3.1. Exploring the Relationship between Electrical Conductivity (EC) and Spectral Indices: Initial Analysis
In this phase of the study, we analyze the spectral properties of soil depending on the salinity, specifically the visible and near-infrared regions provided by the Landsat data. In the first step of the analysis, the corresponding values of the ground points were extracted from the calculated indices. These values were then stacked for every index, encompassing all the values from all the images. The correlation equation was then applied to the EC corresponding values and the calculated indices’ values.
Table 3 shows the correlation between some of the ground-measured electrical conductivity (EC) values and values extracted from the eleven spectral indices for each point across the time series of Landsat images, in descending order from the highest to the lowest. The highest correlation coefficient was found for the SI index, at 0.38. Linear R
2 and polynomial R
2 values of 0.14 and 0.15, respectively, were calculated between the EC values and each vegetation index value across the time series of images. The SI and SI1 indices had the highest correlation, respectively, and the NDVI and S2 indices showed the lowest correlation with EC.
Figure 5 displays the first four highest correlation and regression values relating to EC and salinity indices for the first approach. However, after extracting the tested linear regression and polynomial regression for each index, respectively, the results showed weak correlation values in this approach due to the high difference in the salinity values.
Upon conducting linear regression and polynomial regression analyses for each index, it was observed that this approach yielded weak correlation values. This can be attributed to the significant variation in salinity values, which ranged from 0.56 to 100, excluding EC values exceeding 100, as they had a noticeable impact on the results. To mitigate the impact of salinity values exceeding 100 on the regression results, we excluded the salinity values that exceeded 100 from the dataset before conducting the regression analysis. By removing these extreme values, the analysis would focus on the range of salinity values that are more representative and less likely to have a disproportionate influence on the results. This analysis aimed to determine which indices are most strongly correlated with ground EC values, indicating their potential effectiveness for estimating soil salinity without any complex analysis or ranging. However, the results showed low correlations and regressions, likely due to the significant variability in salinity ranges. Without applying any classification between the values, the usage of the indices will not have a high value or significant use.
Figure 6 shows the resulting relations of the first approach between the correlation, regression values, and the eleven indices’ values extracted from the calculations.
3.2. Quantifying the Relationship between Salinity Levels and Spectral Indices
To explore a better relationship, the second analysis scenario categorized EC–vegetation index correlations based on different salinity levels. The ground EC measurements were classified into seven different salinity categories: none, slightly saline, moderately saline, high, and three different degrees of severe.
Table 4 illustrates the correlation values calculated for the second approach, wherein the correlation coefficient was calculated between EC values and each index for each salinity category.
This analysis is designed to determine which indices show the strongest correlations with EC at different salinity levels, indicating their potential suitability for distinguishing between salinity levels. This finding revealed a modest relationship between vegetation indices and EC values, but a more significant relationship between soil salinity indices and bare soil. This is consistent with the findings of [
35,
36].
Figure 7 shows the highest correlation was S2 = 0.96, with R
2 linear = 0.89 and R
2 polynomial = 0.91 for moderate saline samples. As we added more details about the classification and separated the samples, the results improved gradually.
The findings presented in
Figure 8 demonstrate that the second approach yielded improved correlation and regression values as the range of electrical conductivity (EC) was expanded. Specifically, the performance of the eleven indices examined in the study was enhanced across all salinity ranges. This approach investigated the relationship between vegetation indices and electrical conductivity (EC) values for soil samples with varying degrees of salinity. The results indicated that non-saline samples had a fair correlation with the indices, with S6 showing the highest correlation value. In slightly saline samples, the correlation improved, with S2 exhibiting the highest correlation value. For moderately saline samples, all the salinity indices showed a good relationship with EC, but S2 had the highest correlation value. However, for highly saline samples, the correlation was weaker than that observed for moderately saline samples, and S2 continued to exhibit the highest correlation value. These findings suggest that the correlation between EC and vegetation indices weakens as salinity levels increase.
3.3. Temporal Analysis of Salinity Indices: Exploring Relationships over Time
In the third analysis scenario, we conducted a comprehensive investigation of the relationship between electrical conductivity (EC) and various indices. This analysis was performed separately for each index, taking into account individual images and different salinity levels. To obtain the indices, we calculated them for each image based on the corresponding salinity degree. This allowed us to establish specific relationships for each scene, treating them as separate areas of interest. Subsequently, we employed a skills model to evaluate the behavior of each index in relation to the varying salinity degrees. This evaluation enabled us to assess how the indices responded to different levels of salinity and identify any patterns or trends. By accumulating the matched indices, we were able to analyze the collective behavior and gain insights into their overall performance and suitability for salinity assessment. Correlation coefficients were determined separately for each index and for different salinity levels, as presented in
Table 5.
