Abstract: (4 Views)
Introduction and Objectives: Climate change is currently considered a serious threat to many species and recognized as one of the most important factors in the global biodiversity loss. Therefore, understanding how the spatial distribution and species composition are affected by climate change is very important in protecting natural ecosystems and achieving sustainable development. Species distribution models are the most widely used methods to predict the effect of climate change on plant species distribution changes. Considering the ecological, economic and commercial values of the chestnut-leaved oak (Quercus castaneifolia C. A. Mey) in the Hyrcanian forests, the aim of this study is to use different modeling algorithms in order to simulate suitable climatic ranges to determine the habitat suitability of Q. castaneifolia in the current climatic conditions and its potential changes in the years 2070 and 2100 AD.
Material and Methods: After determining the species presence using the statistics data of forestry projects in the north of Iran and the detailed plans of Golestan, Mazandaran and Gilan provinces, the bioclimatic variables of 1979-2013 were extracted from the CHELSA global database. The bioclimatic variables related to the years 2014-2019 were also produced in the Idrisi TerrSet software using raster images of monthly precipitation and monthly maximum and minimum temperature available in the same database. Then, the weighted average of these two series of bioclimatic variables (1979-2019) was included in the modeling process. In addition, the physiographic variables of elevation, slope and solar-radiation aspect index (TRASP) extracted from the Digital Elevation Model (DEM) were also used as input to the modeling process. After selecting the environmental variables with Variance Inflation Factor (VIF), the relationship between the species occurrence data and the map of environmental variables was mathematically defined using the R statistical-programming software. Regression, machine learning, and classification modeling algorithms including Artificial Neural Network (ANN), Classification Tree Analysis (CTA), Generalized Linear Model (GLM), Multiple Adaptive Regression Spline (MARS), Maximum Entropy (MaxEnt), and Random Forest (RF) were evaluated using the Biomod2 package. In order to reduce uncertainty, a unified framework including six species distribution models was used. Area Under the Curve (AUC) index, True Skill Statistic (TSS), Sensitivity and Specificity were used to evaluate the performance of the models. After determining the importance of the participating variables in the modeling with the VarImp function, the species response curves to the most important variables were drawn based on the outperformed individual model. To predict the effect of climate change on species distribution, MRI-ESM2-0 model of the sixth phase of climate change models (CMIP6) under two optimistic (SSP1-2.6) and pessimistic (SSP5-8.5) climate change scenarios over the near future (2041-2070) and distant future (2071-2100) was used.
Results: Based on the evaluation criteria, the individual models have good performance and were considered to create an ensemble model. Among models, the ensemble model with TSS and AUC equal to 0.904 and 0.988, respectively, and then the random forest model had the highest efficiency. Based on the contribution percentage values, precipitation of the driest month (Bio14), the slope and precipitation seasonality (Bio15) had the largest contribution in the distribution of Q. castaneifolia and determining its habitat suitability, respectively. According to the ensemble model, the suitable habitat areas of the species in the current climatic conditions cover 38.13% of the study area. The produced maps show the high suitability of Q. castaneifolia in the western and central parts of Hyrcanian region. According to the species response curves to the most important environmental variables, in suitable habitats, the precipitation of the driest month is at least 10 mm, the precipitation seasonality is less than 50 mm, and the average slope is 2-22%. The results showed that in both time periods and under both climate change scenarios, there will be changes in the spatial distribution of Q. castaneifolia and the most severe one would be a 4.9 percent loss in the species suitable climate range in 2100 under the pessimistic scenario (RCP8.5). By comparing the ensemble map of current habitat suitability and habitat suitability under the effect of climate change, it was predicted that the most change in habitat suitability will occur in the eastern and southern parts of Hyrcanian region. Probably, with the increase in temperature, the habitat of species will shift from lower latitudes or altitudes to higher latitudes.
Conclusion: Despite the difference in the nature of different modeling algorithms, the resulting predictions were almost similar for Q. castaneifolia. Meanwhile, the random forest had the highest accuracy and the generalized linear model had the lowest accuracy among the individual models. Only the random forest model could have a performance equivalent to the average output of several modeling methods. Suitable habitat of Q. castaneifolia will decrease under the pessimistic scenario in both future time frame and under optimistic scenario until 2070. Examining the potential effects of climate change on the spatial distribution of this valuable plant species as an important species of Hyrcanian forests seems to be an essential tool for its planning, conservation and management. Habitat suitability maps can be proposed as a basis for reforestation plans in threatened areas. Therefore, preparing comprehensive conservation plans with the aim of reducing the effects of climate change on this valuable species seems necessary.
