect from influencing the full fruit export trade [10,11].
BR was reported in China as early as the 1920s and is generally common in China, with six Monilinia spp.  [9,11,[20][21][22][23]. M. yunnanensis was the most widely distributed species in 12 provincial-level administrative regions, mainly harming stone fruit and fruit plants [7]. M. fructicola was followed by M. yunnanensis, which mainly affects stone fruit plants in Beijing, Shandong, and Hebei Provinces [7]. It was followed by M. polystroma, which was mainly distributed in Hebei, Heilongjiang, and Shandong Provinces, and mainly infected fruit plants. M. mumeicola has been detected only in peaches, apricot, and Chongqing sour cherry in Hubei Province [5,24,25]. In this study, we preinvestigated 12 plum plantations in Mianyang city, Nanchong city, Luzhou city, Zigong city, Dazhou city, and Yibin city of Sichuan Province from May to June 2022 and detected M. fructicola at all sites. However, there were no report of the detection of M. fructicola from plum BR in Sichuan province [7,23,25]. We chose this specie as the pathogen marker to explore the geographical distribution of BR in plums by MaxEnt model.
MaxEnt is a widely used software for predicting species geographical distributions, especially when the number of distribution points is uncertain and the correlation between climate and environmental factors is unclear. The MaxEnt model can obtain a high prediction accuracy based only on distribution data [26][27][28], making it possible to analyze and predict the suitable habitat of the pathogen and its host plants. Recently, the MaxEnt model has been widely adopted in predicting plant potential planting areas, animal and plant habitats, invasive plant distribution areas, and quarantine pest prediction [29][30][31][32]. There are a few applications in plant protection, mainly used to analyze the climatic suitability of pests and diseases, predict the invasion possibility of quarantined pests and diseases, and simulate the impact of climate change on the distribution areas of pests and diseases [27,33]. For plant pathology, Wang et al. [34] combined MaxEnt and GIS tools to predict the potentially suitable areas of Diaphorina citri under climate change scenarios in China. Galdino's team first mapped the global scale of the potential risk of the mango's sudden decline by MaxEnt [35]. In addition to the current climatic scenarios, Ruheili et al. [36] utilized future projected climatic scenarios to eliminate the hotspots and proportions of the areas of witches' broom disease in Oman. In our pre-investigation, we detected M. fructicola from plum fruits in Sichuan province, which is not included in the locations of the previous studies of plum BR. We hypothesized there were some locations with the potential to overlap P. salicina and M. fructicola in China, but they had not been acknowledged due to the lack of a mature model in the existing literature. Therefore, we collected geographic location information of P. salicina and M. fructicola by searching databases and the literature, downloaded climate variables from the WorldClim website, and used MaxEnt to simulate the potential suitable distribution of each in China. We evaluated the dominant environmental variables restricting the geographical distribution of P. salicina and M. fructicola, and assessed their overlap to provide evidence for future research and protection against BR.

Climatic Variables Related to P. salicina and M. fructicola
The historical climate data were downloaded from the WorldClim website (https: //www.worldclim.org/, accessed on 3 September 2022) with a spatial resolution of 5 . According to the literature review and the pre-investigation of our study, the climate data included 19 bioclimate factors, reflecting the characteristics and seasonal variation of temperature and precipitation with strong biological significance, monthly average precipitation, monthly average maximum, minimum temperature, etc., and the climate index from 1970 to 2000 (Table 1) [8,13,15,16,24,25,[36][37][38][39][40][41]. Then, the climate variables were extracted from the administrative zoning map of China as the base map. For screening modeling variables, correlation analysis of climate data was implemented using ENMtools software to calculate the Pearson coefficient. Initial climate variables and species distribution data were imported into MaxEnt to calculate the initial contribution rate, and variables with very low contribution rates were removed. Suitable environmental variables were screened based on a Pearson coefficient higher than |0.8| (very significant correlation) and the contribution rate.

