Identification and Analysis of Changes in Sand Dunes and the Boundary of the Rig-e Shotoran Desert in the Northwest of Tabas Geopark

Extended Abstract

Introduction

Sandy seas or Erg in Iran, Afghanistan, Pakistan and Tajikistan are also known as (Reg), (Rig) or (Rek) and the name (Registan) means accumulated sand (Abassi et al., 2019). The most important issue in identifying and changing Regs is the way of expansion and the main axis of movement of Regs, determining the direction and extent of their development or limitation. Most of the Regs in the areas where the high air temperature and intense evaporation have caused a drop in moisture on the soil surface, and as a result, have weakened the bond between the soil grains in the horizontal surface of sandy soils and increasing the activity of sands and flowing sands are formed (Dang et al., 2004). Shotoran Reg, one of the largest sandy areas of Iran, has also undergone geomorphological changes under the influence of natural and human factors. Wind processes, including the movement of sand dunes and wind erosion, are one of the most important natural factors in moving sands and changing the shape of landforms in this region. The purpose of this research is to investigate the spatial-temporal changes of the Shotoran Reg in a period of 30 years using the Landsat satellite images and ground evidence. The results show that the identification of changes can be the basis for prediction future changes.

Methodology

The investigation of the border changes of Shotoran Reg based on Landsat satellite images (TM, ETM+ and OLI gauges) during 3 time periods with 10-year intervals (from 1994 to 2024) was analyzed by remote sensing in geographic information system software. The satellite images used were extracted from the United States Geological Survey website. Validation of the data was done in accordance with the reference pixels of the satellite image, validation and estimation of error and accuracy was done using the Kappa coefficient method, which in short was calculated to be 90% accurate and the Kappa coefficient was 0.85, which is a very acceptable evaluation.



Results and Discussion

In the classification of sand dunes using Landsat images and the Random Forest algorithm, the satellite images must first be preprocessed for data preparation. The identified range using Landsat images and the Random Forest model has been provided for four periods, including the years 1994, 2004, 2014, and 2024. Temporal changes due to weather conditions are observed in the images from 2014 and 2024. Specifically, in the 2014 image, due to drought, small parts of the salt flat have dried up, and sand movement caused by wind has covered portions of the surface. As a result, the model identified these areas as sand dunes. However, in the 2024 image, due to rainfall in the study area and considering the elevation of the salt flat regions, flooding has caused those areas to become wet. Therefore, it can be concluded that the boundaries of the sand dunes may change temporarily due to weather conditions, but in the long term, a distinct boundary exists between the sand dunes and the salt flat in this region. The analysis of changes in the boundaries of the Shotoran sand dunes based on climatic data shows that the annual wind speed charts, categorized by different speed ranges, from 1990 to 2024, exhibit no significant change and lack a trend. Landsat satellite imagery and the Random Forest machine learning algorithm have been utilized. Firstly, the Landsat satellite series (5, 7, 8, and 9), with global coverage and freely available data, are highly suitable for monitoring land cover changes on a large scale. With a spatial resolution ranging from 15 to 30 meters (depending on the band) and multispectral imaging capabilities, Landsat is reliable for detecting surface features such as sand, soil, salt, vegetation, and water. Secondly, the supervised Random Forest machine learning algorithm is highly effective and robust for classifying satellite imagery, particularly in handling noisy data and spectral overlap. Analysis of satellite data indicates that the overall extent of the Shotoran sand dunes has remained largely stable over the past 30 years. Changes observed in the boundaries of the sand dunes between images do not necessarily reflect actual changes in the extent of the sand dunes. They are related to the following technical and environmental factors: Landsat 5 (TM), Landsat 7 (ETM+), and Landsat 8 (OLI) have differences in spatial resolution, spectral resolution, and signal-to-noise ratio. Landsat 8 offers improved spectral quality and accuracy compared to previous generations, enabling better differentiation between surface covers such as sand, salt, and soil. These factors contribute to the observed apparent changes in the sand dune extent. During drought periods, these areas may dry out and become covered by sand, but with the first rainfall or runoff, the white, reflective salt surfaces or moist areas reappear.

Conclusion

The extent of the Shotoran sand dunes is geomorphologically stabilized, and the observed changes in satellite imagery are primarily due to technical factors (differences in sensors and imaging conditions) and environmental factors (variations in surface moisture and annual climate). Natural boundaries, such as salt flats and surrounding highlands, act as natural barriers, preventing the horizontal expansion of the sands. To accurately investigate the reasons for the expansion or lack thereof of the Shotoran sand dune boundaries, field methods and morphometric analysis of the sand dunes over several years are required. However, this was beyond the scope of this study due to time and cost constraints.