Next Article in Journal
Influence of Perceived Sensory Dimensions on Cultural Ecosystem Benefits of National Forest Parks Based on Public Participation: The Case of Fuzhou National Forest Park
Previous Article in Journal
Impact and Spatial Effect of Government Environmental Policy on Forestry Eco-Efficiency—Examining China’s National Ecological Civilization Pilot Zone Policy
 
 
Search for Articles:
Title / Keyword
Author / Affiliation / Email
Journal
Article Type
 
 
Advanced Search
 
Section
Special Issue
Volume
Issue
Number
Page
 
Journals
Forests
Volume 15
Issue 8
10.3390/f15081313
forests-logo
Submit to this Journal Review for this Journal Propose a Special Issue
► Article Menu

Article Menu

share Share announcement Help format_quote Cite question_answer Discuss in SciProfiles thumb_up
...
Endorse textsms
...
Comment
first_page
settings
Order Article Reprints
Font Type:
Arial Georgia Verdana
Font Size:
Aa Aa Aa
Line Spacing:
Column Width:
Background:
Article

Analysis of the Potential Distribution of Shoot Blight of Larch in China Based on the Optimized MaxEnt and Biomod2 Ensemble Models

by
Xiuyun Zhang
1,
Wenhui Wu
1 and
Yingmei Liang
2,*
1
Key Laboratory for Silviculture and Conservation of Ministry of Education, Beijing Forestry University, Beijing 100083, China
2
Museum of Beijing Forestry University, Beijing Forestry University, Beijing 100083, China
*
Author to whom correspondence should be addressed.
Forests 2024, 15(8), 1313; https://doi.org/10.3390/f15081313
Submission received: 10 June 2024 / Revised: 7 July 2024 / Accepted: 25 July 2024 / Published: 27 July 2024
(This article belongs to the Section Forest Health)
Browse Figures
Versions Notes

Abstract

:
Shoot blight of larch caused by Neofusicoccum laricinum (Sawada) Y. Hattori & C. Nakash poses a significant threat to the growth and development of larch plantations and is among the most devastating diseases of forest trees. Its consecutive occurrence can cause serious damage and even death of the host plant. Analyzing the geographical distribution patterns of shoot blight of larch in China based on the optimized maximum entropy (MaxEnt) and Biomod2 ensemble (EM) models and recognizing the environmental factors limiting the spread of this disease could provide a reasonable basis for its control. The potential geographical distribution areas of shoot blight of larch were predicted using occurrence data and environmental variables. The area under the receiver operating characteristic (ROC) curve (AUC) was employed to compare the predictive performance of the optimized MaxEnt and Biomod2 ensemble models. Our results showed that both models had a prominent performance in predicting the potential distribution of shoot blight of larch, with the latter performing slightly better based on the AUC than the former. The potentially suitable areas for shoot blight of larch, as predicted by both the MaxEnt and Biomod2 ensemble models, were similarly distributed, mainly in Northern China, including Heilongjiang, Jilin, Liaoning, and Northeastern Inner Mongolia. The environmental variables significantly limiting the distribution of shoot blight of larch identified using the jackknife method and Pearson’s correlation analysis included the annual mean temperature, annual precipitation, precipitation of the wettest quarter, mean temperature of the warmest quarter, and elevation. This research offers a theoretical basis for rationally delineating potential sites invaded by shoot blight of larch, strengthening the detection and quarantine in critical areas, formulating timely effective control measures, and establishing conservation measures for larch resources.