This allowed us to establish specific relationships for each scene based on temporal analysis. Subsequently, we employed a skills model to evaluate the behavior of each index with varying salinity degrees.
Figure 9 displays the correlation and regression models for each index, with the highest correlation found for moderate saline samples S1 (0.99), R
2 linear (0.88), and R
2 polynomial (0.93) for eight out of the nine images. The graph demonstrates how the correlation values change depending on the picture/time and index. The objective of this analysis was to examine the relationship between EC salinity levels and the fluctuations observed in each index across multiple satellite image scenes, representing different seasons and temporal changes. The aim was to differentiate between the changes occurring in vegetated areas and bare lands over time, as the studies of [
2,
37,
38] revealed. Considering that the study area exhibits a mixed vegetation pattern that poses challenges in distinguishing between the two, the approach provided valuable insights into the potential of each index (highest and lowest) to detect changes in salinity or distinguish between different salinity levels. This approach is particularly useful for conducting time series analysis aimed at identifying temporal changes in salinity levels within a given scene. Hence, the scene is considered the primary unit of analysis, and the inclusion of time-based analysis can provide valuable information regarding the temporal changes in salinity levels.
Conducting a temporal analysis involves comparing images captured at different times to identify changes in land cover and spectral values. By analyzing the temporal patterns, it is possible to distinguish between changes caused by vegetation growth and those caused by other factors, such as seasonality or land cover variations. The highly dynamic nature of surface salinity processes necessitates a dynamic and temporally sensitive approach to detecting soil salinity. The use of multi-temporal remote sensing imagery is a particularly suitable method for monitoring changes in salinity levels, particularly in irrigated areas. This approach enables a comprehensive and longitudinal assessment of the evolving salinity conditions of the soil, which is imperative for understanding the complex interactions between irrigation practices and soil salinity. Given the significant impact of water management practices on soil salinity levels, the use of multi-temporal remote sensing imagery provides a valuable tool for researchers and practitioners seeking to monitor and manage soil salinity in irrigated areas [
17].
The inclusion of the image-/time-based classification in
Figure 10 highlights the significance of time-based analysis in understanding the relationships between EC and salinity indices in the third methodology. In summary, this analysis involved a meticulous examination of the correlations between EC and indices. We considered the indices separately for each image and salinity level, created scene-specific relationships, and evaluated their behavior using a skills model. This approach allowed us to explore the relationship between salinity and indices, providing valuable insights into their applicability for salinity assessment.
3.4. Assessing Spatial Correlation and Salinity Indices: Investigating the Influence of Spatial Distribution and Land Cover Change on Salinity Indices
In the fourth scenario, we conducted a spatial correlation analysis on the electrical conductivity (EC) values by grouping sample points for each pond and calculating the corresponding index values. Sample locations, as presented in
Table 6, were used to perform correlation and regression analyses. The results, illustrated in
Figure 11, revealed that the salinity index (SI) had the strongest correlation with EC values among the four ponds (Bahi, Siwa, Aghormi, and Zaitoun), with correlation coefficients of 0.23, 0.23, 0.18, and 0.61, respectively. This finding suggests that the SI index is useful for evaluating salinity levels in aquatic environments. This approach is suitable for analyzing various morphological locations and elevations, especially those with distinct site characteristics.
Furthermore, our analysis of pond elevations demonstrated a positive correlation between elevation and salinity levels. The highest concentrations of saline soil samples were observed in the Bahi and Siwa pond areas with shallower water depths. Thus, the SI index can be a good tool for monitoring and managing the salinity levels in areas with varying morphological characteristics and elevations.
The findings of our fourth analysis approach involve establishing a relationship between salinity indices, spatial distribution, and the digital elevation model (DEM), which is visually presented in
Figure 12. The analysis revealed that the areas surrounding the shallower ponds exhibited the highest levels of soil salinity.
