Mohadeseh Amiri; Mohammad Shafiezadeh; Mostafa Tarkesh; Seyyed Mostafa Moslemi
Abstract
Introduction: Invasive species are currently the concern of ecologists, conservationists and natural resource managers, and decrease biodiversity due to their rapid spread. These species cause changes in ecological processes, function and structure of community in natural ecosystems. The most obvious ...
Read More
Introduction: Invasive species are currently the concern of ecologists, conservationists and natural resource managers, and decrease biodiversity due to their rapid spread. These species cause changes in ecological processes, function and structure of community in natural ecosystems. The most obvious change in the invaded areas is the reduction of biodiversity and the creation of a pure community of invasive plants. One of the invasive species in our country is Prosopis juliflora, which is important in the field of combat desertification, biological control and stabilization of quicksand dunes in the southern regions of Iran due to its resistance to adverse environmental conditions. Material and methods: In the present study, the efficiency of five discrimination group models (GLM, GBM, ANN, SRE, RF) and a profile model (MAXENT) and their ensemble with the weighted average approach in spatial distribution of this species in Makuran region and determining the most important environmental factors affecting the invasion distribution were investigated. By recording 63 occurrence points and 100 absence points, using climatic, physiographic and human variables as environmental variables, and evaluating the performance of models by Area under Curve (AUC), True Skill Statistics (TSS), Sensitivity and Specificity, the species invasion potential was determined. Results and discussion: Among the single algorithms, according to the threshold-independent and threshold-dependent evaluation criteria, two machine learning techniques, i.e. RF and GBM, predicted the climatic habitat of this invasive species with higher accuracy. Also, the evaluation criteria in the ensemble prediction were higher than the average of all single modelling algorithms. According to the ensemble model, P. juliflora habitats occupy about 15% of the total study area. After generalization of the models to the geographical space, it was found that the invaded areas have spread in a uniform strip on the shores of Oman Sea and Persian Gulf. Evaluation of variable importance indicated that Altitude was the most important independent variable justifying about half of the changes in the ensemble model and has the greatest effect on species distribution. The variable of distance from road was the next important variable. But, aspect was mentioned the least important environmental variable affecting the scattering of the invasion. Based on response curves, the maximum probability of the species presence was observed at the altitude of 50 m above sea level and a distance less than 50 m from the road. Also, the species is most likely to be present, if the temperature in the warmest month and the coldest season of the year is more than 34 and 14 °C, respectively, and the precipitation seasonality is 100-150. Conclusion: It was found that the integrated algorithms in the framework of ensemble modelling showed higher accuracy and the maps derived from potential distribution of species invasion make it possible to restrict and manage the distribution range of invasive species by providing management solution and conserving plans to protect native species. In fact, the results of this study can be used as a basis for subsequent monitoring to prevent further spread of invasive species and to create a balance between the native vegetation protection programs of the region and desertification management measures.
Samira Sadat Fatemi Azarkhavarani; Mohammad Rahimi; Mostafa Tarkesh; Hooman Ravanbakhsh
Abstract
Introduction: Species distribution modeling is a common method for understanding the relationships between a species and its environment and is used to predict the changes in distribution due to environmental changes. A lot of research has been done around the world based on these models. In this study, ...
Read More
Introduction: Species distribution modeling is a common method for understanding the relationships between a species and its environment and is used to predict the changes in distribution due to environmental changes. A lot of research has been done around the world based on these models. In this study, the optimal habitat of Juniperus excelsa, an indicator species, and the most important tree species in Irano-Turanian forests in the Alborz Mountains have been investigated using 38 environmental parameters and domain and logistic regression models. Material and methods: The study area consists of parts of the Alborz Mountains located in the north of Iran with an area of 14656 km2. In the present study, a stratified random sampling method was used to determine the sampling sites. Finally, 390 occurrence sites (240 presence sites and 150 absentee sites) of J. excelsa at a 30-second scale (approximately 1 km×1 km) as the input model was studied. Also, 38 environmental parameters were considered as predictive variables for implementing two models of the domain and logistic regression. Results and discussion: The results showed that the domain model had a high performance for predicting the habitat of J. excelsa with AUC =0.97, Kappa =0.730, and TSS = 0.91. Based on the results, the areas with the lowest potential for the presence of J. excelsa were 5665.95 km2, the moderate potential was 2033.1 km2, the good potential was 3076.38 km2, the very good potential was 3063.42 km2, and the high potential was 817.29 km2. The results obtained from the implementation of the logistic regression model indicated that 5084.37 km2 of the studied area was in the class of least potential, 2539.35 km2 had moderate potential, 1410.21 km2 had good potential, 1104. 84 km2 had very good potential, and 4517.37 km2 had high potential. Also, the results showed that the suitable habitats for J. excelsa were regions with a mean diurnal range of 13 °C and 15.5 °C, annual precipitation of 120-220 mm, precipitation of 64-90 mm, precipitation of coldest quarter of 35-60 mm, the altitude from the sea level of 3100-1800 m, and a slope of 30-10% in the southern, southeast, and east directions. Also, the probability of species occurrence was more in areas near the snow and water line as well as on calcic rocks. Also, in the habitat of J. excelsa the NDVI index varied between 0.38 and 0.20. The results of the performance evaluation of the models showed that the domain model had higher performance in predicting the suitable habitat of J. excelsa than the logistic regression model in the study area. Conclusion: Species distribution models can provide a suitable management plan for different parts of the habitat. Areas with low potential of suitable habitat are usually partial habitats and populations in these areas are more vulnerable than the others, whereas areas with very good potential are excellent for a protected area. Areas with good to the excellent potential that do not have significant tree cover are suitable for forestry and native ecosystem restoration. The present study showed that the domain model, despite using only presence data, can be a useful method for predicting the suitable habitat of J. excelsa in the Alborz Mountains. Therefore, it can be stated that species distribution models, with acceptable accuracy, have enough performance in the evaluation of species distribution and can be used to execute conservation strategies.
