Salinity Determination Of Soil Through Machine Learning And Remote Sensing Techniques

Salinity Determination Of Soil Through Machine Learning And Remote Sensing Techniques | Authorea try { document.documentElement.classList.add('js'); } catch (e) { } var _gaq = _gaq || []; _gaq.push(['_setAccount', 'G-8VDV14Y67G']); _gaq.push(['_trackPageview']); (function() { var ga = document.createElement('script'); ga.type = 'text/javascript'; ga.async = true; ga.src = ('https:' == document.location.protocol ? 'https://ssl' : 'http://www') + '.google-analytics.com/ga.js'; var s = document.getElementsByTagName('script')[0]; s.parentNode.insertBefore(ga, s); })(); Skip to main content Preprints Collections Wiley Open Research IET Open Research Ecological Society of Japan All Collections About About Authorea FAQs Contact Us Quick Search anywhere Search for preprint articles, keywords, etc. Search Search ADVANCED SEARCH SCROLL This is a preprint and has not been peer reviewed. Data may be preliminary. 4 February 2025 V1 Latest version Share on Salinity Determination Of Soil Through Machine Learning And Remote Sensing Techniques Authors : Fatma KAPLAN 0000-0002-4873-3997 [email protected] and Ali Volkan BİLGİLİ Authors Info & Affiliations https://doi.org/10.22541/au.173865076.63148165/v1 886 views 246 downloads Contents Abstract Information & Authors Metrics & Citations View Options References Figures Tables Media Share Abstract Global soil salinity is a problem that jeopardizes ecosystem health and agricultural productivity. Applying traditional soil salinity analysis techniques over large areas can be challenging, time-consuming, and costly because they rely on laboratory-based measurements. Thus, machine learning techniques and remote sensing are being utilized more and more to determine soil salinity quickly and accurately. By using satellites and aerial sensors to record the spectral characteristics of the soil surface, remote sensing technologies can identify indirect markers of soil salt buildup. Specifically, the salinity of soil is often mapped using spectral data from the visible, near-infrared (VNIR), and thermal bands. In order to forecast soil salinity, machine learning algorithms analyze these spectral data to model intricate and non-linear relationships. Methods such as, Support Vector Machines (SVM), Artificial Neural Networks (ANN), Random Forest, and Deep Learning provide high accuracy rates in predicting soil salinity. These methods have a lot of promise for tracking changes in soil salinity over time and space, improving farming methods, and creating potent anti-salinity plans. Additionally, by assisting with decisions about soil management, particularly in large-scale farming areas, these techniques support the growth of sustainable agricultural practices. Consequently, it is recognized that a promising approach to controlling and tracking soil salinity is the combination of machine learning and remote sensing. Salinity Determination Of Soil Through Machine Learning And Remote Sensing Techniques Fatma KAPLAN 1 e-mail: [email protected] 1 Harran University, Faculty of Agriculture, Soil Science and Plan Nutrition, Sanliurfa,Turkiye 1 Ali Volkan BİLGİLİ 2 e-mail: [email protected] 2 Harran University, Faculty of Agriculture, Soil Science and Plan Nutrition, Sanliurfa,Turkiye 2 Abstract Global soil salinity is a problem that jeopardizes ecosystem health and agricultural productivity. Applying traditional soil salinity analysis techniques over large areas can be challenging, time-consuming, and costly because they rely on laboratory-based measurements. Thus, machine learning techniques and remote sensing are being utilized more and more to determine soil salinity quickly and accurately. By using satellites and aerial sensors to record the spectral characteristics of the soil surface, remote sensing technologies can identify indirect markers of soil salt buildup. Specifically, the salinity of soil is often mapped using spectral data from the visible, near-infrared (VNIR), and thermal bands. In order to forecast soil salinity, machine learning algorithms analyze these spectral data to model intricate and non-linear relationships. Methods such as, Support Vector Machines (SVM), Artificial Neural Networks (ANN), Random Forest, and Deep Learning provide high accuracy rates in predicting soil salinity. These methods have a lot of promise for tracking changes in soil salinity over time and space, improving farming methods, and creating potent anti-salinity plans. Additionally, by assisting with decisions about soil management, particularly in large-scale farming areas, these techniques support the growth of sustainable agricultural practices. Consequently, it is recognized that a promising approach to controlling and tracking soil salinity is the combination of machine learning and remote sensing. Keywords: soil salinity, remote sensing, machine learning algorithms, sustainable agriculture Introduction Degradation of agricultural lands is caused in part by increasing salinity, which is a result of both human activity and climate change. The reduction of pasture and arable land areas may result from this circumstance. While agricultural areas are growing, climate factors accelerate salinization; the problem is made worse by increased use of land and water resources (Savin, 2023). The rise in salt in the top soil layers is caused in part by irrigation and drainage system failures. Crop yields are among the adverse effects of soil salinity on agricultural production. The removal of secondary salinization from lands and the management of water use are critical to food security and agricultural productivity. Soil salinity plays a major role in desertification, land degradation, and a decline in agricultural productivity. It is most prevalent in semi-arid and arid regions (Wen et al., 2020; AbdelRahman et al., 2019; Aboelsoud et al., 2022), and it is predicted that roughly half of the world’s agricultural land may become saline by 2050 (Abdelaziz et al., 2019). Thus, the salinity of soil has become a crucial concern for global ecological conservation (Jia et al., 2023; Erdoğan et al., 2021), agricultural management, and development (AbdelRahman et al., 2022). Conventional methods of monitoring and predicting soil salinity frequently involve extensive fieldwork and sample collection. In this regard, it is crucial for food production (Mukhopadhyay et al., 2021), ecological security (Sun et al., 2022), and sustainable agricultural development (Zhang et al., 2020) to quickly and accurately observe soil salinity on a big scale and to systematically understand its temporal-spatial variations. Remote sensing methods are commonly used to predict soil salinity in arid and semi-arid regions due to their efficacy and ability to cover large areas (He et al., 2023; Aboelsoud and AbdelRahman, 2017). These days, satellite image texture features are useful for agricultural monitoring and classification, such as identifying wetlands (Chatziantoniou et al., 2017), classifying land cover and tree species (Ferreira et al., 2019; Zhu et al., 2019), estimating soil organic carbon (Nguyen et al., 2022), and assessing plant biomass (Dube and Mutanga, 2015). Furthermore, it is simpler to recognize patterns of salt distribution in space because texture features capture the spatial distribution of the soil surface. Finally, texture information can show the roughness and texture of the soil surface, which are characteristics associated with salt accumulation and soil salinity. Traditional methods of tracking soil salinization, like as fixed-point surveys, require a lot of work and time. Furthermore, large-scale salinization data cannot be obtained quickly due to the long time required for updates. Therefore, a common technique for tracking soil salinization data is remote sensing technology, which is quick, efficient, and able to gather a lot of data (Li et al., 2021; Peng et al., 2019; Wang et al., 2019). Previous studies have looked at the dynamics and distribution of soil salinity to distinguish between saline and non-salinated soils qualitatively. There has been a recent trend, however, to map soil salinity using advanced modeling, remote sensing, and geographic information systems (Maxwell et al., 2018). From an agricultural and environmental perspective, these technologies are crucial due to their wide coverage capabilities and effective use of remote sensing data to observe and identify soil salinity. Using medium to high resolution satellite data, modern remote sensing technology has greatly improved the mapping and monitoring of different soil properties, including salinity (Zheng et al., 2009). Sentinel-2 has been especially useful for mapping soil salinization (Tripathi and Tiwari, 2021; Yahiaoui et al., 2021; Maki et al., 2022; Ismaili et al., 2023). A recent study overcame climatic challenges by accurately estimating soil salinity using Sentinel-2 data with Google Earth Engine using machine learning techniques (Kaplan et al., 2023). Weather-based monitoring using optical sensors, such as Sentinel-2, may be limited, though. The ability to monitor in any weather condition and record subsurface features is provided by Synthetic Aperture Radar (SAR) remote sensing (Baghdadi et al., 2009). The correlations between electrical conductivity (EC) and remote sensing data has been modelled with promising results using machine learning algorithms, which include methods like support vector machines (Guan et al., 2013), random forest regression (Fathizad et al., 2020), simple linear regression (Hihi et al., 2019), and other regression models (Allbed et al., 2014). Using Sentinel-1 imagery to estimate salinity in arid and semiarid regions, these statistical models have demonstrated encouraging outcomes (Tripathi and Tiwari, 2021; Hao et al., 2019; Li et al., 2023; Merembayev et al., 2022; Taghadosi et al., 2019; Zhang et al., 2020; Zhang et al., 2022). These advancements open the door to better methods for managing and estimating soil salinity. The ability of machine learning algorithms to identify both linear and nonlinear relationships between input variables and soil salinity improves the accuracy of soil salinity assessment. This has led to the use of various machine learning-based regression algorithms in the study of soil salinization in recent years (Fathizad et al., 2020; Ge et al., 2022; Li et al., 2022; Mahajan et al., 2021; Periasamy et al., 2022; Wang et al., 2022). Because of the large geographical and time-based variability in soil salinity, it is strong to find an algorithm that can generate optimal estimations for all data sets. Thus, it is important to choose an appropriate algorithm based on various research areas (Shi et al., 2022). In Kuka Oasis, Xinjiang, China, for example, Wang et al. estimated soil salinity using 13 machine learning techniques (including RF, SVR, and GPR). According to Wang et al. (2020), they found that RF outperformed other machine learning algorithms in arid regions. The effectiveness of the GBM, XGboost, and RF methods for estimating soil salinity was assessed by Zarei et al. in Eshtehad Salt Lake, Alborz