Construction and Evaluation of the MaxEnt Model
To construct the MaxEnt model, the species distribution data were transferred into csv files, and the tiff variable layer was converted by the format conversion function of ArcGIS into the ASC layer required by MaxEnt software. The species distribution data and climate variables were imported into the software "Sample" and "Environmental layers", respectively. Response curves for climate variables were created by checking "Create response curves", the predictions were drawn by "Make pictures of predictions", and variable importance was measured through Jackknife analysis. Output format and Output file type are set to default values. In the initial model, the "Random test percentage" of test data was set as 25%. Then, the reconstructed models were set to improve the accuracy. "Random seed" was set as a random proportion, "Regularization Multiplier" was set to 1, and the number of " Replicates" was entered as 10, indicating the model would run 10 times. Other parameters were set to the default software. According to the UN's Intergovernmental Panel on Climate Change (IPCC)'s explanation of the probability of species presence along with the results of previous research, the suitability grades were divided into four categories and displayed in different colors on the map: highly suitable area (p ≥ 0.66, red), moderately suitable area (0.33 ≤ p < 0.66, orange), poorly suitable area (0.05 ≤ p < 0.33, yellow), and unsuitable area (p < 0.05, white) [42,43].
The receiver operating characteristic (ROC) curve output by MaxEnt is one of the most effective and widely-used methods for evaluating the accuracy of niche models by excluding false positive and false negative distribution results [35,[44][45][46][47]. The ROC curve is plotted with a false positive rate (1-specific rate) and true positive rate (1-omission rate) as the horizontal and vertical coordinates according to a series of dichotomies. The area under the curve (AUC) is not affected by the incidence of distribution points and the judgment threshold. The value range of AUC is [0, 1]. The closer the AUC is to 1, the greater the correlation between environmental variables and the distribution model and the higher the accuracy of prediction results. AUC values of 0.5-0.7 indicate poor performance. Values of 0.7-0.9 indicate moderate performance, and a value greater than 0.9 indicates high performance [43,48].

Extraction and Analysis of Overlapping Areas in Suitable Areas
The local analysis function of ArcGIS was utilized to extract the grids of overlapping areas of the total suitable areas of P. salicina and M. fructicola. The distribution in provinces (regions and cities) was calculated according to the grid attributes.

