Comparison of Geostatistical and Random Forest Methods in Mapping Soil Salinity in the Lands of Zahak County of Sistan Plain

Authors

1 Assistant Professor, Soil and Water Research Department, Sistan Agricultural and Natural Resources Research and Education Center, AREEO, Zabol, Iran

2 MSc., Soil and Water Research Department, Sistan Agricultural and Natural Resources Research and Education Center, AREEO, Zabol, Iran

3 Assistant Professor, Soil and Water Research Institute, Agricultural Research, Education and Extension Organization(AREEO), Karaj, Iran

Abstract

Soil maps are the major source of information for land management, natural resources, and environment. Soil salinity is one of the most important factors limiting crop production in Sistan plain. In this research, geostatistical and random forest methods were compared to produce soil salinity maps. Preparation of accurate maps for soil salinity conditions is of great help in proper management of lands in this area. For this purpose, 460 composite soil samples were collected from 0-30 cm depth in 41000 ha of Zahak region of Sistan plain, using 750 m grid network. Then, the electrical conductivity of saturation paste was measured. Afterwards, 361 samples were used for training and 99 for testing. In the geostatistical model, different semi-variogram including circular, spherical, exponential, and Gaussian and different interpolation methods including inverse distance weighting, ordinary kriging, simple kriging, universal kriging and co-kriging were fitted and the best models were selected. In random forest model, digital soil mapping technique was used and environmental covariates were derived from digital elevation model (DEM) map and A Landsat 8 ETM+ image. The results of geostatistical method showed that soil salinity had a medium spatial correlation in the study area and the best semi-variogram and interpolation model were spherical and ordinary kriging, respectively. In random forest model, aspect, NDVI, and NDSI were the most important covariate in predicting soil salinity. The results Root Mean Square Error and Mean Error for the training and testing data showed that random forest method was slightly better than geostatistics. The use of other covariates such as land use and soil series maps can increase accuracy of the maps based on random forest methods, thereby improving decision-making. Using other environmental covariates that were not used in this study such as land unit and soil series map can also improve the accuracy of the map.

