Next Article in Journal
Leaf Functional Traits of Zanthoxylum planispinum ‘Dintanensis’ Plantations with Different Planting Combinations and Their Responses to Soil
Next Article in Special Issue
A Novel Simplified Protocol for Pre-Processing Whole Wood Samples for Stable Isotope Analysis in Tree Rings
Previous Article in Journal
Do Ectomycorrhizal Trees Select Ectomycorrhizal Fungi That Enhance Phosphorus Uptake under Nitrogen Enrichment?
Previous Article in Special Issue
Different Taxonomic and Functional Indices Complement the Understanding of Herb-Layer Community Assembly Patterns in a Southern-Limit Temperate Forest
 
 
Font Type:
Arial Georgia Verdana
Font Size:
Aa Aa Aa
Line Spacing:
Column Width:
Background:
Article

Modeling Climate Change Effects on the Distribution of Oak Forests with Machine Learning

1
Department of Forest Science, Faculty of Agriculture, Ilam University, Ilam 516-69315, Iran
2
General Department of Natural Resources and Watershed Management of Ilam Province, Ilam 516-69315, Iran
3
Department of Forestry, Faculty of Natural Resources and Marine Sciences, Tarbiat Modares University, Tehran 111-14115, Iran
4
German Research Center for Artificial Intelligence, 26129 Oldenburg, Germany
5
John von Neumann Faculty of Informatics, Obuda University, 1034 Budapest, Hungary
6
Institute of the Information Society, University of Public Service, 1083 Budapest, Hungary
*
Author to whom correspondence should be addressed.
Forests 2023, 14(3), 469; https://doi.org/10.3390/f14030469
Submission received: 16 January 2023 / Revised: 14 February 2023 / Accepted: 20 February 2023 / Published: 24 February 2023
(This article belongs to the Special Issue Forest Resilience and Resistance to Climate Change)

Abstract

:
The present study models the effect of climate change on the distribution of Persian oak (Quercus brantii Lindl.) in the Zagros forests, located in the west of Iran. The modeling is conducted under the current and future climatic conditions by fitting the machine learning method of the Bayesian additive regression tree (BART). For the anticipation of the potential habitats for the Persian oak, two general circulation models (GCMs) of CCSM4 and HADGEM2-ES under the representative concentration pathways (RCPs) of 2.6 and 8.5 for 2050 and 2070 are used. The mean temperature (MT) of the wettest quarter (bio8), solar radiation, slope and precipitation of the wettest month (bio13) are respectively reported as the most important variables in the modeling. The results indicate that the suitable habitat of Persian oak will significantly decrease in the future under both climate change scenarios as much as 75.06% by 2070. The proposed study brings insight into the current condition and further projects the future conditions of the local forests for proper management and protection of endangered ecosystems.

1. Introduction

Forest ecosystems have been globally affected by climate change, geomorphic features and human impact [1,2]. Climate change is widely introduced as a main threat to diversity, species survival and ecosystem stability in most biomes [3,4]. In fact, climate change has been introduced as one of the main causes of the emergence of new species, change in a species range, species extinction, destruction of biodiversity, loss of ecosystem resilience and disturbance regimes [4,5,6]. A species’ distribution is an essential spatial feature influenced by the environment and human impact. On a large scale, the climate has been suggested as a fundamental factor determining the distribution of species worldwide [7]. Various plant species in different geographical regions of the world are affected by climate change in different ways [8]. Due to the fact that climate change has a major role in species distribution [9], predicting the distribution of plant species under climate change scenarios is important for ecosystem sustainability and management [10,11]. Recently, several studies have predicted the potential effects of climate change on plant species distribution [12,13,14,15,16,17,18,19,20], and generally, these studies show that climate change causes changes in the range of plant species.
Species distribution models (SDMs) are widely used to predict species response to climate change [9,21,22,23,24]. SDMs were found to be reliable models to anticipate current and future habitats as demonstrated by, e.g., Guisan and Thuiller (2005), Joseph et al. (2009) and Levin et al. (2013) [25,26,27]. Such models are also used to demonstrate the relationship between species distribution and environmental conditions as investigated by, e.g., Booth et al. (2014) [28]. In addition, the literature includes a diverse range of various statistical and machine learning methods to predict the distribution of species under climate change, conservation planning and predicting the future suitable habitat of a species [9]. The use of machine learning algorithms to predict the geographical distribution of species is a relatively new field of study. Machine learning algorithms can significantly speed up ecological processing and analysis and successfully solve classification problems [29]. Among advanced machine methods, the Bayesian additive regression tree (BART) as a new ensemble-based method has gained significant popularity due to its accuracy in the field [30]. Consequently, in the present study, we aimed at using BART.
Zagros mountain forests, with an area of 5 million hectares in western Iran located in a semi-arid climate, are considered the most important yet sensitive ecosystems of Iran. The Zagros mountain forests are identified as the only ancient and continuous habitat of Persian oak (Quercus brantii Lindl.) [31]. The role of these forests is crucial for soil and water conservation and people’s livelihood in this region [32]. In addition, the forests are a repository of various plants and animal species. It is believed that the area of these forests in the past was wider than today [33]. These forests have undergone drastic changes in cover and structure in recent decades, which has had many negative consequences for the communities that depend on them [34]. Climate change has negatively affected these forests by increasing temperature and decreasing rainfall [35]. The occurrence of oak decline and secondary pests and diseases in this region in recent decades, according to many researchers, has been associated with the phenomenon of climate change [36,37,38]. In addition, human activities such as land use change, wood harvesting and livestock grazing have destroyed large parts of these forest ecosystems [34,39,40]. Former studies stated that climate change has already affected the habitat of Persian oak in the Zagros forests. For instance, Safaei et al. (2021) modeled the effect of climate change on the distribution of Persian oak in Zagros forests, and the results revealed that in the future, there will be changes in the spatial distribution of this species [41]. Valavi et al. (2018) investigated the effect of climate change on the habitat of Persian oak in the Zagros forests and anticipated that the potential habitat of this species will decrease and will be further limited to higher altitudes in the future [14]. Malekian and Sadeghi (2019) predicted the effect of climate change on the habitat of Persian oak in western Iran and concluded that the suitable habitat of this species will decrease in some areas and shift to higher altitudes in the future [42]. Malekian and Sadeghi (2019) used statistical projection, and later Safaei et al. (2021) showed the importance and applicability of optimization and evolutionary algorithms. Despite the study by Valavi, et al. (2018) that used an ensemble learning method, there is still a research gap in using various machine learning methods to improve modeling capability. In this study, the applicability of a Bayesian learning method is explored. Our study was limited to one study area and one species. Such limitations must be addressed in future studies by exploring various case studies and different species.
Despite limited studies on the effect of climate change on the habitat of Persian oak in the Zagros forests, the relationship between changes in the distribution of Persian oak in relation to various environmental variables has not been well considered. Therefore, in this study, in order to model the distribution of Persian oak in Ilam province as a specific habitat for this species, the BART method has been used. The BART method is a tree-based machine learning method that has been successfully used to address regression and classification issues. This method was proposed by Chipman et al. (2010) [43] and has become common in recent years due to its high efficiency in regression and classification problems compared to other machine learning methods. Other Bayesian methods have been used for modeling species distribution, but the BART method has not been used for modeling the distribution of plant species, especially not tree species. This research seeks to answer the following questions: Firstly, what are the main environmental attributes that affect the distribution of Persian oak? Secondly, how much of the area of Persian oak habitat will change in the future under different climate change scenarios? Due to the deteriorating situation of Zagros forests and considering that few studies have been done in the field of modeling the distribution of Persian oak in these forests, the results of the present study can help to understand the current condition and predict the future situation of these forests and provide a guideline for management and protection of these semi-arid ecosystems.