1. Introduction

Larch (Larix Mill.), a vital conifer species for afforestation in China, has the advantages of rapid forest formation, high adaptability, and effective soil and water conservation [1]. Shoot blight of larch, caused by a fungal pathogen recently renamed Neofusicoccum laricinum (Sawada) Y. Hattori & C. Nakash [2], is one of the most severe plant diseases of branches in larch plantations, and it first occurred in the early 1970s in Jilin, China [3]. In the past ten years, it has spread throughout the vast forest areas in the three provinces of Northeast China [3]. In addition, the disease has been distributed to varying degrees in North Korea, the Philippines, the United Kingdom, Canada, and other countries [4]. The disease has thus far destroyed larch forests in Northern China to varying degrees. It also poses a serious threat to larch trees, forests, and nurseries, especially young and middle-aged 6–15-year-old trees. In the early stages of disease occurrence, slight withering and bending are observed, while in successive years, it causes crown baldness. The occurrence of this disease gradually reduces the diameter at breast height, height, and growth of larch trees, and it even leads to their death in severe cases. The occurrence, distribution, and prevalence of this disease are vital factors that severely limit the development and growth of larch plantations in Northern China [4]. Since shoot blight of larch is a dangerous disease that causes serious harm and spreads rapidly, the country has already put it on the list of quarantine objects. In the latest List of Key Management Invasive Species published by the Ministry of Agriculture and Rural Development and six other departments, shoot blight of larch has been identified as one of the five main plant diseases (http://beijing.customs.gov.cn).
Species distribution models (SDMs) are indeed critical techniques in various research fields, such as ecology and conservation biology. They are used to quantify the realized environmental niches of species and can be projected onto geographic space to estimate the potentially suitable distribution of a species based on the relationships between species occurrence data and a set of environmental factors [5]. SDMs have gained significant importance in recent years, particularly in assessing the potential impacts of climate change on species, assisting in conserving rare and endangered species, assessing species invasions, and forecasting disease spread [6]. In recent decades, SDMs have rapidly developed, with various methods emerging, such as random forest (RF), generalized linear models (GLMs), and the maximum entropy model (MaxEnt) [7,8]. MaxEnt has gained popularity as an SDM method due to its numerous benefits, including its accessibility and user-friendly interface. Compared to other prediction models, MaxEnt exhibits higher simulation accuracy, requires a smaller sample size, and is less affected by sampling bias [9,10]. Moreover, the MaxEnt model allows both continuous and categorical data to be used as inputs [9]. However, some researchers have criticized that complex species–environment relationships may fit well with a specific dataset but perform poorly on others, indicating potential overfitting [11,12,13]. MaxEnt can address this complexity by adjusting its settings appropriately [8,14]. It has been shown that optimized MaxEnt models can notably improve predictive performance and produce more precise and sound species distribution [15,16,17].
Due to the different principles of algorithms, each model has inimitable advantages and disadvantages. There are many factors to consider when selecting optimal SDMs for any given species, including data availability and resolution, a species’ niche characteristics, and the complexity of the environment [5]. Users of SDMs cannot necessarily straightforwardly determine the optimality of the type of algorithm used for their situation. Without application-specific metrics for evaluating model performance, SDM researchers tend to employ algorithms that generate the most precise prediction results. Algorithm performance differs widely depending on the situation, and previous comparisons have not consistently identified a superior algorithm in any class [5,18]. Ensemble-modeling techniques have been introduced as potential solutions to address this issue. This approach integrates information from individual models fitted with different techniques and has been proposed [19,20]. SDM users widely apply ensemble modeling, believing that it enhances prediction compared to individual models [21]. However, empirical studies on how well ensemble models make predictions in comparison to individual SDMs are still lacking. One study [22] reported no improvement in the performance of ensemble models over that of individual models [23]. By comparison, another study [24] found that ensemble models outperformed individual models. Since the structure and source of validation data can significantly influence model performance assessment [25] and the testing of ensemble models remains limited, our understanding of ensemble performance in different settings is still fairly limited.
The identification of the geographical region affected by shoot blight of larch is fundamental and important for effective and timely control. The objectives of our study were as follows: (1) to create the distribution map and define the key geographical habitats of shoot blight of larch in China, (2) to specify the vital environmental factors affecting the distribution of shoot blight of larch, and (3) to verify and compare the performance of the optimized MaxEnt and ensemble Biomod2 models in anticipating the geographical region of shoot blight of larch. This research offers a crucial theoretical perspective that can significantly contribute to the implementation of monitoring mechanisms, early warning systems, and preventative measures. An additional objective of this study is to bridge the existing knowledge gap by testing the predictive capabilities of the popular Biomod2 ensemble and optimized MaxEnt models utilizing available data on the shoot blight of larch, providing valuable advice on the selection of the appropriate method from the Biomod2 ensemble and optimized models for the precise prediction of plant disease distribution.

2. Materials and Methods

2.1. Occurrence Data of the Shoot Blight of Larch

A set of records of shoot blight of larch was collected from the following several staple pathways: (1) data collected by the members of the research group who went to Tahe County in Heilongjiang Province, Alihe Forest Farm in Inner Mongolia, Chengdei in Hebei Province, Haicheng and Fushun cities in Liaoning Province, Dunhua and Yanji cities in Jilin Province, LiuPanShan in Ningxia Province, Ningshan County in Shaanxi Province, and other places to carry out the field investigation of the occurrence of shoot blight of larch. (2) Published journal articles and books from the Web of Science and CNKI. (3) Documents published by various quarantine authorities and contacting the relevant forestry authorities.
In order to mitigate the sampling bias of the data, we further dealt with the data on the geographic distribution of our target species using the ENMTools method [26], which selects distribution records so that only one observation is kept per 2.5 arc min grid cell. Finally, 67 sites with records of shoot blight of larch were retained.

2.2. Environmental Factors

The raster of the environmental factors employed, including 1 topographical variable (elevation) and 19 bioclimatic variables (Table 1), were obtained from the WorldClim website [27] for 1970–2000. The suitability of climate for the growth and development of various species, including plant pathogens, can be reflected by these variables [28,29]. Moreover, China’s nationwide environmental data were obtained using ArcGIS software 10.4.
To diminish the multicollinearity between environmental factors that leads to model overfitting, a correlation analysis of 19 environmental and topographical variables was implemented with R 3.6.3 When two environmental variables were too strongly correlated (r > 0.8), the one with the greater impact contribution to the response was retained for subsequent model simulations [15]. Among the 20 default environment variables, only 8 were ultimately selected for entering into the model, including mean annual temperature (bio1), isothermality (bio3), temperature seasonality (bio4), mean annual temperature (bio10), annual precipitation (bio12), precipitation of wettest quarter (bio16), precipitation of warmest quarter (bio18), and elevation.

