برآورد هدایت هیدرولیکی اشباع خاک‌های منتخب از دشت اردبیل با استفاده از مدل‌های رگرسیونی و شبکه‌های عصبی مصنوعی

نوع مقاله: مقاله پژوهشی

نویسندگان

1 دانش­ آموخته کارشناسی ارشد، گروه علوم و مهندسی خاک، دانشکده کشاورزی و منابع طبیعی، دانشگاه محقق اردبیلی

2 دانشیار گروه علوم و مهندسی خاک، دانشکده کشاورزی و منابع طبیعی، دانشگاه محقق اردبیلی

3 دانشیار گروه مهندسی بیوسیستم، دانشکده کشاورزی و منابع طبیعی، دانشگاه محقق اردبیلی

4 دانش‌آموخته کارشناس ارشد علوم خاک، دانشگاه ارومیه

5 دانش‌آموخته کارشناس ارشد گروه آب، دانشگاه ارومیه

چکیده

هدایت هیدرولیکی اشباع به­عنوان یک ویژگی دیریافت می­تواند از ویژ­گی­های زودیافت خاک شامل جرم ویژه ظاهری، بافت خاک، کربن آلی، کربنات کلسیم معادل با استفاده از توابع انتقالی رگرسیونی و شبکه­های عصبی مصنوعی برآورد شود. هدایت هیدرولیکی اشباع خاک به روش بار افتان در 100 نمونه خاک جمع­آوری شده از دشت اردبیل تعیین شد. بعد از انجام تجزیه­های شیمیایی و فیزیکی روی نمونه­های خاک، داده­ها به دو سری داده­های آموزشی (80 نمونه) و داده­های اعتبارسنجی (20 نمونه) تقسیم شدند. مدل­های رگرسیونی توسط نرم­افزار SPSS و به روش گام­به­گام و مدل­های شبکه عصبی توسط نرم­افزارNeurosolution  شکل گرفتند. برای انجام تجزیه­های آماری از ضریب تبیین (R2)، جذر میانگین مربعات خطا (RMSE) و ضریب آکائیک (AIC) استفاده شد. بهترین مدل رگرسیونی دارای متغیرهای شن، سیلت و جرم مخصوص ظاهری بود و بهترین مدل شبکه عصبی از متغیرهای ورودی میانگین هندسی قطر ذرات خاک، انحراف معیار هندسی قطر ذرات خاک و جرم مخصوص ظاهری به­دست آمد. مقادیر R2، (cm min-1)RMSE در فاز آموزش و اعتبارسنجی برای بهترین مدل­ رگرسیونی به­ترتیب برابر (53/0، 074/0 و 51/0، 052/0) و برای بهترین مدل شبکه عصبی به­ترتیب برابر (84/0، 04/0 و 73/0، 06/0) بود. در این پژوهش به­صورت جداگانه از تمامی پارامترهای مستقل شامل جرم مخصوص ظاهری، جرم مخصوص حقیقی، درصد آهک، میانگین هندسی قطر و انحراف معیار هندسی قطر ذرات خاک به­عنوان ورودی در تکنیک شبکه عصبی استفاده شد. مقادیر R2 و (cm min-1) RMSE در مرحله آموزش و آزمون به­ترتیب برابر (87/0، 036/0 و 58/0، 076/0) بود. نتایج تحقیق در این مورد نشان داد شبکه­­های عصبی با داده­های ورودی یکسان هدایت هیدرولیکی اشباع خاک را با دقت بیشتری (84/0=R2) نسبت به مدل­های رگرسیونی (53/0=R2) برآورد می­کنند. همچنین مشاهده شد زمانی که تعداد داده­های ورودی در روش شبکه عصبی افزایش می­یابد دقت برآورد در داده­های آموزشی بیشتر می­شود.

کلیدواژه‌ها


عنوان مقاله [English]

Estimating the Soil Saturated Hydraulic Conductivity in Ardabil Plain Soils Using Artificial Neural Networks and Regression Models

نویسندگان [English]

  • hamed amirabedi 1
  • shorollahi asghari 2
  • tarahom mesri 3
  • Naser Balandeh 4
  • ebrahim johari 5
1 MSc. Graduate, Department of Soil Science and Engineering, Faculty of Agriculture and Natural Resources, University of Mohaghegh Ardabili, Ardabil, Iran
2 Associate Professor, Department of Soil Science and Engineering, Faculty of Agriculture and Natural Resources, University of Mohaghegh Ardabili, Ardabil, Iran.
3 3Associate Professor, Department of Biosystem Engineering, Faculty of Agriculture and Natural Resources, University of Mohaghegh Ardabili, Ardabil, Iran
4 4MSc. Graduate, Department of Soil Science, Faculty of Agriculture and Natural Resources, Urmia University, Urmia, Iran
5 5MSc. Graduate, Department of Water Engineering, Faculty of Agriculture and Natural Resources, Urmia University, Urmia, Iran
چکیده [English]