2. Material and Methods

2.1. Study Area

The study area is Ilam province with an area of 20,133 km2 located in western Iran. This region is located at 32°03′ N to 34°02′ N latitude and 45°40′ E to 48°03′ E longitude [35]. Altitude in this region varies from 34 m to 2632 m. The average annual precipitation in Ilam province is 578 mm. The area of forests in Ilam province is about 641,000 ha, and the predominant species of these forests is Persian oak (Quercus brantii Lindl.), which constitutes about 90% of these forests [44] (Figure 1).

2.2. Species Occurrence Data

The species occurrence locations were collected using field surveys and then thinned with a distance of 1 km using the spThin package in R software [45]. The spThin package randomly thins species occurrence points using the minimum user-defined distance. We thinned the occurrence points with a distance of 1 km because we used climate data with a 1 km spatial resolution [18]. A total of 966 presence records of the species were obtained in the study area. In addition, pseudo-absence points were obtained with the random point method using GIS software.

2.3. Environmental Variables

The environmental variables used in this research include different bioclimatic variables with a resolution of 1 km that were downloaded from the WorldClim database (www.worldclim.org accessed on 10 October 2022) [46]. To detect collinearity among variables, we used the Variance Inflation Factor (VIF) and removed variables with VIF greater than 10 [47,48]. As a result, we used 5 bioclimatic variables to predict the habitat of Persian oak. In addition, slope and solar radiation were selected as predictors for species distribution modeling (Figure 2). A description of the 19 bioclimatic variables is given in Table 1. To predict the potential habitat of Persian oak in the future, we used two General Circulation Models (GCMs) namely CCSM4 and HADGEM2-ES under the Representative Concentration Pathways (RCPs) 2.6 and 8.5 for 2050 and 2070 [49].

2.4. Species Distribution Modeling

In order to model the distribution of Persian oak, the BART method has been used. BART proposed by Chipman et al. (2010) [43] is a tree-based data driven method suitable for regression and classification. BART has recently increased popularity due to its high efficiency compared to other machine learning methods. The difference between BART and other tree-based methods is that it controls the structure of each tree using a prior distribution. BART performs modeling using the MCMC process that accepts or rejects trees during iterations. BART is a non-parametric Bayesian algorithm that uses the tree ensembles method [50]. We also used a Generalized Linear Model (GLM) as a benchmark to compare and ensure the performance of BART.

2.5. Model Evaluation

To evaluate the models, the area under the curve (AUC), the true skill statistic (TSS), type I error and type II error were used. AUC is a suitable index to evaluate the accuracy and performance of the model. The closer the AUC value is to 1, the better the accuracy and performance of the model [51]. The threshold values of AUC are as follows: AUC ≥ 0.9 = very good, 0.9 > AUC ≥ 0.8 = good and AUC < 0.8 = weak [52,53]. The value of TSS is from +1 to −1, in which the value +1 indicates a high performance of the model, and the values 0 and less than 0 indicate a low performance [54].

3. Results

3.1. Performance Evaluation of Models

In this study, we used the area under the curve (AUC), the true skill statistic (TSS), type I error and type II error to evaluate model performance. The results of the model evaluation are shown in Table 2. The evaluation using the AUC index shows that BART has a very good performance in predicting the habitat of the Persian oak (AUC = 0.93). In addition, the value of TSS indicates the good accuracy of the model in forecasting (TSS = 0.75). Type I error (0.08) and type II error (0.16) also show that BART is a suitable model for predicting the habitat of Persian oak, and it has a higher accuracy in forecasting compared to the GLM (AUC = 0.86, TSS = 0.68) (Figure 3).