2.3. Optimized MaxEnt Model

We used the R package Kuenm [30] to optimize the model parameters, such as regularization multipliers (RM) and feature classes (FC), which have a great impact on the accuracy of MaxEnt. This model offers five features, including a fragmented feature (H), a linear feature (L), a product feature (P), a threshold feature (T), and a quadratic feature (Q). The Kuenm package tested 1160 candidate models and selected the final parameters based on the delta AICs value in the Akaike minimum information criterion (AIC) method. The model with the lowest delta AICc value was considered the optimal model [31]. MaxEnt 3.4.3 was used to construct the optimal MaxEnt model with the lowest delta AICc values. In the model, all presence data were randomly divided into a training set containing 75% of the presence data and a test set containing the remaining 25%. Ten replicated runs were employed in the construction of the model to decrease the uncertainty of the MaxEnt model, and the final output was the average of these repeated runs. The respective and corresponding contributions of each environmental factor were assessed through the jackknife test in the MaxEnt model [14].

2.4. The Biomod2 Ensemble Model

In the modeling processes, we used the effective number of disease occurrences and environmental variables as inputs into the models with 10 iterations. After evaluating the accuracy of the 10 models, we finally selected 8 models with average AUC values greater than 0.9, including classification tree analysis (CTA), GLM, flexible discriminant analysis (FDA), MaxEnt, artificial neural networks (ANN), generalized boosted models (GBM), one rectilinear envelope similar to BIOCLM (SRE), and RF, to construct the ensemble model. All model construction procedures were performed in the Biomod2 package. The specific steps were as follows: 75% of the 67 samples of shoot blight of larch were randomly chosen as the training data, and the remaining 25% of the samples were employed as the test data. This process of dividing the two parts of the data was repeated 10 times. Each model employed the default settings of the Biomod2 model, except for the MaxEnt model, which employed combinations of Kuenm-defined feature classes and regularization multipliers. Moreover, to decrease the uncertainty of random sampling for model-building, 1000 pseudoabsence points were stochastically selected, with 3 replications. Eventually, 240 various models were developed for this purpose.

2.5. Evaluation of the Models

The area under the receiver operating characteristic (ROC) curve (AUC) was used to assess the credibility of the models. The AUC generally ranges from 0.5 to 1. A higher AUC close to 1 indicates greater credibility of the model. However, the model is denoted by having no predictive accuracy, poor predictive accuracy, normal predictive accuracy, moderate predictive accuracy, and high predictive accuracy when the AUC values range from 0.5 to 0.6, 0.6 to 0.7, 0.7 to 0.8, 0.8 to 0.9, and 0.9 to 1.0, respectively [32].

2.6. Classification of the Potentially Suitable Distribution

The probability of the output value of each grid cell in the model being between 0 and 1 can reflect the relative suitability. We reclassified the model output with the Jenks natural breaks method: grid cells with suitability values below 0.2 were considered unsuitable regions, and the remaining grid cells were categorized into three groups, in order of low-suitability regions (0.2–0.4), medium-suitability regions (0.4–0.6), and high-suitability regions (0.6–1.0). ArcGIS software was used to calculate the proportion of potentially suitable areas (PSAs) and create visualizing distribution maps of the final results [33].

3. Results

3.1. Model Parameter Optimization and Performance

The optimal MaxEnt model was used to predict the potential distribution areas of shoot blight of larch based on 67 points of shoot blight if larch was detected and eight environmental variables. With the default settings of the MaxEnt parameters, the regularization multiplier RM = 1, the feature class FC = LQHPT, and the delta AICc value = 243.70 were set. The model is optimum depending on the AIC value when RM = 3.2, FC = LQH, and the delta AICc value = 0. Therefore, we chose RM = 3.2 and FC = LQH as the optimal parameter settings for this model.
The accuracy of the model was assessed using AUC values. We performed a prediction accuracy test on the distribution of shoot blight of larch, and the AUC values of the MaxEnt model and Biomod2 ensemble model were 0.989 and 0.931, respectively, which showed that these two models had good prediction accuracy.

3.2. The Importance of Environmental Variables

The potential distribution of the shoot blight of larch was mostly influenced by six variables used in the MaxEnt model, which together contributed to 95.2% of the model’s prediction accuracy. The analysis of the contribution of a single factor revealed that the precipitation of the wettest quarter bio16 (25.6%), temperature seasonality bio4 (25.3%), mean annual temperature (14.8%), elevation (12%), annual precipitation (10.8%), and mean temperature of the warmest quarter (6.7%) were the main variables affecting model performance. Precipitation of the warmest quarter (3.4%) and isothermality (1.3%) made a total contribution of 4.7%. The total contribution rate of temperature-dependent factors was 73.7%, while that of precipitation-dependent factors was 14.2%. The jackknife test of the regularized training gain and test gain for variable importance indicated that bio12, bio10, and bio16 were the most important variables (Figure 1).
When the other variables were set to average, the dependence of the probability from the prediction model was on one factor. The response curve shows the quantitative relationship between the logistic probability of occurrence and the appropriate ranges of environmental factors. The shoot blight of larch had a higher risk of occurrence (>0.5) in areas at an elevation of 150–780 m, with an annual mean temperature of 0.5–8 °C, mean temperature of the warmest quarter of 16–21 °C, annual precipitation of 510–830 mm, and precipitation of the wettest quarter of 323–500 mm (Figure 2).