Material and Methods: After determining the species presence using the statistics data of forestry projects in the north of Iran and the detailed plans of Golestan, Mazandaran and Gilan provinces, the bioclimatic variables of 1979-2013 were extracted from the CHELSA global database. The bioclimatic variables related to the years 2014-2019 were also produced in the Idrisi TerrSet software using raster images of monthly precipitation and monthly maximum and minimum temperature available in the same database. Then, the weighted average of these two series of bioclimatic variables (1979-2019) was included in the modeling process. In addition, the physiographic variables of elevation, slope and solar-radiation aspect index (TRASP) extracted from the Digital Elevation Model (DEM) were also used as input to the modeling process. After selecting the environmental variables with Variance Inflation Factor (VIF), the relationship between the species occurrence data and the map of environmental variables was mathematically defined using the R statistical-programming software. Regression, machine learning, and classification modeling algorithms including Artificial Neural Network (ANN), Classification Tree Analysis (CTA), Generalized Linear Model (GLM), Multiple Adaptive Regression Spline (MARS), Maximum Entropy (MaxEnt), and Random Forest (RF) were evaluated using the Biomod2 package. In order to reduce uncertainty, a unified framework including six species distribution models was used. Area Under the Curve (AUC) index, True Skill Statistic (TSS), Sensitivity and Specificity were used to evaluate the performance of the models. After determining the importance of the participating variables in the modeling with the VarImp function, the species response curves to the most important variables were drawn based on the outperformed individual model. To predict the effect of climate change on species distribution, MRI-ESM2-0 model of the sixth phase of climate change models (CMIP6) under two optimistic (SSP1-2.6) and pessimistic (SSP5-8.5) climate change scenarios over the near future (2041-2070) and distant future (2071-2100) was used.
Results: Based on the evaluation criteria, the individual models have good performance and were considered to create an ensemble model. Among models, the ensemble model with TSS and AUC equal to 0.904 and 0.988, respectively, and then the random forest model had the highest efficiency. Based on the contribution percentage values, precipitation of the driest month (Bio14), the slope and precipitation seasonality (Bio15) had the largest contribution in the distribution of Q. castaneifolia and determining its habitat suitability, respectively. According to the ensemble model, the suitable habitat areas of the species in the current climatic conditions cover 38.13% of the study area. The produced maps show the high suitability of Q. castaneifolia in the western and central parts of Hyrcanian region. According to the species response curves to the most important environmental variables, in suitable habitats, the precipitation of the driest month is at least 10 mm, the precipitation seasonality is less than 50 mm, and the average slope is 2-22%. The results showed that in both time periods and under both climate change scenarios, there will be changes in the spatial distribution of Q. castaneifolia and the most severe one would be a 4.9 percent loss in the species suitable climate range in 2100 under the pessimistic scenario (RCP8.5). By comparing the ensemble map of current habitat suitability and habitat suitability under the effect of climate change, it was predicted that the most change in habitat suitability will occur in the eastern and southern parts of Hyrcanian region. Probably, with the increase in temperature, the habitat of species will shift from lower latitudes or altitudes to higher latitudes.
Conclusion: Despite the difference in the nature of different modeling algorithms, the resulting predictions were almost similar for Q. castaneifolia. Meanwhile, the random forest had the highest accuracy and the generalized linear model had the lowest accuracy among the individual models. Only the random forest model could have a performance equivalent to the average output of several modeling methods. Suitable habitat of Q. castaneifolia will decrease under the pessimistic scenario in both future time frame and under optimistic scenario until 2070. Examining the potential effects of climate change on the spatial distribution of this valuable plant species as an important species of Hyrcanian forests seems to be an essential tool for its planning, conservation and management. Habitat suitability maps can be proposed as a basis for reforestation plans in threatened areas. Therefore, preparing comprehensive conservation plans with the aim of reducing the effects of climate change on this valuable species seems necessary.
Send email to the article author
| Rights and permissions | |
 |
This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License. |