3.2. Importance of Environmental Variables

The importance of a variable in BART is usually measured by counting the number of times a given variable is divided by a tree across subsequent trees [55]. In models with more trees, the difference in variable importance is less pronounced. Therefore, variable selection can be performed by implementing models with fewer trees (m = 10 or m = 20) and observing which variables are not included in the trees [43]. Figure 4 shows the importance of environmental variables in modeling the distribution of Persian oak. This diagram indicates that the mean temperature of the wettest quarter (bio8), solar radiation, slope and precipitation of the wettest month (bio13) are the most important variables in modeling the distribution of Persian oak, and precipitation seasonality (bio15) is the least important.

3.3. Response Curves

Examining the response curves of Persian oak to environmental variables showed that with increasing slope and bio13, the occurrence of Persian oak in the study area increases. However, with increasing bio4 and bio8, the probability of species presence decreases. In addition, with increasing bio2, the species presence increases with a gentle slope. Overall, bio15 was the least important variable for the presence of Persian oak in the study area (Figure 5).

3.4. Current and Future Potential Distribution of Persian Oak

The habitat suitability map of Persian oak in Ilam province was predicted under current and future climatic conditions (Figure 6 and Figure 7). The results revealed that under current climatic conditions, 2524.36 km2, which is 12.59% of the study area, has high suitability, and 1603.58 km2, which is 7.99% of the study area, has moderate suitability for the presence of Persian oak. In addition, 14,211.66 km2, which is 70.89% of the study area, was categorized as not suitable habitat (Table 3).
The results of modeling the distribution of Persian oak under future climatic conditions showed that under RCP2.6 in 2050, 1561.80 km2, which is 7.79% of the study area, and under RCP8.5 in 2050, 1980.43 km2, which is 9.88% of the study area, would be suitable for species presence. In addition, under RCP2.6 in 2070, 2123.45 km2 (10.59%) of the study area, and under RCP8.5 in 2070, 424.44 km2 (2%) of the area would be suitable for the presence of Persian oak. According to the results, under RCP8.5 in 2070, 4827 km2 (75.06%) of the species’ habitat will be lost (Table 3).

3.5. Elevation and Distribution of Persian Oak

We investigated the relationship between elevation and the distribution of Persian oak under current and future climatic conditions. The findings indicted that the habitat of Persian oak will shift to higher altitudes in the future. Figure 8 shows that Persian oak is currently the most present in altitudes of about 1100 to 1800 m. By 2050, under the RCP 2.6 and RCP 8.5 scenarios, and also by 2070 under the RCP 2.6 scenario, the species’ habitat will be slightly shifted to higher altitudes. By 2070, under the RCP 8.5 scenario, the species’ habitat will move significantly to higher altitudes and will have the highest density at altitudes of about 1400 to 2300 m.

4. Discussion

This research aimed to investigate the effect of climate change on the distribution of Persian oak (Quercus brantii Lindl.) in Ilam province, western Iran, using an advanced machine learning method. So far, the BART has not been used widely in ecology topics. For example, Yen et al. (2011) used BART to select the habitat of birds in Australia [56]. However, it has not been used for modeling the distribution of plant species. In the present study, we used the area under the curve (AUC), the true skill statistic (TSS), type I error and type II error to evaluate the performance of BART. The results showed that BART has a high performance in predicting the habitat of Persian oak. In terms of the importance of environmental variables, the mean temperature of the wettest quarter (bio8) was the main important attribute in the distribution of Persian oak followed by solar radiation, slope and precipitation of the wettest month (bio13). Valavi et al. (2018) showed that the Persian oak was more sensitive to temperature-related factors and in the future, the habitat of this species would shift to higher altitudes with lower temperatures [14]. Safaei et al. (2021) showed that temperature had a significant effect on the spatial distribution of Persian oak, so that the response of this plant species to climate change would shift to higher altitudes with lower temperatures and higher precipitation [41]. Another environmental factor which influenced the distribution of Persian oak was slope. Topography influences the microclimate of plants by affecting the amount of solar radiation received by mountain forests [57]. Generally, topography plays an important role in the distribution of plant species by affecting on the microclimate [58]. Our results confirmed that with increasing slope, the occurrence of Persian oak in the study area increases. One reason for this could be lower grazing and lower destruction in steep slopes. Moradi et al. (2022) reported that livestock grazing reduces tree regeneration in Zagros forests [40].
The findings of the present research showed that the suitable habitat of Persian oak would decrease in the future under both climate change scenarios (RCP 2.6 and RCP 8.5) compared to the current suitable habitat. According to the results, by 2050 under the RCP 2.6 and RCP 8.5 scenarios, 7.79% and 9.88% of the research area, respectively, would be suitable for the presence of Persian oak. In addition, by 2070, under the RCP 2.6 scenario, 10.59%, and under RCP 8.5 scenario, only 2.12% of the study area would be suitable for the presence of Persian oak. The results also showed that under RCP 8.5 in 2070, 75.06% of the species’ habitat will be lost. The results of previous studies also show that the habitat of the Persian oak will decrease in the future due to climate change compared to its current habitat [14,41,42]. Other oak species such as Q. coccifera, Q. libani and Q. aegilops were also predicted to reduce their potential habitat in the future [16,59,60,61]. Examining the relationship between elevation and the distribution of Persian oak showed that the species’ habitat will be shifted to higher altitudes in the future. The results showed that by 2070, the species’ habitat will move to altitudes above 1400 m. At low altitudes, for various reasons including human activities, degradation and livestock grazing, as well as proximity to residential areas and land use changes, we see more destruction of oak habitats. Furthermore, decreasing temperature and increasing precipitation at higher altitudes are the other factors that will cause the habitat to move to higher altitudes. In response to climate change, many plant species in mountainous areas migrate to higher altitudes [62,63,64]. Previous studies have shown that different species of oak prefer higher altitudes with lower temperatures and move to higher altitudes under climate change [16,41,61,65], which is consistent with the findings of this study.
In recent years, many trees, especially Persian oak, have been destroyed in large parts of Zagros forests. In Ilam province, climate change, drought and dust stress are among the factors that have caused the decline in oak trees. Among these factors, temperature and precipitation are the most important factors in the decline of oak forests [35]. Shiravand and Hosseini (2020) reported that increasing temperature and decreasing precipitation play a significant role in the decline of the Zagros oak ecosystem [32]. Increasing temperature by affecting evapotranspiration and humidity causes drought stress in Persian oak trees [41]. This species usually occupies higher altitudes with lower temperatures and grows hardly in warmer regions [14]. In addition, human activities such as land use changes and livestock grazing, especially in low altitudes where human activities are more intense, have caused the destruction of large parts of these forests and have increased the vulnerability of this species to climate change. As a result, it is predicted that this species will be shifted to higher altitudes with fewer human activities. Therefore, by controlling human activities, for example, reducing livestock grazing pressure and expanding protected areas, the rate of destruction in these forests can be reduced to some extent.