Province, Iran. They found that both GBM and RF were outperformed by the XGboost method (Zarei et al., 2021). Prior research has demonstrated that machine learning algorithms’ performance will change depending on the field, remote sensing, and research areas (Zarei et al., 2021). In order to create high-accuracy salinity maps, adaptive machine learning techniques and suitable remote sensing imagery must be combined. Secondary salinasation Agricultural sustainability depends on evaluating the risk of secondary soil salinization brought on by human land management. A challenge to agricultural sustainability is secondary soil salinization (Hillel, 2000). When irrigation or other farming practices cause the salinity of the surface soil to rise from a saline level to another saline level, this is known as secondary soil salinization (Peck and Hatton, 2003). Fundamental soil salinization, on the other hand, happens naturally when salt that has been stored in the soil or groundwater is carried to the surface of the land during development (Spies and Woodgate, 2005). Irrigation and Dryland salinity are two instances of secondary soil salinity (Peck and Hatton, 2003; Spies and Woodgate, 2005). The two main causes of dryland salinity are either above-normal rainfall, as in the Argentine province of Buenos Aires (Panell and Ewing, 2006), or rising water levels brought on by the substitution of shallow-rooted crops for native vegetation, as in the Canadian prairies (Wiebe et al., 2003) and southern Australia (Webb, 2002). Sometimes, as in the northern part of the US Great Plains, these factors may coexist (Lobell et al., 2011). Irrigation salinity is primarily affected by rising groundwater tables caused by excessive irrigation and inadequate drainage to remove leaching and salts (e.g., Corwin et al., 2007; Herrero and Pérez-Coveta, 2005; Houk et al., 2006). The main distinction between the two types of salinity is the management options available to address salinity risk, despite the similarities in the hydrological processes that result in salt mobilization in each (Spies and Woodgate, 2005). Restoring deep-rooted plants and stable land cover reduces the salinity of drylands, while improved irrigation management systems specifically address irrigation salinity (Spies and Woodgate, 2005). Often, soil salinity has a negative effect on environmental conditions and land use. Shore salinity Arid and semiarid regions are undergoing significant change due to climate change, which has an impact on communities and the environment (Brochier and Ramieri, 2001). The equilibrium of ecological systems and the well-being of society are impacted by the complicated issues posed by shifting weather patterns, warming temperatures, and shifting ecosystems (Shivanna, 2022). Sea levels are rising in coastal areas, and weather patterns are altering, affecting the availability of water and agricultural land (Griggs and Requero, 2021). Using remote sensing techniques extensively, Etienne is the principal investigator of soil salinization (Etienne, 2012; Etienne, 2014; Etienne, 2017). The most recent research on this topic, however, was released in 2021 and used spectral indices like the salinity index (SI), automatic water extraction index (AWEI), and normalized difference vegetation index (NDVI) to map soil salinity and identify areas with halophytic vegetation (Kurbatova et al., 2021). Traditional methods of monitoring and estimating soil salinity usually entail work field work and sampling. Previous studies primarily examined the allocation and dynamics of soil salinity in order to qualitatively differentiate between saline and non-saline soils. However, there has recently been a revolution in the use of geographic information systems (GIS), advanced modeling, and remote sensing to map soil salinity (Maxwell et al., 2018). These technologies efficiently observe and identify soil salinity using remote sensing data, providing the broad spatial coverage required for both agricultural and environmental outlooks (Hihi et al., 2023). Figure 1. Formation mechanism of soil salinity Figure 1 illustrates the mechanism of soil salinity, which can be summed up as follows: Salts are dissolved by irrigation and rainfall that seeps into the ground. While some of this water seeps deeply into the soil (deep infiltration), some is absorbed by plants (evapotranspiration). As water evaporates due to high evaporation rates, salts build up on the soil’s surface. Soil salinity results from excessive irrigation or poor drainage, which causes salt buildup. This process creates issues in agricultural areas and has a detrimental effect on soil fertility. Remote sensing techniques Soil salinity must be measured using remote sensing techniques for soil quality assessment and tracking, especially in agricultural areas. Salt concentration in soil can be mapped using satellite imagery and data from aerial vehicles thanks to remote sensing technologies. These techniques examine the reflectance characteristics of soil in various electromagnetic spectrum bands, particularly in the visible, infrared, and microwave regions. In contrast to non-saline soils, saline soils exhibit distinct spectral signatures that enable the assessment of salinity levels. For the sensitive assessment of soil salinity, multispectral and hyperspectral imaging methods are particularly popular. When compared to conventional methods, these technologies are favored because they are quicker, more cost-effective, and more widely applicable. While Metternicht and Zinck (2003) examined the potential and constraints of utilizing remote sensing to determine soil salinity, Allbed and Kumar (2013) emphasized the significance of remote sensing and GIS in this process. Research in this field was greatly advanced by Farifteh et al. (2008), who closely examined the spectral characteristics of saline soils. These investigations show the possibility of remote sensing technologies in this area and emphasize the significance of spectral analysis and GIS integration in monitoring and mapping soil salinity (Metternicht and Zinck, 2003; Farifteh et al., 2008; Allbed and Kumar 2013). Tables 1 and 2 include an understanding of the methodology used for remote sensing soil salinity measurements. Table 1 offers an summary of the principles of remote sensing technologies, data collection techniques, and data uses. Table 2 lists various salinity indices used in previous studies to measure soil salinity. The comprehensive overview provided by these tables makes it easier to understand how remote sensing methods are applied in soil salinity analysis. Table 1. Working principle of remote sensing technology, its purpose and studies using these technologies 1 Spektroradiometer A light source is sent to the ground and data is obtained from the reflected light. Obtaining spectral reflectance and its values Soil characterization (Kaplan and Bilgili, 2024). 2 EM-38 Electromagnetic waves are sent into the soil and measurements are made directly. Obtained directly from the soil surface in the field Determination of salinity in soil (Bilgili et al., 2015). 3 Sentinel Processing of satellite images Soil indices Soil characterization (He., et al. 2024). 4 Landsat Processing of satellite images Soil indices Soil characterization (Han, et al. 2023). 5 Google Earth Engine Processing of images Soil indices Soil characterization (Kaplan, et al. 2023). S # : Studies using remote sensing techniques Table 2. Salinity indices 1 Salinity Index1 SI1 (Blue*Red) 0.5 (Khan et al.,2005) 2 Salinity Index2 SI2 (Green*Red) 0.5 (Khan et al.,2005) 3 Salinity Index3 SI3 (Green 2 +Red 2 ) 0.5 (Nicolas et al.,2006) 4 Salinity Index4 SI4 (Blue*Red)/Green (Abbas and Khan, 2007) 5 Salinity Index5 SI5 (Green+Red)/2 (Nicolas et al.,2006) 6 Normalized diference salinity index NDSI (Red−NIR)/(red+NIR) Khan et al. (2001) 7 Salinity index SI1 Sqrt (green 2 + red 2 ) Douaoui et al. (2006) 8 Salinity index SI2 Sqrt (green × red) Douaoui et al. (2006) 9 Salinity index SI3 √G 2 +R 2 + NIR 2 Douaoui et al. (2006) 10 Salinity index SI4 Sqrt (blue × red) Khan et al. (2001) 11 Salinity index SI5 (Red×NIR)/green Abbas and Khan (2007) 12 Salinity index SI6 Blue/red Abbas and Khan (2007) 13 Normalized Difference Salinity Index NDSI (R-NIR)/(R+NIR)= (B4−B8a)/(B4+B8a) Wang et al., 2019 14 Salinity index I S1 B/R=B2/B4 Wang et al., 2019 15 Salinity index II S2 (B-R)/(B +R)= (B3−B4)/(B3+B4) Wang et al., 2019 16 Salinity index III S3 (G ×R)/B=(B3×B4)/B2 Wang et al., 2019 17 Salinity index V S5 (B ×R)/G=(B2×B4)/B3 Wang et al., 2019 18 Salinity index VI S6 (R ×NIR)/G=(B4×B8a)/B3 Wang et al., 2019 19 Salinity index SI (B +R) 0.5 =(B2+B4) 0.5 Wang et al., 2019 20 Salinity index 1 SI1 (G ×R) 0.5 =(B3×B4) 0.5 Wang et al., 2019 21 Salinity index 2 SI2 [(G) 2 +(R) 2 +(NIR) 2 ] 0.5 =[(B3)2+(B4)2+(B8a)2] 0.5 Wang et al., 2019 22 Salinity index 3 SI3 [(R) 2 +(G) 2 )] 0.5 =[(B4)2+(B3)2] 0.5 Wang et al., 2019 23 Normalized Difference Salinity Index red-edge 1 NDSI re1 (Red-edge 1-NIR)/(Red-edge 1 +NIR) =(B5−B8a)/(B5+B8a) Wang et al., 2019 24 Normalized Difference Salinity Index red-edge 2 NDSI re2 (Red-edge 2-NIR)/(Red-edge 2 +NIR) =(B6−B8a)/(B6+B8a) Wang et al., 2019 25 Normalized Difference Salinity Index red-edge 3 NDSI re3 (Red-edge 3-NIR)/(Red-edge 3 +NIR) =(B7−B8a)/(B7+B8a) Wang et al., 2019 26 Salinity index I red-edge 1 S1 re1 B/Red-edge 1=B2/B5 Wang et al., 2019 27 Salinity index I red-edge 2 S1 re2 B/Red-edge 2=B2/B6 Wang et al., 2019 28 Salinity index I red-edge 3 S1 re3 B/Red-edge 3=B2/B7 Wang et al., 2019 29 Salinity index II red-edge 1 S2 re1 (B-Red-edge 1)/(B +Red-edge 1) =(B3−B5)/(B3+B5) Wang et al., 2019 30 Salinity index II red-edge 2 S2 re2 (B-Red-edge 2)/(B +Red-edge 2) =(B3−B6)/(B3+B6) Wang et al., 2019 31 Salinity index II red-edge 3 S2 re3 (B-Red-edge 3)/(B +Red-edge 3) =(B3−B7)/(B3+B7) Wang et al., 2019 32 Salinity index III red-edge 1 S3 re1 (G ×Red-edge 1)/B=(B3×B5)/B2 Wang et al., 2019 33 Salinity index III red-edge 2 S3 re2 (G ×Red-edge 2)/B=(B3×B6)/B2 Wang et al., 2019 34 Salinity index III red-edge 3 S3 re3 (G ×Red-edge 3)/B=(B3×B7)/B2 Wang et al., 2019 35 Salinity index V red-edge 1 S5 re1 (B ×Red-edge 1)/G=(B2×B5)/B3 Wang et al., 2019 36 Salinity index V red-edge 2 S5 re2 (B ×Red-edge 2)/G=(B2×B6)/B3 Wang et al., 2019 37 Salinity index V red-edge 3 S5 re3 (B ×Red-edge 3)/G=(B2×B7)/B3 Wang et al., 2019 38 Salinity index VI red-edge 1 S6 re1 (Red-edge 1 ×NIR)/G=(B5×B8a)/B3 Wang et al., 2019 39 Salinity index VI red-edge 2 S6 re2 (Red-edge 2 ×NIR)/G=(B6×B8a)/B3 Wang et al., 2019 40 Salinity index VI red-edge 3 S6 re3 (Red-edge 3 ×NIR)/G=(B7×B8a)/B3 Wang et al., 2019 41 Salinity index red-edge 1 SI re1 (B +Red-edge 1) 0.5 =(B2+B5) 0.5 Wang et al., 2019 42 Salinity index red-edge 2 SI re2 (B +Red-edge 2) 0.5 =(B2+B6) 0.5 Wang et al., 2019 43 Salinity index red-edge 3 SI re3 (B +Red-edge 3) 0.5 =(B2+B7) 0.5 Wang et al., 2019 44 Salinity index 1 red-edge 1 SI1 re1 (G ×Red-edge 1) 0.5 =(B3×B5) 0.5 Wang et al., 2019 45 Salinity index 1 red-edge 2 SI1 re2 (G ×Red-edge 2) 0.5 =(B3×B6) 0.5 Wang et al., 2019 46 Salinity index 1 red-edge 3 SI1 re3 (G ×Red-edge 3) 0.5 =(B3×B7) 0.5 Wang et al., 2019 47 