How long will the four seasons be by 2100? Increasing evidence suggests that the length of a single season or in regional scales has changed under global warming, but a hemispherical‐scale response of the four seasons in the past and future remains unknown. We find that summer in the Northern Hemisphere midlatitudes has lengthened, whereas winter has shortened, owing to shifts in their onsets and withdrawals, accompanied by shorter spring and autumn. Such changes in lengths and onsets can be mainly attributed to greenhouse‐warming. Even if the current warming rate does not accelerate, changes in seasons will still be exacerbated in the future. Under the business‐as‐usual scenario, summer is projected to last nearly half a year, but winter less than 2 months by 2100. The changing seasonal clock signifies disturbed agriculture seasons and rhythm of species activities, more frequent heat waves, storms and wildfires, amounting to increased risks to humanity.

The record‐long 2020 Meiyu season since 1961 caused severe floods over the Yangtze and Huaihe River valleys (YHRV). Why the Meiyu duration doubled in 2020 remains a puzzle. We show that the long‐lasting Meiyu can be divided into three stages: advanced‐onset, strong‐persisting, and delayed‐withdrawal. The advanced‐onset was associated with an extremely negative‐phase East Atlantic/West Russia teleconnection. The strong‐persisting was attributed to a positive‐phase Pacific–Japan (PJ) pattern sustained by La Niña's rapid development. The delayed withdrawal was related to the combined effect of a positive PJ pattern and a mid‐troposphere “two ridge–one trough” pattern over Asia. Two subseasonal forecasting systems predicted the positive rainfall anomalies over the YHRV in the first two stages, but not the third stage, which may be associated with poor prediction of the two ridge–one trough pattern. These results highlight the role of subseasonal evolution in extreme climate events.

Both mean and extreme rainfall decreased over India and Northern China during 1979–2005 at a rate of 0.2%/decade. The aerosol dampening effects on rainfall has also been suggested as a main driver of mean rainfall shift in India and China. Conflicting views, however, exist on whether aerosols enhance or suppress hazardous extreme heavy rainfall. Using Coupled Model Intercomparison Project phase 5 (CMIP5) multimodel ensemble, here we show that only a subset of models realistically reproduces the late‐20th‐century trend of extreme rainfall for the three major regions in Asia: drying in India and Northern China and wetting in Southern China, all consistent with mean rainfall change. As a common feature, this subset of models includes an explicit treatment of the complex physical processes of aerosol‐cloud interaction (i.e., both cloud‐albedo and cloud‐lifetime effects), while simulation performance deteriorates in models that include only aerosol direct effect or cloud‐albedo effect. The enhanced aerosol pollution during this rapid industrialization era is the leading cause of the spatially heterogeneous extreme rainfall change by dimming surface solar radiation, cooling adjacent ocean water, and weakening moisture transport into the continental region, while GHG warming or natural variability alone cannot explain the observed changes. Our results indicate that the projected intensification of regional extreme rainfall during the early‐to‐mid 21st‐century, in response to the anticipated aerosol reduction, may be underestimated in global climate models without detailed treatment of complex aerosol‐cloud interaction.

To forecast the terrestrial carbon cycle and monitor food security, vegetation growth must be accurately predicted; however, current process-based ecosystem and crop-growth models are limited in their effectiveness. This study developed a machine learning model using the extreme gradient boosting method to predict vegetation growth throughout the growing season in China from 2001 to 2018. The model used satellite-derived vegetation data for the first month of each growing season, CO₂ concentration, and several meteorological factors as data sources for the explanatory variables. Results showed that the model could reproduce the spatiotemporal distribution of vegetation growth as represented by the satellite-derived normalized difference vegetation index (NDVI). The predictive error for the growing season NDVI was less than 5% for more than 98% of vegetated areas in China; the model represented seasonal variations in NDVI well. The coefficient of determination (R²) between the monthly observed and predicted NDVI was 0.83, and more than 69% of vegetated areas had an R² > 0.8. The effectiveness of the model was examined for a severe drought year (2009), and results showed that the model could reproduce the spatiotemporal distribution of NDVI even under extreme conditions. This model provides an alternative method for predicting vegetation growth and has great potential for monitoring vegetation dynamics and crop growth.