Extended Abstract
 
Introduction
Desert pavements, which are one of the drylands landforms, play a crucial role in the processes of these ecosystems. Desert surfaces that are covered in stones, rubble, and sand that can't be carried by wind are known as desert pavements. The identification of areas that have desert pavement and the estimated percentage of pavement coverage, as well as the sensitivity of the land unit to wind erosion, can lead to the identification of critical centers of wind erosion. Creating a detailed surface map is the initial step towards quantitative analysis of this role. Remote sensing technology is an effective tool for identifying and classifying landforms because it produces satellite images with spectral and spatial resolution, and it can also be a valuable tool for classifying desert pavements. Satellite images have been used in research to reveal the desert pavement. In these studies, the land cover classification method has been used to distinguish between the desert pavement surface and other types of land cover. Satellite image classification methods can adequately distinguish between desert pavement surfaces and other land covers. Knowing the percentage of paved stones in the region is of special importance in the study of natural resources, particularly in desert zones. Remote sensing techniques can be used to separate desert pavements, and their integration with field observations and statistical methods can help improve the estimation of the percentage of desert pavements, especially in vast, inaccessible and remote areas. This study is aimed at defining a model to estimate the percentage of desert pavement by employing statistical methods and satellite images to explore the possibility of estimating the amount of desert pavement coverage. The correlation between Tasseled Cap Transformation (TCT) bands and ground sampling points of the pavement percentage was utilized for this purpose. Finally, a map of the amount of desert pavement coverage in the study area was prepared.
 
Material and Methods
The study area falls within a 20-kilometer distance of Yazd city, situated at north latitude 3526361 to 3551713 and east longitude 215108 to 238866. A long strip runs from Khezrabad mountain upstream to Ashkezar and Zarch cities downstream, and it is composed of coarse, medium, and fine glacial plains. The study area's boundary was determined using topographic maps, aerial photos, Landsat ETM images, and Google Earth images. Geological maps, lithology units, and geomorphological facies were prepared based on the UTM coordinate system of the study area. The percentage of desert pavement coverage was measured on 27/12/2023 using the aforementioned maps and stratified-random sampling method. The OLI sensor image of the Landsat satellite was obtained simultaneously with the field sampling date. The initial process involved performing pre-processing operations on the satellite image, including radiometric and geometric corrections, to prepare it for processing operations. In the image processing stage, Tasseled Cap Transformation (TCT) was used and images of brightness, greenness, humidity, fourth, fifth and sixth bands were extracted based on standard coefficients for Landsat 8 satellite. After that, the TCT bands' numerical values, which were related to the locations of the sampled points on the ground, were retrieved. Pearson's correlation coefficient was used to calculate the statistical correlation between ground pavement coverage and TCT bands. Using a factor analysis method, the suitable variables were determined to prepare the desert pavement percentage estimation model. The desert pavement model was created using a stepwise multiple regression method after determining the appropriate variables. In this method, the values ​​of one variable (dependent variable Y) were estimated from the values ​​of two or more variables. (independent variables X1, X2, ..., XP).
 
Results and Discussion
To determine the most suitable model for estimating the percentage of desert pavement, factor analysis was used. Tasseled cap transformation functions were used to present the characteristics of soil brightness, greenness, and moisture in six distinct maps. The indicated indices are a single-band index that ranges from zero to 1. The problem is more easily distinguished when the values are close to zero. Statistical correlation between the percentage of pavement and six components of TCT by Pearson's method, indicate a significant correlation at 1% between the components of Brightness, Greenness, Tasseled Cap Transformation 5, and the percentage of pavement coverage. The highest correlation coefficients between the percentage of pavement coverage and the Brightness components and TCT (respectively 0.64 and 0.62), and the lowest one related to the humidity component, were estimated at 0.02. The principal component analysis algorithm in the factor analysis method detected the selected indicators of Greenness and Brightness by analyzing the input data output. Greenness and brightness variables were considered independent variables. Their relationship with desert pavement was investigated as a dependent variable using the stepwise regression method. The selected model shows the relationship between the amount of desert pavement and the Greenness and Brightness variables, with a correlation coefficient of 0.61 and a standard error of 23.2. The obtained model can predict approximately 61% of surface pavement changes in the study area. The connection between the bands obtained from Tasseled Cap Transformation 5 and the ground collection points of pavement percentage was employed for this purpose. At last, a map of the desert pavement coverage in the study area was created. The research's findings indicated that a suitable model for preparing a map of desert pavement percentage can be created by combining ground observations with statistical methods and satellite images.