Assessing the vulnerability of the Central Black Earth Region to gully erosion using machine learning methods

Aim. Demonstrate a new approach for assessing the susceptibility of an area to gully erosion based on machine learning algorithms.

Methodology. The key research methods were: machine learning-based modeling and geo­information modeling. A literary method was used in the analysis of existing solutions to assess the susceptibility of the territory to gully erosion. In the practical field of research, one of the most modern methods of machine learning, CatBoost, is widely used, taken as the basis of the developed approach. One of the main ideas of the proposed approach is the ensembling of machine learning models.

Results. A new method is proposed — a smoothed multilevel assessment of a territory's predisposition to gully erosion, using the Vorobyevsky District of the Voronezh Region as an example. The similarities and differences between the proposed approach and existing methods based on the idea of ensemble modeling are considered. Two new metrics for as­sessing the accuracy of the proposed method, RF1 and NDF, are justified. The concepts of soft, hard, and weighted modeling levels are introduced, allowing the contribution of relief morphometry to gully erosion development to be assessed. It has been established that the greatest influence on this process is exerted by the absolute and relative heights of the terrain, the LS factor, the catchment area, and the slope exposure. Together, they explain 95% of the areas of existing gully erosion in the region. Based on the results of modeling in the Vorobyevsky district of the Voronezh region, 2,853 hectares of land with a high and very high predisposition to gully erosion were identified. In terms of landscape, these cor­respond to the steppe-like valley slopes of southern exposures, which are distinguished by their greatest length, steepness, and height, and have a concave cross-section and significant catchment areas.

Research implications. The significance of the study lies in the proposal of a new approach to assessing the predisposition of the territory to gully erosion based on machine learning methods. From a practical point of view, the ideas proposed in the work and the method itself can be used to obtain a higher-quality result in assessing the predisposition of a terri­tory to gully erosion, compared with a number of classical methods and machine learning technologies, especially when analyzing large territories with high uneven distribution of gully erosion.

1. Zorina E. F., ed. Geografiya ovrazhnoj erozii [Geography of gully erosion]. Moscow: MSU Publ., 2006, 324 p.

2. Drozdov K. A. [Geography of gully erosion in the Central Cernozems Region]. In: Regionalnye landshaftnye issledovaniya [Regional landscape researches], 2022, vol. 6, pp. 65–75.

3. Yermolayev O. P., Rysin I. I., Golosov V. N. [Mapping assessment of gully erosion in the East of the Russian Plain]. In: Geomorfologiya [Geomorphology], 2017, vol. 2, pp. 8–51. DOI: 10.15356/0435-4281-2017-2-38-51

4. Korecky N. A., Bevz V. N. [Natural factors of gully formation and their geographical analysis on the territory of Vorobyevsky district of Voronezh region]. In: Regionalnye landshaftnye issledovaniya [Regional landscape researches], 2022, vol. 6, pp. 20–38.

5. Koretsky N. A., Bevz V. N. Methodological Issues of the Gully Erosion Probability Models in the Central Chernozem Region Using Machine Learning. In: Izvestiya Dagestanskogo gosudarstvennogo pedagogicheskogo universiteta. Yestestvennyye i tekhnicheskiye nauki [News of the Dagestan State Pedagogical University. Natural and Exact Sciences], 2023, vol. 17, pp. 41–55. DOI: 10.31161/1995-0675-2023-17-3-41-55

6. Korecky N. A, Bevz V. N. [Some patterns of gully erosion on the territory of Vorobyovka district, Voronezh region]. In: Regionalnye landshaftnye issledovaniya [Regional landscape researches], 2020, vol. 4, pp. 113–122.

7. Sidorchuk A. Yu. Express method of estimation of gully erosion potential. In: Lomonosov Geography Journal, 2021, vol. 1, pp. 54-61.

8. Hawker L., Uhe P., Paulo L., Sosa J. A 30 m global map of elevation with forests and buildings removed. In: Environmental Research Letters, 2022, vol. 17, p. 024016. DOI: 10.1088/1748-9326/ac4d4f

9. Breiman L. Random Forests. In: Machine Learning, 2001, vol. 45, pp. 5–32. DOI: 10.1023/A:1010933404324

10. Prokhorenkova L., Gusev G., Vorobev A., Dorogush A. V., Gulin A. CatBoost: unbiased boosting with categorical features. In: Journal of Functional Analysis, 2017, vol. 1706, pp. 09516.

11. Arabameri A., Wei Chen, Loche M., Xia Zhao, Yang Li, Lombardo L., Cerda A., et al. Comparison of machine learning models for gully erosion susceptibility mapping. In: Geoscience Frontiers, 2020, vol. 11, pp. 1609–1620. DOI: 10.1016/j.gsf.2019.11.009

12. Igwe O., John U. I., Solomon O., Ozioko O. GIS-based gully erosion susceptibility modeling, adapting bivariate statistical method and AHP approach in Gombe town and environs Northeast Nigeria. In: Geoenviron Disasters, 2020, vol. 7. DOI: 10.1186/s40677-020-00166-8

13. Florinsky I. V. Digital Terrain Analysis in Soil Science and Geology. Amsterdam: Elsevier Inc., 2016. 486 p.

14. Fengjie Fan, Xingli Gu, Jun Luo, Bin Zhang, Hui Liu, Haiqing Yang, Lei Wang. Identification of gully erosion activity and its influencing factors: A case study of the Sunshui River Basin. In: PLoS ON, 2024, vol. 19. DOI: 10.1371/journal.pone.0309672

15. Gelete T. B., Pasala P., Abay N. G., Weldu G. Integrated machine learning and geospatial analysis enhanced gully erosion susceptibility modeling in the Erer watershed in Eastern Ethiopia. In: Environmental Informatics and Remote Sensing, 2024, vol. 12, pp. 1–20. DOI: 10.3389/fenvs.2024.1410741

16. Saha S., Roy J., Arabameri A., Roy J., Blaschke T. Machine Learning-Based Gully Erosion Susceptibility Mapping: A Case Study of Eastern India. In: Sensors, 2020, vol. 20, pp. 1313. DOI: 10.3390/s20051313

17. Shary P. A., Sharaya L. S., Mitusov A. V. Fundamental quantitative methods of land surface analysis. In: Geoderma, 2002, vol. 107, pp. 1–32. DOI: 10.1016/S0016-7061(01)00136-7

18. Sidorchuk A. The Theoretical Approach to the Modelling of Gully Erosion in Cohesive Soil. In: Earth, 2022, vol. 3, pp. 228–244. DOI: 10.3390/earth3010015

19. Wang Y., Zhang Y., Chen H. Gully erosion susceptibility prediction in Mollisols using machine learning models. In: Journal of Soil and Water Conservation, 2023, vol. 78, pp. 385–396. DOI: 10.2489/jswc.2023.00019