Salinity index 2 red-edge 1 SI2 re1 [(G) 2 +(Red-edge 1) 2 +(NIR) 2 ] 0.5 =(B3 2 +B5 2 +B8a 2 ) 0.5 Wang et al., 2019 48 Salinity index 2 red-edge 2 SI2 re2 [(G) 2 +(Red-edge 2) 2 +(NIR) 2 ] 0.5 =(B3 2 +B6 2 +B8a 2 ) 0.5 Wang et al., 2019 49 Salinity index 2 red-edge 3 SI2 re3 [(G) 2 +(Red-edge 3) 2 +(NIR) 2 ] 0.5 =(B3 2 +B7 2 +B8a 2 ) 0.5 Wang et al., 2019 50 Salinity index 3 red-edge 1 SI3 re1 [(Red-edge 1) 2 +(G) 2 ] 0.5 =[(B5) 2 +(B3) 2 ] 0.5 Wang et al., 2019 51 Salinity index 3 red-edge 2 SI3 re2 [(Red-edge 2) 2 +(G) 2 ] 0.5 =[(B6) 2 +(B3) 2 ] 0.5 Wang et al., 2019 52 Salinity index 3 red-edge 3 SI3 re3 [(Red-edge 3) 2 +(G) 2 ] 0.5 =[(B7) 2 +(B3) 2 ] 0.5 Wang et al., 2019 *The reflectance values for blue (B, Band 2, 490 nm); green (G, Band 3; 560 nm); red (R, Band 4; 665 nm); and near-infrared (NIR, Band 8; 842 nm) in sentinel -2, wave bands. (Explanation for samples 1-12 in Table 2). * B2 (B: blue), B3 (G: green), B4 (R: red), B5 (red-edge – RE1), B6 (RE2), B7 (RE3), B8a (narrow NIR) (Explanation for samples 13-52 in Table 2). Machine learning models Using machine learning techniques to measure soil salinity is a recent trend that has shown promise. Machine learning algorithms that incorporate information from multiple sources, such as soil samples, remote sensing data, and environmental factors, are used to predict the salinity of soil. The three most broadly used algorithms for simulating soil salinity are Artificial Neural Networks (ANN), Support Vector Machines (SVM), and Random Forest (RF). The success of machine learning techniques in mapping soil salinity was highlighted in groundbreaking research by Taghizadeh-Mehrjardi et al. (2021). Farifteh et al. examined the spectral properties of saline soils to provide insight into the analysis of data that feeds into machine learning models (Farifteh et al., 2008). These resources offer thorough details on using machine learning to determine soil salinity. The machine learning models and soil parameters used in these models to calculate soil salinity are summarized in Tables 3 and 4. The basic features of the machine learning models (including Support Vector Machines, Random Forest, and Artificial Neural Networks) are described in Table 3. Along with their purposes, Table 4 lists the soil parameters that were used in the models, including moisture content, electrical conductivity, and organic matter. These tables offer a thorough explanation of how machine learning techniques are used to forecast soil salinity. Table 3. Machine learning models and their basic features 1 Random forest (RF) Multiple decision trees are used in Breiman’s ensemble learning technique to produce incredibly accurate predictions. (He., et al. 2024). 2 Linear Regression A simple statistical technique for modeling a two-variable linear relationship (Kaplan, et al. 2023). 3 Support Vector Machine (SVM) An algorithm for machine learning that focuses on identifying the best hyperplane to divide data for classification and regression (He., et al. 2024). 4 Extreme Gradient Boosting Regression (XGR) A powerful regression method based on Gradient Boosting, using sequential trees to minimize error rates (Han, et al. 2023). 5 Support vector regression (SVR) SVM modification for regression issues that seeks to forecast data points within a given tolerance (Han, et al. 2023). 6 Random forest regression (RFR) An application of the Random Forest algorithm to regression problems, using the average of multiple decision trees for predictions (Han, et al. 2023). 7 Exponential Regression A type of regression used to model exponential relationships between independent and dependent changeables (Hihi, et al. 2023). 8 Cubist A machine learning model based on regression trees, making rule-based predictions (He., et al. 2024). 9 Backpropagation A learning algorithm used in artificial neural networks for error minimization, based on gradient descent (He., et al. 2024). 10 Polynomialand Linear Regression Regression methods used to model both linear and polynomial relationships (Hihi, et al. 2023). 11 Instance-bases learning with parameter k(IBk) A learning technique that uses similar instances to make predictions based on the k-Nearest Neighbors (k-NN) algorithm (Kaplan, et al. 2023). 12 Generalized Linear Model A generalization of linear regression, used to model data with different distributions (Das, et al. 2023). 13 Multivariate adaptive regression splines Model A flexible regression method using piecewise linear functions to model nonlinear relationships (Erkin, et al. 2019). 14 Deep Learning (multilayer feed-forward ANN) Model A deep learning method using multilayer artificial neural networks to learn complex data relationships (Das, et al. 2023). 15 Backpropagation Neural Network (BPNN) A backpropagation algorithm-trained model of a multilayer artificial neural network (Cui, et al. 2023). Table 4. For modeling of soil salinity using ML techqniques, various secondary variables have been used 1 EC Predict, Classification, Mapping, Aksoy et al., 2022 2 Topography Predict, Classification, Mapping, Aksoy et al., 2022 3 Total salt Predict, Classification, Mapping, Erkin, et al. 2019 4 Moisture Predict, Classification, Mapping, Erkin, et al. 2019 5 Vegetation Predict, Classification, Mapping, Aksoy et al., 2022 6 pH Predict, Classification, Mapping, Erkin, et al. 2019 7 Organic matter Predict, Classification, Mapping, Erkin, et al. 2019 8 Spatial location Predict, Classification, Mapping, Aksoy et al., 2022 9 Climate Predict, Classification, Mapping, Aksoy et al., 2022 10 Total and available nitrogen Predict, Classification, Mapping, Erkin, et al. 2019 11 Phosphorus Predict, Classification, Mapping, Erkin, et al. 2019 12 Potassium Predict, Classification, Mapping, Erkin, et al. 2019 13 Bulk density Predict, Classification, Mapping, Erkin, et al. 2019 14 Soil texture Predict, Classification, Mapping, Erkin, et al. 2019 15 Salinization type and level Predict, Classification, Mapping, Erkin, et al. 2019 16 SAR Predict Xiao, et al. 2023 17 Soil temperature Predict Xiao, et al. 2023 18 Gravimetric moisture content Predict Xiao, et al. 2023 19 ESP Predict Xiao, et al. 2023 20 RSC Predict Xiao, et al. 2023 21 PS Predict Xiao, et al. 2023 22 MAR Predict Xiao, et al. 2023 23 K + Predict Xiao, et al. 2023 24 Ca 2+ Predict Xiao, et al. 2023 25 Na + Predict Xiao, et al. 2023 26 Mg 2+ Predict Xiao, et al. 2023 27 Cl Predict Xiao, et al. 2023 28 HCO 3– Predict Xiao, et al. 2023 29 CO 3 2– Predict Xiao, et al. 2023 30 SO 4 2 Predict Xiao, et al. 2023 Performance Evaluation Metrics in Machine Learning A range of accuracy metrics are used to assess how well machine learning models perform. Mean Absolute Error (MAE) and Root Mean Square Error (RMSE) were used by Cui et al. (2023) to quantify the size of prediction errors. Root Relative Squared Error (RRSE) and Relative Absolute Error (RAE), which were employed in model comparisons by Kaplan et al. (2023), provide the ratio of errors to reference models. He et al. (2024) used the Coefficient of Determination (R2), which gauges how well a model can explain the variance in the data. Xiao et al. (2023) adopted the Residual Prediction Deviation (RPD) method, which uses standard deviation to assess prediction quality. Amirgaliyev et al. (2024) evaluated the model’s performance utilizing the Mean Absolute Percentage Error (MAPE), which reveals errors as a percentage. Each of these parameters has a distinct function and enables models to be assessed from various angles. Table 5. Metrics for evaluating machine learning models accuracy 1 Mean Absolute Error (MAE) (Cui, et al. 2023). 2 Root Mean Square Error (RMSE) (Cui, et al. 2023). 3 Relative Absolute Error (RAE) (Kaplan, et al. 2023). 4 Root Relative Squared Error (RRSE) (Kaplan, et al. 2023). 5 Coefficient of determination (R²) (He., et al. 2024). 6 Residual prediction deviation (RPD) (Xiao, et al. 2023). 7 Mean Absolute Percentage Error (MAPE) Amirgaliyev, et al. 2024 The accuracy metrics used to assess machine learning models’ performance are listed in Table 5. These metrics assess the models’ predictive accuracy and dependability. Commonly used metrics like Residual Prediction Deviation (RPD), Mean Absolute Error (MAE), Mean Absolute Percentage Error (MAPE), Coefficient of Determination (R2), Root Relative Squared Error (RRSE), Root Mean Square Error (RMSE), and Relative Absolute Error (RAE) are included in the table. Each metric assesses a distinct feature of the model; for instance, MAE and RMSE show the size of the prediction errors, R2 shows how well the model explains the data’s variance, and MAPE shows the percentage errors. These metrics offer a crucial framework for evaluating and enhancing model performance. Remote sensing and machine learning methods determine soil salinity by utilizing spectral data and soil properties. The parameters used in previous studies for this purpose are summarized in Table 6. Table 6. In previous studies, a variety of factors were used to determine soil salinity using remote sensing and machine learning methods. 1 Inner Mongolia, China EC, TUZ 240 bare soil, 360 soil samples in the vegetation period 0-10 Sentinel-1/2 Data Contrast, dissimilarity,homogeneity, angular secondmoment, correlation, entropy, mean, and variance OOB Importance Analysis, Pearson Correlation Analysis, Backpropagation, Support Vector Machine, RanForest, Cubist, bare soil (R2 = 0,688 ve RMSE = 0,207) in vegetated soil (R2 = 0,494 ve RMSE = 0,304) Estimation He., et al. 2024 2 Kerkennah Archipelago, Tunisia EC 59 0-30 Sentinel-1 C-SAR and Sentinel-2 MSI Satellite Dataset VV, VH (Sentinel-1 renkli kompozit: kırmızı (VV), yeşil (VH), mavi (VH/VV) Correlation Analysis Random Forest Regression, PolynomialandLinearRegression, Exponential Regression, R2 = 0.75 and RMSE = 0.47 Classification, estimation Hihi, et al. 2023 3 southern Kazakhstan EC 203 Sentinel-2 bands Band 2,3,4,8 ALOS-PALSAR’dan dört farklı polarizasyon (HH, HV, VH ve VV) Correlation Analysis LightGBM and Ridge Linear Regression RSME, R 2 , MAPE estimation Amirgaliyev, et al. 2024 4 Songnen Plain, China EC 291 0-20 Landsat 8(Spectral Band) Coastal/Aerosol, Blue, Green, Red, Near Infrared, Short Wavelength Infrared, Panchromatic, Cirrus, Long Wavelength Infrared, Correlation Analysis Cubist, Support vector regression (SVR), Random forest regression (RFR), Extreme Gradient Boosting Regression (XGR), RSME:0.31, R 2 :0.80, MAE:0.21 Classification, Estimation, Mapping Ge, et al. 2023 5 northwest China PS, SAR, ESP, RSC, MAR, soil temperature,gravimetric moisture content, pH, K+, Ca2, Na+, Mg2+, Cl, HCO3–, CO3 2– , SO4 2 467 0-10, 10-20, 20-40, 40-60, 60-80, 80-100 - - Pearson Correlation Analysis Random forest (RF), Support vector machine (SVM), Extreme gradient boosting (XGB), Training-testing; RSME: 0.598, R 2 : 0.989, MAE: 0.466, RPD: 9.475 Validation; RSME:0.990, R 2 :0.614, MAE:0.484, RPD:10.119 Estimation Xiao, et al. 2023 6 Abu Dhabi coast in the west of the United Arab Emirates EC, 393 0-10 Sentinel-2 Data and Google Earth Enginev B2 (blue), B3 (green), B4 (red), B8 (near-infrared – NIR), B5 (red-edge – RE1), B6 (RE2), B7 (RE3), B8a (narrow NIR), B11 (short wave