نوع مقاله : مقاله پژوهشی
نویسندگان
1 دانشجوی دکتری گروه مهندسی علوم خاک، دانشکده کشاورزی، دانشگاه شهید چمران اهواز، خوزستان، ایران
2 استاد تمام گروه مهندسی علوم خاک، دانشکده کشاورزی، دانشگاه شهید چمران اهواز، خوزستان، ایران
3 استاد تمام مدرسه محیط زیست، کالج کشاورزی آنتاریو، دانشگاه گوئلف، آنتاریو، کانادا
چکیده
کلیدواژهها
عنوان مقاله [English]
نویسندگان [English]
Introduction: Soil salinity significantly contributes to the degradation of natural ecosystems in arid and semi-arid regions. Timely and effective monitoring and mapping of soil salinity are essential to prevent land degradation and promote sustainable soil management. Remote sensing data is recognized as an effective and accurate tool for identifying soil salinity. Khuzestan Province, one of Iran's agriculturally significant areas, is facing salinity challenges. Therefore, the objective of this research is to map and monitor soil salinity in Bavi County, located in Khuzestan Province, using feature space models based on remote sensing.
Materials and Methods: First, the locations of 350 sample points were determined using the conditioned Latin Hypercube Sampling (cLHS) method, and the electrical conductivity (EC) of the soil was measured at depths of 0–10, 10–20, and 20–30 cm. In this study, Landsat-9 data were used to extract optical indices and Sentinel-1 data were used to extract radar indices. After initial corrections were applied to the satellite images, the indices were calculated and normalized using raster calculation tool in ArcGIS 10.8.2 software. In the next step, the Digital Number (DN) values of the corresponding pixels were extracted from each index and the linear regression relationship between the indices was obtained. Accordingly, 12 two-dimensional feature space models consisting of optical and radar indices were constructed in SAGAGIS 9.5 software. The linear classification model of salinization of Hong et al. (2022) was used to prepare soil salinity maps. Finally, the salinity maps were separated into five soil salinity classes based on the classification of Brown et al. (1954) and the Support Vector Machine (SVM) algorithm in ENVI 5.6 software and were validated using the Overall Accuracy, Kappa Coefficient, and Bias parameters.
Results and Discussions: Descriptive statistics for the field data indicated a decrease in both the mean salinity and its variability with increasing soil depth. Analysis of the regression relationships between remote sensing indices within the feature space models revealed high coefficients of determination (R²), ranging from 0.814 to 0.973 for optical models and from 0.790 to 0.794 for radar models. Additionally, remote sensing indices in optical feature space models exhibited a stronger correlation compared to those in radar feature space models. After creating the density scatter plots and determining the equation of the best-fit line for each feature space model, the slope of the line perpendicular to the fitted straight line (K) was calculated. After creating the density scatter plots and determining the equation of the best-fit line for each feature space model, the slope of the line perpendicular to the fitted straight line (K) was calculated. Subsequently, linear equations for to determine soil salinity levels in feature space models were derived. Based on these equations, soil salinity maps for each model were generated using ArcGIS 10.8.2. The soil salinity maps obtained from the optical and radar feature space models were then classified into five categories: Non-Saline soils (0-2 dS/m), Slightly Saline soils (2-4 dS/m), Saline soils (4-8 dS/m), Strongly Saline soils (8-16 dS/m), and Extremely Saline soils (>16 dS/m). The classification results of the soil salinity maps using the Support Vector Machine algorithm demonstrated that the optical SI-Albedo feature space model performed best at the 0–10 cm depth, while the radar NRPB-RVI and RVI-NDPI feature space models performed best at the 10–20 cm and 20–30 cm depths, respectively, for estimating soil salinity. The estimations made by these models showed some level of overestimation for EC values through the study area.
Conclusions: The salinity classes in the maps generated from the SI-Albedo, NRPB-RVI, and RVI-NDPI models indicated that extremely saline soils, with EC greater than 16 dS/m, were the most frequent and predominantly located in the western parts of the study area. This study demonstrated that using optical and radar feature space models are an effective approach for monitoring soil salinity in arid and semi-arid regions. The maps of soil salinity levels can be utilized to optimize irrigation patterns, develop drainage systems, cultivate salt-tolerant plants, rehabilitate vegetation cover, and plan land use in the affected areas.
کلیدواژهها [English]
)