Determining the Potential Ecological Niche of Festuca ovina L. using Genetic Algorithm (GARP) and DOMAIN Models in Rangelands of Isfahan Province

Understanding the spatial distribution of plants, a crucial aspect of plant ecology, plays a significant role in describing the sustainability of rangeland ecosystems and designing appropriate management programs. A wide range of environmental factors, from atmospheric to land surface characteristics, influence species distribution ranges at local to global scales. Among these factors, bioclimatic variables are indispensable in species distribution modeling. This study used two modeling techniques, Genetic Algorithm for Rule Set Production (GARP) and DOMAIN, to model the potential ecological niche of sheep's fescue (Festuca ovina L.) in the rangelands of Isfahan Province.

The occurrence data of the studied species were collected using a stratified-random method. Twenty-two environmental layers, including three physiographic variables and 19 climatic variables derived from precipitation and temperature, were used in the modeling process. These variables were converted to a pixel size of one square kilometer in the ArcMap 10.3 software and then entered into the modeling process as digital maps. The genetic algorithm model was implemented in Open Modeler 1.1.0 software, and the domain model was implemented in DOMAIN 32 software. To select environmental variables, the Pearson correlation test was performed. One variable was selected, and the other was removed from pairs of variables with a correlation of more than 0.8. In the domain model, habitat suitability was predicted based on Gower's dissimilarity metric. To validate and evaluate the potential habitat maps obtained from the two models, confusion matrices and statistical indices derived from them, such as AUC (the area under the ROC curve), Kappa coefficient, and overall accuracy (CCR), were used. In the sensitivity analysis, environmental variables that had the largest drop in model performance were selected as the most important environmental variables.

The sensitivity analysis showed that the most effective environmental variables in the distribution of the species were height above sea level, precipitation of the warmest quarter, mean diurnal range, isothermality, annual precipitation, annual mean temperature, slope, and temperature annual range. The genetic algorithm (AUC = 0.99) performed better than DOMAIN (AUC = 0.67). In the study of the geographical range of the species under the models, it was predicted that favorable habitats (occurrence probability greater than 75%) in GARP and DOMAIN models occupy about 7069 and 10187 square kilometers of the province, respectively. The results of GARP and DOMAIN models also indicated that the main habitat of F. ovina is in Fereydunshahr and Semirom counties.

Certain areas of Fereydunshahr and Semirom counties, which encompass summer rangelands of the province, along with the heights of Karkas and Muteh as the cold steppe rangelands of the province, have a high potential for the presence of F. ovina. The impact of climate change, combined with excessive grazing, can gradually reduce and eliminate this palatable species, replacing it with other species. The findings of this study can be applied to determine areas prone to seeding and pit-seeding projects using this species. It is suggested that biological and management activities, such as land use change, grazing intensity, and inter-species competition, be included in the modeling process to enhance the prediction of favorable species habitats.