Patterns and Determinants of Post-Soviet Cropland Abandonment in the Western Siberian Grain Belt

Abstract

The transition from a command to a market economy resulted in widespread cropland abandonment across the former Soviet Union during the 1990s. Spatial patterns and determinants of abandonment are comparatively well understood for European Russia, but have not yet been assessed for the vast grain belt of Western Siberia, situated in the Eurasian forest steppe. This is unfortunate, as land-use change in Western Siberia is of global significance: Fertile black earth soils and vast mires store large amounts of organic carbon, and both undisturbed and traditional cultural landscapes harbor threatened biodiversity. We compared Landsat images from ca. 1990 (before the break-up of the Soviet Union) and ca. 2015 (current situation) with a supervised classification to estimate the extent and spatial distribution of abandoned cropland. We used logistic regression models to reveal important determinants of cropland abandonment. Ca. 135,000 ha classified as cropland around 1990 were classified as grassland around 2015. This suggests that ca. 20% of all cropland remain abandoned ca. 25 years after the end of the Soviet Union. Abandonment occurred mostly at poorly drained sites. The likelihood of cropland abandonment increased with decreasing soil quality, and increasing distance to medium-sized settlements, roads and railroads. We conclude that soil suitability, access to transport infrastructure and availability of workforce are key determinants of cropland abandonment in Western Siberia.

Mapping Cropland Abandonment in Mountainous Areas Using an Annual Land-Use Trajectory Approach

“…This study developed a new approach to map abandoned cropland by monitoring the annual land-use trajectory of the study area. Comparing to the previous research [1,5,8,13,14,20,26,28,46], there are two advantages and one weakness of the developed method. First, different from most of the previous research, this research detected the abandoned cropland by monitoring the consecutive land-use conversions during a certain period.…”

“…According to previous research, the vegetation succession after land abandonment in forest vegetation areas should be subject to annual and biennial weed, perennial grass and other grass, ligneous plants, and late succession species in sequence [45]. The changes from bare land to weeds or other grasses can take one to three years, to dwarf shrubs at least five years, and to trees at least ten years without human interference [46][47][48]. Therefore, spontaneous cropland abandonment should result in weeds or grasses as the dominant vegetation grown within the first three years of abandonment.…”

Spatial Distribution and Simulation of Cropland Abandonment in Wushan County, Chongqing, China

“…Previous research has shown that the species diversity of abandoned cropland under the natural state cannot significantly be improved within a period of 5 years, making it difficult to establish forest systems on such land [3,[32][33][34]. Therefore, this study considered abandoned cropland during the 5 years from 2011 to 2016 converted into grassland and unused land, as well as abandoned cropland converted into grassland, forest land, and unused land in the next 20 years.…”

“…However, this interpretation mode tends to be limited by the phenomenon of "same object but different spectra, same spectrum but different objects" [10,11,23]. Excessive dependence on the ground object spectral characteristics in remote sensing interpretation often leads to incorrect or omissive extraction [10,11,23,32]. In contrast to the traditional interpretation mode, the decision tree classification method makes full use of the remote sensing image data and other spatial data and can effectively avoid the phenomenon of "same object but different spectra, same spectrum but different objects", thereby obtaining more precise interpretation results when compared to traditional interpretation methods [10,11].…”

“…Besides, the more complex the data and the more variables, the better the algorithm will be. The CART model builds the prediction criterion in the form of a binary tree and uses the Gini Index in economics as the criterion of selecting the best testing variables by using the sample with the highest "purity degree" [10,23,32]. The Gini Index is defined as follows [36]:…”