The study addresses the problem of identifying abandoned agricultural lands in the northern part of the Zeya-Bureya plain (Oktyabrsky district, Amur region) using Landsat satellite imagery and the Random Forest machine learning algorithm. The approach involves the use of multitemporal composite images generated by averaging cloud-free scenes from May to September for the periods 1984–1991 and 2019–2024, in order to reduce the influence of shadows and cloud cover, short-term disturbances, and seasonal anomalies. Through the analysis of the seasonal dynamics of vegetation and water indices (SAVI and MNDWI), Tasseled Cap transformation components, and differential index values (ΔSAVI) across various landscape types (abandoned lands, croplands, forests, and waterlogged areas), the study substantiates a combined multiseasonal and multi-index classification approach. The effectiveness of using the MNDWI index and the TCW component in classification was compared. It was found that replacing MNDWI with TCW increased the accuracy of the "abandoned land" class from 76.1% to 81.3% (Producer’s Accuracy). The superiority of TCW is demonstrated due to its higher sensitivity to vegetation moisture rather than solely to open water surfaces. The overall classification accuracy reached 94.8%, with a Kappa coefÏcient of 0.93. Based on the classification results, the total area of abandoned lands was estimated at approximately 32,000 ha (18% of the total agricultural land area). The distribution of abandoned lands by soil type was analyzed; revealing that 57.3% of all abandoned lands are concentrated on low-productive soils, confirming that the least suitable lands were predominantly taken out of cultivation. The obtained results provide a transition from approximate estimates to precise spatial accounting of abandoned lands.

Masyutina Yulia A.

1. Meyfroidt P., Schierhorn F., Prishchepov A. V. Drivers, constraints and trade-offs associated with recultivating abandoned cropland in Russia, Ukraine and Kazakhstan. Global Environmental Change, 2016;37:1–15. doi: 10.1016/j.gloenvcha.2016.01.003.
2. Nechaeva T. V. Abandoned lands of Russia: distribution, agroecological condition and prospects of use (review). Pochvy i okruzhayushchaya sreda, 2023;6;2. doi: 10.31251/pos.v6i2.215 (in Russ.).
3. Romanenko G. A. (Eds.). Agroecological condition and prospects for the use of Russian lands that have been eliminated from active agricultural turnover, Moscow, Rosinformagrotekh, 2008, 64 p. (in Russ.).
4. Cheng M., Xue Z., Xiang Y. Soil organic carbon sequestration in relation to revegetation on the Loess Plateau, China. Plant Soil, 2015;397:31–42. doi: 10.1007/s11104-015-2486-5.
5. Kuemmerle T., Olofsson P., Chaskovskyy O. Post-Soviet farmland abandonment, forest recovery, and carbon sequestration in western Ukraine. Global Change Biology, 2011. doi: 10.1111/j.1365-2486.2010.02333.x.
6. Zhou Sh., Shengjie Zh., Tiexi Ch., Ning Z. The impact of cropland abandonment of post-Soviet countries on the terrestrial carbon cycle based on optimizing the cropland distribution map. Biology, 2022;11:620. doi: 10.3390/biology11050620.
7. Lavrova N. A., Kidyaeva N. A. Land use efÏciency in the Amur region. Vestnik Altaiskoi akademii ekonomiki i prava, 2019;4–2:228–233. EDN BPKJQH (in Russ.).
8. Gorelick N., Hancher M., Dixon M. Google Earth Engine: Planetary-scale geospatial analysis for everyone. Remote Sensing of Environment, 2017;202. doi: 10.1016/j.rse.2017.06.031.
9. Huang C., Wylie B., Yang L. Derivation of a Tasseled Cap transformation based on Landsat 7 At-satellite reflectance. International Journal of Remote Sensing, 2002;23:1741–1748. doi: 10.1080/01431160110106113.
10. Şimşek F. F., Durduran S. Land cover classification using Land Parcel Identification System (LPIS) data and open source Eo-Learn library. Geocarto International, 2022;38;1:19. doi: 10.1080/10106049.2022.2146760.
11. Zhai Y., Roy D. P., Martins V. S. Conterminous United States Landsat-8 top of atmosphere and surface reflectance tasseled cap transformation coefÏcients. Remote Sensing of Environment, 2022;274:112992. doi: 10.1016/j.rse.2022.112992.
12. Congedo L. Semi-Automatic Classification Plugin: A Python tool for the download and processing of remote sensing images in QGIS. Journal of Open Source Software, 2021;6(64):3172. doi: 10.21105/joss.03172.
13. Ticehurst C., Teng J., Sengupta A. Development of a multi-index method based on Landsat reflectance data to map open water in a complex environment. Remote Sensing, 2022;14;5:1158. doi: 10.3390/rs14051158.
14. Stepanov A. S., Dubrovin K. N., Verkhoturov A. L., Aseeva T. A. Features of the classification of agricultural lands of the Khabarovsk krai using satellite data. Dal'nevostochnyi agrarnyi vestnik, 2022;16;4:54–62. doi: 10.22450/199996837_2022_4_54. EDN TNLGXV (in Russ.).
15. Illarionova L., Dubrovin K., Fomina E. Multi-sensor NDVI time series for crop and fallow land classification in Khabarovsk krai, Russia. Sensors, 2025;25;18:5746. doi: 10.3390/s25185746. EDN AUVZXR.
16. Lesiv M., Shchepaschenko D., Moltchanova E. Spatial distribution of arable and abandoned land across former Soviet Union countries. Scientific Data, 2018;5:180056. doi: 10.1038/sdata.2018.56.
17. Yin H., Prishchepov A. V., Kuemmerle T. Mapping agricultural land abandonment from spatial and temporal segmentation of Landsat time series. Remote Sensing of Environment, 2018;210:12–24. doi: 10.1016/j.rse.2018.02.050. EDN UYLGCB.