Unveiling land degradation neutrality: a nonlinear perspective framework for unexpected land degradation in Kazakhstan

The unexpected land degradation in dryland regions has become a focal issue of global climate change and human activity since the onset of the Anthropocene. Therefore, conducting land degradation assessments from a nonlinear perspective is crucial to achieve the goal of Land Degradation Neutrality and Sustainable Development Goal (SDG 15.3.1). The monitoring framework that integrates land cover, land productivity, and soil organic carbon has emerged as the primary model for assessing land degradation neutrality, relying on the evolving trends from time-series data. However, these frameworks currently lack time-series data for soil organic carbon and overlook nonlinear state transition thresholds in the evolution of land systems, resulting in considerable uncertainty in land degradation assessments. Here, basing on the space-for-time substitution theory, we establish a framework for assessing land degradation states primarily using static data that are derived from soil organic carbon, net primary productivity, and our obtained physically meaningful endmember fraction using spectral mixture analysis model. This framework emphasizes the nonlinear evolutionary characteristics and thresholds among these three variables to quantify stable and unstable states during land degradation and their corresponding thresholds. Utilizing three widely used data sources—Soilgrid 250m, MOD17A3, and internally produced bare soil fractions products—and focusing on Kazakhstan as our study area. Soil organic carbon represents the system state variable and net primary productivity alongside bare soil fraction serve as environmental variables, we identify two state transition thresholds of soil organic carbon as slow variable of complex system (4.1% and 1.6%). This allows us to delineate the study area into two stable states: one representing an ideal state dominated by forest and agricultural land cover (23.72%), and the other depicting a completely degraded area dominated by desert cover (19.71%). The remaining areas between these two stable states constitute the unstable state, posing certain degradation risks. Building upon this, we further utilize a self-organizing neural network to integrate these three variables and assess the degree of degradation in the unstable areas, classifying them into severe, moderate, and mild degradation, accounting for 16.51%, 55.82%, and 27.67%, respectively. This space-for-time substitution method overcomes the challenges posed by the absence of time-series data, assisting a more spatially diversified approach to achieve land degradation neutrality assessment during the Anthropocene.