COMPARATIVE ESTIMATION OF THE ACCURACY OF SIMULATION MODELING OF SOIL COVER AND FORECAST OF CARTOGRAMS OF AGRO-INDUSTRIAL GROUPS OF SOILS
DOI:
https://doi.org/10.31861/biosystems2017.02.298Keywords:
soil map, cartogram of agro-industrial groups of soils, training data set, simulation, morphometric parameters, DEM, predicative algorithmsAbstract
The main goal of the mathematical experiment was to compare the accuracy of the construction of predicative maps, depending on the type of input data, in particular the soil map and the complete or abbreviated (without definitions by composition of grain size) variants of the cartograms of agro-industrial soil groups. The tasks were solved: by building a digital relief model (DEM); digitization of cartographic materials; generation of a set of maps of morphometric and other derived characteristics; the analysis of the connections and the role of the mentioned parameters in the variability of the soil cover; creation of predicative map-versions of soils and cartograms of agro-industrial groups of soils. Object of research: a fragment of the territory of the Chernivtsi region with complex geomorphological conditions. Main methods used: correlation analysis; the principal component method; predicative algorithms Decision Trees, Random Forests and K-Nearest Neighbors. On the basis of the correlation analysis, the tightness of the connection and the role of predictors (independent variables) in the variability of the soil cover were assessed, and the analysis of the main components involved the selection of 9 basic ones: absolute altitude; topographic moisture index; the amount of solar radiation per unit area; steepness of slopes; longitudinal and maximum curvature of the topographic surface; accumulation, length and distance to water flow. The quality of predicted cartographic materials was estimated using the Cohen’s kappa coefficient. Differences in the qualitative characteristics of the obtained simulated map-versions are established and it is shown that the morphometric parameters of the relief and its derivatives are a reliable basis for predicative modeling. An extended assessment of the quality of the map-models is made, depending on the type of input data and it is shown that the most accurate predictor cartogram of complete agro-industrial soil groups is used with the set of predictors used. Differences in the quality of predictive soil maps were established by using 3 types of predicative algorithms and it was shown that classification models, in particular, Decision Trees and Random Forests, which allowed obtaining up to 93% of the coincidence of real and model data, were the most suitable for such tasks. The possibilities of constructing forecast maps of soils using a standard set of materials that can be accessed by soil scientists in modern Ukrainian realities are shown: soil and topographic maps in conjunction with free full-featured software - GRASS and Quantum geoinformation systems, Easy Trace vectorizer and R-Statistic, language and environment for statistical computing.
References
A. B. Achasov Dani dystancijnogho zonduvannja jak osnova kartoghrafuvannja gruntiv: ekonomichnyj aspekt. / A. B. Achasov, Gh. V. Titenko, V. I. Kurilov // Visnyk Kharkivsjkogho nacionaljnogho universytetu imeni V. N. Karazina. — Serija: Ekologhija. – 2015. Vyp. 10. – S. 60–66. URL http://journals.uran.ua/visnukkhnu_ecology/article/download/25458/33191
Breiman L. Random forests. / L. Breiman // Machine learning. - 2001. – Vol. 45, № 1. – Р. 5-32. URL https://doi.org/10.1023/A:1010933404324
Browning, D. M. Digital soil mapping in the absence of field training data: A case study using terrain attributes and semiautomated soil signature derivation to distinguish ecological potential / D. M. Browning, M. C. Duniway // Applied and Environmental Soil Science. – 2011. URL https://doi.org/10.1155/2011/421904
Machine learning for predicting soil classes in three semi-arid landscapes / C. W. Brungard, J. L. Boettinger, M. C. Duniway [et al.] // Geoderma. – 2015. – Vol. 239. – P. 68–83. URL https://doi.org/10.1016/j.geoderma.2014.09.019
