The answer to how much carbon land ecosystems are projected to remove from the atmosphere until 2100 is different for each Earth system model. We find that differences across models are primarily explained by the annual land carbon sink dependence on temperature and soil moisture, followed by the dependence on CO₂ air concentration, and by average climate conditions. Our insights on why each model projects a relatively high or low land carbon sink can help to reduce the underlying uncertainty.

Earth observations have many missing values. They are often filled using information from spatial and temporal contexts that mostly ignore information from related observed variables. We propose the gap-filling method CLIMFILL that additionally uses information from related variables. We test CLIMFILL using gap-free reanalysis data of variables related to soil–moisture climate interactions. CLIMFILL creates estimates for the missing values that recover the original dependence structure.

We introduce the first chain of computationally efficient Earth system model (ESM) emulators to translate user-defined greenhouse gas emission pathways into regional temperature change time series accounting for all major sources of climate change projection uncertainty. By combining the global mean emulator MAGICC with the spatially resolved emulator MESMER, we can derive ESM-specific and constrained probabilistic emulations to rapidly provide targeted climate information at the local scale.

We presented a global comparison between observed and simulated trends in a flood index over the 1971–2005 period using the Global Streamflow Indices and Metadata archive and six global hydrological models available through The Inter-Sectoral Impact Model Intercomparison Project. Streamflow simulations over 2006–2099 period robustly project high flood hazard in several regions. These high-flood-risk areas, however, are under-sampled by the current global streamflow databases.

We focus on the net exchange of water between land and air via evapotranspiration and dew during the night. We provide, for the first time, an overview of the magnitude and variability of this flux across the globe from observations and climate models. Nocturnal water loss from land is 7 % of total evapotranspiration on average and can be greater than 15 % locally. Our results highlight the relevance of this often overlooked flux, with implications for water resources and climate studies.

Earth system models (ESMs) are invaluable to study the climate system but expensive to run. Here, we present a statistical tool which emulates ESMs at a negligible computational cost by creating stochastic realizations of yearly land temperature field time series. Thereby, 40 ESMs are considered, and for each ESM, a single simulation is required to train the tool. The resulting ESM-specific realizations closely resemble ESM simulations not employed during training at point to regional scales.

Freshwater resources are of high societal relevance and understanding their past variability is vital to water management in the context of current and future climatic change. This study introduces GRUN: the first global gridded monthly reconstruction of runoff covering the period from 1902 to 2014. The dataset agrees on average much better with the streamflow observations than an ensemble of 13 state-of-the-art global hydrological models and will foster the understanding of freshwater dynamics.

Because changes in freshwater availability can impact many natural ecosystems and human activities, it is crucial to better understand long-term changes in the water cycle. This dataset is a reconstruction of past changes in land water storage over the last century, obtained by combining satellite observations with historical weather data. It can be used to investigate both regional changes in freshwater availability or drought frequency and long-term changes in the global water cycle.

The production of 30 959 daily streamflow time series in the Global Streamflow and Metadata Archive (GSIM) project is presented. The paper also describes the development of three metadata products that are freely available. Having collated an unprecedented number of stations and associated metadata, GSIM can be used to advance large-scale hydrological research and improve understanding of the global water cycle.

Time-series indices characterizing streamflow at annual, seasonal and monthly resolution at more than 30 000 stations around the world are presented. The data belong to the Global Streamflow and Metadata Archive (GSIM) and allow for an assessment of water balance components, hydrological extremes and the seasonality of water availability. The quality of the data is tested using automated methods to aid potential users to gauge the suitability of the data for specific applications.

Assessing projected hydroclimatological changes is crucial, but associated with large uncertainties. We statistically assess here the response of precipitation and water availability to global temperature change, enabling us to estimate the significance of drying/wetting tendencies under anthropogenic climate change. We further show that opting for a 1.5 K warming target just slightly influences the mean response but could substantially reduce the risk of experiencing extreme changes.

Despite the scientific and societal relevance of freshwater, there are to date no observation-based pan-European runoff estimates available. Here we employ state-of-the-art techniques to estimate monthly runoff rates in Europe. The new data product is based on an unprecedented collection of river flow observations which are combined with atmospheric variables using machine learning. Potential applications of the presented product include climatological assessments and drought monitoring.

The widely used Budyko framework is by definition limited to steady-state conditions. In this study we analytically derive a new, two-parameter formulation of the Budyko framework that represents conditions under which evapotranspiration exceeds precipitation. This is technically achieved by rotating the water supply limit within the Budyko space. The new formulation is shown to be capable to represent first-order seasonal dynamics within the hydroclimatological system.

Water storages and fluxes on land are key variables in the Earth system. To provide context for local investigations and to understand phenomena that emerge at large spatial scales, information on continental freshwater dynamics is needed. This paper presents a methodology to estimate continental-scale runoff on a 0.5° spatial grid, which combines the advantages of in situ observations with the power of machine learning regression. The resulting runoff estimates compare well with observations.