Evolution of NDVI secular trends and responses to climate change: A perspective from nonlinearity and nonstationarity characteristics

Abstract

The evolution of vegetation patterns plays a crucial role in monitoring ecosystem dynamics in face of global warming. Notably, the time-series vegetation trends vary unevenly over time. However, conventional linear methods based on the stationarity assumption, cannot detect the implicit characteristics of nonlinear and gradual vegetation changes. Meanwhile, characterizing the inherent features of nonlinearity and nonstationarity of climatic drivers remains a challenge. This study applied multidimensional ensemble empirical mode decomposition (MEEMD) and Breaks For Additive Seasonal and Trend (BFAST) algorithm to diagnose spatiotemporal evolution and abrupt change in vegetation secular trends based on normalized difference vegetation index (NDVI) data of the Hexi Corridor during 1982–2015. Geographically and temporally weighted regression (GTWR) was used to address the spatiotemporal nonlinearity and nonstationarity of climatic drivers. A wide range of browning trends gradually evolved into greening trends from 1982 to 2015. Compared with ordinary least squares regression, MEEMD could adaptively decompose short-term trends of noise and seasonality and elucidate the entire evolutionary process of NDVI trends. Browning trends prevailed before abrupt changes, and greening trends widely expanded after 2006 due to drought before and intense precipitation during this year. GTWR addressed the nonequilibrium effect of the time dimension on climate drivers, and the R-squared achieved 0.86. The response processes between vegetation and climatic drivers showed significant spatiotemporal nonstationarity and aggregation characteristics based on three-dimensional visualization. In addition to the drivers of temperature, precipitation, solar radiation, and potential evapotranspiration, the drying effect and mechanical stimuli of wind speed on vegetation could not be underestimated. This study provides a novel framework to solve nonlinearity and nonstationarity problems related to vegetation trends and their response mechanisms and promotes the application of remote sensing to solve practical problems.