5. Conclusions

The aim of this study was to predict the effect of climate change on the distribution of Persian oak (Quercus brantii Lindl.) in Ilam province, western Iran, using a new machine learning method. The findings of this study revealed that the proposed BART model has a high performance in predicting the habitat of Persian oak. The modeling is conducted under current and future climatic conditions by fitting the machine learning method. For the anticipation of the potential habitats of the Persian oak, two GCMs of CCSM4 and HADGEM2-ES under the RCPs of 2.6 and 8.5 for 2050 and 2070 are used. The mean temperature of the wettest quarter (bio8), solar radiation, slope and precipitation of the wettest month (bio13) are reported as the most important variables in modeling. The results further indicate that the suitable habitat of Persian oak will significantly decrease in the future under both climate change scenarios as much as 75.06% by 2070. The proposed study brings insight into the current condition and further projects the future conditions of the local forests for proper management and protection of endangered ecosystems. The model further anticipated that with an increasing slope, the occurrence of Persian oak in the study area will increase. The reasons for this can be lower grazing practice, human activities and also improving the ability of oak to produce sprouts in these conditions. The results confirmed that the suitable habitat of Persian oak would decrease in the future compared to the current habitat and the habitat destruction will continue. Due to the importance of Persian oak in the Zagros forests, implementation of unified management and conservation actions, particularly at low altitudes, to help seedling regeneration of this species and planting with drought-resistant native tree and shrub species seems essential for this region. The valuation of oak forest proper management is required for these semi-arid ecosystems, especially where climate change impact coupled with increasing anthropogenic pressure have essential negative impacts on forest ecosystem geography. The results of this study can help future researchers carry out research on the protection and management of these forests using previous knowledge about this region and this species. For future studies, it is recommended to apply the proposed method for modeling further species and other regions.

Author Contributions

Conceptualization, H.M., M.H., O.K., K.A. and A.M.; methodology, H.M., M.H., O.K., K.A. and A.M.; software, H.M., M.H., O.K., K.A. and A.M.; formal analysis, H.M., M.H., O.K., K.A. and A.M.; investigation, H.M., M.H., O.K., K.A. and A.M.; data curation, H.M., M.H., O.K., K.A. and A.M.; writing—original draft preparation, H.M., M.H., O.K., K.A. and A.M.; writing—review and editing, H.M., M.H., O.K., K.A. and A.M.; visualization, H.M., M.H., O.K., K.A. and A.M.; supervision, M.H. and A.M.; project administration, M.H.; funding, H.M., M.H., O.K., K.A. and A.M. All authors have read and agreed to the published version of the manuscript.

Funding

This research is supported by Ilam University.

Institutional Review Board Statement

Not applicable.

Informed Consent Statement

Not applicable.

Data Availability Statement

Data are available through the corresponding or last author upon reasonable request for academic purposes.

Acknowledgments

A.M. wishes to acknowledge DFKI IML (iml.dfki.de, accessed on 10 October 2022) and the Ministry of Science and Culture of the German State of Lower Saxony. For more information, visit cst.dfki.de. The responsibility lies with the authors.

Conflicts of Interest

Authors declare no conflict of interest.

