نوع مقاله : مقاله پژوهشی
عنوان مقاله English
نویسنده English
Aim: The purpose of this research is to identify the area of sand dunes and their changes in desert areas using spectral indicators in Landsat 8 and Sentinel 2 satellites.
Material & Method: In this research, using four bands of Sentinel 2 data, a new spectral index named NDESI has been presented for the identification and recognition of sand dunes in the Rig Jen area. This index uses the blue, red, or red edge of vegetation and two short-wave infrared bands, SWIR1 and SWIR2, to produce the image. A threshold calculation method was used to create unique thresholds for each image.
Finding: Based on this, the threshold values for equations 1 and 2 in March and July 2023 of the Sentinel 2 satellite were obtained as 0.261 and 0.217, respectively. This amount for Landsat 8 satellite in 2013, 2018, and 2023 was equal to 0.063, 0.0735, and 0.071, respectively. According to equation 1 of the Sentinel 2 satellite in July 2023, the extent of sand dunes in this area was equal to 2262 square kilometers. For Landsat 8 satellite in the same year, it was 2638 square kilometers. In the discussion of Pearson correlation, it was also observed that the highest correlation of 0.63 between the NDESI index and band 7 of the Landsat 8 satellite and the lowest correlation of -0.14 between this index and band 2 in equation 1 of Sentinel satellite 2 has been.
Conclusion: Finally, the accuracy evaluation of the images obtained from equations 1 and 2 showed an overall accuracy of 87.4 and 83.7 percent, respectively. This index is also compatible with Landsat 8 data.
Innovation: The results of this research have been used in the investigation of sand dunes and their identification.
کلیدواژهها English
Extended Abstract
1. Introduction
Coastal and desert sand dunes cover about 20% of the dry areas of the world. Active sand deserts occupy about 10% of the land between 30 degrees north and south latitudes. In some African countries, such as Morocco, 45 to 55 percent of the land is desert, which is spread in its southeastern parts. Since the 1960s, optical images have been widely used for land cover mapping. For land cover detection, processing techniques such as false color composite (FCC), principal component analysis (PCA), minimum noise fraction (MNF), the ratio between bands, and spectral indices have been used. The technique of spectral indices has been widely used to identify land covers. To research and monitor hills and sandy surfaces, remote sensing techniques were developed using the OLI sensor of the Landsat 8 satellite. According to the mentioned cases, the sand dunes of Rig Jen were investigated in this research. In this research, the 20-meter Sentinel 2 data was used to identify desert sand dunes in order to provide a new index called the Normalized Excess Sand Difference Index (NDESI). This index can distinguish between barren soil and sand dunes. This index was adapted to the Landsat 8 OLI meter to track the changes in sand dunes over time. The OLI sensor in Landsat 8 has more image archives than Sentinel 2.
2. Materials and Methods
In this research, two types of images are used. One of these images is related to Sentinel 2 level 2A reflectivity images of the lower atmosphere (BOA), which was produced from Level IC data and refers to March and July 2023. This image has 13 bands with a resolution of 10 meters, 20 meters, and 60 meters. This research used only four bands to provide a new index. In order to adapt the new index to Landsat 8 images and its effectiveness in examining the changes in sand dunes over time, Landsat 8 OLI sensor images were used in July 2013, 2018, and March 2023. These images have passage number 162 and row 37. These images have 11 bands, but only 4 bands are used. All images are available for free and cover the entire study area, which is free of vegetation and clouds. The last image used in the research was examined on Google Earth to evaluate its accuracy.
3. Results and Discussion
A small change in the number of thresholds can show the area of dunes more or less than usual. The equation for thresholds in Landsat 8 images shows more realistic values. However, this is not seen in the Sentinel images. Therefore, the peak point of spectral reflectance was used to determine the thresholds in Sentinel 2 images. The areas determined based on these thresholds in equation 1 and 2 are very close. For example, in the Sentinel images of March 2023, the area according to equation 1 equals 1101.1 square kilometers, and according to equation 2, it is 1343.6 square kilometers. The total area of the study area in Rig Jen is equal to 3548.27 square kilometers. Therefore, 37.8 percent of the area is covered with sand dunes, which equals 31.01 percent based on equation 1. This shows that the difference between equations 1 and 2 is 79.6%, which is negligible. However, the difference between Sentinel 2 and Landsat images has been significant. According to the images of this satellite, the area of sand dunes in 2023 is equal to 1633.1; in other words, sand dunes occupy 46% of the area of the region, and the difference between Landsat 8 and Sentinel 2 in equation 2 is 8.2%.
4. Conclusions
The normalized difference of excess sand index (NDESI) is a new index for drawing sand dunes, and for this purpose, it uses Sentinel 2 and Landsat 8 images. Due to confounding factors, sand dunes stabilized by dense vegetation may not be recognized and may not be universally applicable. However, this index has performed better in dry and barren lands with active sand dunes. Equations 1 and 2 in Sentinel 2 images assign the highest pixel values to sand, making these areas easily recognizable with the brightest color. Using equation 2 in areas with dense vegetation is better because it can separate vegetation and water from sand dunes. Based on the presented equations, the NDESI index was matched with Landsat 8 data. Using equation number 3 and Landsat 8 bands, sand dunes were detected in 2013, 2018, and 2023. The limited movement of sand dunes in the studied area was due to the surrounding of this area with mountainous areas. The sand threshold calculation technique allows choosing the correct threshold for each equation and each image. The thresholds obtained from the data of Sentinel 2 and Landsat 8 images have achieved acceptable results.
5. Acknowledgement & Funding
The authors are thankful to all interview participants for supporting this research.
The manuscript did not receive a grant from any organization
Conflict of Interest
The authors declare no conflict of interest.