infra-red – SWIR1), B12 (SWIR2, B1 (costal aerosol), B9 (water vapour) and B10 (cirrus) Correlation Analysis M5P, Random forest (RF), Linear Regression, instance-bases learning with parameter k (IBk) Mean Absolute Error (MAE) and Root Mean Square Error (RMSE), Relative Absolute Error (RAE) and Root Relative Squared Error (RRSR) Estimation Kaplan, et al. 2023 7 ANA, BRM, CHI, MED, SRN EC 360 0-5 AVIRIS-NG surface reflectance spectra over Continuum removal (CR); BDi, BDRi, NBDIi, IBD, BNAi, ANMB, Asymmetry, visible near-infrared region (VNIR), shortwave-infrared (SWIR) Pearson Correlation Analysis Random Forest, Gradient Boosting Machines, Deep Learning (multilayer feed-forward ANN), Generalized Linear Model, RSME:0.15-0.16, R 2 :0.89-0.55 Estimation Das, et al. 2023 8 Xinjiang Uyghur Autonomous Region (XUAR), China total salt, moisture, pH, organic matter; total and available nitrogen, phosphorus, potassium; bulk density, soil texture, and salinization type and level 139 0-20 Landsat 8 Operational Land Imager (OLI), Moderate Resolution Imaging Spectroradiometer (MODIS) One was composed offour scenes of Landsat 8 OLI images with seven bands and a spatial resolution of 30 m, and the other was one scene of MODIS data (MOD09GA with two bands and MOD09GQ with eight bands) with a spatial resolution of 250 m. Principal component analysis, Pearson correlation analysis Random forest regression, Multivariate adaptive regression splines RF modeling RSME: 1.83 R 2 : 0.86 RPD: 2.7 MARS modeling R2: 0.81 RMSE: 4.8 RPD: 1.96 Estimation, Classification, Mapping, Erkin, et al. 2019 9 West of Lake Urmia in Iran EC, iklim, bitki örtüsü, topografya, mekansal konum 70 0-20 Landsat-8 OLI, Sentinel-2A satellite Blue, Green, Red, NIR, SWIR1,and SWIR2 Principal Component Analysis Classification and regression trees (CART), random forest (RF), and support vector regression (SVR) R2 RMSE Estimation, Classification, Mapping, Aksoy et al., 2022 10 Huludao City in Liaoning Province, northeastern China DEM, EC 310 0-20 Sentinel-2 satellite images, Google Earth Engine Soil Spectral Indi ces (SSIs), Vegetation Spectral Indices (VSIs) - RF, SVM, and ANN RMSE=0.03, AIC=-919, BIS=-891, and R 2 =0.84 Estimation, Classification, Mapping, Wang and Sun, 2024 Figure 2. Procedures for utilizing remote sensing and machine learning techniques to evaluate soil salinity The processes of background subtraction, data collection, data cleaning, model training, and validation are all step of of assessing soil salinity using machine learning and remote sensing techniques. To map and assess soil salinity, models trained on spectral images and soil samples are used. Soil salinity can be quickly and accurately determined using these techniques. Figure 2 provides a detailed illustration of this mechanism. Review of Previous Studies on Determination of Soil Salinity by Machine Learning and Remote Sensing In monitoring soil salinity using remote sensing, the integration of Sentinel-1/2 textural and spectral data significantly improved salinity prediction, particularly when using the random forest model, for both vegetated and bare soils. Sentinel-1 textural features were more effective for bare soils, while Sentinel-2 performed better for vegetated soils, and this combination enhanced the accuracy of soil salinity estimation (He et al., 2024). The details of this study are given in Table 7. Table 7. He et al., detail of the study conducted in 2024 Study Area Shahaoqu Irrigation Area, Inner Mongolia, China Study Period April to August 2019 Data Sources Measured soil salinity data, Sentinel-1/2 textural and spectral data Key Findings Sentinel-1 Texture Features (More sensitive to bare soil salinity Top four features: HOM, ENT, COR, CON). Sentinel-2 Texture Features (More sensitive to vegetated soil salinity Top four features: VAR, CON, HOM, ENT). Model Performance Combining texture features and spectral data improved model performance Models Used Random Forest, Cubist, Support Vector Machines (SVM), Backpropagation (BP). Best Model: Random Forest Performance Metrics Vegetated Soil R=0.494, RMSE=0.308 Bare Soil R² = 0.688, RMSE=0.207 Reference He et al., 2024 This study addresses soil salinity and land degradation in southern Kazakhstan by leveraging GIS, remote sensing, and machine learning to map agricultural land salinity. While complex models like LightGBM showed limited accuracy gains over simpler models, the research highlights the potential of high-resolution radar data and suggests further improvements with deep learning and ground-based measurements for precise salinity mapping (Amirgaliyev et al., 2024). The details of this study are given in Table 8. Table 8. Amirgaliyev et al., detail of the study conducted in 2024 Topic Soil Salinization and Land Degradation in Southern Kazakhstan Significance Significant issues in southern Kazakhstan, Caused by climate change, human activity, and unequal water distribution Primary Problem Transboundary flow of rivers affecting water resources Solution Approach Development of new techniques for managing soil salinity, Relies on remote sensing monitoring and GIS Study Objective Effective for continuous surveillance and performs well in cloudy conditions Methodology Machine learning techniques for automatic mapping of soil salinity Importance of Mapping Precise mapping helps prevent or reduce salinity effects on agriculture Experimental Findings Complex models (e.g., LightGBM) do not significantly outperform simple models (e.g., Ridge regression) on small datasets Recommendations Mapping approach for agricultural lands’ salinity, Further enhancement with deep-learning techniques and ground-based data Reference Amirgaliyev et al., 2024 This study aims to determine soil salinity parameters by using random forest (RF), support vector machine (SVM), and extreme gradient boosting (XGB) models to predict soil ion composition. The XGB model outperformed the others in overall performance, particularly with electrical conductivity (EC), soil water content (SWC), and temperature (T) as inputs, accurately predicting all ion compositions with high precision (Xiao et al., 2023). The details of this study are given in Table 9. Table 9. Xiao et al., detail of the study conducted in 2023 Study Location Shihezi region, northwest China. Objective Determine soil ion composition for irrigation control and salinization prevention Challenges Costly and challenging long-term in-situ measurement. Machine Learning Models Extreme Gradient Boosting (XGB), Support Vector Machine (SVM), Random Forest (RF). Predicted Parameters Residual Sodium Carbonate (RSC), Magnesium Adsorption Ratio (MAR), Exchangeable Sodium Percentage (ESP), Total Dissolved Ionic Matter (TDI), Potential Salinity (PS), Sodium Adsorption Ratio (SAR). Input Variables Soil Water Content (SWC), Electrical Conductivity (EC), Potential Hydrogen (pH), Soil Temperature (T). Data Used 467 soil samples for training, assessment, and validation. Model Performance Enhancements SVM and RF: Enhanced by EC, T, and pH. XGB: Enhanced by EC, SWC, and T. Model Comparison - XGB outperformed SVM and RF in generalization. - XGB reduced RMSE by 24.2% to 54.8% compared to SVM. RF and SVM Performance - TDI: R² > 4.83. - PS: R² > 0.772, RMSE 0.957. - ESP: R² > 0.67, RMSE 1.74. XGB Performance Predicted all soil ion compositions with R² > 0.770 and RPD > 1.98. Reference Xiao et al., 2023. Soil salinity, exacerbated by climate change and human activities, necessitates accurate monitoring, which can be improved using remote sensing data and machine learning. This study utilizes Sentinel-2 imagery and open-source tools to predict soil salinity in hyper-arid regions, achieving a high correlation (0.84) between test and modeling results, while highlighting the need for advanced machine learning approaches for future, more precise salinity mapping (Kaplan et al., 2023). The details of this study are given in Table 10. Table 10. Kaplan et al., detail of the study conducted in 2023 Study Area United Arab Emirates Issue Climate change or environmental contamination from excessive industry and agriculture are the main causes of the sharp rise in soil salinity. Need for Solution To solve this issue, precise and current measurements of soil salinity are required. Method Remote sensing data can expedite and enhance soil salinity mapping Objective of the Study Utilizing Sentinel-2 satellite imagery, forecast soil salinity in hyper-arid regions while investigating different machine learning and modeling methods. Data Collected 393 soil samples were gathered for testing and modeling Tools Used Weka and Google Earth Engine are examples of open-source tools and data. Results A high correlation (0.84) was found between test and modeling results. Reference Kaplan et al., 2023. This study combines hyperspectral remote sensing (HRS) and machine learning to predict soil salinity in low to moderately salinized croplands, using data from AVIRIS-NG and a hybrid feature selection algorithm. An ensemble of random forest (RF) and deep learning (DL) models achieved high accuracy, and a newly proposed hyperspectral soil salinity index, based on Shannon entropy, outperformed existing indices and showed strong correlation with measured and predicted soil electrical conductivity (EC), offering a cost-effective alternative to field sampling (Das et al., 2023). The details of this study are given in Table 11. Table 11. Das et al., detail of the study conducted in 2023 Issue Land degradation affects millions of hectares worldwide, with soil salinization being a major cause. Solution Modern data mining techniques in conjunction with hyperspectral remote sensing (HRS) allow for the cost-effective and real-time monitoring of soils affected by salt. Study Objective To forecast soil salinity using AVIRIS-NG data in cropland soils that are low to moderately salinized. Study Locations Five areas in India. Data Source Next Generation Airborne Visible-Infrared Imaging Spectrometer (AVIRIS-NG). Feature Selection Four spectral absorption features that are sensitive to soil salinity were found using a hybrid feature selection algorithm. Machine Learning Models Gradient Boosting Machines (GBM), Random Forest (RF), Deep Learning (DL). Target Variable Soil Electrical Conductivity (EC). Best Performing Model Ensemble of RF and DL models: - Training: R² = 0.89, NRMSE = 0.15. - Testing: R² = 0.55, NRMSE = 0.16. New Hyperspectral Salinity Index Proposed index aggregates specific absorption features using Shannon entropy. Performance of New Index - Exceeded alternative indices based on remote sensing. - ML-predicted EC (r = 0.78), and measured EC (r = 0.68), showed a strong correlation at the 1% significance level. Application Successfully applied to six salinity classes in HRS images. Spectral-Spatial Resampling The best spectral and spatial resolution for upcoming hyperspectral missions was assessed. Practical Use The hyperspectral salinity index can replace costly and time-consuming field sampling to determine soil salinity. Reference Das et al., 2023. This study evaluated Landsat-8 and Sentinel-2A data with machine learning models (ANN, DT, RF, SVM) to predict soil