Introduction
Soil erosion is one of the most important aspects of soil and land degradation in agricultural ecosystems and natural resources. The soil erodibility factor (K) in the universal soil loss equation is one of the main influencing factors, indicating the sensitivity of soil to erosion. Today, digital soil mapping has become a powerful method for modeling and presenting accurate soil information and helping to make constructive decisions in the field of precision agriculture and environmental activities. One of the common methods of DSM are the geostatistical methods that have been developed and expanded in the past few decades and researchers have paid a lot of attention to them to describe the spatial changes of soil properties. Therefore, preparing a map of this factor using digital mapping techniques can provide very important information regarding the identification of areas at risk of erosion for users and land managers. Thus, the present study was conducted with the aim of preparing spatial distribution maps of K-factor in part of the Sistan plain.
Methodology
The studied area, with an area of about 235,000 hectares, includes a part of the Sistan plain in the east of Iran and in the north of Sistan and Baluchistan province, located on the Hirmand river delta. The soils are from young development (Entisols) to low to medium development (Aridisols) and in terms of moisture and thermal regime, they are aridic. and hyperthermic. Field studies included sampling from 460 points of the surface soil layer (0-15 cm) using a supervised random method covering the regions of Zabol, Hirmand, Zahak, Nimruz and Hamoun. The characteristics related to the estimation of K-factor (clay, silt, sand, very fine sand, organic matter, structural class and permeability class) were measured based on standard laboratory methods. To prepare digital maps of K-factor, four scenarios along with two geostatistical methods, ordinary co-kriging (OCK) and ordinary kriging (OK), were used. The four scenarios included S1: OCK model + topographic parameters, S2: OCK model + remote sensing parameters (I), S3: OCK model + remote sensing parameters (II) and S4: only using the OK model. It is worth mentioning that the topography and remote sensing variables are extracted from the digital height model (DEM) with a resolution of 12.5 meters from ALOS-PALSAR satellite and images from Sentinel 1 and Sentinel 2 as auxiliary variables.
Results and Discussion
The results showed that the average of sand, very fine sand, silt and clay in the soils is 41.7, 17, 41.6 and 16.67%, respectively. Also, the average organic matter in the soil was measured at 0.98%, which is in the low content class. The highest amount of K factor calculated in the studied area was 0.11 and the lowest was 0.007 (with an average of 0.06) ton/MJ h/mm, which indicates the high erodibility of the soil in the area. The main textural classes of the studied soils were in three classes: Loam, Silt Loam and Sandy Loam (more than 80% of the samples) and other classes had a small share. It seems that in the soil texture classes with the dominant amount of silt, mainly the amount of clay and organic matter decreases, and this increases the erodibility of the soil. The results of fitting the experimental semivariograms showed that the exponential model had the highest efficiency compared to the others, and the spatial autocorrelation class for the four scenarios under review was strong except for OK. The results of the accuracy assessment of predictions showed that the first scenario (S1), which combined OCK and topographic attributes (S1: diffuse insolation, relative slope position, valley depth), had higher accuracy and lower error (0.016 = RMSE) compared to the other three scenarios (S2: combination of Sentinel-1 bands and NDVI; S3: NDVI, SAVI, IPVI; S4: OK). Statistical description of the data indicated that more than 60% of the data fell into the very high erodibility class, followed by 23% in the high erodibility class, indicating that the spatial distribution maps show that in most areas of the region, especially the north, northeast, west, and southwest, soils have very critical conditions in terms of erosion sensitivity, which should be prioritized in soil management and conservation planning.
Conclusion
The present research was conducted with the aim of investigating the role of environmental variables in the preparation of the K-factor prediction map along with the use of COK and OK geostatistical models in a large part of the Sistan Plain. The results showed that the COK model combined with topographic auxiliary variables (extracted from DEM) had the highest accuracy value based on the index provided statistics. Also, the results showed that the changes in the K-factor are in line with the changes in the soil particles size in the region, and their changes in the Sistan plain depend on its geomorphic features (floodplain and delta) and coincide with the changes in the sedimentation regime of the Hirmand River and its branches in the plain. Additionally, due to the unstable geomorphological conditions of the Sistan plain, conducting periodic studies aimed at monitoring the temporal changes of this factor using other spatial modeling approaches such as machine learning algorithms is recommended for the future.