عنوان مقاله [English]
Habitat Determination of Ferula gummosa Boiss. Using Generalized Additive Model in Lar Rangeland, Tehran Province
Knowledge of plants spatial distribution involves understanding and management of ecological factors affecting them which a prominent role in the evaluation, protection, development and regional planning. Due to the ecological and economic importance of Ferula gummosa Boiss, this study aimed to identify potential habitat using GAM model, distribution of this species with environmental factors and economic valuation was selected Lar rangeland custom unit at Tehran province. Identify and collect basic information of the study area, determination of the custom unit boundary by using a topographic and other map, typing of habitat, random sampling and estimation of soil surface cover properties and soil sampling to 30 cm depth, modeling of potential habitat, the prediction of the economic value of market value in the study area has been process of implementation of this study. The results showed that phosphorus factor with decreasing linear relationship and moisture, acidity and silt with 2 degree of freedom were the most important factors affecting density of Ferula gummosa Boiss (R2 is 0.44 and RMSE equals 1.32). Also, aspect, moisture, silt, potassium, organic matter and nitrogen with 1 degree of freedom and elevation with 2 degrees of freedom were major factors affecting canopy cover (R2 is 0.68 and RMSE equals 0.25). Also, from the thirteen factors examined, aspect with 1 degree of freedom and elevation and silt were major factors affecting on distribution Ferula gummosa Boiss 1 degree of freedom (AUC = 0.72 and Kappa coefficient = 0.49). Potential habitat map shows that 7.8 percent of the total study area has a perfect match, 37.9% fit, 44.9% are low and 8.5% inappropriate to predict the distribution of this species.
“Potential habitat”, “soil”, “Aspect”, “Lar Rangeland”, “GAM model”.
Determining the field of plant species development as an integral part of natural ecosystems and identifying different habitat characteristics of plants has always been considered by experts (Azarnivand et al., 2007). The basis of plant species distribution modeling is plant species data. This data can be based on presence, presence and absence only, or on abundance data, and modeling techniques are available for each of these types of data. Inside the country, some researchers have conducted studies in the field of distribution model models that can be studied (Ghazi Moradi et al., 2016; Heidari et al., 2017; Jafarian and Kargar, 2017; Jafarian et al., 2019 ).In this regard, according to unofficial reports about the presence of the plant in the rangelands of Lar Dam watershed in Tehran province, in order to determine and identify the potential habitat of the plant using mathematical and modeling methods, the distribution of this species with environmental variables and Possibility of economic exploitation for livestock farmers, Saman Orfi Lar of Shemiranat city in Tehran province was selected.
Random-classification method was used for sampling (Hirzel and Guisan,2002). For this purpose, after separating the boundaries of the types and determining the representative areas in the sampling sites, the characteristics of surface cover including the percentage of total species cover and the percentage of Ferula gummosa cover, determining the number of plant species including the density of total plant species and The load density in each plot was done by recording the position of the plots. For sampling, the vegetation of the region was collected during the months of May to early July 2017 (during the growing season). A total of 7 plant types were identified in the study area. In each type, 10 plots of 2 × 2 m2 were established and a total of 70 plots of vegetation were sampled. In each plot, the percentage of canopy cover of all species and the percentage of canopy cover of Ferula gummosa, percentage of rocks and pebbles, bare soil and litter amount, determining the number of plant species including total plant species density and Barijeh density by recording the location of plots. In each plant type, 3 soil samples were randomly collected in the vicinity of the plot from a depth of 0-30 cm and a total of 21 soil samples. The basis of the analysis used in this part of the research is actually the use of the generalized GAM additive model. The 10-fold validation calculation is performed in the caret package (Artensen et al., 2010). This model was implemented in R 2.9.2 software and GRASP package.
Results of evaluation of affecting factors on the distribution of Ferula gummosa Boiss. Using GAM model and the error matrix showed that the prediction of model had a kappa coefficient of 0.49 and good agreement with the area under the plot curve (AUC) method with value of 0.72 that is acceptable. In this study, using the minimum distance algorithm, was prepared map of the potential habitat of Ferula gummosa Boiss. with a spatial resolution of 90 meters. According to the Boyce index, 4 classes were identified for classification of potential habitat map. According to the results, the maximum and minimum areas of habitat were 3078.6 and 581.4 hectares, respectively. So that 8.7% of the total study area had excellent proportion and 37.9% was suitable for the potential habitat of the mentioned species.
Changes in chemical and physical properties of soil are the most important factors affecting the frequency and vegetation changes, which is consistent with the findings. In this study, the distribution of Ferula gummosa plant and the preparation of a map of potential habitats for only one customary system at the level of 6852 hectares were investigated. Since one of the goals of modeling is to study ecological hypotheses, it is suggested that in future studies, while examining ecological nests and determining the potential habitats of species associated with Ferula gummosa, the extent of their habitat overlap with this species should be determined. Using a secure database removes many of the initial barriers to modeling research. Therefore, it is suggested that raw data be collected and made available by creating a database so that these studies can be carried out both locally and on a large scale.