Coupling mechanism between vegetation and multi-depth soil moisture in arid–semiarid area: Shift of dominant role from vegetation to soil moisture

Soil moisture (SM) is an essential water source for vegetation, while vegetation also exerts importance on SM during succession. It is crucial to investigate the coupling mechanism between SM and Normalized Difference Vegetation Index (NDVI), which can drive terrestrial carbon sinks while maintaining water sustainability. The Granger causality was used to analyze the unidirectional and bidirectional coupling relationship of SM and NDVI in the 0–289 cm soil layers. The windowed cross correlation was utilized to quantify the bidirectional time lag effect. The findings indicate that the vegetation improved significantly (92.47%) in 2001–2021 in the arid–semiarid area, while the SM in all four layers showed a decreasing trend (>50%). In most areas, NDVI was bidirectional dependent on SM. Vegetation dominated the loss of shallow SM (<100 cm) through root uptake and transpiration, whereas deep SM (>100 cm) in turn dominated the vegetation growth, structure and spatial pattern. At the spatial scale, the top layer of SM in the arid area demonstrated less sensitivity to NDVI due to coupling with the atmosphere. Instead, the influence of NDVI on SM in the semiarid area was dominant, so more attentiveness to the consumption of SM by vegetation is needed. The bidirectional legacy effect could be up to 6 months, attributed to water transport distance, vegetation root systems and physiological mechanisms. The correlation reversed when the lag time was extended or the soil layer was deepened, which means that the greening of vegetation will intensify deep soil drying. Differences in functional traits among plants lead to varying responses. The changes of deep SM need to focus on when implementing artificial meadow construction. This study provides a theoretical basis for optimizing ecological engineering projects, which is essential for restoring global drylands ecosystems.