By considering a wide range of indices and methods, the comparison becomes applicable across diverse environmental conditions and land cover types. This enables the assessment of soil salinity in various agricultural settings, regardless of specific electrical conductivity ranges or environmental contexts. Various anthropogenic activities can result in secondary salinization. Over-irrigation of crops is one of these activities, which can cause a rise in the water table, bringing salt to the surface and causing salinization [
39]. Deforestation can also contribute to salinization by interfering with the water cycle and altering soil moisture and salinity. Land clearing can enhance soil erosion and organic matter loss, increasing the likelihood of salinization. Water can build up in poor drainage systems, causing soil salinization. Mining activities, through releasing salt into the land and streams, can also contribute to salinization. In conclusion, anthropogenic activities can lead to secondary salinization via a variety of pathways. Due to the uncontrolled flow of irrigation water from wells that the Siwa Oasis’s residents dug to use the water for continuous irrigation, the environment is endangered due to the rise in groundwater near the surface.
The general administration of Matrouh Drainage Projects constructed main combined drains to collect wastewater from primary and secondary drains in the Siwa Oasis. Instead of allowing the wastewater to flow directly to the ponds, a mechanical lifting method using pumps was implemented to transport the water. This change had a positive impact on the drainage system and led to a noticeable increase in the agricultural area of the Oasis after previously flooded areas became suitable for cultivation. The sustainability of the Oasis residents also increased as a result.
Environmental variables such as vegetation cover and soil index are crucial for monitoring soil salinization, and their accuracy must be carefully evaluated [
40]. Subsequently, an analysis was conducted to investigate the relationship between the distribution of soil salinity, pond elevation, land cover, and land use. Estimation of land cover change based on uncorrected images can present unrealistic change rates, which are two to three times higher than those obtained with corrected images. This is due to variations in sensor sensitivity and atmospheric conditions that can influence the accuracy of change detection. To improve the accuracy of change detection, the use of corrected images is critical for obtaining reliable estimates of land cover change over time [
41].
Figure 13 demonstrates the sample density for each class, along with the various ranges of salinity that were considered during the analysis using the Inverse Distance Weighting interpolation IDW algorithm.
Figure 14 shows that the majority of the salinity samples were concentrated in cultivated areas (see also
Table 7).
Management interventions implemented thereafter led to a significant reduction in excess water by about 94.7% from 1998 to 2012, resulting in a 24% decrease in the total area of lakes. The closure of hand-dug wells resulted in an 11% decrease in groundwater withdrawal from 1998 to 2006. Groundwater withdrawal decreased by 33% from 2008 to 2012, leading to a 24% reduction in lake area over 10 years (2000–2010) [
42]. The change in land cover detection from 2003 to 2020, as shown in
Figure 15, occurred after the government replaced old dug wells with newly designed wells and implemented proper management.
This part of the study aimed to investigate the impact of continuous monitoring of soil salinity and land cover changes in Siwa’s area over 17 years. By considering the cultivated soil’s vegetation performance as an indirect indicator of soil salinity, we can infer potential soil salinity levels [
15]. The underlying assumption is that a substantial amount of vegetation signifies successful crop growth and lower levels of detrimental salts in the soil. Using remote sensing techniques, changes were detected regarding land cover, and it was found that the total vegetated area increased from 59.8 sq. km in 2003 to 89 sq. km in 2020, representing a 48% increase, as shown in
Figure 16. Moreover, it may not be suitable for images with complex land cover patterns or mixed pixels. By applying a supervised classification method and band combinations, we found that the cultivated areas had expanded by 19.70 sq. km, as barren lands were reclaimed and converted into cultivable ones. Moreover, the wetlands had shrunk by 8.27 sq. km, as they were transformed into cultivated areas. The results suggest that effective salinity monitoring can contribute to the expansion of cultivated areas and the improvement in land use practices. Collaborating with the irrigation and agriculture authorities in Siwa can help address irrigation shortages and meet agriculture requirements.
Based on the analysis, it can be concluded that the effective management of water resources and continuous monitoring of salinity levels can increase the cultivated area, improve water usage, and enhance water distribution. Ensuring that the timing and predictability of the water supply are adequate is crucial for effective water management, often more so than the mere adequacy of supply [
43]. From a systemic or social perspective, promoting equity in water distribution is essential to prevent certain users from receiving an excess of water at the expense of others. Additionally, optimizing water usage by providing an adequate but not excessive supply and minimizing losses is equally important. By working together, stakeholders can develop effective solutions that balance the needs of different users and promote sustainable water management practices.