References

  1. Jonašova, M.; Prach, K. Central-European mountain spruce (Picea abies (L.) Karst.) forests: Regeneration of tree species after a bark beetle outbreak. Ecol. Eng. 2004, 23, 15–27. [Google Scholar] [CrossRef]
  2. Lindner, M.; Maroschek, M.; Netherer, S.; Kremer, A.; Barbati, A.; Garcia-Gonzalo, J.; Seidl, R.; Delzon, S.; Corona, P.; Kolström, M.; et al. Climate change impacts, adaptive capacity, and vulnerability of European forest ecosystems. For. Ecol. Manag. 2010, 259, 698–709. [Google Scholar] [CrossRef]
  3. Maclean, I.M.D.; Wilson, R.J. Recent ecological responses to climate change support predictions of high extinction risk. Proc. Natl. Acad. Sci. USA 2011, 108, 12337–12342. [Google Scholar] [CrossRef] [PubMed]
  4. Bellard, C.; Bertelsmeier, C.; Leadley, P.; Thuiller, W.; Courchamp, F. Impacts of climate change on the future of biodiversity. Ecol. Lett. 2012, 15, 365–377. [Google Scholar] [CrossRef]
  5. Abatzoglou, J.T.; Williams, A.P. Impact of anthropogenic climate change on wildfire across western US forests. Proc. Natl. Acad. Sci. USA 2016, 113, 11770–11775. [Google Scholar] [CrossRef]
  6. Stevens-Rumann, C.S.; Kemp, K.B.; Higuera, P.E.; Harvey, B.J.; Rother, M.T.; Donato, D.C.; Morgan, P.; Veblen, T.T. Evidence for declining forest resilience to wildfires under climate change. Ecol. Lett. 2018, 21, 243–252. [Google Scholar] [CrossRef]
  7. Manthey, M.; Box, E.O. Realized climatic niches of deciduous trees: Comparing western Eurasia and eastern North America. J. Biogeogr. 2007, 34, 1028–1040. [Google Scholar] [CrossRef]
  8. Sáenz-Romero, C.; Lamy, J.-B.; Ducousso, A.; Musch, B.; Ehrenmann, F.; Delzon, S.; Cavers, S.; Chałupka, W.; Dağdaş, S.; Hansen, J.K.; et al. Adaptive and plastic responses of Quercus petraea populations to climate across Europe. Glob. Chang. Biol. 2017, 23, 2831–2847. [Google Scholar] [CrossRef]
  9. Guisan, A.; Zimmermann, N.E. Predictive habitat distribution models in ecology. Ecol. Model. 2000, 135, 147–186. [Google Scholar] [CrossRef]
  10. Mahmoodi, S.; Heydari, M.; Ahmadi, K.; Khwarahm, N.R.; Karami, O.; Almasieh, K.; Naderi, B.; Bernard, P.; Mosavi, A. The current and future potential geographical distribution of Nepeta crispa Willd., an endemic, rare and threatened aromatic plant of Iran: Implications for ecological conservation and restoration. Ecol. Indic. 2022, 137, 108752. [Google Scholar] [CrossRef]
  11. Wiens, J.A.; Stralberg, D.; Jongsomjit, D.; Howell, C.A.; Snyder, M.A. Niches, models, and climate change: Assessing the assumptions and uncertainties. Proc. Natl. Acad. Sci. USA 2009, 106, 19729–19736. [Google Scholar] [CrossRef] [PubMed]
  12. Chakraborty, A.; Joshi, P.K.; Sachdeva, K. Predicting distribution of major forest tree species to potential impacts of climate change in the central Himalayan region. Ecol. Eng. 2016, 97, 593–609. [Google Scholar] [CrossRef]
  13. Fatemi, S.S.; Rahimi, M.; Tarkesh, M.; Ravanbakhsh, H. Predicting the impacts of climate change on the distribution of Juniperus excelsa M. Bieb. in the central and eastern Alborz Mountains, Iran. iForest 2018, 11, 643–650. [Google Scholar] [CrossRef]
  14. Valavi, R.; Shafizadeh-Moghadam, H.; Matkan, A.; Shakiba, A.; Mirbagheri, B.; Kia, S.H. Modelling climate change effects on Zagros forests in Iran using individual and ensemble forecasting approaches. Theor. Appl. Climatol. 2019, 137, 1015–1025. [Google Scholar] [CrossRef]
  15. Song, Y.G.; Petitpierre, B.; Deng, M.; Wu, J.P.; Kozlowski, G. Predicting climate change impacts on the threatened Quercus arbutifolia in montane cloud forests in southern China and Vietnam: Conservation implications. For. Ecol. Manag. 2019, 444, 269–279. [Google Scholar] [CrossRef]
  16. Khwarahm, N.R. Mapping current and potential future distributions of the oak tree (Quercus aegilops) in the Kurdistan Region, Iraq. Ecol. Process. 2020, 9, 56. [Google Scholar] [CrossRef]
  17. López-Sánchez, C.A.; Castedo-Dorado, F.; Cámara-Obregón, A.; Barrio-Anta, M. Distribution of Eucalyptus globulus Labill. in northern Spain: Contemporary cover, suitable habitat and potential expansion under climate change. For. Ecol. Manag. 2021, 481, 118723. [Google Scholar] [CrossRef]
  18. Ahmadi, K.; Alavi, S.J.; Amiri, G.Z.; Hosseini, S.M.; Serra-Diaz, J.M.; Svenning, J.C. The potential impact of future climate on the distribution of European yew (Taxus baccata L.) in the Hyrcanian Forest region (Iran). Int. J. Biometeorol. 2020, 64, 1451–1462. [Google Scholar] [CrossRef]
  19. Purohit, S.; Rawat, N. MaxEnt modeling to predict the current and future distribution of Clerodendrum infortunatum L. under climate change scenarios in Dehradun district, India. Model. Earth Syst. Environ. 2022, 8, 2051–2063. [Google Scholar] [CrossRef]
  20. Koç, D.E.; Biltekin, D.; Ustaoğlu, B. Modelling potential distribution of Carpinus betulus in Anatolia and its surroundings from the Last Glacial Maximum to the future. Arab. J. Geosci. 2021, 14, 1186. [Google Scholar] [CrossRef]
