Emergent constraints on land surface atmosphere interaction in climate models

Earth System Models (ESMs) have been widely used to predict future climate changes under various greenhouse gas emission pathways. This provides a physical scienctific basis for international climate assessments [e.g., Intergovernmental Panel on Climate Change (IPCC) Assessment Reports] and the generation of national climate scenarios. For instance, accurate estimations of future runoff, temperature-precipitation interactions and snow cover extend are of critical importance for human adaptation to terrestrial water shortage. However, these climate projections from ESMs (e.g., CMIP5/6) contain non-negligible uncertainties. Therefore, there is lack of clarity regarding future rates of climate change and possible associated risks to society, including heat stress, intense rainfall or flooding, and water availability more generally. In this thesis, we introduce a series of emergent relationships (See the Method in Chapter 1.3) to reduce the uncertainties in the predictions of the global land surface runoff (Chapters 2 and 3), CO2 fertilization effect (Chapter 4), temperature-precipitation interactions in Amazon (Chapter 5) and global snow cover extend (Chapter 6) in the CMIP5/6 models. Compared to estimates taken directly from the CMIP5/6 model ensemble, we find that the CMIP6 models underestimate the global runoff sensitivity to temperature (by 36-104%), the decline trend of CO2 fertilization effect (by 13.7–33.2%) and the global snow losses (by 23.5 – 67.6%) during 2015-2100, while the CMIP5 models overestimate the negative sensitivity of temperature to temperature in the Amazon. These new estimates of future climate changes have important implications on the global cycles of carbon, energy and water. For instance, the underestimated future global snow losses may threaten water availability for irrigation, hydropower generation, domestic and industrial water use during the 21st century, especially in snow dominated regions. Before reducing the uncertainties of future climate changes, characteristics and driving factors for the changes of these climate variables in the past and future periods are investigated by using the observations and simulations. This provides a better understanding of the potential mechanisms underpinning the emergent relationships. We find that the runoff in the 1287 hydrological stations that affects catchments occupying 2/5 of the global land area, increases in dry season but decreases in flood season. This homogenization of water discharge between seasons is mainly driven by dams’ operation. We also find that the global CO2 fertilization effect in most of Earth’s land surface has been continuously weakened, thereby potentially increasing the challenges for achieving the Paris Agreement’s long-term temperature goal. These important finding may bring wide-ranging influences on international and national policy-making for water resources management, and the weakening capacity of the terrestrial biosphere in absorbing atmospheric CO2.