3.3. Spatial Distribution of Potential Habitat Areas

Under the current climate scenario, the potential habitat area for shoot blight of larch in China predicted by the Biomod2 ensemble model accounts for 19.54% of the total land area of China, and it is mainly distributed in Liaoning, Jilin, the majority of Heilongjiang, the northeastern part of Inner Mongolia, the northern part of Hebei, Shanxi, and part of Shaanxi in China. In addition, it is also distributed in Shandong, Shaanxi, Gansu, Beijing, Hubei, Gansu, Qinghai, Ningxia, Sichuan, and Guizhou (Figure 3). The potentially suitable area for shoot blight of larch in China accounts for 15.09% of the total area of the medium- and highly suitable areas, concentrated in the eastern part of Jilin; Eastern Liaoning; most of Heilongjiang; localized areas in Eastern Inner Mongolia; and localized areas in Hebei, Shanxi, and Shaanxi. The low-suitability areas account for 4.45% of the total area and are mainly located in Central and Western Liaoning, Southwestern Jilin and part of Heilongjiang, part of Northeastern Inner Mongolia, Central and Eastern Shandong, part of Shanxi, part of Shandong, part of Shaanxi, Northern Hebei, and Southeastern Gansu.
Under the current climate scenario, the results predicted by the MaxEnt model indicate that the potential habitat for shoot blight of larch in China covers 20.33% of the total national area of China, and it is mainly distributed in Liaoning, Jilin, Eastern Heilongjiang, Northeastern Inner Mongolia, Western Shanxi, and Southern Gansu in China. In addition, it is also sporadically distributed in Hebei, Beijing, Shanxi, Hubei, Gansu, Qinghai, Shaanxi, Ningxia, Sichuan, and Guizhou (Figure 4). The potentially suitable area for shoot blight of larch in China accounts for 10.22% of the total area of the highly and moderately suitable distribution region and is concentrated in the eastern part of Jilin, Eastern Liaoning, the southeastern and central parts of Heilongjiang, and the local area of Eastern Inner Mongolia. The low-suitability area accounts for 10.11% of the total area, mainly located in Central and Western Liaoning, Southwestern Jilin and part of Western Heilongjiang, Northeastern Inner Mongolia, Eastern Shandong, the local area of Shanxi, Northern Hebei, Eastern Shandong, the local area of Northern Hebei, Shandong, localized areas in Shaanxi, Northern Hebei, and Southeastern Gansu.
Combining the predictions of the two models, it was found that the potential distribution maps of shoot blight of larch generated by the Biomod2 ensemble and MaxEnt models were similar, while the proportions of potentially suitable regions, including the highly suitable, moderately suitable, and low-suitability areas, differed (Figure 5). The suitable habitats for shoot blight of larch were largely distributed in Northern China, including Heilongjiang, Jilin, Liaoning, and Northeastern Inner Mongolia, as well as parts of Shandong, Hebei, Beijing, Shanxi, Hubei, Gansu, Qinghai, Shaanxi, Ningxia, Sichuan, and Guizhou. The potentially suitable regions predicted by the former were smaller than those predicted by the latter. The MaxEnt model predicted potentially suitable areas that accounted for 20.33% of the total land area of China, of which 4.08% was highly suitable, 6.14% was moderately suitable, and 10.11% had low suitability, whereas the Biomod2 ensemble model identified potentially suitable areas constituting 19.54% of the total region of China, of which 10.31% was highly suitable, 4.78% was moderately suitable, and 4.45% had low suitability.

4. Discussion

4.1. Environmental Variables and Changes in the Potential Distribution of the Shoot Blight of Larch

The results demonstrated that mean annual temperature, annual precipitation, precipitation of the wettest quarter, mean temperature of the warmest quarter, and elevation were the main environmental variables affecting the distribution of the shoot blight. By creating marginal response curves, we examined how the predicted probability of species presence could change with each significant factor. This study revealed that the suitable annual mean temperature threshold corresponding to an occurrence probability greater than 0.5 was in the range of 0.5~7.8 °C (Figure 2). The annual mean temperature in the suitable distribution area for shoot blight of larch in Northeast China is −5–11 °C [34], while in the Greater Khingan region, where larch resources are abundant, it is −2–6 °C, which is consistent with the results of this study. Therefore, the regions with a low annual mean temperature in China where shoot blight of larch hosts occurs are more suitable for its distribution, and more attention is given to preventing this disease.

4.2. Model Comparisons and Selection

Our study is the first to use two distinct correlative modeling approaches, the MaxEnt and Biomod2 ensemble models, to predict the potential distribution of shoot blight of larch. The AUC values for both models were close to 1 and exceeded 0.9, suggesting that the predictive performance of these two models for shoot blight of larch in China was precise and met the desired level of accuracy. The Biomod2 Ensemble model, on the other hand, performed better in prediction and showed a higher TSS value, indicating greater model reliability.
Because of the major principles of the individual model algorithms, each model has its own strengths and weaknesses, and its performance can become uneven if the input data change. The MaxEnt model has recently gained popularity among SDM models due to its advantages, and it remains relatively robust even with small sample sizes [35]. Using multiple algorithms can enable researchers to find proper ecological signals from the “noise” associated with single-model uncertainties [36]. SDM users extensively employ ensemble modeling, which has also been proposed to enhance the accuracy of species distribution predictions [24]. In contrast, some studies conducted by Crimmins et al. [22], Hao et al. [37], and Zhu and Peterson [23] found that ensemble models did not perform better than individual models. No technique can rescue species that are difficult to predict [38]. A few ecological characteristics are known to affect SDM projection results, such as life span [39], growth rate, successional status [38], habitat specialization, and dispersal ability [40]. Therefore, to forecast the potentially suitable distribution regions of the species, it is essential to select a suitable ecological niche model according to the characteristics of the species itself. It is suggested that a suitable approach should be dependent on the objectives and types of the modeled distributions.
Here, the potential distribution areas of shoot blight of larch in China predicted by the two MaxEnt and Biomod2 ensemble models were similar in terms of area ratio and region of distribution and were mainly located in Northeast China, North China, Southwest China, and Northwest China. However, the prediction results showed that the moderately and highly suitable regions predicted by the Biomod2 ensemble model were significantly greater than those forecasted by the MaxEnt model (Figure 5), but in terms of the total potentially suitable distribution areas, there was little difference in the proportion. This further revealed that the two models have high reliability and accuracy in predicting the potential distribution area of the shoot blight of larch in China. This is highly important for further rational division of potential invasion risk areas of larch shoot blight, strengthening quarantine monitoring in key areas, timely formulation of effective control measures, monitoring and control of occurrence areas, and early warning and supervision in nonoccurrence regions.