salinity in southeast Iran, finding that Sentinel-2A and ANN performed best in winter, while RF was most accurate in summer, highlighting the importance of selecting appropriate salinity indices and satellite data for reliable predictions (Golestani et al., 2023). The details of this study are given in Table 12. Table 12. Golestani et al., detail of the study conducted in 2023 Issue Soil salinization leads to large deserts and disrupts ecological balance. Solution Satellite imaging and remote sensing methods can assess, map, and monitor saline lands globally. Study Objective To determine important salinity indices for forecasting and assess Sentinel-2A and Landsat-8 imagery for monitoring spatiotemporal changes in soil electrical conductivity (EC). Study Area Marginal lands of Sirjan Playa, southeast Iran. Data Collection In two seasons (summer and winter), 90 soil samples were systematically sampled from a depth of 0 to 20 cm. Satellite Data Sentinel-2A and Landsat-8 images obtained near the sampling time for both seasons. Machine Learning Algorithms Random Forests (RF), Decision Trees (DT), Artificial Neural Networks (ANN), Support Vector Machines (SVM). Best Performing Model - Winter: Sentinel-2A-derived ANN model (R² = 0.77, RMSE = 16.1, NRMSE = 27.1). - Summer: RF showed the lowest error for ECe prediction. Comparison of Satellite Data In practically all models, Sentinel-2A data produced lower RMSE and NRMSE values than Landsat-8 in both seasons. Most Reliable Salinity Index The most accurate indicator of soil salinity was the Vegetation Soil Salinity Index (VSSI). Key Findings - Satellite imagery and machine learning can effectively track and predict soil salinity. - Proper selection of salinity indices and satellite data is crucial for accurate predictions. Practical Implications helps arid and semi-arid regions manage their marginal lands sustainably. Reference Golestani et al., 2023. Soil salinization significantly impacts agricultural sustainability, and this study explores the use of UAV-based multispectral remote sensing combined with machine learning models (SVM, RF, BPNN) to accurately estimate soil salt content (SSC). The results show that combining soil moisture data with spectral indices improves model accuracy, with the best model achieving high precision (R² = 0.775) for 0–20 cm soil depth, demonstrating the potential of UAV and machine learning for effective farmland soil salinity monitoring (Cui et al., 2023). The details of this study are given in Table 13. Table 13. Cui et al., detail of the study conducted in 2023 Issue Soil salinization significantly impacts sustainable agricultural growth, making accurate estimation of soil salt content (SSC) essential. Solution A UAV-based multi-spectral remote sensing platform combined with machine learning models to estimate SSC. Study Objective To investigate the viability of retrieving SSC using machine learning models and create a soil salt distribution map. Data Collection Spectral data collected using a UAV multi-spectral remote sensing platform. Variable Screening Techniques Grey Relational Analysis (GRA) and Pearson Correlation Analysis (PCA) used to screen 20 spectral indices. Spectral Variable Groups - Vegetation Index - Salt Index - Combination Variable (includes soil moisture data). Machine Learning Models Support Vector Machine (SVM), Random Forest (RF), Backpropagation Neural Network (BPNN). Soil Depths Analyzed 0–20 cm and 20–40 cm. Best Performing Model Combination variable group with soil moisture data performed best. Model Performance Metrics - R²: 0.775 - MAE: 0.055 - RMSE: 0.038 Prediction Accuracy 0–20 cm depth predictions were more accurate and stable than 20–40 cm. Output Soil salt spatial map created for 0–20 cm depth, showing salinization distribution in the study area. Key Findings - GRA outperformed PCA in improving model accuracy. - Machine learning models combined with UAV data enhance SSC monitoring. Reference Cui et al., 2023. Soil salinity significantly impacts agriculture in southern Kazakhstan, and this study uses satellite data and machine learning algorithms (Decision Tree, Gaussian Process, Random Forest) to classify and predict soil salinity, with results evaluated using accuracy, recall, and f1 metrics. The analysis, supported by the SHAP framework, highlights the influence of dataset features on classification and provides insights for effective salinity management in the Turkestan region (Merembayev et al., 2022). The details of this study are given in Table 14. Table 14. Merembayev et al., detail of the study conducted in 2022 Issue Soil salinity is a major factor affecting agriculture in southern Kazakhstan. Objective To estimate and forecast soil salinity before planting seasons to plan for salt leaching. Method Machine learning algorithms and satellite data (including radar data) were used to classify soil salinity. Study Area Turkestan region, Kazakhstan (over 102 data points). Machine Learning Algorithms Decision Tree, Gaussian Process, Random Forest. Evaluation Metrics Accuracy, Recall, F1-score. Feature Influence Analysis Examined the influence of dataset features on classification using SHAP (Shapley Additive exPlanations) framework. Key Findings - Comparative analysis of machine learning algorithms was conducted. - Results aligned with the SHAP framework. Reference Merembayev et al., 2022. Soil salinity, a major environmental issue, can be effectively monitored using hyperspectral satellite imagery, as demonstrated in this study focusing on Zaghouan, Tunisia. The combination of AutoEncoder (AE) for feature representation and Support Vector Machines (SVM) for classification outperformed other methods, providing accurate predictions of soil salinity and supporting sustainable land management in arid regions (Klibi et al., 2020). The details of this study are given in Table 15. Table 15. Klibi et al., detail of the study conducted in 2020 Issue Soil salinity is a significant environmental hazard caused by natural and human-induced processes. Importance Monitoring soil salinity is crucial for sustainable land use and management. Solution Hyperspectral satellite imagery can greatly aid in soil salinity detection. Study Area Zaghouan, northeastern Tunisia (a semi-arid and arid region). Objective To predict soil salinity using the spectral signature and feature vector of Hyperion hyperspectral imagery. Feature Representation AutoEncoder (AE) neural network architecture was used for feature representation. Classification Algorithms K-Nearest Neighbors (KNN), Decision Trees (DT), Support Vector Machines (SVM). Best Performing Model AE-SVM combination outperformed other methods in predicting soil salinity. Key Findings - AE-SVM showed superior performance compared to KNN and DT. - Hyperspectral imagery is effective for soil salinity prediction. Reference Klibi et al., 2020. Toprak tuzluluğu, kurak ve yarı kurak bölgelerde arazi bozulmasının önemli bir tehdididir ve bu çalışma, Da’an Şehri’ndeki toprak tuzluluğunu tahmin etmek için Landsat 8 OLI görüntüleri ve makine öğrenimini kullanmıştır. Cubist modeli en yüksek tahmin doğruluğuna (RMSE=0.31 mS/cm) ulaşmış olup, bitki örtüsü tuzluluk indeksi (CRSI) elektriksel iletkenlik ile en güçlü korelasyonu göstermiş ve Cubist yönteminin bu tür bölgelerde toprak tuzluluğunu izlemede etkili olduğunu kanıtlamıştır (Ge et al., 2023). The details of this study are given in Table 16. Table 16. Ge et al., detail of the study conducted in 2023 Issue Soil salinization is a major threat to land degradation in arid and semi-arid regions. Study Area Da’an City. Objective To estimate soil salinity using Landsat 8 OLI imagery and machine learning. Data Used Landsat 8 OLI imagery (blue, green, red, and near-infrared bands). Spectral Indexes 19 indexes calculated, including: - 3 vegetation indexes - 15 salinity indexes - 1 brightness index. Machine Learning Algorithms - Cubist - Support Vector Regression (SVR) - Random Forest Regression (RFR) - Extreme Gradient Boosting Regression (XGBoost). Best Performing Model Cubist model: - RMSE = 0.31 mS/cm (highest prediction accuracy). Key Findings - Cubist model provided the best spatial distribution of soil salinity. - Canopy Salinity Index (CRSI) showed the strongest correlation with electrical conductivity (r = -0.44). - Cubist model based on 9 spectral indexes achieved RMSE = 0.34 mS/cm after variable screening. Recommendation Cubist method is recommended for soil salinity monitoring in arid and semi-arid regions. Reference Ge et al., 2023. Soil salinization poses a significant threat to agriculture in arid and semi-arid regions, and this study utilizes Sentinel-2A, Landsat-8 OLI, and ground-based EC measurements on the Google Earth Engine platform to map soil salinity using machine learning algorithms (SVR, RF, CART). The RF algorithm produced the most accurate spatial distribution of soil salinity classes, while CART showed slightly better prediction accuracy, demonstrating the effectiveness of remote sensing and machine learning for precise soil salinity monitoring (Aksoy et al., 2022). The details of this study are given in Table 17. Table 17. Aksoy et al., detail of the study conducted in 2022 Issue Soil salinization is a major environmental hazard in arid and semi-arid regions, threatening crop productivity and global agriculture. Objective To map soil salinity using Sentinel-2A, Landsat-8 OLI, and ground-based EC measurements on the Google Earth Engine (GEE) platform. Data Sources - Sentinel-2A satellite imagery - Landsat-8 OLI - Ground-based electrical conductivity (EC) measurements. Machine Learning Algorithms - Support Vector Regression (SVR) - Random Forest (RF) - Classification and Regression Trees (CART). Key Variables Top 5 variables used: - Wetness band - 3 soil salinity indices - 1 vegetation index. Best Performing Model - RF algorithm: Most precise spatial distribution of soil salinity classes. - CART: Slightly better prediction accuracy (R² = 0.98 for Sentinel-2A, 0.96 for Landsat-8). Accuracy Results - RF: R² = 0.96 (Sentinel-2A), 0.94 (Landsat-8). - CART: R² = 0.98 (Sentinel-2A), 0.96 (Landsat-8). Key Findings - RF model accurately estimated salinity levels in salt crusts, agricultural lands, drainage areas, and swamps. - SVR algorithms performed poorly in estimating soil EC values compared to RF and CART. Platform Used Google Earth Engine (GEE). Reference Aksoy et al., 2022. This study evaluates the suitability of GF-5, Sentinel-1, and Sentinel-2 remote sensing data for soil salinity research in the Qaidam Basin, China, using machine learning models (PSO-ANN, WOA-ELM, PLSR). Sentinel-2 data showed the strongest correlation with soil salinity, and PSO-ANN outperformed other models, demonstrating the effectiveness of combining high-resolution satellite data and machine learning for large-scale soil salinity monitoring (Jiang et al., 2023). The details of this study are given in Table 18. Table 18. Jiang et al., detail of the study conducted in 2023 Issue Soil salinization is a prevalent problem associated with land degradation, threatening the soil ecosystem. Objective To examine the suitability of GF-5, Sentinel-1, and Sentinel-2 