Bui E. N. A strategy to fill gaps in soil survey over large spatial extents: an example from the Murray– Darling basin of Australia / E. N. Bui, C. J. Moran // Geoderma. – 2003. – Vol. 111 (1). – P. 21–44. URL https://doi.org/10.1016/s0016-7061(02)00238-0
An appropriate data set size for digital soil mapping in Erechim, Rio Grande do Sul, Brazil / A. T. Caten, R. S. D. Dalmolin, F. d. A. Pedron [et al.] // Revista Brasileira de Ciência do Solo. – 2013. – Vol. 37 (2). – P. 359–366. URL https://doi.org/10.1590/s0100-06832013000200007
Cherlinka V. R. Problemy stvorennja, gheorektyfikaciji ta vykorystannja krupnomasshtabnykh cyfrovykh modelej reljjefu / V. R. Cherlinka, Ju. M. Dmytruk // Gheopolytyka y эkogheodynamyka reghyonov. – 2014. – Vol. 10 (1). – P. 239-244. URL http://geopolitika.crimea.edu/arhiv/2014/tom10-v-1/040cherlin.pdf
CherlinkaV. R. Adaptacija velykomasshtabnykh kart gruntiv do jikh praktychnogho vykorystannja u GhIS. In: Aghrokhimija i gruntoznavstvo. Mizhvidomchyj tematychnyj naukovyj zbirnyk. – 2015. – Vyp. 84. TOV «Smughasta typoghrafija», Kharkiv, pp. 20–28. URL http://agrosoil.yolasite.com/resources/2015-AiG-84-pp20-28.pdf
Cherlinka V. Using Geostatistics, DEM and Remote Sensing to Clarify Soil Cover Maps of Ukraine. In: Dent, D., Dmytruk, Y. (Eds.), Soil Science Working for a Living: Applications of soil science to presentday problems. Springer-Verlag GmbH, Cham, Switzerland, 2017. – Ch. 7, pp. 89–100. URL https://link.springer.com/chapter/10.1007/978-3-319-45417-7_7
Cutler A. Random Forests / A. Cutler, D. R. Cutler, J. R. Stevens. – Springer US, Boston, MA, 2012. – pp. 157–175. URL https://doi.org/10.1007/978-1-4419-9326-7_5
Debella-Gilo M. Spatial prediction of soil classes using digital terrain analysis and multinomial logistic regression modeling integrated in GIS / M. DebellaGilo, B. Etzelmüller // Examples from Vestfold County, Norway. Catena. – 2009. – Vol. 77 (1). – P. 8–18. URL https://doi.org/10.1016/j.catena.2008.12.001
Dobos, E., Hengl, T., 2009. Soil mapping applications. In: Hengl, T., Reuter, H. I. (Eds.), Geomorphometry: Concepts, Software, Applications. Vol. 33 of Developments in Soil Science. Elsevier, Amsterdam, Ch. 20, pp. 461–479. URL https://doi.org/10.1016/s0166-2481(08)00020-2
EasyTrace group, 2015. Easy Trace 7.99. Digitizing software. URL http://www.easytrace.com
Feng, C., Michie, D., 1994. Machine learning of rules and trees. Machine learning, neural and statistical classification, 50–83. URL http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.27.355&rep=rep1&type=pdf
Florinsky, I. V., 2016. Digital Terrain Analysis in Soil Science and Geology, 2nd Edition. ACADEMIC PRESS / Elsevier, Amsterdam. URL https://doi.org/10.1016/c2015-0-02363-2
Giasson, E., Figueiredo, S. R., Tornquist, C. G., Clarke, R. T., 2008. Digital soil mapping using logistic regression on terrain parameters for several cological regions in Southern Brazil. In: Hartemink, A. E., McBratney, A. B., de Lourdes MendonçaSantos, M. (Eds.), Digital Soil Mapping with Limited Data. Springer Netherlands, Amsterdam, Ch. 19, pp. 225–232. URL https://doi.org/10.1007/978-1-4020-8592-5_19
GRASS Development Team, 2017. Geographic Resources Analysis Support System (GRASS GIS) Software. Version 7.2. URL http://grass.osgeo.org
Grinand C. Extrapolating regional soil landscapes from an existing soil map: Sampling intensity, validation procedures, and integration of spatial context / C. Grinand, D. Arrouays, B. Laroche, M. P. Martin // Geoderma. – 2008. – Vol. 143 (1). – P. 180–190. URL https://doi.org/10.1016/j.geoderma.2007.11.004
Hengl, T., 2009. A practical guide to geostatistical mapping, 2nd Edition. Office for Official Publications of the European Communities, Luxembourg. URL http://www.academia.edu/download/40396676/A_Practical_Guide_to_Geostatistical_Mapping.pdf
An overview and comparison of machine-learning techniques for classification purposes in digital soil mapping / B. Heung, H. C. Ho, J. Zhang, A. Knudby, C. E. Bulmer, M. G. Schmidt // Geoderma. – 2016. – 265. – P. 62–77. URL https://doi.org/10.1016/j.geoderma.2015.11.014