  21. Shafer, S.L.; Bartlein, P.J.; Thompson, R.S. Potential changes in the distributions of western north America tree and shrub taxa under future climate scenarios. Ecosystems 2001, 4, 200–215. [Google Scholar] [CrossRef]
  22. Austin, M.P.; Van Niel, K.P. Improving species distribution models for climate change studies: Variable selection and scale. J. Biogeogr. 2011, 38, 1–8. [Google Scholar] [CrossRef]
  23. Rehfeldt, G.E.; Crookston, N.L.; Sáenz-Romero, C.; Campbell, E.M. North American vegetation model for land-use planning in a changing climate: A solution to large classification problems. Ecol. Appl. 2012, 22, 119–141. [Google Scholar] [CrossRef] [PubMed]
  24. Chang, T.; Hansen, A.J.; Piekielek, N. Patterns and variability of projected bioclimatic habitat for Pinus albicaulis in the Greater Yellowstone Area. PLoS ONE 2014, 9, e111669. [Google Scholar] [CrossRef]
  25. Guisan, A.; Thuiller, W. Predicting species distribution: Offering more than simple habitat models. Ecol. Lett. 2005, 8, 993–1009. [Google Scholar] [CrossRef]
  26. Joseph, L.N.; Maloney, R.F.; Possingham, H.P. Optimal allocation of resources among threatened species: A project prioritization protocol. Conserv. Biol. 2009, 23, 328–338. [Google Scholar] [CrossRef]
  27. Levin, N.; Watson, J.E.M.; Joseph, L.N.; Grantham, H.S.; Hadar, L.; Apel, N.; Perevolotsky, A.; DeMalach, N.; Possingham, H.P.; Kark, S. A framework for systematic conservation planning and management of Mediterranean landscapes. Biol. Conserv. 2013, 158, 371–383. [Google Scholar] [CrossRef]
  28. Booth, T.H.; Nix, H.A.; Busby, J.R.; Hutchinson, M.F. BIOCLIM: The first species distribution modelling package, its early applications and relevance to most current MaxEnt studies. Divers. Distrib. 2014, 20, 1–9. [Google Scholar] [CrossRef]
  29. Zhao, Q.; Yu, S.; Zhao, F.; Tian, L.; Zhao, Z. Comparison of machine learning algorithms for forest parameter estimations and application for forest quality assessments. For. Ecol. Manag. 2019, 434, 224–234. [Google Scholar] [CrossRef]
  30. Hernández, B.; Raftery, A.E.; Pennington, S.R.; Parnell, A.C. Bayesian additive regression trees using Bayesian model averaging. Stat. Comput. 2018, 28, 869–890. [Google Scholar] [CrossRef]
  31. Sagheb-Talebi, K.; Sajedi, T.; Pourhashemi, M. Forests of Iran: A Treasure from the Past, a Hope for the Future; Springer: Berlin/Heidelberg, Germany, 2014. [Google Scholar]
  32. Shiravand, H.; Hosseini, S.A. A new evaluation of the influence of climate change on Zagros oak forest dieback in Iran. Theor. Appl. Climatol. 2020, 141, 685–697. [Google Scholar] [CrossRef]
  33. Pourhashemi, M.; Marvi Mohajer, M.R.; Zobeiri, M.; Zahedi Amiri, G.; Panahi, P. Identification of forest vegetation units in support of government management objectives in Zagros forests, Iran. Scand. J. For. Res. 2004, 19, 72–77. [Google Scholar] [CrossRef]
  34. Khalyani, A.H.; Mayer, A.L. Spatial and temporal deforestation dynamics of Zagros forests (Iran) from 1972 to 2009. Landsc. Urban Plan. 2013, 117, 1–12. [Google Scholar] [CrossRef]
  35. Ahmadi, R.; Kiadaliri, H.; Mataji, A.; Kafaki, S. Oak forest decline zonation using AHP model and GIS technique in Zagros forests of Ilam province. J. Biodivers. Environ. Sci. 2014, 4, 141–150. [Google Scholar]
  36. Hosseini, A.; Hosseini, S.M.; Linares, J.C. Site factors and stand conditions associated with Persian oak decline in Zagros mountain forests. For. Syst. 2018, 26, e014. [Google Scholar] [CrossRef]
  37. Karami, M.; Heydari, M.; Sheykholeslami, A.; Eshagh Nimvari, M.; Omidipour, R.; Yuan, Z.; Prevosto, B. Dieback intensity but not functional and taxonomic diversity indices predict forest productivity in different management conditions: Evidence from a semi-arid oak forest ecosystem. J. Arid Land 2022, 14, 225–244. [Google Scholar] [CrossRef]
  38. Najafi-Harsini, F.; Oladi, R.; Pourtahmasi, K.; Souto-Herrero, M.; García-González, I. Using tree-ring width and earlywood vessel features to study the decline of Quercus brantii Lindl in Zagros forests of Iran. Eur. J. For. Res. 2022, 141, 379–393. [Google Scholar] [CrossRef]
  39. Eskandari, S.; Jaafari, M.R.; Oliva, P.; Ghorbanzadeh, O.; Blaschke, T. Mapping land cover and tree canopy cover in Zagros forests of Iran: Application of Sentinel-2, Google Earth, and field data. Remote Sens. 2020, 12, 1912. [Google Scholar] [CrossRef]
  40. Moradi, M.; Jorfi, M.R.; Basiri, R.; Yusef Naanaei, S.; Heydari, M. Beneficial effects of livestock exclusion on tree regeneration, understory plant diversity, and soil properties in semiarid forests in Iran. Land Degrad. Dev. 2022, 33, 324–332. [Google Scholar] [CrossRef]
  41. Safaei, M.; Rezayan, H.; Firouzabadi, P.Z.; Sadidi, J. Optimization of species distribution models using a genetic algorithm for simulating climate change effects on Zagros forests in Iran. Ecol. Inform. 2021, 63, 101288. [Google Scholar] [CrossRef]