remote sensing data for soil salinity research. Study Area Golmud, Qaidam Basin, China. Remote Sensing Data GF-5, Sentinel-1, and Sentinel-2. Machine Learning Models - Particle Swarm Optimization-Artificial Neural Network (PSO-ANN) - Whale Optimization Algorithm-Extreme Learning Machine (WOA-ELM) - Partial Least Squares Regression (PLSR). Key Findings - Sentinel-2 showed a stronger correlation with electric conductivity (EC) than Sentinel-1 and GF-5. - Sentinel-2-based models outperformed Sentinel-1 and GF-5-based models in simulation accuracy. - PSO-ANN performed the best, while PLSR performed the worst. Model Performance Metrics - Mean Absolute Error (MAE) - R² (Coefficient of Determination) - Root Mean Square Error (RMSE). Comparison of Remote Sensing Data - Sentinel-1: High sensitivity and penetrability. - GF-5: High spectral resolution. - Sentinel-2: High spatial resolution. Conclusion Sentinel-1, GF-5, and Sentinel-2 are suitable for monitoring soil salinity. Machine learning methods (e.g., PSO-ANN) combined with spaceborne data (e.g., Sentinel-2) are effective for large-scale soil salinization studies. Reference Jiang et al., 2023. Soil salinity is a critical environmental issue that affects crop productivity and land degradation, and this study uses machine learning techniques (RF, SVM, ANN) to accurately map and monitor soil salinity in Huludao City, China. The RF model demonstrated the highest accuracy (R²=0.84), with environmental parameters like NDVI and GNDVI playing key roles in salinity patterns, highlighting the need for further research to address human-induced salinity and improve estimation methods (Wang and Sun, 2024). The details of this study are given in Table 19. Table 19. Wang and Sun detail of the study conducted in 2024 Issue Soil salinity is a major environmental issue that impedes crop growth, reduces productivity, and leads to land degradation. Objective To estimate the spatial distribution of soil salinity in Huludao City, northeastern China, using machine learning techniques. Study Area Huludao City, northeastern China. Data Collection 310 soil samples were collected to determine soil salinity. Environmental Parameters Derived from remote sensing data (e.g., NDVI, GNDVI, standh, BI). Machine Learning Models - Random Forests (RF) - Support Vector Machines (SVM) - Artificial Neural Networks (ANN). Uncertainty Estimation Upper limit (95%) and lower limit (5%) prediction intervals were used. Best Performing Model RF model: - RMSE = 0.03 - AIC = -919 - BIC = -891 - R² = 0.84. Uncertainty Metrics Prediction Interval Coverage Probability (PICP) showed high predictive accuracy. Key Environmental Parameters NDVI, GNDVI, standh, and BI were the main regulators of soil salinity patterns. Conclusion - Machine learning techniques (especially RF) are effective for accurate soil salinity mapping. - Human activity has increased soil salinity, requiring further research for more precise estimation methods. Reference Wang and Sun, 2024. The Harran Plain faces soil salinization due to poor irrigation practices, and this study investigates the use of visible and near-infrared reflectance spectroscopy (VNIRS) and electromagnetic induction (EM) techniques to assess soil salinity levels. Both methods showed moderate success, with VNIRS achieving higher accuracy (R²=0.80–0.91) compared to EM (R²=0.47–0.79), demonstrating their potential for efficient soil salinity monitoring in precision agriculture (Bilgili et al., 2015). The details of this study are given in Table 20. Table 20. Bilgili et al., detail of the study conducted in 2015 Issue Soil salinization in the Harran Plain (17,000 hectares) due to careless irrigation practices, topography, climate, soil properties, and inadequate drainage. Challenges Salinity varies spatially and temporally due to groundwater levels, topography, and agricultural practices (irrigation, fertilization). Extensive sampling and laboratory analyses are costly and time-consuming. Solution Precision agriculture methods, including Visible and Near-Infrared Reflectance Spectroscopy (VNIRS) and Electromagnetic Induction (EM) techniques. Study Area Harran Plain, Turkey. Data Collection 90 salinity-affected locations were investigated. Techniques Used - EM technique: EM-38 instrument. - VNIRS: Spectroradiometer for 2 mm air-dry soils under laboratory conditions. Calibration Method Partial Least Squares Regression (PLSR) for electrical conductivity (ECe) and reflectance. Results - VNIRS: R² values ranged from 0.80 to 0.91 for estimating soil salinity. - EM technique: R² values between ECa and ECe ranged from 0.47 to 0.79. Conclusion Both VNIRS and EM techniques showed moderate success in estimating soil salinity. Precision agriculture methods can overcome traditional sampling challenges. Reference Bilgili et al., 2015. This study explores the fusion of Sentinel-1 SAR and Sentinel-2 multispectral images to improve soil salinity inversion accuracy in bare soil, testing the approach in China’s Hetao Irrigation Area. The Random Forest (RF) model achieved the best results (R²=0.801), demonstrating that combining SAR and multispectral data through Gram-Schmidt fusion enhances salinity monitoring accuracy under varying climatic conditions (He et al., 2023). The details of this study are given in Table 21. Table 21. He et al., detail of the study conducted in 2023 Issue Multispectral satellites have richer wavelength bands but are limited by cloudy weather, while SAR has higher penetration and is less affected by clouds. Objective To improve soil salinity inversion accuracy by fusing Sentinel-1 (SAR) and Sentinel-2 (multispectral) images and testing in the Hetao Irrigation Area, Inner Mongolia, China. Study Area Hetao Irrigation Area, Inner Mongolia, China. Data Fusion Method Gram-Schmidt (GS) fusion of Sentinel-1 (VV band) and Sentinel-2 images. Preprocessing Maximum-normalization applied to the fused image’s DN value. Machine Learning Models - Back Propagation (BP) - Support Vector Machine (SVM) - Random Forest (RF). Input Variables Remote indices and highly relevant fused single bands after two-tailed significance test and variable selection. Evaluation Metrics R², RMSE, RPD, RSS, RPIQ. Results - RF model performed best: R² = 0.801, RMSE = 0.686. - SVM: R² = 0.624, RMSE = 0.875. - BP: R² = 0.613, RMSE = 0.918. Key Findings - Fusion of Sentinel-1 and Sentinel-2 images improved bare soil salinity inversion accuracy. - RF outperformed SVM and BP in accuracy. Conclusion Combining Sentinel-1 radar images with Sentinel-2 multispectral images using GS fusion and machine learning is feasible and improves soil salinity inversion accuracy. Reference He et al., 2023. Conclusion Soil salinity is the main factor contributing to the degradation of agricultural land brought on by human activity and climate change. Due to decreased agricultural productivity and the difficulty of implementing sustainable agricultural practices, this situation jeopardizes food security. Traditional methods of monitoring soil salinity can be costly and time-consuming; however, remote sensing technologies and machine learning algorithms make the process faster, more efficient, and more comprehensive. Machine learning techniques (e.g., RF, SVR, XGboost, etc.) provide extremely accurate salinity estimates, and Sentinel-1 and Sentinel-2 satellite images are essential for mapping soil salinity. For managing and monitoring soil salinity, these techniques have a lot of promise, particularly in arid and semiarid areas. For precise estimations, it is crucial to choose the best algorithm across various regions and data sets. Future developments of more efficient methods to regulate soil salinity and boost agricultural productivity will be aided by the research done. References Abbas, A., & Khan, S. (2007). Using remote sensing techniques for appraisal of irrigated soil salinity. Modeling and Simulation Society of Australia and New Zealand, 2632–2638. https://www.researchgate.net/publication/237421639 Abdel Rahman, M. A. E., Afifi, A. A., D’Antonio, P., Gabr, S. S., & Scopa, A. (2022). Detecting and mapping salt-affected soil with arid integrated indices in feature space using multi-temporal Landsat imagery. Remote Sensing, 14(11), 2599. https://doi.org/10.3390/rs14112599 Abdel Rahman, M. A. E., Metwaly, M. M., & Shalaby, A. (2019). Quantitative assessment of soil saline degradation using remote sensing indices in Siwa Oasis. Remote Sensing Applications: Society and Environment, 13, 53–60. https://doi.org/10.1016/j.rsase.2018.10.008 Abdelaziz, M. E., Abdelsattar, M., Abdeldaym, E. A., Atia, M. A., Mahmoud, A. W. M., Saad, M. M., & Hirt, H. (2019). Piriformospora indica alters Na+/K+ homeostasis, antioxidant enzymes, and LeNHX1 expression of greenhouse tomato grown under salt stress. Scientia Horticulturae, 256, 108532. https://doi.org/10.1016/j.scienta.2019.108532 Aboelsoud, H. M., AbdelRahman, M. A., Kheir, A. M., Eid, M. S., Ammar, K. A., Khalifa, T. H., & Scopa, A. (2022). Quantitative estimation of saline-soil amelioration using remote-sensing indices in arid land for better management. Land, 11(7), 1041. https://doi.org/10.3390/land11071041 Aboelsoud, H. M., & AbdelRahman, M. A. (2017). Rapid field technique for soil salinity appraisal in north Nile delta using EM38 through some empirical relations. International Journal of Plant & Soil Science, 14(5), 1–9. https://doi.org/10.9734/IJPSS/2017/31806 Aksoy, S., Yildirim, A., Gorji, T., Hamzehpour, N., Tanik, A., & Sertel, E. (2022). Sentinel-2A ve Landsat-8 OLI verilerini kullanarak Google Earth Engine platformunda toprak tuzluluğu haritalama için makine öğrenimi algoritmalarının performansının değerlendirilmesi. Uzay Araştırmalarındaki Gelişmeler, 69(2), 1072–1086. Allbed, A., & Kumar, L. (2013). Soil salinity mapping and monitoring in arid and semi-arid regions using remote sensing technology: A review. Advances in Remote Sensing, 2013. Allbed, A., Kumar, L., & Aldakheel, Y. Y. (2014). Assessing soil salinity using soil salinity and vegetation indices derived from IKONOS high-spatial resolution imageries: Applications in a date palm dominated region. Geoderma, 230–231, 1–8. https://doi.org/10.1016/j.geoderma.2014.03.025 Amirgaliyev, Y., Mukhamediev, R., Merembayev, T., Kuchin, Y., Ataniyazova, A., & Omarova, P. (2024). Remote sensing and machine learning algorithms to predict soil salinity in southern Kazakhstan. Discover Sustainability , 5 (1), 1-14. Baghdadi, N., Boyer, N., Todoroff, P., El Hajj, M., & Bégué, A. (2009). Potential of SAR sensors TerraSAR-X, ASAR/ENVISAT, and PALSAR/ALOS for monitoring sugarcane crops on Reunion Island. Remote Sensing of Environment, 113(9), 1724–1738. https://doi.org/10.1016/j.rse.2009.04.005 Bilgili, A., Çullu, M., & Aydemir, S. (2015). Tuzdan Etkilenmiş Toprakların Yakın Kızılötesi Yansıma Spektroradyometre ve Elektromanyetik İndüksiyon Tekniği Yardımıyla Karakterize Edilebilme Potansiyelinin Araştırılması. Harran Tarım ve Gıda Bilimleri Dergisi, 18(1), 33–46. Brochier, F., & Ramieri, E. (2001). Climate change impacts on the Mediterranean coastal zones. SSRN Journal, 27, 82. https://doi.org/10.2139/ssrn.270927 