4.3. Model Predictive Performance

The MaxEnt model offers several strengths for species distribution anticipation, including an easily interpreted modeling mechanism and no requirement for background data [28]. However, it is important to note that optimizing the parameter settings of the MaxEnt model is necessary to achieve accurate forecasting results [41,42,43]. Many researchers have shown that the default parameters of the MaxEnt model exhibit complexity and high fitting degrees, which can be managed through adjustments to the regularization multiplier and feature class combinations [14]. In species distribution model studies, most scholars have primarily utilized climate factors alone or a combination of climate, topography, and soil variables for analysis, which can be either with default parameters or optimized MaxEnt parameters specifically for climate variable modeling. Consequently, most researchers have assessed species distributions using default parameters [44]. However, such analyses may lead to overfitting and increased complexity, potentially reducing the accuracy of research results and generating interpretations that are challenging, such as response curves of environmental factors exhibiting significant volatility. Thus, it is essential to optimize the model [7]. The MaxEnt model with optimized parameters allows the effective avoidance of overfitting and can better predict the distribution of species. This study established 40 numerical regularization multipliers (ranging from 0 to 4) and employed 6 characteristic combinations, namely, L, Q, P, T, LQ, LP, LT, LH, QP, QT, QH, PT, PH, TH, LQP, LQT, LQH, LPT, LPH, QPT, QPH, QTH, PTH, LQPT, LQPH, LQTH, LPT, and LQPTH, resulting in 1160 parameter combinations imported into the Kuenm package in R software to reduce inaccuracies. Previous studies have seldom combined the Kuenm package with the Biomod2 package. This study revealed that the MaxEnt model optimized with the parameters FC = LQH and RM = 3.5 yielded optimal results. Furthermore, the optimized MaxEnt model effectively simulated the potential geographical distribution of shoot blight of larch with high reliability.

5. Conclusions

Combining the predictions of the two ecological niche models, it can be seen that under the current climate scenario, the suitable areas for larch dieback disease are located mainly in the northeastern, northern, southwestern, and localized areas in Northwestern China. Shoot blight of larch is a serious plant disease, and its host is widely distributed in China. This study predicts and analyzes the potential habitat of shoot blight of larch in China, which can provide a reference and guidance for the prevention and control of early warning and quarantine in relevant forestry industry quarantine departments and is highly important for the early warning and prevention of shoot blight of larch in China.
The results showed that both ecological niche models had high prediction accuracy. The Biomod2 ensemble model performed better in predicting the potentially suitable distribution areas of shoot blight of larch in China, but this does not mean that the Biomod2 ensemble model can achieve better prediction performance and accuracy for all the research objects. Therefore, when predicting the distribution areas of species, researchers should choose a model according to the characteristics of the research subjects, and only in this way can more accurate and reliable conclusions be drawn. In this study, we compared and analyzed the differences between the two models in predicting the potential distribution area of shoot blight of larch to provide a reference for the selection of prediction models for the potential habitat distribution area affected by plant diseases.

Author Contributions

Conceptualization, X.Z. and Y.L.; methodology, X.Z., and Y.L.; software, X.Z.; investigation, X.Z., Y.L., and W.W.; writing—original draft preparation, X.Z.; writing—review and editing, X.Z., Y.L., and W.W. All authors have read and agreed to the published version of the manuscript.

Funding

This research was funded by the National Key Research and Development Program, grant number 2021YFD1400300.

Data Availability Statement

The disease occurrence data are available on request from the corresponding author.

Conflicts of Interest

The authors declare no conflicts of interest.