  42. Malekian, M.; Sadeghi, M. Predicting impacts of climate change on the potential distribution of two interacting species in the forests of western Iran. Meteorol. Appl. 2019, 27, e1800. [Google Scholar] [CrossRef]
  43. Chipman, H.A.; George, E.I.; McCulloch, R.E. BART: Bayesian additive regression trees. Ann. Appl. Stat. 2010, 4, 266–298. [Google Scholar] [CrossRef]
  44. Fattahi, M. Zagros Oak Forests and the Most Important Degradation Factors; Research Institute of Forests and Rangelands: Tehran, Iran, 1994; p. 64. [Google Scholar]
  45. Team, R.C. R: A Language and Environment for Statistical Computing; R Foundation for Statistical Computing: Vienna, Austria, 2021. [Google Scholar]
  46. Hijmans, R.J.; Cameron, S.E.; Parra, J.L.; Jones, P.G.; Jarvis, A. Very high resolution interpolated climate surfaces for global land areas. Int. J. Climatol. 2005, 25, 1965–1978. [Google Scholar] [CrossRef]
  47. Chatterjee, S.; Hadi, A.S. Regression Analysis by Example; John Wiley and Sons: Hoboken, NJ, USA, 2006. [Google Scholar]
  48. Dormann, C.F.; Elith, J.; Bacher, S.; Buchmann, C.; Carl, G.; Carré, G.; Marquéz, J.R.G.; Gruber, B.; Lafourcade, B.; Leitão, P.J.; et al. Collinearity: A review of methods to deal with it and a simulation study evaluating their performance. Ecography 2013, 36, 27–46. [Google Scholar] [CrossRef]
  49. IPCC. Climate Change 2014: Synthesis Report. Contribution of Working Groups I, II and III to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change; Pachauri, R.K., Meyer, L.A., Eds.; IPCC: Geneva, Switzerland, 2014. [Google Scholar]
  50. Prado, E.B.; Moral, R.A.; Parnell, A.C. Bayesian additive regression trees with model trees. Stat. Comput. 2021, 31, 20. [Google Scholar] [CrossRef]
  51. Phillips, S.J.; Anderson, R.P.; Schapire, R.E. Maximum entropy modeling of species geographic distributions. Ecol. Model. 2006, 190, 231–259. [Google Scholar] [CrossRef]
  52. Gassó, N.; Thuiller, W.; Pino, J.; Vilà, M. Potential distribution range of invasive plant species in Spain. NeoBiota 2012, 12, 25–40. [Google Scholar]
  53. Hosmer, D.W.; Lemeshow, S.; Sturdivant, R.X. Applied Logistic Regression; John Wiley and Sons:: Hoboken, NJ, USA, 2013. [Google Scholar]
  54. Allouche, O.; Tsoar, A.; Kadmon, R. Assessing the accuracy of species distribution models: Prevalence, kappa and the true skill statistic (TSS). J. Appl. Ecol. 2006, 43, 1223–1232. [Google Scholar] [CrossRef]
  55. Carlson, C.J. embarcadero: Species distribution modelling with Bayesian additive regression trees in R. Methods Ecol. Evol. 2020, 11, 850–858. [Google Scholar] [CrossRef]
  56. Yen, J.D.L.; Thomson, J.R.; Vesk, P.A.; Mac Nally, R. To what are woodland birds responding? Inference on relative importance of in-site habitat variables using several ensemble habitat modelling techniques. Ecography 2011, 34, 946–954. [Google Scholar] [CrossRef]
  57. Desta, F.; Colbert, J.J.; Rentch, J.S.; Gottschalk, K.W. Aspect induced differences in vegetation, soil, and microclimatic characteristics of an Appalachian watershed. Castanea 2004, 69, 92–108. [Google Scholar] [CrossRef]
  58. Coban, H.O.; Cosgun, S. The role of topography in the spatial distribution of tree species in the Mediterranean region of Turkey. Fresenius Environ. Bull. 2020, 29, 1369–1378. [Google Scholar]
  59. Babalik, A.A.; Sarikaya, O.; Orucu, O.K. The current and future compliance areas of Kermes oak (Quercus coccifera L.) under climate change in Turkey. Fresenius Environ. Bull. 2021, 30, 406–413. [Google Scholar]
  60. Çoban, H.O.; Örücü, Ö.K.; Arslan, E.S. MaxEnt modeling for predicting the current and future potential geographical distribution of Quercus libani Olivier. Sustainability 2020, 12, 2671. [Google Scholar] [CrossRef]
  61. Al-Qaddi, N.; Vessella, F.; Stephan, J.; Al-Eisawi, D.; Schirone, B. Current and future suitability areas of kermes oak (Quercus coccifera L.) in the Levant under climate change. Reg. Environ. Chang. 2017, 17, 143–156. [Google Scholar] [CrossRef]
  62. Shirk, A.J.; Cushman, S.A.; Waring, K.M.; Wehenkel, C.A.; Leal-Sáenz, A.; Toney, C.; Lopez-Sanchez, C.A. Southwestern white pine (Pinus strobiformis) species distribution models project a large range shift and contraction due to regional climatic changes. For. Ecol. Manag. 2018, 411, 176–186. [Google Scholar] [CrossRef]
  63. Alavi, S.J.; Ahmadi, K.; Hosseini, S.M.; Tabari, M.; Nouri, Z. The response of English yew (Taxus baccata L.) to climate change in the Caspian Hyrcanian Mixed Forest ecoregion. Reg. Environ. Chang. 2019, 19, 1495–1506. [Google Scholar] [CrossRef]
  64. Akyol, A.; Örücü, Ö.K.; Arslan, E.S. Habitat suitability mapping of stone pine (Pinus pinea L.) under the effects of climate change. Biologia 2020, 75, 2175–2187. [Google Scholar] [CrossRef]
