Disentangling exogenous and endogenous memory in global vegetation productivity: dominant role and future strengthening of vegetation carryover effects

Vegetation productivity is not only determined by current environmental conditions but also reflects the lagged influence of past climate and vegetation states. This temporal dependency, often referred to as ecological memory, arises from both antecedent climate conditions (exogenous lagged climatic effects, LCE) and prior vegetation states (endogenous vegetation growth carryover effects, VGC). However, their spatiotemporal variability, relative importance, and responses to future climate change remain poorly understood at the global scale. Here, we develop a unified analytical framework by integrating vector autoregressive model and impulse response functions to disentangle the roles of LCE and VGC in regulating global gross primary productivity (GPP) across space, time, and future climate scenarios. Both components exhibit rapid initial responses followed by gradual decay within approximately five months, yet differ markedly in magnitude, direction, and persistence. LCE show strong hemispheric asymmetry: in the Northern Hemisphere, increases in temperature exert the strongest positive effect on GPP, whereas increases in atmospheric dryness (vapor pressure deficit) produce the strongest negative effect. In contrast, in the Southern Hemisphere, increased precipitation is the dominant positive driver of GPP, while negative responses exhibit greater spatial heterogeneity and show no clear dominant controlling factor, suggesting more complex and regionally varying climatic influences. In contrast, VGC display globally consistent positive responses with substantially greater intensity (17.35 gC m −2 month −1 at a 1-month lag) and minimal hemispheric differences. Across all lag periods, VGC dominate productivity variability, contributing over 84% at short lags and remaining above 77% after 12 months, whereas LCE play a secondary but regionally differentiated role. Future projections further reveal a pronounced intensification of VGC under all emission scenarios, with peak responses increasing to 25–35 gC m −2 month −1 and strengthening toward the end of the century. These findings demonstrate that global vegetation productivity is primarily regulated by endogenous vegetation memory, while exogenous climatic effects provide shorter-lived and spatially heterogeneous modulation. The increasing strength of vegetation memory under future climate change highlights its critical role in shaping ecosystem responses, underscoring the need to explicitly incorporate both memory components into Earth system models to improve predictions of terrestrial carbon dynamics and ecosystem resilience.