References

  1. Liu, Y.; Miao, C.; Wang, H. Influence of climate change on distribution of suitable areas of Larix plantation in China. Acta Ecol. Sin. 2023, 43, 9686–9698. [Google Scholar] [CrossRef]
  2. Hattori, Y.; Ando, Y.; Nakashima, C. Taxonomical re-examination of the genus Neofusicoccum in Japan. Mycoscience 2021, 62, 250–259. [Google Scholar] [CrossRef] [PubMed]
  3. Liu, X. Biocontrol and Molecular Diagnosis of the Larch Shoot Blight. Master’s Thesis, Northeast Forestry University, Harbin, China, 2009. [Google Scholar]
  4. Shi, C. Popularizing of Integrated Control Technology to Larch Blight. Master’s Thesis, Northeast Forestry University, Harbin, China, 2003. [Google Scholar]
  5. Elith, J.; Leathwick, J.R. Species distribution models: Ecological explanation and prediction across space and time. Ann. Rev. Ecol. Evol. Systemat. 2009, 40, 677–697. [Google Scholar] [CrossRef]
  6. Escobar, L.E.; Romero-Alvarez, D.; Leon, R.; Lepe-Lopez, M.A.; Craft, M.E.; Borbor-Cordova, M.J.; Svenning, J.-C. Declining prevalence of disease vectors under climate change. Sci. Rep. 2016, 6, 39150. [Google Scholar] [CrossRef] [PubMed]
  7. Zhao, G.; Cui, X.; Sun, J.; Li, T.; Wang, Q.I.; Ye, X.; Fan, B. Analysis of the distribution pattern of Chinese Ziziphus jujuba under climate change based on optimized biomod2 and MaxEntMaxEnt models. Ecol. Indic. 2021, 132, 108256. [Google Scholar] [CrossRef]
  8. Zhao, Z.; Xiao, N.; Shen, M.; Li, J. Comparison between optimized MaxEntMaxEnt and random forest modeling in predicting potential distribution: A case study with Quasipaa boulengeri in China. Sci. Total Environ. 2022, 842, 156867. [Google Scholar] [CrossRef] [PubMed]
  9. Merow, C.; Smith, M.J.; Silander, J.A., Jr. A practical guide to MaxEntMaxEnt for modeling species’ distributions: What it does, and why inputs and settings matter. Ecography 2013, 36, 1058–1069. [Google Scholar] [CrossRef]
  10. Phillips, S.J.; Anderson, R.P.; Dudík, M.; Schapire, R.E.; Blair, M.E. Opening the black box: An open-source release of E. Ecography 2017, 40, 887–893. [Google Scholar] [CrossRef]
  11. Čengić, M.; Rost, J.; Remenska, D.; Janse, J.H.; Huijbregts, M.A.J.; Schipper, A.M. On the importance of predictor choice, modelling technique, and number of pseudo-absences for bioclimatic envelope model performance. Ecol. Evol. 2020, 10, 12307–12317. [Google Scholar] [CrossRef]
  12. Heikkinen, R.K.; Marmion, M.; Luoto, M. Does the interpolation accuracy of species distribution models come at the expense of transferability? Ecography 2012, 35, 276–288. [Google Scholar] [CrossRef]
  13. Wenger, S.J.; Olden, J.D. Assessing transferability of ecological models: An underappreciated aspect of statistical validation. Methods Ecol. Evol. 2012, 3, 260–267. [Google Scholar] [CrossRef]
  14. Phillips, S.J.; Anderson, R.P.; Schapire, R.E. Maximum entropy modeling of species geographic distributions. Ecol. Modell. 2006, 190, 231–259. [Google Scholar] [CrossRef]
  15. Bowen, A.K.M.; Stevens, M.H.H. Temperature, topography, soil characteristics, and NDVI drive habitat preferences of a shade-tolerant invasive grass. Ecol. Evol. 2020, 10, 10785–10797. [Google Scholar] [CrossRef]
  16. Freeman, B.; Jiménez-García, D.; Barca, B.; Grainger, M. Using remotely sensed and climate data to predict the current and potential future geographic distribution of a bird at multiple scales: The case of Agelastes meleagrides, a western African forest endemic. Avian Res. 2019, 10, 1–9. [Google Scholar] [CrossRef]
  17. Holder, A.M.; Markarian, A.; Doyle, J.M.; Olson, J.R. Predicting geographic distributions of fishes in remote stream networks using maximum entropy modeling and landscape characterizations. Ecol. Modell. 2020, 433, 109231. [Google Scholar] [CrossRef]
  18. Segurado, P.; Araujo, M.B. An evaluation of methods for modelling species distributions. J. Biogeogr. 2004, 31, 1555–1568. [Google Scholar] [CrossRef]
  19. Strubbe, D.; Jackson, H.; Groombridge, J.; Matthysen, E. Invasion success of a global avian invader is explained by within-taxon niche structure and association with humans in the native range. Divers. Distrib. 2015, 21, 675–685. [Google Scholar] [CrossRef]
  20. Thuiller, W.; Lafourcade, B.; Engler, R.; Araújo, M.B. BIOMOD–A platform for ensemble forecasting of species distributions. Ecography 2009, 32, 369–373. [Google Scholar] [CrossRef]
  21. Hao, T.; Elith, J.; Guillera-Arroita, G.; Lahoz-Monfort, J.J. A review of evidence about use and performance of species distribution modelling ensembles like BIOMOD. Divers. Distrib. 2019, 25, 839–852. [Google Scholar] [CrossRef]
  22. Crimmins, S.M.; Dobrowski, S.Z.; Mynsberge, A.R. Evaluating ensemble forecasts of plant species distributions under climate change. Ecol. Modell. 2013, 266, 126–130. [Google Scholar] [CrossRef]
  23. Zhu, G.P.; Peterson, A.T. Do consensus models outperform individual models? Transferability evaluations of diverse modeling approaches for an invasive moth. Biol. Invasions 2017, 19, 2519–2532. [Google Scholar] [CrossRef]
  24. Marmion, M.; Parviainen, M.; Luoto, M.; Heikkinen, R.K.; Thuiller, W. Evaluation of consensus methods in predictive species distribution modelling. Divers. Distrib. 2009, 15, 59–69. [Google Scholar] [CrossRef]
  25. Roberts, D.R.; Bahn, V.; Ciuti, S.; Boyce, M.S.; Elith, J.; Guillera-Arroita, G.; Hauenstein, S.; Lahoz-Monfort, J.J.; Schröder, B.; Thuiller, W. Cross-validation strategies for data with temporal, spatial, hierarchical, or phylogenetic structure. Ecography 2017, 40, 913–929. [Google Scholar] [CrossRef]
  26. Warren, D.L.; Glor, R.E.; Turelli, M. ENMTools: A toolbox for comparative studies of environmental niche models. Ecography 2010, 33, 607–611. [Google Scholar] [CrossRef]
  27. Fick, S.E.; Hijmans, R.J. WorldClim 2: New 1-km spatial resolution climate surfaces for global land areas. Int. J. Climatol. 2017, 37, 4302–4315. [Google Scholar] [CrossRef]
  28. Elith, J.; Simpson, J.; Hirsch, M.; Burgman, M.A. Taxonomic uncertainty and decision making for biosecurity: Spatial models for myrtle/guava rust. Austral. Plant Pathol. 2013, 42, 43–51. [Google Scholar] [CrossRef]
  29. Narouei-Khandan, H.A.; Halbert, S.E.; Worner, S.P.; van Bruggen, A.H.C. Global climate suitability of citrus huanglongbing and its vector, the Asian citrus psyllid, using two correlative species distribution modeling approaches, with emphasis on the USA. Eur. J. Plant Pathol. 2016, 144, 655–670. [Google Scholar] [CrossRef]
  30. Cobos, M.E.; Peterson, A.T.; Barve, N.; Osorio-Olvera, L. kuenm: An R package for detailed development of ecological niche models using E. PeerJ 2019, 7, e6281. [Google Scholar] [CrossRef]