Abstract
Soil saturated hydraulic conductivity (Ks) can be estimated from surrogate data such as soil texture, bulk density and organic carbon and CaCO3 contents using regression (Reg-PTFs) and artificial neural networks (ANN-PTFs) pedotransfer function (PTF). Saturated hydraulic conductivity was measured by falling head method in 100 soil samples that obtained from Ardabil plain, Iran. After performing physicals and chemicals analysis on soil samples, the data were divided into two sets of training (80 samples) and validation data (20 samples). Regression models were created by SPSS software, stepwise method and neural networks models were created by Neurosolution software. Statistics criteria such as coefficient of determination (R2), root mean square deviation (RMSE) and Akaike information Criterion (AIC) were determined. Input variables in the best regression models were sand, silt and bulk density. The best neural network models were obtained from the input variables that include bulk density, geometric mean and standard deviation of soil particle size distribution. The values for R2 and RMSE in training and testing data set for the Reg-PTF were 0.53, 0.074 and 0.51, 0.052 and for the ANN-PTF they were 0.84, 0.04 and 0.73, 0.06, respectively. In this research all independent variables such as bulk density, particle density, CaCO3, geometric mean and standard deviation of the particle size distribution included as inputs for development of Reg-PTFs and ANN-PTFs. The amount of R2 and RMSE for training and testing data set equal 0.87, 0.036 and 0.58, 0.076, respectively. Results showed that the ANN-PTF (R2= 0.84) performs better than the Reg-PTF (R2= 0.53) in this case. It was also found that when all independent variables were used as inputs in the neural ANN-PTF the values of R2 and RMSE (0.87 and 0.036) have been improved in the training data set.

کلیدواژه‌ها [English]

  • Hydraulic conductivity
  • neural networks
  • Pedotransfer function
Reference

Alijanpour Shalmani A., Shabanpour M., Asadi H., and Bagheri F. 2011. Estimation of soil aggregate stability in forest`s soils of Guilan Province by artificial neural networks and regression pedotransfer functions. Journal of Soil and Water,21 (3):153-162. (in Persian)

Amirabedi H., Asghari Sh., Mesri T., and Balandeh N. 2016. Prediction of mean weight diameter of aggregates using artificial neural network and regression models. Applied Soil Research, 4(1): 39-53. (in Persian)

Amirabedi H., Asghari Sh., Mesri T., and Keivan behjo F. 2013.Estimating of field capacity, permanent wilting point and available water content in Ardabil plain using regression and artificial neural network models. Applied Soil Research, 1(1):60-72. (in Persian).

Barzegar A.R. 2008. Advanced Soil Physics. Shahid Chamran University Press.309 pp.

Boadu, F. K. 2000. Hydraulic conductivity of soils from grain-size distribution: new models. Journal of Geotechnical and Geoenvironmental Engineering.126(8), 739-746.

Brakensiek D.L., Rawls W.J., and Stephenson G.R. 1984. Modifying SCS hydrologic soil groups and curve numbers for rangeland soils.ASAE Paper (PNR-84203)

Campbell G.S. 1985. Soil Physics with Basic: Transport Models for Soil–Plant System. Elsevier, New York. 150 pp.

Doai M., Shabanpaour Shahrestani M., Bagheri F. 2005. Modelling of saturated hydraulic conductivity of Gilan province artificial neural networks. The Agriculture Science Research Report, 1(6): 41-48. (in Persian)

Ghanbarian-Alavigeh B., Liaghat A., and Sohrabi S. 2009. Application of artificial neural networks in prediction of saturated hydraulic conductivity using soil physical parameters. Journal of Agricultural Engineering Research, 10(1):1. 97-112.

Ghorbani Dashtaki Sh., Homaee M., and Mahdian M. 2009. Estimating soil water infiltration parameters using Artificial Neural Networks. Journal of Water and Soil, 23(1):185-198. (in Persian)

Hassan Shah A., Lone M., Stephen I., and Anderson H. 1997. Regression model to predict hydraulic conductivity from simple soil physical and chemical properties. 7th ICID international drainage workshop.Malaysia.

Jabro J.A., 1992. Estimation of Saturated Hydraulic Conductivity of Soils from Particle Size Distribution and Bulk Density Data. Trans ASAE. 35:557-560.

Jain A., and Kumar A. 2006. An evaluation of artificial neural network technique for the determination of infiltration model parameters. B.V. Amsterdam, Netherlands. Pp. 272-282..

Kadu P.R., Vaidya P.H., Balpande S.S., Satyavathi P.L.A., and Pal D.K. 2003. Use of hydraulic conductivity to evaluate the suitability of vertisols for deep-rooted crops in semiarid parts of central india. Soil Use and Management,19: 208-216.

Kelishadi1 H., Mosaddeghi M.R., Hajabbasi M.A., and Ayoubi S. 2013. Evaluating and developing pedotransfer functions to predict soil saturated hydraulic conductivity at landscape scale in Central Zagros. Applied Soil Research, 1(2): 16-33.

