| Abstrakt: |
The models for predicting the geographic distribution of plant species are static and probabilistic models and specify the mathematical relationships governing the geographical distribution of species with their current environment and important environmental factors affecting the distribution of species. The aim of the present research was to preparing a habitat prediction map of the Artemisia chamaemelifolia including the important medicinal species in the rangelands of Ardabil province. In the rangelands of Ardabil province, 449 sampling sites of the presence and absence of the studied species were considered from 2018 to 2021. Two categories of environmental factors including bioclimatic variables (19 cases) and topographic variables including primary indicators (3 cases) and secondary indicators (6 cases) were investigated in relation to the presence of the species. Maps of all environmental factors in the environment of the geographic information system were prepared and overlapped with 70% of the data. Prediction of species presence with four different modeling methods including; generalized linear model (GLM), generalized additive model (GAM), random forest model (RF) and generalized boosted regression model (GBM) were performed in the R software environment. The analysis of the importance of environmental variables for the models was done in the Biomode2 package. To evaluate the models, 30% of the species data and three statistics of Area under Curve (AUC), Kappa and True Skill Statistics (TSS) were used. The results of the modeling showed that the elevation was the most effective variable in the distribution of A. chamaemelifolia species in all four studied methods. In the GAM method, the variables of precipitation of the driest month, precipitation of the hottest season, and percentage of slope were also obtained as effective factors on the presence of this species. In addition to the elevation and the precipitation of the hottest season, the RF model also added the average daily temperature range to the factors affecting the presence of the species. Comparing the performance of the models showed that the GAM model with AUC 0.993, Kappa index 0.969, and TSS 0.985, is the best model among the studied models and was able to predict the habitat of the species at an excellent level. After that, the RF model with the AUC of 0.996, kappa 0.936, and TSS 0.941, with a small difference, is the second approved model in this connection. The results of this study showed that the species of A. chamaemelifolia has a relatively limited ecological niche and tends to grow in its own habitat conditions. Therefore, the prepared prediction maps can introduce the geographical areas in which the species is present to design and announce the protection areas of the species and can be used to suggest the species in the modification and restoration of areas with similar ecological conditions as the species' presence areas. The prediction model in the current study is expected to be effective for future conservation strategies. [ABSTRACT FROM AUTHOR] |
