FORECASTING THE CONTENT OF ABNORMALLY DISTRIBUTED CHROME IN SOIL BY HYBRID MODELS BASED ON ARTIFICIAL NEURAL NETWORKS

A.V. Shichkin, A.G. Buevich, A.P. Sergeev, E.M. Baglaeva, I.E. Subbotina

Institute of Industrial Ecology Ural Branch of RAS, Ekaterinburg, Russia, ul. S. Kovalevskoi 20, Yekaterinburg, 620219 Russia

The work is devoted to the use of a hybrid model combining artificial neural networks (ANNs) and kriging to predict an anomalously distributed chromium (Cr). It is known that the combination of geostatistical interpolation approaches (kriging) and neural networks in the model leads to a better accuracy in forecasting and performance. Generalized regression neural networks (GRNN) and multi- layer perceptron (MLP) are classes of neural networks widely used for modeling in environmental studies. In this paper, we have compared two neural networks, i.e., GRNN and MLP, as well as two combined methods: GRNN with residual scoring (GRNNRK) and MLP with residual killing (MLPRK). The study is based on the actual data sets on surface contamination of soil in Novy Urengoy, Russia, with chromium obtained from earlier perfromed screening. The network structures were chosen during the computer modeling based on the minimization of RMSE. MLP and MLPRK showed the best prognostic accuracy in comparison with kriging, GRNN and GRNNRK.

Key words: hybrid models, artificial neural networks, chrome, residual kriging.

REFERENCES

1. Saet, Yu.E., Revich, B.A., Yanin, E.P. Geokhimiya okru- zhayushchei sredy [Environment geochemistry], Moscow, Nedra Publ., 1990, pp. 84–108 (in Russian).

2. Anagu, I., Ingwersen, J., Utermann, J., Streck, T. Esti- mation of heavy metal sorption in German soils using artificial neural networks. Geoderma, 2009, 152, pp.104– 112.

3. Chukanov, V.N., Sergeev, A.P., Ovchinnikov, S.M., Medvedev, A.N. Diagnostics of snow-cover contamina- tion with soluble and insoluble metal impurities. Russian Journal of Nondestructive Testing, 2006, no. 42, pp. 630– 636.

4. Dai, F., Zhou, O., Lva, Z., Wang, X., Liu, G. Spatial prediction of soil organic matter content integrating ar- tificial neural network and ordinary kriging in Tibetan Plateau. Ecological indicators, 2014, no. 45, pp. 184–194.

5. Demyanov, V., Kanevsky, M., Chernov, S., Savelieva, E., Timonin, V. Neural Network Residual Kriging Applica- tion for Climatic Data. Journal of Geographic Information and Decision Analysis, 1998, no. 2, pp. 215–232.

6. Falamaki, A. Artificial neural network application for predicting soil distribution coefficient of nickel. Journal of Environmental Radioactivity, 2013, no. 115, pp. 6–12.

7. Goovaerts, P. Geostatistics in soil science: state of the art and perspectives. Geoderma, 1999, no. 89, pp.1–45.

8. Guo, G.H., Wu, F., Xie, F., & Zhang, R. Spatial distri- bution and pollution assessment of heavy metals in ur- ban soils from southwest China. Journal of Environmen- tal Sciences, 2012, vol. 24, issue 3, pp. 410–418.

9. Kanevski, M., Arutyunyan,R.,Bolshov,L.,Demyanov,V., Maignan, M. Artificial neural networks and spatial es- timations of Chernobyl fallout. Geoinformatics, 1995, no. 7(1–2), pp. 5–11.

10. Kanevski, M. Spatial predictions of soil contamination using general regression neural networks. Internation- al Journal of Systems Research and Information Systems, 1999, vol. 8, issue 4. pp. 241–256.

11. Kanevski, M., Pozdnoukhov, A., Timonin, V. Machine Learning for Spatial Environmental Data. Theory. Ap- plications and Software, EPFL Press, 2009.

12. Koike, K., Matsuda, S., Suzuki, T., Ohmi, M. Neural network-based estimation of principal metal contents in the Ho-kuroku district, Northern Japan, for explor- ing Kuroko-type Deposits. Natural Resources Research, 2002, no. 11 (2), pp. 135–156.

13. Lakes, T., Müller, D., Krüger, C. Cropland change in southern Romania: A comparison of logistic regressions and artificial neural networks. Landscape Ecology, 2009, 24(9), pp. 1195–1206.

14. Li, Y., Li, C., Tao, J.J., Wang, L.D. Study on Spa- tial Distribution of Soil Heavy Metals in Huizhou City Based on BP-ANN Modeling and GIS. Procedia Envi- ronmental Sciences, 2011, no. 10, p. 1953–1960.

15. Liu, F., He, X., Zhou, L. Application of generalized re- gression neural network residual kriging for terrain sur- face interpolation. Proc. SPIE7492, International Sym- posium on Spatial Analysis, Spatial-Temporal Data Modeling, and Data Mining, 2009, 74925F.

16. Mohanty, K., Majumdar, T.J. Using artificial neural net- works for synthetic surface fitting and the classification of remotely sensed data. International Journal of Applied Earth Observation and Geoinformation, 1999, no. 1(1), pp. 78–84.

17. Samanta, B., Ganguli, R., Bandopadhyay, S. Transac- tions of the Institution of Mining and Metallurgy, 2005, no. 114, pp. 129–139.

18. Sergeev, A.P., Baglaeva, E.M., Shichkin, A.V. Case of soil surface chromium anomaly of a northern urban ter- ritory – preliminary results. Atmospheric Pollution Research, 2010, vol. 1, pp. 44–49.

19. Sergeev, A.P., Baglaeva, E.M., Antonov, K.L., Medve- dev, A.N., Rakhmatova A. Yu. Anomalies of chromium surface distribution in urban soils from subarctic region of Russia. 15th International multidisciplinary scientif- ic geoconference SGEM 2015. Water Resources. Forest, Marine and Ocean Ecosystems. Conference proceedings, V. II Soils, Forest Ecosystems, Marine and Ocean Eco- systems. 18–24 June, 2015, Bulgaria, pp. 27–34.

20. Shaker, R., Tofan, L., Bucur, M., Costache, S., Sava, D., Ehlinger T. Land cover and landscape as predictors of groundwater contamination: a neural-network model- ling approach applied to Dobrogea, Romania. Journal of Environmental Protection and Ecology, 2010, no. 11(1), pp. 337–348.

21. Shaker, R.R., Ehlinger, T.J. Exploring non-linear re- lationships between landscape and aquatic ecological condition in southern Wisconsin: A GWR and ANN ap- proach. International Journal of Applied Geospatial Re- search, 2014, no. 5(4), pp. 1–20.