Evaluation of desertification in Qazvin Plain using Sentinel 2 images and spectral indices and degree of desertification (DDI)

1. Introduction

Different researches provide a wide range of remote sensing-based techniques for desertification monitoring. Among them, we can mention the techniques of spectral composition analysis and spectral indices. Approaches based on different indicators are widely used to investigate land degradation. For example, Normalized Difference Vegetation Index (NDVI) and Modified Soil Vegetation Index (MSAVI) are used to monitor vegetation conditions in desertification assessment. Indices such as albedo and surface soil particle size index (TGSI) can detect the spatial heterogeneity of soil texture. In fact, many researchers around the world have successfully used these indicators to evaluate desertification. In some researches, albedo-NDVI and albedo-MSAVI models have been proposed for desertification analysis. Other researchers also used albedo-TGSI based model. However, among these indices, the albedo-NDVI index has been widely considered. This index shows a strong negative correlation and reflects the state of desertification. This research is based on TGSI, NDVI, albedo and MSAVI spectral indices. The combination of these indicators is tested through Pearson correlation. The aim of this research is to map the different classes of desertification in the Qazvin Plain in 2023 using Sentinel 2 images, to integrate the indicators obtained from remote sensing data to identify areas affected by desertification and to establish correlation between MSAVI indicators., albedo, NDVI and TGSI and create a linear regression to determine the DDI index.



2. Methodology

In this research, using the Sentinel 2 satellite image in 2023, the degree of desertification is evaluated. This image covers the entire region. The spatial resolution of the images provides the opportunity to observe the ground accurately. Sentinel 2's multispectral images consist of 13 spectral bands with a resolution of 10 meters (4 bands), 20 meters (6 bands) and 60 meters (3 bands). These 13 bands cover a wide range of wavelengths from 440 to 2200 nm. In this research, red, infrared, blue, green, SWIR1 and SWIR2 bands were used. The satellite image of July 11, 2023 for a day without cloud cover was used to perform the analysis. In addition, July images allow the separation of agricultural land from natural vegetation. The method adopted in this research is based on the extraction of spectral indices NDVI, TGSI, albedo and MSAVI, which provides the possibility to evaluate and monitor desertification. This research included acquiring satellite images, calculating spectral indices, analyzing the correlation between spectral indices, calculating the degree of desertification index (DDI) and regression analysis. The estimation of four indices NDVI, albedo, MSAVI and TGSI is done using Sentinel 2 image. The extraction of different degrees of desertification is based on the analysis of several combinations of spectral indices in order to select a combination that shows the best classes of desertification. To extract the degree of desertification intensity and determine the best correlation, four combinations of albedo-NDVI, albedo-TGSI, albedo-MSAVI and TGSI-MSAVI were considered. A linear regression relationship was established between these indices. To draw the intensity of desertification, regression relationships between the above indicators were used.



3. Results and Discussion

In this research, desertification was analyzed by building two albedo-NDVI and albedo-MSAVI models. The results obtained from these models achieve better results than the traditional models based on vegetation. Among these two models, the albedo-MSAVI model can be the most suitable for the studied area, which is mainly composed of areas with low vegetation cover. The albedo-NDVI model obtains better results for areas with dense vegetation. However, these areas include only a small part of the studied territory. Considering the effectiveness of albedo-MSAVI and albedo-NDVI models in the study area, these models can be used as a reference for decision makers in natural resource management. This model can be used for other areas with similar characteristics. Based on the results, special operational plans should be implemented to overcome the desertification situation and reduce land degradation. At first, TGSI, albedo, NDVI and MSAVI spectral indices were extracted. Then, for linear regression analysis, different combinations of indices were considered and Pearson correlation relationship was established between NDVI and albedo, MSAVI and albedo, TGSI and albedo, TGSI and MSAVI indices. Then, by using the regression relationships between NDVI-albedo and MSAVI-albedo, DDI desertification degree index was estimated. Finally, the map of desertification was drawn for the entire studied area in five classes: very severe, severe, moderate, low and very low.

4. Conclusion

The obtained results showed that there is a negative correlation between albedo and MSAVI and NDVI indices, and a positive correlation between this index and TGSI. MSAVI-albedo combination has the best correlation coefficients with a value of -0.417 and the lowest correlation is estimated at 0.33 between albedo and TGSI. According to the DDI index, the desertification situation in the region has been in the alarming range. In fact, according to Nest Albedo with MSAVI index, 55.6% of the region is classified as very severe desertification. While only 1.9% was very low in the area of desertification. According to albedo-NDVI ratio, 68% of the region is in a state of extreme desertification. The values obtained for the two models are very close to each other and similar estimates have been obtained in estimating extreme and very low values.