  31. Wu, Z.; Gao, T.; Luo, Y.; Shi, J. Prediction of the global potential geographical distribution of Hylurgus ligniperda using a maximum entropy model. For. Ecosyst. 2022, 9, 100042. [Google Scholar] [CrossRef]
  32. Lobo, J.M.; Jiménez-Valverde, A.; Real, R. AUC: A misleading measure of the performance of predictive distribution models. Glob. Ecol. Biogeogr. 2008, 17, 145–151. [Google Scholar] [CrossRef]
  33. Li, D.; Li, Z.; Liu, Z.; Yang, Y.; Khoso, A.G.; Wang, L.; Liu, D. Climate change simulations revealed potentially drastic shifts in insect community structure and crop yields in China’s farmland. J. Pest Sci. 2023, 96, 55–69. [Google Scholar] [CrossRef]
  34. Yan, H. Spatio-Temporal Epidemic Characteristics and Risk Analysis of Poplar Canker of Cytospora chrysosperma in Northeast China. Ph.D. Thesis, Beijing Forestry University, Beijing, China, 2020. [Google Scholar]
  35. Townsend Peterson, A.; Papeş, M.; Eaton, M. Transferability and model evaluation in ecological niche modeling: A comparison of GARP and E. Ecography 2007, 30, 550–560. [Google Scholar] [CrossRef]
  36. Araujo, M.; New, M. Ensemble forecasting of species distributions. Trends Ecol. Evol. 2007, 22, 42–47. [Google Scholar] [CrossRef] [PubMed]
  37. Hao, T.; Elith, J.; Lahoz-Monfort, J.J.; Guillera-Arroita, G. Testing whether ensemble modelling is advantageous for maximising predictive performance of species distribution models. Ecography 2020, 43, 549–558. [Google Scholar] [CrossRef]
  38. Zimmermann, N.E.; Elith, J.; Graham, C.H.; Phillips, S.; Peterson, A.T. What matters for predicting the occurrences of trees: Techniques, data, or species’ characteristics? Ecol. Monogr. 2007, 77, 615–630. [Google Scholar] [CrossRef]
  39. Hanspach, J.; Kühn, I.; Pompe, S.; Klotz, S. Predictive performance of plant species distribution models depends on species traits. Pers. Plant Ecol. Evol. Systemat. 2010, 12, 219–225. [Google Scholar] [CrossRef]
  40. McCune, J.L.; Rosner-Katz, H.; Bennett, J.R.; Schuster, R.; Kharouba, H.M. Do traits of plant species predict the efficacy of species distribution models for finding new occurrences? Ecol. Evol. 2020, 10, 5001–5014. [Google Scholar] [CrossRef] [PubMed]
  41. Lissovsky, A.A.; Dudov, S.V. Species-distribution modeling: Advantages and limitations of its application. 2. MaxEntMaxEnt. Biol. Bull. Rev. 2021, 11, 265–275. [Google Scholar] [CrossRef]
  42. Morales, N.S.; Fernández, I.C.; Baca-González, V. MaxEnt’sMaxEnt parameter configuration and small samples: Are we paying attention to recommendations? A systematic review. PeerJ 2017, 5, e3093. [Google Scholar] [CrossRef]
  43. Radosavljevic, A.; Anderson, R.P. Making better E models of species distributions: Complexity, overfitting and evaluation. J. Biogeogr. 2014, 41, 629–643. [Google Scholar] [CrossRef]
  44. Pueyo-Herrera, P.; Tang, C.Q.; Matsui, T.; Ohashi, H.; Qian, S.; Yang, Y.; Herrando-Moraira, S.; Nualart, N.; López-Pujol, J. Ecological niche modeling applied to the conservation of the East Asian relict endemism Glyptostrobus pensilis (Cupressaceae). New For. 2023, 54, 1131–1152. [Google Scholar] [CrossRef]
Forests 15 01313 g001
Figure 1. The jackknife test was used to assess the significance of major environmental factors to determine the distribution of shoot blight of larch. bio1: annual mean temperature; bio10: mean temperature of the warmest quarter; bio12: annual precipitation; bio16: precipitation of the wettest quarter; bio18: precipitation of the warmest quarter; bio3: isothermality; bio4: temperature seasonality; and elev: elevation.
Figure 1. The jackknife test was used to assess the significance of major environmental factors to determine the distribution of shoot blight of larch. bio1: annual mean temperature; bio10: mean temperature of the warmest quarter; bio12: annual precipitation; bio16: precipitation of the wettest quarter; bio18: precipitation of the warmest quarter; bio3: isothermality; bio4: temperature seasonality; and elev: elevation.
Forests 15 01313 g001
Forests 15 01313 g002
Figure 2. Response curves for the most vital environmental factors in the ecological niche model for the shoot blight of larch.
Figure 2. Response curves for the most vital environmental factors in the ecological niche model for the shoot blight of larch.
Forests 15 01313 g002
Forests 15 01313 g003
Figure 3. Potentially suitable distribution of shoot blight of larch in China predicted by the Biomod2 Ensemble model. The map was obtained from the Resource and Environment Science and Data Center, Chinese Academy of Sciences (https://www.resdc.cn/ (accessed on 10 October 2023)).
Figure 3. Potentially suitable distribution of shoot blight of larch in China predicted by the Biomod2 Ensemble model. The map was obtained from the Resource and Environment Science and Data Center, Chinese Academy of Sciences (https://www.resdc.cn/ (accessed on 10 October 2023)).
Forests 15 01313 g003
Forests 15 01313 g004
Figure 4. Potentially suitable distribution of shoot blight of larch in China predicted by the MaxEnt model.
Figure 4. Potentially suitable distribution of shoot blight of larch in China predicted by the MaxEnt model.
Forests 15 01313 g004
Forests 15 01313 g005
Figure 5. Potentially suitable distribution areas for shoot blight of larch predicted by the Biomod2 Ensemble and MaxEnt models.
Figure 5. Potentially suitable distribution areas for shoot blight of larch predicted by the Biomod2 Ensemble and MaxEnt models.
Forests 15 01313 g005
Table 1. Environmental data used in this study.
Table 1. Environmental data used in this study.
AbbreviationDescriptionWhether to Use
for Modeling
bio1Mean Annual Temperature (°C)Yes
bio2Mean Diurnal Range (Mean of monthly (max temp − min temp)) (°C)No
bio3Isothermality (bio2/bio7) × 100Yes
bio4Temperature Seasonality (standard deviation × 100)Yes
bio5Max Temperature of Warmest Month (°C)No
bio6Min Temperature of Coldest Month (°C)No
bio7Annual Temperature Range (bio5–bio6) (°C)No
bio8Mean Temperature of Wettest Quarter (°C)No
bio9Mean Temperature of Driest QuarterNo
bio10Mean Temperature of Warmest QuarterYes
bio11Mean Temperature of Coldest Quarter (°C)No
bio12Annual Precipitation (mm)Yes
bio13Precipitation of Wettest Month (mm)No
bio14Precipitation of Driest Month (mm)No
bio15Precipitation Seasonality (mm)No
bio16Precipitation of Wettest QuarterYes
bio17Precipitation of Driest Quarter (mm)No
bio18Precipitation of Warmest Quarter (mm)Yes
bio19Precipitation of Coldest Quarter (mm)No
elevElevation (metric)Yes
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

