ارزیابی مدل‌های مختلف برآورد فرسایش‌پذیری خاک در خاک‌های لسی استان گلستان

Introduction
Currently, soil erosion is considered one of the important environmental problems in Iran's watersheds. Accurate estimation of sedimentation and soil erodibility is very important for the protection and management of natural resources. Soil erodibility indicates the potential ability of soil to erode, and its increase is considered a serious threat to the stability and production capacity of agricultural lands. Measuring soil erodibility by direct method is expensive and time-consuming, and nowadays indirect methods are used to determine it. One of the most common indirect methods for estimating this parameter is the use of the Wischmeier and Smith nomograph, which was developed for non-calcareous soils. However, the majority of the soils in Iran is calcareous and the use of the nomograph method does not provide a correct estimate of the amount of erodibility. Loess is one of the most important Quaternary sedimentary formations in northeastern Iran, which consists of a relatively high percentage of lime. Therefore, the main goal of current research is to determine the most suitable indirect model to determine soil erodibility in Loess lands with the origin of calcareous sediments.

Methodology
In the current study, Arab-Qara-Haji watershed with an area of 2596 hectares located in the east of Golestan province was selected as a typical loess area. The average annual precipitation of this watershed is 489 mm and it has a moderate semi-arid climate. This area is one of the sedimentary basins of Kopeh-Dagh zone related to the Jurassic period and consists of limestone and loess sediments. Based on the field studies, three main types of soil including hills (67.1%), loess plateaus (20.6%) and river terraces (12.3%) were identified in this watershed and the land unit was selected as working units. For this purpose, 48 surface samples (0-10 cm depth) were collected from each part of the land unit. In this research, two series of samples were taken from each section. The first series of samples was used to determine the soil physicochemical properties and the second series was sampled by a special cylinder to determine the soil saturated hydraulic conductivity using the falling head method. In the laboratory, soil texture was determined by hydrometric method (D<0.075), soil granularity by sieve analysis (D>0.075), soil organic matter by wet burning method and neutral lime percentage by Nelson method. The experience of the soil expert of this research was used to determine the soil structure in field. Then, soil erodibility was estimated using four indirect methods, including Weishmeier and Smith nomograph, Torri, Shirazi, and Vaezi. In this study, the soil actual erodibility was measured using rain simulator in the field (in plots of 1 m2). Then the validation of the models was performed using three standard indices MAD, RMSE and MAPE and the most appropriate model was selected.

Results
Soil texture assessment in this watershed revealed that around 90 percent of soil samples have silt loam texture and the rest of samples has loam one. Also, the average lime in study area was estimated about 29%. The high percentage of lime (29%) and silt (60%) in this watershed has greatly increased the tunnel erosion probability. The validation results of the used models revealed that the use of the Vaezi method to estimate the soil erodibility factor in the loess lands of the study area has the highest accuracy (RMSE=0.00225). However, it is less estimated than the actual values. These results were confirmed in all 3 soil types of
Arab-Qara-Haji watershed. After Vaezi model, Shirazi model (RMSE=0.03600) is more suitable than other models to estimate the soil erodibility factor in the loess lands of the area. This model has overestimated in all three types of soil including hill, loess plateau, and alluvial terrace, which is 11.2 times higher than the actual data from the rain simulator in the Arab-Qara-Haji watershed. After the Vaezi and Shirazi models, Wischmeier nomograph and Torri models, respectively, have a weaker performance in determining the soil erodibility factor in the limestone lands of the study area. Like the Shirazi method, Wischmeier nomograph and Tori methods have overestimated the soil erodibility factor. In the study area, this increase has been18.6 and 21.9 times higher than the actual data for Wischmeier nomograph and Torri models, respectively.

Conclusion
The present research results revealed that the use of Vaezi model to estimate soil erodibility in loamy lands with a high percentage of neutral lime is more appropriate compared to other used models. Based on the Vaezi and colleagues model, if the amount of neutral lime in the soil is less than 13%, it has a controlling role on the soil erodibility factor. While, this model is not able to estimate the soil erodibility in areas with neutral lime more than 30%. Therefore, for a number of soil samples in which the neutral lime was above 30%, the Vaezi model was not able to calculate the soil erodibility factor. It seems that currently, for the areas where the percentage of neutral lime in the surface soil is less than 30%, the use of the Vaezi model is more suitable than other models. It is suggested to recalibrate and update the Vaezi model for areas where the percentage of lime is more than 30%.
The present research results revealed that the use of Vaezi model to estimate soil erodibility in loamy lands with a high percentage of neutral lime is more appropriate compared to other used models. Based on the Vaezi and colleagues model, if the amount of neutral lime in the soil is less than 13%, it has a controlling role on the soil erodibility factor. While, this model is not able to estimate the soil erodibility in areas with neutral lime more than 30%. Therefore, for a number of soil samples in which the neutral lime was above 30%, the Vaezi model was not able to calculate the soil erodibility factor. It seems that currently, for the areas where the percentage of neutral lime in the surface soil is less than 30%, the use of the Vaezi model is more suitable than other models. It is suggested to recalibrate and update the Vaezi model for areas where the percentage of lime is more than 30%.