Diverging Northern Hemisphere Trends in Meteorological Versus Ecological Indicators of Spring Onset in CMIP6

Plant phenology regulates the carbon cycle and land‐atmosphere coupling. Currently, climate models often disagree with observations on the seasonal cycle of vegetation growth, partially due to how spring onset is measured and simulated. Here we use both thermal and leaf area index (LAI) based indicators to characterize spring onset in CMIP6 models. Although the historical timing varies considerably across models, most agree that spring has advanced in recent decades and will continue to arrive earlier with future warming. Across the Northern Hemisphere for the periods 1950–2014, 1981–2014, and 2015–2099 in the historical and SSP5‐8.5 simulations, thermal‐based indicators estimate spring advances of −0.7 ± 0.2, −1.4 ± 0.4, and −2.4 ± 0.7 days/decade, while LAI‐based indicators estimate −0.4 ± 0.3, −0.1 ± 0.3, and −1±1.1 days/decade. Thereby, LAI‐based indicators exhibit weaker trends toward earlier onset, leading to uncertainties from different indices being as large or larger than model uncertainty. Reconciling these discrepancies is critical for understanding future changes in spring onset. Plain Language Summary The timing of spring onset as indicated by green‐up affects plants, bird and insect populations, rivers, and agriculture. However, state‐of‐the‐art land surface models disagree with satellite‐derived records on the seasonal cycles of vegetation growth, making it difficult to accurately predict green‐up, its response to climate, and the ecological consequences. Here we calculate two sets of spring onset indicators using climate model outputs to characterize spring onset variations and trends in the recent past and future. We find spring has been advancing in recent decades and will continue to arrive earlier with future warming. Thermal‐based indicators show that spring onset advances by −0.7, −1.4, and −2.4 days/decade in the Northern Hemisphere during 1950–2014, 1981–2014, and 2015–2099, respectively. This result suggests that spring onset today is on average four days earlier than spring onset 30 years ago and this rate will nearly double in the future. However, compared to meteorological‐based indicators, vegetation growth‐based indicators exhibit weaker trends toward earlier onset. Therefore, how we define and measure spring onset, as well as the models we use to predict changes in the environmental factors, influence future changes in the start of spring. Key Points Divergence between thermal‐ and growth‐based spring onset indicators grows w