Chatziantoniou, A., Psomiadis, E., & Petropoulos, G. (2017). Co-orbital Sentinel 1 and 2 for LULC mapping with emphasis on wetlands in a Mediterranean setting based on machine learning. Remote Sensing, 9(12), 1259. https://doi.org/10.3390/rs9121259 Corwin, D. L., Rhoades, J. D., & Šimůnek, J. (2007). Leaching requirement for soil salinity control: Steady-state versus transient models. Agricultural Water Management, 90(3), 165–180. https://doi.org/10.1016/j.agwat.2007.03.004 Cui, J., Chen, X., Han, W., Cui, X., Ma, W., & Li, G. (2023). Estimation of soil salt content at different depths using UAV multi-spectral remote sensing combined with machine learning algorithms. Remote Sensing, 15(21), 5254. https://doi.org/10.3390/rs15215254 Das, A., Bhattacharya, B. K., Setia, R., Jayasree, G., & Das, B. S. (2023). A novel method for detecting soil salinity using AVIRIS-NG imaging spectroscopy and ensemble machine learning. ISPRS Journal of Photogrammetry and Remote Sensing, 200, 191–212. https://doi.org/10.1016/j.isprsjprs.2023.04.012 Douaoui, A. E. K., Nicolas, H., & Walter, C. (2006). Detecting salinity hazards within a semiarid context employing combining soil and remote sensing data. Geoderma, 134(1–2), 217–230. https://doi.org/10.1016/j.geoderma.2005.10.009 Dube, T., & Mutanga, O. (2015). Investigating the robustness of the new Landsat-8 operational land imager derived texture metrics in estimating plantation forest aboveground biomass in resource-constrained areas. ISPRS Journal of Photogrammetry and Remote Sensing, 108, 12–32. https://doi.org/10.1016/j.isprsjprs.2015.06.002 Étienne, L. (2017). La Salinisation des Sols dans L’archipel de Kerkennah, Tunisie. Mappemonde, 119. https://doi.org/10.4000/mappemonde.2658 Étienne, L., Dahech, S., Beltrando, G., & Daoud, A. (2012). Dynamiques Récentes Des Sebkhas Littorales De L’archipel Des Kerkennah (Tunisie Centro-Méridionale): Apport De La Télédétection. Revue de Télédétection, 11, 274–281. Etienne, L. (2014). Accentuation récente de la vulnérabilité liée à la mobilité du trait de côte et à la salinisation des sols dans l’archipel de Kerkennah (Tunisie). Thèse de doctorat, Université Paris Diderot (Paris 7) Sorbonne Paris Cité; Université de Sfax (Faculté des Lettres et Sciences Humaines). https://theses.hal.science/tel-01075029 Farifteh, J., Van der Meer, F., Van der Meijde, M., & Atzberger, C. (2008). Spectral characteristics of salt-affected soils: A laboratory experiment. Geoderma, 145(3–4), 196–206. https://doi.org/10.1016/j.geoderma.2007.12.002 Fathizad, H., Ardakani, M. A. H., Heung, B., Sodaiezadeh, H., Rahmani, A., Fathabadi, A., … & Taghizadeh-Mehrjardi, R. (2020). Spatio-temporal dynamic of soil quality in the central Iranian desert modeled with machine learning and digital soil assessment techniques. Ecological Indicators, 118, 106736. https://doi.org/10.1016/j.ecolind.2020.106736 Fathizad, H., Ardakani, M. A. H., Sodaiezadeh, H., Kerry, R., & Taghizadeh-Mehrjardi, R. (2020). Investigation of the spatial and temporal variation of soil salinity using random forests in the central desert of Iran. Geoderma, 365, 114233. https://doi.org/10.1016/j.geoderma.2020.114233 Ferreira, M. P., Wagner, F. H., Aragão, L. E. O. C., Shimabukuro, Y. E., & de Souza Filho, C. R. (2019). Tree species classification in tropical forests using visible to shortwave infrared WorldView-3 images and texture analysis. ISPRS Journal of Photogrammetry and Remote Sensing, 149, 119–131. https://doi.org/10.1016/j.isprsjprs.2018.12.013 Ge, X., Ding, J., Teng, D., Wang, J., Huo, T., Jin, X., … & Han, L. (2022). Updated soil salinity with fine spatial resolution and high accuracy: The synergy of Sentinel-2 MSI, environmental covariates, and hybrid machine learning approaches. Catena, 212, 106054. https://doi.org/10.1016/j.catena.2022.106054 Golestani, M., Ghahfarokhi, Z. M., Esfandiarpour-Boroujeni, I., & Shirani, H. (2023). Evaluating the spatiotemporal variations of soil salinity in Sirjan Playa, Iran using Sentinel-2A and Landsat-8 OLI imagery. Catena, 231, 107375. https://doi.org/10.1016/j.catena.2023.107375 Griggs, G., & Reguero, B. G. (2021). Coastal adaptation to climate change and sea-level rise. Water, 13(16), 2151. https://doi.org/10.3390/w13162151 Guan, X., Wang, S., Gao, Z., & Lv, Y. (2013). Dynamic prediction of soil salinization in an irrigation district based on the support vector machine. Mathematical and Computer Modelling, 58(3–4), 719–724. https://doi.org/10.1016/j.mcm.2012.12.011 Han, Y., Ge, H., Xu, Y., Zhuang, L., Wang, F., Gu, Q., & Li, X. (2023). Estimating soil salinity using multiple spectral indexes and machine learning algorithm in Songnen Plain, China. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 16, 7041–7050. https://doi.org/10.1109/JSTARS.2023.3298986 He, Y., Yin, H., Chen, Y., Xiang, R., Zhang, Z., & Chen, H. (2024). Soil salinity estimation based on Sentinel-1/2 texture features and machine learning. IEEE Sensors Journal, 24(9), 1–12. https://doi.org/10.1109/JSEN.2024.3384567 He, Y., Zhang, Z., Xiang, R., Ding, B., Du, R., Yin, H., … & Ba, Y. (2023). Monitoring salinity in bare soil based on Sentinel-1/2 image fusion and machine learning. Infrared Physics & Technology, 131, 104656. https://doi.org/10.1016/j.infrared.2023.104656 Herrero, J., & Pérez-Coveta, O. (2005). Soil salinity changes over 24 years in a Mediterranean irrigated district. Geoderma, 125(3–4), 287–308. https://doi.org/10.1016/j.geoderma.2004.09.003 Hihi, S., Rabah, Z. B., Bouaziz, M., Chtourou, M. Y., & Bouaziz, S. (2019). Prediction of soil salinity using remote sensing tools and linear regression model. Advances in Remote Sensing, 8(2), 77–88. https://doi.org/10.4236/ars.2019.82005 Hihi, S., Katlane, R., Kilani, B., Zekri, M. W., Bensalah, R., Siewert, C., & Kallel, M. (2023). Evaluating drought effects on soil: Innovative soil salinity monitoring via SAR data, Sentinel-2 imagery, and machine learning algorithms in Kerkennah Archipelago. Atmosphere, 14(10), 1514. https://doi.org/10.3390/atmos14101514 Hillel, D. (2000). Salinity management for sustainable irrigation: Integrating science, environment, and economics. World Bank Publications. Houk, E., Frasier, M., & Schuck, E. (2006). The agricultural impacts of irrigation-induced waterlogging and soil salinity in the Arkansas Basin. Agricultural Water Management, 85(1–2), 175–183. https://doi.org/10.1016/j.agwat.2005.07.003 Ismaili, M., Krimissa, S., Namous, M., Htitiou, A., Abdelrahman, K., Fnais, M. S., … & Benabdelouahab, T. (2023). Assessment of soil suitability using machine learning in arid and semi-arid regions. Agronomy, 13(1), 165. https://doi.org/10.3390/agronomy13010165 Jafarpoor, A., Sadeghi, S. H., Darki, B. Z., & Homaee, M. (2022). Changes in hydrologic components from a mid-sized plot induced by rill erosion due to cyanobacterization. International Soil and Water Conservation Research, 10(1), 143–148. https://doi.org/10.1016/j.iswcr.2021.10.003 Jia, Y., Wu, J., Cheng, M., & Xia, X. (2023). Global transfer of salinization on irrigated land: Complex network and endogenous structure. Journal of Environmental Management, 336, 117592. https://doi.org/10.1016/j.jenvman.2023.117592 Jiang, X., Duan, H., Liao, J., Guo, P., Huang, C., & Xue, X. (2023). Comparative research on multi-algorithm of soil salinity monitoring based on Gaofen-5, Sentinel-1, and Sentinel-2. International Journal of Remote Sensing, 44(15), 4704–4726. https://doi.org/10.1080/01431161.2023.2230478 Kaplan, F., & Bilgili, A. V. (2024). Characterization of some properties of soils formed on basalt parent material using spectroradiometric and geostatistical techniques. ISPEC Journal of Agricultural Sciences, 8(3), 583–603. Kaplan, G., Gašparović, M., Alqasemi, A. S., Aldhaheri, A., Abuelgasim, A., & Ibrahim, M. (2023). Soil salinity prediction using machine learning and Sentinel-2 remote sensing data in hyper-arid areas. Physics and Chemistry of the Earth, Parts A/B/C, 130, 103400. https://doi.org/10.1016/j.pce.2023.103400 Khan, N. M., Rastoskuev, V., Sato, Y., & Shiozawa, S. (2005). Assessment of hydro-saline land degradation by using a simple approach of remote sensing indicators. Agricultural Water Management, 77(1–3), 96–109. https://doi.org/10.1016/j.agwat.2004.09.038 Khan, N. M., Rastoskuev, V. V., Shalina, E. V., & Sato, Y. (2001). Mapping salt-affected soils using remote sensing indicators—A simple approach with the use of GIS IDRISI. In Paper presented at the 22nd Asian Conference on Remote Sensing, Singapore, 5–9. Klibi, S., Tounsi, K., Rebah, Z. B., Solaiman, B., & Farah, I. R. (2020, September). Soil salinity prediction using a machine learning approach through hyperspectral satellite image. In 2020 5th International Conference on Advanced Technologies for Signal and Image Processing (ATSIP) (pp. 1–6). IEEE. https://doi.org/10.1109/ATSIP49331.2020.9231876 Kurbatova, A. I., Bouchhima, R. A., Grigorets, E. A., Tsymbarovich, P. R., & Ksibi, M. (2021). Methodology for mapping soil salinity and halophyte cover using remote sensing data in Kerkennah, Tunisia. Euro-Mediterranean Journal for Environmental Integration , 6 (2), 51. Lobell, D. B., Schlenker, W., & Costa-Roberts, J. (2011). Climate trends and global crop production since 1980. Science , 333 (6042), 616-620. Li, Y., Wang, C., Wright, A., Liu, H., Zhang, H., & Zong, Y. (2021). Combination of GF-2 high spatial resolution imagery and land surface factors for predicting soil salinity of muddy coasts. CATENA, 202, 105304. https://doi.org/10.1016/j.catena.2021.105304. Li, J., Zhang, T., Shao, Y., & Ju, Z. (2023). Comparing machine learning algorithms for soil salinity mapping using topographic factors and Sentinel-1/2 data: A case study in the Yellow River Delta of China. Remote Sensing, 15(9), 2332. https://doi.org/10.3390/rs15092332. Li, X., Li, Y., Wang, B., Sun, Y., Cui, G., & Liang, Z. (2022). Analysis of spatial-temporal variation of the saline-sodic soil in the west of Jilin Province from 1989 to 2019 and influencing factors. CATENA, 217, 106492. https://doi.org/10.1016/j.catena.2022.106492. Mahajan, G. R., Das, B., Gaikwad, B., Murgaonkar, D., Desai, A., Morajkar, S., … & Kulkarni, R. M. (2021). Monitoring properties of the salt-affected soils by multivariate analysis of the visible and near-infrared hyperspectral data. Catena , 198 , 105041. Maki,M.; Sritumboon, S.; Srisutham, M.; Yoshida, K.; Homma, K.; Sukchan, S. Impact of changes in the relationship between salinity and soil moisture on remote sensing data usage in northeast Thailand. Hydrol. Res. Lett. 2022, 16, 54–58. [CrossRef] Maxwell, A.E.; Warner, T.A.; Fang, F. Implementation of machine-learning classification in remote sensing: An applied review. Int. J. Remote Sens. 2018, 39, 2784–2817. [CrossRef] Merembayev, T.; Amirgaliyev, Y.; Saurov, S.; Wójcik, W. Soil Salinity Classification Using Machine Learning Algorithms and Radar Data in the Case from the South of Kazakhstan. J. Ecol. Eng. 2022, 23, 61 67. [CrossRef] Metternicht, G. I., & Zinck, J. A. (2003). Remote sensing of soil salinity: potentials and constraints. Remote sensing of environment , 85 (1), 1-20. Mukhopadhyay, R., Sarkar, B., Jat, H. S., Sharma, P. C., & Bolan, N. S. (2021). Soil salinity under climate