  65. Sun, S.; Zhang, Y.; Huang, D.; Wang, H.; Cao, Q.; Fan, P.; Yang, N.; Zheng, P.; Wang, R. The effect of climate change on the richness distribution pattern of oaks (Quercus L.) in China. Sci. Total Environ. 2020, 744, 140786. [Google Scholar] [CrossRef]
Figure 1. The study area in the west of Iran (www.rsgisc.com accessed on 10 October 2022).
Figure 1. The study area in the west of Iran (www.rsgisc.com accessed on 10 October 2022).
Forests 14 00469 g001
Figure 2. Environmental variables used for modeling the distribution of Quercus brantii in the west of Iran.
Figure 2. Environmental variables used for modeling the distribution of Quercus brantii in the west of Iran.
Forests 14 00469 g002
Figure 3. Performance evaluation curves based on the BART model.
Figure 3. Performance evaluation curves based on the BART model.
Forests 14 00469 g003
Figure 4. Importance of environmental variables in modeling the distribution of Quercus brantii in the west of Iran.
Figure 4. Importance of environmental variables in modeling the distribution of Quercus brantii in the west of Iran.
Forests 14 00469 g004
Figure 5. Response curves of Quercus brantii to environmental variables.
Figure 5. Response curves of Quercus brantii to environmental variables.
Forests 14 00469 g005
Figure 6. Habitat suitability map for Persian oak (Quercus brantii) in the west of Iran based on the BART model under current climatic conditions.
Figure 6. Habitat suitability map for Persian oak (Quercus brantii) in the west of Iran based on the BART model under current climatic conditions.
Forests 14 00469 g006
Figure 7. Habitat suitability maps for Persian oak (Quercus brantii) in the west of Iran based on the BART model under future climatic conditions.
Figure 7. Habitat suitability maps for Persian oak (Quercus brantii) in the west of Iran based on the BART model under future climatic conditions.
Forests 14 00469 g007
Figure 8. The relationship between elevation and the distribution of Persian oak under current and future climatic conditions.
Figure 8. The relationship between elevation and the distribution of Persian oak under current and future climatic conditions.
Forests 14 00469 g008
Table 1. Environmental attributes including mean temperature (MT), maximum temperature, minimum temperature and precipitation considered for modeling the distribution of Quercus brantii.
Table 1. Environmental attributes including mean temperature (MT), maximum temperature, minimum temperature and precipitation considered for modeling the distribution of Quercus brantii.
VariablesDescriptionUnitUse for Modeling
BIO1Annual Mean Temperature°C🗴
BIO2Mean Diurnal Range°C
BIO3Isothermality%🗴
BIO4Temperature Seasonality%
BIO5Maximum Temperature of Warmest Month°C🗴
BIO6Minimum Temperature of Coldest Month°C🗴
BIO7Temperature Annual Range (BIO5-BIO6)°C🗴
BIO8MT of Wettest Quarter°C
BIO9MT of Driest Quarter°C🗴
BIO10MT of Warmest Quarter°C🗴
BIO11MT of Coldest Quarter°C🗴
BIO12Annual Precipitationmm/m2🗴
BIO13Precipitation of Wettest Monthmm/m2
BIO14Precipitation of Driest Monthmm/m2🗴
BIO15Precipitation Seasonality%
BIO16Precipitation of Wettest Quartermm/m2🗴
BIO17Precipitation of Driest Quartermm/m2🗴
BIO18Precipitation of Warmest Quartermm/m2🗴
BIO19Precipitation of Coldest Quartermm/m2🗴
Slope %
Solar Radiation kJ m−2 day−1
Table 2. Results of the performance evaluation of the models.
Table 2. Results of the performance evaluation of the models.
ModelAUCCutoffTSSType I ErrorType II Error
BART0.930.340.750.080.16
GLM0.86-0.68--
Table 3. Areas (km2) of habitat suitability for Persian oak (Quercus brantii) in the west of Iran under current and future climatic circumstances.
Table 3. Areas (km2) of habitat suitability for Persian oak (Quercus brantii) in the west of Iran under current and future climatic circumstances.
Not SuitableLow SuitabilityModerate SuitabilityHigh SuitabilityLossStableGain
Current14,211.661708.571603.582524.36
RCP 2.6–205015,234.41935.681319.531561.818094622115
RCP 2.6–207015,108.561605.391212.762123.4516554776417
RCP 8.5–205013,704.132553.581809.151980.4312675164993
RCP 8.5–207018,296.81811.68524.19424.44482716042
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

Mirhashemi, H.; Heydari, M.; Karami, O.; Ahmadi, K.; Mosavi, A. Modeling Climate Change Effects on the Distribution of Oak Forests with Machine Learning. Forests 2023, 14, 469. https://doi.org/10.3390/f14030469

AMA Style

Mirhashemi H, Heydari M, Karami O, Ahmadi K, Mosavi A. Modeling Climate Change Effects on the Distribution of Oak Forests with Machine Learning. Forests. 2023; 14(3):469. https://doi.org/10.3390/f14030469

Chicago/Turabian Style

Mirhashemi, Hengameh, Mehdi Heydari, Omid Karami, Kourosh Ahmadi, and Amir Mosavi. 2023. "Modeling Climate Change Effects on the Distribution of Oak Forests with Machine Learning" Forests 14, no. 3: 469. https://doi.org/10.3390/f14030469

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