Evaluation of Simulation Results by the MaxEnt Model
According to the local bureau of agricultural statistics, we sampled, identified, and marked the longitude and latitude of the plum plantations using the unit of a county administrative region. After importing the geographic distribution data into ArcGIS, we calculated the distance between the distribution points in the unit grid and the grid centroid and retained the distribution point closest to the centroid. ArcGIS was used to extract the fitness index corresponding to the distribution points in the field survey. The accuracy of the definition of the distribution points in the high-suitability area was 100%, the accuracy of the definition points in the medium-suitability area was 66%, the accuracy of the definition points in the low-suitability area was 33%, and the accuracy of the definition points in the unsuitable area was 0 ( Table 2). The evaluation formula is as follows: A: accuracy; i: suitability level; N: number of field investigation points; X i : number of grade i distribution points; a i : corresponding accuracy of the grade i suitability area. The AUC indexes of the reconstructed model for P. salicina and M. fructicola were 0.954 and 0.961, respectively, indicating their high precision and credibility ( Figure 2). According to the jackknife test, the AUC values of five environmental variables (bio11, bio18, prec07, tmin01, tmin11) were all >0.8, indicating that they were the main factors affecting the potential distribution area of P. salicina (Figure 3a). The AUC values of bio18 and prec07 were the highest, indicating that the warmest season precipitation and the mean precipitation in July were the most important variables affecting the geographical distribution of P. salicina ( Figure 3a). Conversely, seven primary factors affected the distribution of M. fructicola with AUC values over 0.8 (Figure 3b). The AUC of both tmax10 and tmax11 exceeded 0.92, indicating that the maximum temperatures in October and November were the most essential variables influencing the geographical distribution of M. fructicola (Figure 3b). In addition, the mean temperature of the coldest quarter (bio11), precipitation of the warmest quarter (bio18), precipitation of July (prec07), and minimum temperatures of January (tmin01) and November (tmin11) were the main climatic variables affecting the potential distribution of P. salicina, while the coldest quarter (bio11), precipitation of the driest month (bio14), precipitation of March (prec03), precipitation of October (prec10), maximum temperatures of February (tmax02), October (tmax10), and November (tmax11), and minimum temperature of January (tmin01) were the factors most related to the location of M. fructicola (Figure 3).
The AUC indexes of the reconstructed model for P. salicina and M. fructicola were 0.954 and 0.961, respectively, indicating their high precision and credibility (Figure 2). According to the jackknife test, the AUC values of five environmental variables (bio11, bio18, prec07, tmin01, tmin11) were all >0.8, indicating that they were the main factors affecting the potential distribution area of P. salicina (Figure 3a). The AUC values of bio18 and prec07 were the highest, indicating that the warmest season precipitation and the mean precipitation in July were the most important variables affecting the geographical distribution of P. salicina (Figure 3a). Conversely, seven primary factors affected the distribution of M. fructicola with AUC values over 0.8 (Figure 3b). The AUC of both tmax10 and tmax11 exceeded 0.92, indicating that the maximum temperatures in October and November were the most essential variables influencing the geographical distribution of M. fructicola (Figure 3b). In addition, the mean temperature of the coldest quarter (bio11), precipitation of the warmest quarter (bio18), precipitation of July (prec07), and minimum temperatures of January (tmin01) and November (tmin11) were the main climatic variables affecting the potential distribution of P. salicina, while the coldest quarter (bio11), precipitation of the driest month (bio14), precipitation of March (prec03), precipitation of October (prec10), maximum temperatures of February (tmax02), October (tmax10), and November (tmax11), and minimum temperature of January (tmin01) were the factors most related to the location of M. fructicola (Figure 3). The response curves presented the relationship between P. salicina and the aboveselected environmental variables. Filtered by the response probability >0.66, the average temperature of the coldest quarter (bio11) was from −3.82 to 10.36 • C, the warmest quarterly precipitation (bio18) was from 404.8 to 2200 mm, the average precipitation in July (prec07) was from 131.936 to 450.368 mm, the average minimum temperature in January (tmin01) was from −9.158 to 7.538 • C, and the average minimum temperature in November (tmin11) was from −1.251 to 10.78 • C. These ranges indicated the suitable conditions for P. salicina occurrence (Figure 4).  The response curves presented the relationship between P. salicina and the aboveselected environmental variables. Filtered by the response probability >0.66, the average temperature of the coldest quarter (bio11) was from −3.82 to 10.36 °C, the warmest quarterly precipitation (bio18) was from 404.8 to 2200 mm, the average precipitation in July (prec07) was from 131.936 to 450.368 mm, the average minimum temperature in January (tmin01) was from −9.158 to 7.538 °C, and the average minimum temperature in November (tmin11) was from −1.251 to 10.78 °C. These ranges indicated the suitable conditions for P. salicina occurrence (Figure 4).  Conversely, the response curves of M. fructicola are shown in Figure 5. The most suitable variables and ranges for the distribution of M. fructicola were the average temperature of the coldest quarter (bio11) from −3.937 to 11.557 °C, the precipitation of the driest month   Figure 6. The highly suitable areas for P. salicina were mainly located in southern China, including Chongqing, Guizhou, Jiangsu, most of Zhejiang, most of Anhui, most of Guangxi, Yunnan, Fujian, southeastern Sichuan, northwestern Hunan, southern Henan, and Shandong, as well as in Tibet in northern Guangdong and Jiangxi, and in central Taiwan (Figure 6a). The highly suitable area was 148.54 × 104 km 2 , while the total suitable area was 554.14 × 104 km 2 , accounting for 15.41% and 57.50% of China's land, respectively.

EVIEW
10 of 17 Guizhou, Jiangsu, most of Zhejiang, most of Anhui, most of Guangxi, Yunnan, Fujian, southeastern Sichuan, northwestern Hunan, southern Henan, and Shandong, as well as in Tibet in northern Guangdong and Jiangxi, and in central Taiwan (Figure 6a). The highly suitable area was 148.54 × 104 km 2 , while the total suitable area was 554.14 × 104 km 2 , accounting for 15.41% and 57.50% of China's land, respectively. However, there was a smaller, highly suitable area for M. fructicola, which was mainly located in the Yunguichuan Plateau and Chongqing (Figure 6b). The other areas of southern China were marked as moderately suitable for M. fructicola, including Henan, Jiangxi, Jiangsu, Zhejiang, Hunan, Guangxi, and Anhui (Figure 6b). There was a 58.96 × 104 km 2 area marked as highly suitable for M. fructicola. Its total suitable area covered 382.63 × 104 km 2 , accounting for 6.12% and 39.71% of China's land area (Figure 6b), respectively.