change: Challenges for sustainable agriculture and food security. Journal of Environmental Management, 280, 111736. https://doi.org/10.1016/j.jenvman.2020.111736 Nguyen, T. T., Pham, T. D., Nguyen, C. T., Delfos, J., Archibald, R., Dang, K. B., … & Ngo, H. H. (2022). A novel intelligence approach based active and ensemble learning for agricultural soil organic carbon prediction using multispectral and SAR data fusion. Science of the Total Environment , 804 , 150187. Nicolas, H., Walter, C., 2006. Detecting salinity hazards within a semiarid context by means of combining soil and remote-sensing data. Geoderma 134 (1–2), j.geoderma.2005.10.009. Nurmemet Erkin, Lei Zhu, Haibin Gu, Alimujiang Tusiyiti, “Method for predicting soil salinity concentrations in croplands based on machine learning and remote sensing techniques,” J. Appl. Remote Sens. 13(3), 034520 (2019), doi: 10.1117/1.JRS.13.034520. Pannell, D. J., & Ewing, M. A. (2006). Managing secondary dryland salinity: options and challenges. Agricultural water management , 80 (1-3), 41-56. Peng, J., Biswas, A., Jiang, Q., Zhao, R., Hu, J., Hu, B., & Shi, Z. (2019). Estimating soil salinity from remote sensing and terrain data in southern Xinjiang Province, China. Geoderma , 337 , 1309-1319. Periasamy, S., Ravi, K. P., & Tansey, K. (2022). Identification of saline landscapes from an integrated SVM approach from a novel 3-D classification schema using Sentinel-1 dual-polarized SAR data. Remote Sensing of Environment, 279, 113144. https://doi.org/10.1016/j.rse.2022.113144 Peck, A. J., & Hatton, T. (2003). Salinity and the discharge of salts from catchments in Australia. Journal of hydrology , 272 (1-4), 191-202. Savin IY, Terekhov A, Amirgaliyev Y, Sagatdinova G. Satellite monitoring of salinization of irrigated soils in southern kazakhstan. Eurasian Soil Sci. 2023;56(10):1498–506. Shi, H., Hellwich, O., Luo, G., Chen, C., He, H., Ochege, F. U., … & De Maeyer, P. (2021). A global meta-analysis of soil salinity prediction integrating satellite remote sensing, soil sampling, and machine learning. IEEE Transactions on Geoscience and Remote Sensing , 60 , 1-15. Shivanna, K.R. Climate change and its impact on biodiversity and human welfare. Proc. Indian Natl. Sci. Acad. 2022, 88, 160–171. [CrossRef] Spies, B., & Woodgate, P. (2005). Salinity mapping methods in the Australian context (pp. 23-24). Canberra, Australia: Department of the Environment and Heritage. Sun, Q., Sun, J., Baidurela, A., Li, L., Hu, X., & Song, T. (2022). Ecological landscape pattern changes and security from 1990 to 2021 in Ebinur Lake Wetland Reserve, China. Ecological Indicators, 145, 109648. https://doi.org/10.1016/j.ecolind.2022.109648 Taghadosi, M.M.; Hasanlou, M.; Eftekhari, K. Soil salinity mapping using dual-polarized SAR Sentinel-1 imagery. Int. J. Remote Sens. 2019, 40, 237–252. [CrossRef] Taghizadeh-Mehrjardi, R., Schmidt, K., Toomanian, N., Heung, B., Behrens, T., Mosavi, A., … & Scholten, T. (2021). Improving the spatial prediction of soil salinity in arid regions using wavelet transformation and support vector regression models. Geoderma , 383 , 114793. Tripathi, A.; Tiwari, R.K. A simplified subsurface soil salinity estimation using synergy of SENTINEL-1 SAR and SENTINEL-2 multispectral satellite data, for early stages of wheat crop growth in Rupnagar, Punjab, India. Land Degrad. Dev. 2021, 32, 3905–3919. [CrossRef] Wang, J., Ding, J., Yu, D., Ma, X., Zhang, Z., Ge, X., … & Guo, Y. (2019). Capability of Sentinel-2 MSI data for monitoring and mapping of soil salinity in dry and wet seasons in the Ebinur Lake region, Xinjiang, China. Geoderma , 353 , 172-187. Wang, J., Ding, J., Yu, D., Teng, D., He, B., Chen, X., … & Su, F. (2020). Machine learning-based detection of soil salinity in an arid desert region, Northwest China: A comparison between Landsat-8 OLI and Sentinel-2 MSI. Science of the Total Environment , 707 , 136092. Wang, N., Peng, J., Xue, J., Zhang, X., Huang, J., Biswas, A., … & Shi, Z. (2022). A framework for determining the total salt content of soil profiles using time-series Sentinel-2 images and a random forest-temporal convolution network. Geoderma , 409 , 115656. Wang, W., & Sun, J. (2024). Estimation of soil salinity using satellite-based variables and machine learning methods. Earth Science Informatics , 1-13. Wen, W., Timmermans, J., Chen, Q., & van Bodegom, P. M. (2020). A review of remote sensing challenges for food security with respect to salinity and drought threats. Remote Sensing, 13(1), 6. https://doi.org/10.3390/rs13010006 Wiebe, K., Soule, M. J., Narrod, C., & Breneman, V. E. (2003). Resource quality and agricultural productivity: a multi-country comparison. Land Quality, Agricultural Productivity and Food Security: Biophysical Processes and Economic Choices at Local, Regional and Global Levels. Edward Elgar, Northampton, MA , 147-165. Xiao, C., Ji, Q., Chen, J., Zhang, F., Li, Y., Fan, J., … & Wang, H. (2023). Prediction of soil salinity parameters using machine learning models in an arid region of northwest China. Computers and Electronics in Agriculture , 204 , 107512. Yahiaoui, I.; Bradaï, A.; Douaoui, A.; Abdennour, M.A. Performance of random forest and buffer analysis of Sentinel-2 data for modelling soil salinity in the Lower-Cheliff plain (Algeria). Int. J. Remote Sens. 2021, 42, 148–171. Yujie He, Haoyuan Yin, Yinwen Chen, Ru Xiang, Zhitao Zhang, and Haiying Chen, 2024. Soil Salinity Estimation Based on Sentinel-1/2 Texture Features and Machine Learning, IEEE SENSORS JOURNAL, VOL. 24, NO. 9, 1 MAY 2024. Zarei, A., Hasanlou, M., & Mahdianpari, M. (2021). A comparison of machine learning models for soil salinity estimation using multi-spectral Earth observation data. ISPRS Annals of the Photogrammetry, Remote Sensing and Spatial Information Sciences, V-3-2021, 257–263. https://doi.org/10.5194/isprs-annals-V-3-2021-257-2021 Zhang, Q.; Li, L.; Sun, R.; Zhu, D.; Zhang, C.; Chen, Q. Retrieval of the Soil Salinity from Sentinel-1 Dual-Polarized SAR Data Based on Deep Neural Network Regression. IEEE Geosci. Remote Sens. Lett. 2022, 19, 1–5. [CrossRef] Zhang, Q.; Zhou, Z.S.; Caccetta, P.; Simons, J.; Li, L. Sentinel-1 Imagery Incorporating Machine Learning for Dryland Salinity Monitoring: A Case Study in Esperance, Western Australia. In Proceedings of the IGARSS 2020–2020 IEEE International Geoscience and Remote Sensing Symposium, Waikoloa, HI, USA, 26 September–2 October 2020; IEEE: Piscataway, NJ, USA, 2020; pp. 4914–4917. Available online: https://ieeexplore.ieee.org/document/9323426/ (accessed on 21 August 2023). Zhang, Y. F., Li, Y. P., Sun, J., & Huang, G. H. (2020). Optimizing water resources allocation and soil salinity control for supporting agricultural and environmental sustainable development in Central Asia. Science of the Total Environment, 704, 135281. https://doi.org/10.1016/j.scitotenv.2019.135281 Zheng, Z.; Zhang, F.; Ma, F.; Chai, X.; Zhu, Z.; Shi, J.; Zhang, S. Spatiotemporal changes in soil salinity in a drip-irrigated field. Geoderma 2009, 149, 243–248. [CrossRef] Zhu, J., Pan, Z., Wang, H., Huang, P., Sun, J., Qin, F., & Liu, Z. (2019). An improved multi-temporal and multi-feature tea plantation identification method using Sentinel-2 imagery. Sensors , 19 (9), 2087. Information & Authors Information Version history V1 Version 1 04 February 2025 Copyright This work is licensed under a Non Exclusive No Reuse License. Keywords arazi bozulması machine learning algorithms remote sensing soil salinity sustainable agriculture Authors Affiliations Fatma KAPLAN 0000-0002-4873-3997 [email protected] Harran Universitesi Ziraat Fakultesi View all articles by this author Ali Volkan BİLGİLİ Harran Universitesi Ziraat Fakultesi View all articles by this author Metrics & Citations Metrics Article Usage 886 views 246 downloads .FvxKWukQNSOunydq8rnd { width: 100px; } Citations Download citation Fatma KAPLAN, Ali Volkan BİLGİLİ. Salinity Determination Of Soil Through Machine Learning And Remote Sensing Techniques. Authorea . 04 February 2025. DOI: https://doi.org/10.22541/au.173865076.63148165/v1 If you have the appropriate software installed, you can download article citation data to the citation manager of your choice. Simply select your manager software from the list below and click Download. For more information or tips please see 'Downloading to a citation manager' in the Help menu . Format Please select one from the list RIS (ProCite, Reference Manager) EndNote BibTex Medlars RefWorks Direct import Tips for downloading citations document.getElementById('citMgrHelpLink').addEventListener('click', function() { popupHelp(this.href); return false; }); $(".js__slcInclude").on("change", function(e){ if ($(this).val() == 'refworks') $('#direct').prop("checked", false); $('#direct').prop("disabled", ($(this).val() == 'refworks')); }); Cited by Marzieh Negahban, Kamel Msaada, Beyond the crop: the role of medicinal and aromatic plants in soil carbon sequestration and nitrogen cycling, International Journal of Environmental Health Research, (1-24), (2026). https://doi.org/10.1080/09603123.2026.2653197 Crossref Loading... View Options View options PDF View PDF Figures Tables Media Share Share Share article link Copy Link Copied! Copying failed. Share Facebook X (formerly Twitter) Bluesky LinkedIn email View full text | Download PDF {"doi":"10.22541/au.173865076.63148165/v1","type":"Article"} Now Reading: Share Figures Tables Close figure viewer Back to article Figure title goes here Change zoom level Go to figure location within the article Download figure Toggle share panel Toggle share panel Share Toggle information panel Toggle information panel Go to previous graphic Go to next graphic Go to previous table Go to next table All figures All tables View all material View all material xrefBack.goTo xrefBack.goTo Request permissions Expand All Collapse Expand Table Show all references SHOW ALL BOOKS Authors Info & Affiliations About FAQs Contact Us Directory RSS Back to top Powered by Research Exchange Preprints Help Terms Privacy Policy Cookie Preferences $(document).ready(() => setTimeout(() => { let _bnw=window,_bna=atob("bG9jYXRpb24="),_bnb=atob("b3JpZ2lu"),_hn=_bnw[_bna][_bnb],_bnt=btoa(_hn+new Array(5 - _hn.length % 4).join(" ")); $.get("/resource/lodash?t="+_bnt); },4000)); (function(){function c(){var b=a.contentDocument||a.contentWindow.document;if(b){var d=b.createElement('script');d.innerHTML="window.__CF$cv$params={r:'9ff6c4633914c13d',t:'MTc3OTM5OTkxNg=='};var a=document.createElement('script');a.src='/cdn-cgi/challenge-platform/scripts/jsd/main.js';document.getElementsByTagName('head')[0].appendChild(a);";b.getElementsByTagName('head')[0].appendChild(d)}}if(document.body){var a=document.createElement('iframe');a.height=1;a.width=1;a.style.position='absolute';a.style.top=0;a.style.left=0;a.style.border='none';a.style.visibility='hidden';document.body.appendChild(a);if('loading'!==document.readyState)c();else if(window.addEventListener)document.addEventListener('DOMContentLoaded',c);else{var e=document.onreadystatechange||function(){};document.onreadystatechange=function(b){e(b);'loading'!==document.readyState&&(document.onreadystatechange=e,c())}}}})();