پي ريزي توابع انتقالي طيفي نقطه اي در برآورد فرسايش پذيري خاك در گستره VIS-NIR-SWIR

Developing Point Spectro Transfer Functions in Soil Erodibility Prediction in VIS-NIR-SWIR Rang

استفاده از انعكاس¬ طیفی خاك در دامنه 350 تا 2500 نانومتر و توابع طیفی حاصل از آن به عنوان روشی سریع، كم¬مخرب و تاحدی كم¬هزینه در برآورد ویژگی¬های دیریافت خاك مرسوم شده است، ولیكن تاكنون از این روش در برآورد فرسایش¬پذیری استفاده نشده است. لذا هدف از این مطالعه، پی¬ریزی توابع انتقالی طیفی نقطه¬ای خاك و مقایسه آن با توابع انتقالی خاك و رابطه¬ی ویشمایر و اسمیت (1978) در برآورد فرسایش¬پذیری است. برای این منظور فرسایش¬پذیری با استفاده از 40 كرت استاندارد در بالادست سد سیوند و با استفاده از باران طبیعی و منحنی‏های بازتاب طیفی با استفاده از دستگاه اسپكترومتر زمینی و در شرایط نور طبیعی اندازه¬گیری شد. نتایج نشان داد میانگین فرسایش¬پذیری اندازه¬گیری شده (mm-1 t h MJ−1 014/0) حدود 18/2 برابر كمتر از میانگین فرسایش‏پذیری برآورد شده حاصل از رابطه¬ی ویشمایر و اسیمت (mm-1t h MJ−1 030/0) بود. بر خلاف رابطه¬ی ویشمایر و اسمیت كه در آن كربنات كلسیم معادل در نظر گرفته نشده است، در تابع انتقالی پی¬ریزی شده این ویژگی به عنوان متغیر موثر وارد مدل شد. با بررسی همبستگی بین فرسایش¬پذیری و بازتاب¬های طیفی، طیف¬های مرئی (532 ،622)، مادون قرمز كوتاه (14422227، 2327 و 2343 نانومتر) جهت پی¬ریزی توابع انتقالی طیفی¬ نقطه‏ای انتخاب شدند. بر اساس آماره¬های ارزیابی R2 ، RMSE و ME، توابع انتقالی خاك كارآیی بالاتری نسبت به توابع انتقالی طیفی¬ و رابطه¬ی ویشمایر و اسمیت در برآورد فرسایش¬پذیری داشت. تابع انتقالی طیفی نقطه¬ای با داشتن مقداری اریبی در تخمین¬ها كارآیی بالاتری نسبت به رابطه¬¬ی ویشمایر و اسمیت در برآورد فرسایش¬پذیری داشت.

Introduction: Soil erodibility (K factor) is generally considered as soil sensitivity to erosion and is highly affected by different climatic, physical, hydrological, chemical, mineralogical and biological properties. This factor can be directly determined as the mean rate of soil loss from standard plots divided by erosivity factor. Since measuring the erodibility factor in the field especially watershed scale is time-consuming and costly, this factor is commonly estimated by pedotransfer functions (PTFs) using readily available soil properties. Wischmeier and Smith (1978) developed an equation using multiple linear regressions (MLR) to estimate erodibility factor of the USA using some readily available soil properties. This equation has been used to estimate K based on soil properties in many studies. As using PTFs in large sales is limited due to cost and time of collecting samples, recently soil spectroscopy technique has been widely used to predict certain soil properties using Point SpectroTransfer Functions (PSTFs). PSTFs use the correlation between soil spectra in Vis-NIR (350-2500 nm) and certain soil properties. The objective of this study was to develop PSTFs and PTFs for soil erodibility factor prediction in the Simakan watershed Fars, Iran. Materials and Methods: The Semikan watershed, which mainly has calcareous soil with more than 40% lime (total carbonates), is located in the central of Fars province, between 30°06 '-30°18 'N and 53°05 '-53°18 'E (WGS′ 1984, zone 39°N) with an area of about 350 km2. For this study, 40 standard plots, which are 22.1×1.83 m with a uniform ploughed slope of 9% in the upslope/downslope direction, were installed in the slopes of 8-10% and the deposit of each plot was collected after rainfall. From each plot three samples were sampled and some physicochemical properties including soil texture, organic matter, water aggregate stability, soil permeability, pH, EC were analyzed Spectra of the air-dried and sieved soil samples were recorded in the Vis-NIR-SWIR (350 to 2500 nm) range at 1.4- to 2-nm sampling intervals in a standard and controlled dark laboratory environment using a portable spectroradiometer apparatus (FieldSpec 3, Analytical Spectral Device, ASD Inc.). Some bands which had the highest correlation with K factor were chosen as input parameter for developing PSTFs. A stepwise multiple linear regression method was used for developing PTFs and SPTFs. R2, RMSE and ME were used for comparing PTFs and SPTFs. Results and Discussion: The K values varied from 0.005 to 0.023 t h MJ−1 mm−1 with an average standard deviation of 0.014 and of 0.003 t h MJ−1 mm−1, respectively. The K estimated by Wischmeier and Smith (1978) equation varied from 0.015 to 0.045 t h MJ−1 mm−1 with an average of 0.030 t h MJ−1 mm−1. There was a significant difference (p <0.001) between measured soil erodibility factor and those estimated based on Wischmeier and Smith (1978) in the studied area. A comparison between measured and estimated K values revealed that the measured soil erodibility factor values were from 1.08 to 3.57 with average 2.18 times smaller than the estimated values. The K had positive significant correlation with silt content (r= 0.47, p <0.01) and very fine sand content (r=0.43, p <0.01). The results indicated that CaCO3 had negative effect on the K factor because Ca2+ affects flocculation and aggregate stability, and hence decreases erodibility factor. This parameter had, therefore, a significant coefficient in developed PTFs, while it was not considered as input parameter in Wischmeier and Smith (1978) equation. Based on correlation between band reflectance and K factor some band including B532, B622, B1442, B2227, B2327 and B2343 were selected for developing PSTFs. PTFs with R2= 0.84, RMSE= 0.0014 t h MJ−1 mm−1 and ME= 0.000 t h MJ−1 mm−1 were the best method to predict K. After PTFs, SPTFs with R2= 0.53, RMSE= 0.0028 t h MJ−1 mm−1 and ME= 0.0011 t h MJ−1 mm−1 were the second best method to estimate K. Conclusions: The results showed that the annual average of soil loss was 7.90 t h-1 ya-1 and measured K factor was 0.014 t h MJ−1 mm−1. Organic matter (r=-0.60) and permeability (r= -0.77) had high significant correlation with the K factor. Although the content of lime was not considered in Wischmeier-Smith and RUSLE model, we found that this soil property decreased K significantly due to its strong effects on aggregate stability and soil permeability. Overall, PTFs and SPTFs had better accuracy than Wischmeier-Smith function to predict K factor. Wischmeier-Smith function showed an overestimation to predict K factor particularly for higher values of K.