The Overlap Area between P. salicina and M. fructicola
The overlap area between P. salicina and M. fructicola is mostly located southeast of line 91°48′ E 27°38′ N to 126°47' E 41°45' N ( Figure 7a). Except for Xinjiang, Tibet, Gansu, Inner Mongolia, Heilongjiang, and Jilin, it covers almost all of the land in southern China, with a total area of 380.125 × 104 km 2 (Figure 7a). This accounted for 68.60% and 99.35% of the suitable area for P. salicina and M. fructicola (Figure 7b,c), respectively, showing a high degree of coincidence. The distribution area of only P. salicina was marked in Hei- However, there was a smaller, highly suitable area for M. fructicola, which was mainly located in the Yunguichuan Plateau and Chongqing (Figure 6b). The other areas of southern China were marked as moderately suitable for M. fructicola, including Henan, Jiangxi, Jiangsu, Zhejiang, Hunan, Guangxi, and Anhui (Figure 6b). There was a 58.96 × 104 km 2 area marked as highly suitable for M. fructicola. Its total suitable area covered 382.63 × 104 km 2 , accounting for 6.12% and 39.71% of China's land area (Figure 6b), respectively.

The Overlap Area between P. salicina and M. fructicola
The overlap area between P. salicina and M. fructicola is mostly located southeast of line 91 • 48 E 27 • 38 N to 126 • 47 E 41 • 45 N (Figure 7a). Except for Xinjiang, Tibet, Gansu, Inner Mongolia, Heilongjiang, and Jilin, it covers almost all of the land in southern China, with a total area of 380.125 × 104 km 2 (Figure 7a). This accounted for 68.60% and 99.35% of the suitable area for P. salicina and M. fructicola (Figure 7b,c), respectively, showing a high degree of coincidence. The distribution area of only P. salicina was marked in Heilongjiang, Jilin, a small part of Inner Mongolia, Qinghai, and Tibet (Figure 7a). Conversely, the distribution area of only M. fructicola was scattered in Tibet and Xinjiang, with an area of 2.502 × 104 km 2 , accounting for 0.65% of its suitable area (Figure 7a).

Independent Sample Evaluation
Field investigation of species distribution is the most direct and reliable way to verify the model. In this study, the accuracy of simulation results was further verified by sample collection from several plum plantations in Sichuan Province. We filtered 12 plum plantations (Table 3), among which 10 were located in highly suitable areas, 2 in moderately suitable areas, and 0 in poorly suitable and unsuitable areas. The representative figure of the plum fruit with BR and cultured M. fructicola is shown in Figure 8. According to Formula (1), the calculation accuracy was 94.33%, indicating that the model's simulation performance was good.

