s.j. alavi

The aim of this research was to identify the optimal habitat areas of ash species (Fraxinus excelsior L.) in Kheyrud Nowshahr forest, which covers an area of approximately 8,224 hectares. We recorded the geographical coordinates of 1004 ash individuals using a GPS device, along with environmental variables such as climate variables and primary and secondary topographic predictors extracted from DEM. We used the predictors with the least degree of correlation as input in the Generalized Linear Model, Generalized Additive Model, Random Forest, and maxent models. We prepared and analyzed the variables in ArcGIS, SAGA, and R software. To compare the performance of the four models, we considered the input variables in all models to be the same. We evaluated the models using kappa coefficient (K), Area Under Curve (AUC), and True Skill Statistic (TSS) measures. The RF model had the highest K (0.973), AUC (0.997), and TSS (0.973), indicating the best performance among the models. According to the RF model, the most important variables were valley depth, profile curvature, slope, and topographic position index. These variables indicate the areas with rich soil, sufficient moisture, proper drainage (non-stagnant water), slope less than 45%, and sufficient light (suitable for the vegetative stage). We identified the most suitable areas in the Baharbon, Chelir, and Menia-Sang districts due to optimal conditions for these effective variables.

Generally, optical satellite images are used to produce a land use map. Due to spectral mixing, these data can affect the accuracy of land use classifications, especially in areas with diverse vegetation.

In the present study, in order to achieve the correct land use classification in a mountainous-forested basin, four Landsat 8 thermal images were used with a few additional information (Normalized Difference Vegetation Index (NDVI), Digital Elevation Model (DEM), slope angle and slope aspect) along with optical data and data of multi-temporal images.

Results showed that thermal data, slope angle and DEM have a significant role in increasing the accuracy of land use classification, so that they increase the overall accuracy by about 3-10% from late spring to the beginning of autumn. Among the data used, slope angle and elevation data have a significant role in increasing the accuracy of agricultural classes. The total accuracy and Kappa coefficient in land use maps obtained from monotemporal images in the wet season (late spring; 83.93 and 0.82) and early summer (83.79 and 0.81)) are more than the dry season (late summer; 81.25 and 0.79) and early autumn).

Generally, the highest total accuracy among monotemporal images generated from optical data is about 83.95%, while the application of thermal and additional data along with optical data and the combination of monotemporal images of the wet season, the accuracy of the information multitemporal increased to 91.60% of the land use map.