Keywords


  1. پهلوان راد، م.ر.، خرمالی، ف.، تومانیان، ن.، کیانی، ف.، و کمکی، چ.ب.پ.، 1393. پهنه­بندی رقومی واحدهای خاک با استفاده از مدل درختان تصمیم­گیری تصادفی در استان گلستان. پژوهش­های حفاظت آب و خاک. جلد 21 . شماره 6: 73-93.
  2.   تقی­زاده مهرجردی، ر.ا.، سرمدیان، ف.، روستا، م.ج.، رحیمیان، م.ح.، امید، م.، و تومانیان، ن. 1393.پهنه­بندی رقومی قابلیت هدایت الکتریکی ظاهری با استفاده از رگرسیون کریجینگ و واریوگرام محلی در منطقه اردکان. مدیریت خاک و تولید پایدار. جلد چهار. 29:1.
  3. دیانی، م.، جعفری، س.، خلیل مقدم.، ب.، دهقانی، ا. 1391. پهنه­بندی خطر شوری و سدیمی شدن خاک سطحی با استفاده از زمین­آمار (مطالعه موردی: اراضی غرب رودخانه کارون در استان خوزستان). پژوهش و سازندگی. شماره 94: 86-95. 
  4. سکوتی اسکوئی، ر؛ مهدیان، م؛ محمودی، ش، 1386 ،مقایسه کارایی برخی روشهای زمین آماری برای پیش­ینی پراکنش مکانی شوری خاک، مطالعه موردی دشت ارومیه، مجله پژوهش و سازندگی، شماره74 :90-98.
  5. صادقی، س.، کاظمی پشت مساری، ح.، طهماسبی سروستانی. 1393. ارزیابی روش­های درونیابی جهت تعیین پراکندگی برخی عناصر کم­مصرف در اراضی کشاورزی استان گلستان. مدیریت خاک و تولید پایدار.دوره 4 شماره 3: 323-337.
  6. کاظمی پشت مساری، ح.، طهماسبی سروستانی، ز.ا.، بهنام، ک.، شتایی4، ش.، صادقی، س. 1391. ارزیابی روشهای زمین آمار جهت تخمین وپهنه بندی عناصر غذایی پرمصرف اولیه در برخی اراضی کشاورزی استان گلستان. دانش آب و خاک. جلد 22 شماره 1: 201-219.
  7. محمدی، ع. ۱۳۸۵. پدومتری (آمار مکانی)، انتشارات پلک. ۴۵۳ صفحه.
    1. Akramkhanov, A., Martius, C., Park, S.J., and Hendrickx, J.M. 2011. Environmental factors of spatial distribution of soil salinity on flat irrigated terrain. Geoderma. 163: 55-62.
    2. Breiman, L., and Cutler, A. 2004. Random Forests. Department of Statistics, University of  Berkeley. http://www.stat.berkeley.edu/~breiman/RandomForests/cc_home.htm
    3. Breiman, L., Friedman, J.H., Olshen, R.A., and Stone, C.J. 1984. Classification and Regression Trees. Chapman and Hall, New York. 368 pages.
    4. Brungard, C.W. 2009. Alternative Sampling and Analysis Methods for Digital Soil Mapping in Southwestern Utah. Thesis for Master of Science. Utah State University. USA.
    5. Dehni, A., and Lounis, M. 2012. Remote sensing techniques for salt affected soil mapping: application to the Oran region of Algeria. Pro. Engi. 33: 188-198.
    6. Grimm, R., Behrens, T., Marker, M., and Elsenbeer, H. 2008. Soil organic carbon concentrations and stocks on Barro Colorado Island - Digital soil mapping using random forests analysis. Geoderma. 146(1-2):102-113.
    7. Hengl, T., Heuvelink, B. M., Kempen, B., Leenaars, J.G. B., Walsh., M. G., Shepherd, K.D., Sila, A., MacMillan, R.A., Jesus, J. M., Tamene, L., and Tondoh, J.E. 2015. Mapping soil properties of Africa at 250 m resolution: random forests significantly improve current predictions. PLOS ONE 10 (6): e0125814.  doi:10.1371/journal.pone.0125814.
    8. Isaaks, H.E. and Srivastava, R.M. 1989. An introduction to applied geostatistics. Oxford University Press, NY.
    9. Jenny, H. 1941. Factors of Soil Formation: A System of Quantitative Pedology. McGraw-Hill, New York.
    10. Kempen, B. 2013. Comparison of universal kriging and regression tree modeling for soil property mapping. Geophysical Research Abstracts. Vol. 15, EGU2013-6899.
    11. Ließ, M., Glaser, B., and Huwe, B. 2012. Uncertainty in the spatial prediction of soil texture comparison of regression tree and random forest models. Groderma. 170: 70-79.
    12. McBratney, A.B., Mendonça Santos, M.L., and Minasny, B. 2003. On digital soil mapping. Geoderma. 117(1–2):3-52.
    13. Pahlavan Rad, M.R., Khormali, F., Toomanian, N, Kiani, F and Komaki, C.B. 2015. Digital soil mapping using Random Forest model in Golestan province. Journal of Water and Soil Conservation. 21(6): 73-93.
    14. Pahlavan Rad, M.R., Toomanian, N., Khormali, F., Brungard , C.W., Komaki, B and Bogaert. P., 2014. Mapping and Updating Soil Series Using Random Forest and Conditioned Latin Hypercube Sampling in the Loess Soils of Northern Iran. Geoderma. (232-234): 97-106
    15. R Development Core Team. 2013. R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria.
    16. Rivero, R.G., Grunwald, S., and Bruland, G.L. 2007. Incorporation of spectral data into multivariate geostatistical models to map soil phosphorus variability in a Florida wetland. Geoderma. 140: 428– 443.
    17. Taghizadeh-Mehrjardi, R., Sarmadian, F., Rousta, M.J. Rahimian, M.H., Omid, M. and Toomanian N. 2015. Digital mapping of apparent electrical conductivity using regression kriging and local variogram in Ardakan region. Journal of Soil Management and Sustainable. 4(4): 1-29.
    18. Utest, A., Lopez, T. and Diaz L. 2002. A comparison of soil maps, kriging and a combined method for spatially predicting bulk density and field capacity of Ferralsols in the Havana-Matanzas Plain. Geoderma. 96: 199-213.
    19. Uyan, M. and Cay, T. 2010. Geostatistical methods for mapping groundwater nitrate concentrations. 3rd International conference on cartography and GIS, 15-20 June, Nessebar, Bulgaria.
    20. Were, K., Bui, D. T., Disk, B., Singl, B. R.  2015. Acomparative assessment of support vector regression, arttifical nenural networks, and random forest for peredicting soil organic carbon stocks across an afromonkane land scape. Ecological indicator. 394-403
    21. Wilson, J.P., and Gallant, J.C. 2000. Terrain Analysis: Principles and Applications. In: G.J. Wilson JP (Ed.), Digital terrain analysis. John Wiley, New York. 478 pages.
    22. Yang, L. I., Zho, S. H., Fang, W. U., Hong, Y. L. I., and  Feng, L. I. 2005. Improved prediction of sampling density for soil salinity by different geostatistical methods. Agricultural Sciences in China. 6 (7): 632 – 841.
    23. Zhu, A.X., Hudson, B., Burt, J.E., and Lubich K. 2001. Soil mapping using GIS, expert knowledge and fuzzy logic. Soil Science Society of America Journal. 65:1463-1472.