Khalilmoghadam B., Afyuni M., Abbaspour K.C., Jalalian A., Dehgani A., and Schulin R., 2009. Estimation of surface shear strength in Zagros Region of Iran- A comparison of artificial neural networks and multiple- linear regression models. Geoderma, 153: 29-36.

Khashei Siuki A., Jalali Moakhar V., Noferesti A., and Ramazani Y. 2015. Comparing nonparametric k-nearest neighbor technique with ANN models for predicting soil saturated hydraulic conductivity. Journal of Soil Management and Sustainable, 5(3): 81-95.

Khodaverdiloo H., Homaee M.Th., Van Genuchten M., and Ghorbani Dashtaki Sh. 2011. Deriving and validating pedotransfer functions for some calcareous soils. Journal of Hydrology, 399: 93-99.

Klute A., and Dirksen C. 1986. Hydraulic conductivity of saturated soils. In methods of soil analysis. Soil Science Society of America, Madison, Wisconsin, USA. (pp. 694-700).

Klute A., 1986. Methods of Soil Analysis. Part 1. Physical and Mineralogical Methods. 2 nd edition. Agron. Monog. 9. ASA and SSSA, Madison, WI.

Merdun H., Cinar O., Meral R., and Apan M. 2006. Comparison of artificial neural network and regression pedotransfer functions for prediction of soil water retention and saturated hydraulic conductivity. Soil & Tillage Research, 90: 108–116.

Mesri Gundoshmian T. 2009. The use of intelligent systems to optimize drop combine. PhD thesis, Engineering of Agriculture Machinery and Mechanization Department, University of Tabriz. 119 p. (in Persian)

Minasny B., and Mcbartney A.B. 2002. The neuron method for fitting neural network parametric pedotransfer functions. Soil Science Society of America Journal, 66: 352-361.

Mualem Y., 1976. A new model for predicting the hydraulic conductivity of unsaturated porous media. Water Resource Research, 12, 593–622.

Nosrati Karizak1 F., Movahedi Naeni S.A., Hezarjaribi A., Roshani Gh.A., and Dehghani A.A. 2012. Using artificial neural networks to estimate saturated hydraulic conductivity from easily available soil properties. Journal of Soil Management and Sustainable Production, 2(1). 95-110. (in Persian)

Pachepsky Y.A., Timlin D.J., and Varallyay G. 1996. Artificial neural networks to estimate soil water retention from easily measurable data. Soil Science Society of America Journal. 60: 727–773.

Page A.L. (ed.).1985. Methods of Soil Analysis. Part 2. Chemical and Microbiological Methods. Agronomy No. 9. American Society of Agronomy, Madison, WI.

Salarashayeri A.F., and Siosemarde M. 2012. Prediction of soil hydraulic conductivity from particle-size distribution. International Journal of Environmental, Chemical, Ecological, Geological and Geophysical Engineering, 6(1): 16-20.

Sarmadian F., Ghanbarian Alavijeh B., Taghizadeh Mehrjardi R., and Keshavarzi A. 2011. Comparison of linear and nonlinear pedotransfer functions with artificial neural networks in prediction of surface fractal dimension. Journal of Range and Watershed Management, 64(1): 53-64.

Saxton K. E., Rawls W. J., Romberger J.S., and Papendick R.I. 1986. Estimating generalized soil water characteristics from texture. Soil Science Society of American Journal, 50: 1031–1036.

Schaap M.G., and Bouten W. 1996. Modeling water retention ccurves of sandy soils using neural networks. Water Resource Research, 32: 3033–3040.

Van Genuchten M.Th., 1980. A closed-form equation for predicting the hydraulic conductivity of unsaturated soils. Soil Science Society of American Journal, 44: 892–898.

Walkley A.J., and Black I.A. 1934. An examination of degtjareff method for determining soil organic matter and a proposed modification of the chromic acid titration method. Soil Science, 37: 29–38.

Warrick A.W. 2002. Soil Physics Companion. CRC Press. 389 p.

Wösten J.H.M., Lilly A., Nemes A., and Le Bas C. 1999. Development and use of a database of hydraulic properties of European soils. Geoderma, 90: 169-185.

Wösten J.H.M., Pachepsky Ya.A., and Rawls W.J. 2001. Pedotransfer functions: bridging the gap between available basic soil data and missing soil hydraulic characteristics. Journal of Hydrology, 251:123-150.

Yilmaz I., and Yuksek G. 2009. Prediction of the strength and elasticity modulus gypsum using multiple regression, ANN, and ANFIS models. International Journal of Rock Mechanics and Mining Sciences, 46: 803-810.

Zarinfar S., Ghahraman B., and Davary N. 2011. Development of some pedotransfer functions to predict the saturated hydraulic conductivity of gravel soils using partial least square regression method.Journal of Water and Soil, 25(3): 617-624.