Conclusions and Discussion
Based on the MaxEnt model, we

Conclusions and Discussion
Based on the MaxEnt model, we concluded the key environmental variables affecting the distribution of P. salicina and M. fructicola in China. The average temperature of the coldest quarter (−3.82~10.36 • C), the warmest quarterly precipitation (404.8~2200 mm), the average precipitation in July (131.936~450.368 mm), the average minimum temperature in January (−9.158~7.538 • C), and the average minimum temperature in November (−1.251~10.78 • C) were the suitable conditions for P. salicina occurrence. The essential variables of M. fructicola were the average temperature of the coldest quarter (−3.937~11.557 • C), the precipitation of the driest month (24.64~308.56 mm), the average precipitation amounts in March (36.0192~152.1408 mm) and in October (60.2096~163.2128 mm), the maximum temperatures in February (tmax02) from 10.027 to 23.887 • C, in October (13.756~22.875 • C), and in November (8.49~22.821 • C), and the average minimum temperature in January (−8.928~13.714 • C) (Figures 4 and 5). The key environmental variables predicted that P. salicina was highly suitable to southern China, including Chongqing, Guizhou, Jiangsu, most of Zhejiang, most of Anhui, most of Guangxi, Yunnan, and Fujian, southeastern Sichuan, northwestern Hunan, southern Henan, Shandong, and in Tibet in northern Guangdong and Jiangxi, along with central Taiwan, and M. fructicola was mainly located in the Yunguichuan Plateau and Chongqing in China ( Figure 6). Nevertheless, the overlap area of P. salicina and M. fructicola, which was at risk of plum BR infected by M. fructicola, was mostly located southeast of line 91 • (Figure 7).
Monilinia spp. easily colonize on wounds formed by fruit rupture and produce a large number of conidia from the dead fruit or diseased remnant to infect the flowers or young fruit of the tree when there is enough rain in the spring [37,38]. In our study, the suitable temperature and humidity in winter and spring provided the environmental conditions for M. fructicola survival ( Figure 5). M. fructicola was first found in China in 2003 and has been distributed in Beijing, Shandong, Hebei, and other major stone fruitproducing areas [41]. The distribution areas of M. fructicola around China were Gansu, Yunnan, Chongqing, Hubei, Jiangxi, Fujian, Zhejiang, Shanghai, Shandong, Hebei, Beijing, and Liaoning [5,20,21,23,39]. Based on MaxEnt, our prediction of the suitable areas of M. fructicola was consistent with the conclusions of previous research (Figure 6b). In this study, the unsuitable areas of M. fructicola in China included Northeast China, North China, Northwest China, and the Qinghai-Tibet Plateau, which may stem from the extremely cold and long-duration winters and drought in these areas.
To date, most of the research on BR in China has focused on peach BR. Only a few studies have focused on BR in other stone fruit trees, especially the plum [20,21,24,40,41]. M. fructicola was the second wildly distributed species, followed by M. yunnanensis in China, and mainly affects stone fruit plants in Beijing, Shandong, and Hebei Provinces [7,39]. However, the detected M. fructicola from the plum was only in Beijing, Shandong, Chongqing, and Yunnan [7,23,25]. In recent research, two other species of Monilinia spp., M. fructigena and M. polystroma, have been detected in the plum in China [39,41]. Several studies have suggested that M. laxa, M. fructigena, and M. fructicola have a close genetic relationship, which may contribute to the errors in early molecular sequencing identification [14,23,[49][50][51]. However, our research predicted the overlap area of P. salicina and M. fructicola included, and was larger than the existing records, and only detected M. fructicola from the plum fruits, indicating the limitation of sample collection and the deficiency of research on plum BR. According to our prediction, provinces, including Sichuan, Guizhou, Guangxi, Guangdong, Hunan, Hubei, Henan, Anhui, Jiangxi, Jiangsu, Fujian, Zhejiang, and Taiwan, were in the overlap area and had highly suitable area for P. salicina planting, but didn't have the detection of M. fructicola from previous studies (Figures 6a and 7). Above provinces could be recognized as the potential disease areas of the plum BR caused by M. fructicola, suggesting the need of sampling verification in further researches and the prevention during the plum planting in these areas.
The niche model is based on the assumption that the niche demand of a species is conservative. Factors such as sample size, spatial scale, and climate variables will affect the prediction ability and stability of the model [52]. The species distribution data used in this study were mainly from databases and literature reviews, which ensured the operational requirements of the model, but there were no omissions. In the data obtained by database retrieval and literature review, some distribution points lacked latitudes and longitude and were determined by searching place names through coordinate positioning software, resulting in geographical errors. Furthermore, the occurrence and prevalence of plant pests and diseases are not only affected by climate but also closely related to host conditions, the species and quantity of natural enemies, and the frequency of orchard medication. Environmental factors affecting the distribution of host plants include climate, soil type, vegetation type, topographic factors, variety type, human activities, and socioeconomic structure [53][54][55]. Due to unknown changes in many future factors, and to reduce the complexity of the model, other factors were not included in the environmental variables in this study. It can be speculated that the niche predicted by the MaxEnt model is wider than the actual niche it occupies. In the next step, in addition to the influence of climate factors, the interaction between target species and other factors, the lagged phenomenon of climate change on species distribution, the changes in soil type and vegetation type, and the influence of human activities should also be considered to improve the prediction effectiveness of the model.  Institutional Review Board Statement: Not applicable.

Informed Consent Statement: Not applicable.
Data Availability Statement: All data sets generated and/or analyzed in the current study are available from the corresponding author upon reasonable request.

Conflicts of Interest:
The authors declare no conflict of interest.