Heung, B. Comparing the use of training data derived from legacy soil pits and soil survey polygons for mapping soil classes / B. Heung, M. Hodúl, M. G. Schmidt // Geoderma. – 2017. – Vol. 290. – P. 51–68. URL https://doi.org/10.1016/j.geoderma.2016.12.001
Updating the 1:50,000 Dutch soil map using legacy soil data: A multinomial logistic regression approach / B. Kempen, D. J. Brus, G. B. M. Heuvelink [et al.] // Geoderma. – 2009. – Vol. 151 (3). – P. 311–326. URL https://doi.org/10.1016/j.geoderma.2009.04.023
Kuhn M. Building Predictive Models in R Using the caret Package / M. Kuhn // Journal of Statistical Software. – 2008. – Vol. 28 (5). – P. 1-26. URL https://doi.org/10.18637/jss.v028.i05
Lagacherie P. Mapping of reference area representativity using a mathematical soilscape distance / P. Lagacherie, J. M. Robbez-Masson, N. Nguyen-The, J. P. Barthès // Geoderma. – 2001. – Vol. 101 (3-4). – Vol. 105–118. URL https://doi.org/10.1016/s0016-7061(00)00101-4
Landis J. R. The measurement of observer agreement for categorical data / J. R. Landis, G. G. Koch // Biometrics. – 1977. – Vol. 33 (1). – P. 159–174. URL https://doi.org/10.2307/2529310
Li W. A Random-Path Markov Chain Algorithm for Simulating Categorical Soil Variables from Random Point Samples / C. Zhang, W. Li // Soil Science Society of America Journal. – 2007. – Vol. 71 (3). – P. 656–668. URL https://doi.org/10.2136/sssaj2006.0173
Liu, B., 2011. Web Data Mining: Exploring Hyperlinks, Contents and Usage Data, 2nd Edition. Springer-Verlag GmbH, London New York Dordrecht. URL https://doi.org/10.1007/978-3-642-19460-3
MacMillan, R. A., 2008. Experiences with applied DSM: protocol, availability, quality and capacity building. In: Hartemink, A. E., McBratney, A. B., de Lourdes Mendonça-Santos, M. (Eds.), Digital Soil Mapping with Limited Data. Springer Netherlands, Amsterdam, pp. 113–135. URL https://doi.org/10.1007/978-1-4020-8592-5_10
Malone, B. P., Minasny, B., McBratney, A. B., 2016. Using R for Digital Soil Mapping. Progress in Soil Science. Springer International Publishing. URL https://doi.org/10.1007/978-3-319-44327-0
McBratney A. B. On digital soil mapping / A. B. McBratney, M. L. M. Santos, B. Minasny // Geoderma. – 2003. – Vol. 117 (1-2). – P. 3-52. URL https://doi.org/10.1016/s0016-7061(03)00223-4
Poljchyna S. M. Zastosuvannja suchasnoji systemy klasyfikaciji gruntiv FAO/WRB do karty gruntovogho pokryvu Chernivecjkoji oblasti / S. M. Poljchyna, V. A. Nikorych, O. A. Danchu // Gruntoznavstvo. –2004. – Vol. 5 (1–2),. – P. 27–33. URL http://arr.chnu.edu.ua/jspui/bitstream/123456789/471/1/Nikorich.pdf
Postanova Prezydiji Nacionaljnoji akademiji ..., 2017. Orghanizacijna struktura, porjadok formuvannja ta funkcionuvannja Gruntovo-informacijnogho centru Ukrajiny. Postanova Prezydiji Nacionaljnoji akademiji aghrarnykh nauk Ukrajiny. 20.09.2017 r. Protokol #13. URL http://issar.com.ua/downloads/postanova_vid_20_veresnya_2017_protokol_no13_organizaciyna_struktura_poryadok_formuvannya_gic.pdf
QGIS Development Team, 2015. QGIS Geographic Information System. URL http://qgis.osgeo.org
R Development Core Team, 2017. R: A language and environment for statistical computing. R Foundation for Statistical Computing. URL http://www.rproject.org
Scull P. Predictive soil mapping: a review / P. Scull, J. Franklin, O. A. Chadwick, D. McArthur // Progress in Physical Geography. - 2003. – Vol. 27 (2). – P. 171–197. URL https://doi.org/10.1191/0309133303pp366ra
Venables, W. N., Ripley, B. D. 2002. Modern Applied Statistics with S, 4th Edition. Vol. 53 (1) of Statistics and Computing. Springer-Verlag, New York. URL http://dx.doi.org/10.1007/978-0-387-21706-2
Walter, C., Lagacherie, P., Follain, S., 2006. Integrating pedological knowledge into digital soil mapping. In: Lagacherie, P., McBratney, A. B., Voltz, M. (Eds.), Digital Soil Mapping: An Introductory Perspective. Vol. 31 of Developments in Soil Science. Elsevier, Amsterdam, Ch. 22, pp. 281–301. URL https://doi.org/10.1016/s0166-2481(06)31022-7