Share and Cite

MDPI and ACS Style

Zhang, X.; Wu, W.; Liang, Y. Analysis of the Potential Distribution of Shoot Blight of Larch in China Based on the Optimized MaxEnt and Biomod2 Ensemble Models. Forests 2024, 15, 1313. https://doi.org/10.3390/f15081313

AMA Style

Zhang X, Wu W, Liang Y. Analysis of the Potential Distribution of Shoot Blight of Larch in China Based on the Optimized MaxEnt and Biomod2 Ensemble Models. Forests. 2024; 15(8):1313. https://doi.org/10.3390/f15081313

Chicago/Turabian Style

Zhang, Xiuyun, Wenhui Wu, and Yingmei Liang. 2024. "Analysis of the Potential Distribution of Shoot Blight of Larch in China Based on the Optimized MaxEnt and Biomod2 Ensemble Models" Forests 15, no. 8: 1313. https://doi.org/10.3390/f15081313

APA Style

Zhang, X., Wu, W., & Liang, Y. (2024). Analysis of the Potential Distribution of Shoot Blight of Larch in China Based on the Optimized MaxEnt and Biomod2 Ensemble Models. Forests, 15(8), 1313. https://doi.org/10.3390/f15081313

Note that from the first issue of 2016, this journal uses article numbers instead of page numbers. See further details here.

Article Metrics

Back to TopTop