Determinants of Vegetation Restoration on the Loess Plateau: A Comprehensive Analysis across Multiple Temporal Scales

While the spatiotemporal evolution of loess plateau (LP) greening is well-documented, understanding vegetation dynamics and driving forces across different timescales remains urgent. Long-term normalized difference vegetation index (NDVI) data (1982–2020) were analyzed with trend analysis to investigate the spatiotemporal dynamics of vegetation restoration. It was processed with the ensemble empirical mode decomposition (EEMD) to distinguish the constituent temporal scales of vegetation change. Redundancy analysis quantified the key influencing factors by evaluating their explanatory power and correlation strength with the response variables. The findings revealed that NDVI exhibited a fluctuating upward trend (1982–2020), with pronounced shifts around 1989 and marked increases after 2000. Spatially, NDVI decreased from southeast to northwest, varying by ecological zone and showing the highest values in valley plains, with most areas exhibiting an upward trend. The NDVI variations were composed of short-term (4- and 8-year) and long-term (13- and 20-year) cycles, with short-term scales exerting substantial influence: The southeast showed clear increases at the 4-year scale, while patterns were more complex at the 8-year scale. NDVI was positively correlated with temperature, precipitation, solar radiation, and relative humidity, but negatively correlated with wind speed, evapotranspiration, and air pressure. Temperature, precipitation, solar radiation, and relative humidity positively correlated with NDVI, while wind speed, evapotranspiration, and air pressure showed negative correlations. Temperature superseded precipitation as the primary factor driving regional vegetation change. This multi-temporal scale analysis of vegetation dynamics provides scientific insights for consolidating ecological restoration achievements in similar regions.