Abstract. Freshwater resources are of high societal relevance, and understanding their past variability is vital to water management in the context of ongoing climate change. This study introduces a global gridded monthly reconstruction of runoff covering the period from 1902 to 2014. In situ streamflow observations are used to train a machine learning algorithm that predicts monthly runoff rates based on antecedent precipitation and temperature from an atmospheric reanalysis. The accuracy of this reconstruction is assessed with cross-validation and compared with an independent set of discharge observations for large river basins. The presented dataset agrees on average better with the streamflow observations than an ensemble of 13 state-of-the art global hydrological model runoff simulations. We estimate a global long-term mean runoff of 38 452 km3 yr−1 in agreement with previous assessments. The temporal coverage of the reconstruction offers an unprecedented view on large-scale features of runoff variability in regions with limited data coverage, making it an ideal candidate for large-scale hydro-climatic process studies, water resource assessments, and evaluating and refining existing hydrological models. The paper closes with example applications fostering the understanding of global freshwater dynamics, interannual variability, drought propagation and the response of runoff to atmospheric teleconnections. The GRUN dataset is available at https://doi.org/10.6084/m9.figshare.9228176 (Ghiggi et al., 2019).

River discharge is listed as an Essential Climate Variable (ECV) by the World Meteorological Organization (WMO) (Bojinski et al., 2014) and is one of the best monitored variables of the terrestrial water cycle. Nonetheless, in recent decades available observations have decreased significantly, often in relation to a lack of financial resources or political barriers to data access (

Abstract. 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 a global gridded monthly reconstruction of runoff covering the period from 1902 to 2014. In-situ streamflow observations are used to train a machine learning algorithm that predicts monthly runoff rates based on antecedent precipitation and temperature from an atmospheric reanalysis. The accuracy of this reconstruction is assessed with cross-validation and compared with an independent set of discharge observations for large river basins. The presented dataset agrees on average better with the streamflow observations than an ensemble of 13 state-of-the art global hydrological model runoff simulations. We estimate a global long-term mean runoff of 37 419 km3 yr−1 in agreement with previous assessments. The temporal coverage of the reconstruction offers an unprecedented view on large-scale features of runoff variability also in regions with limited data coverage, making it an ideal candidate for large-scale hydro-climatic process studies, water resources assessments and for evaluating and refining existing hydrological models. The paper closes with example applications fostering the understanding of global freshwater dynamics, interannual variability, drought propagation and the response of runoff to atmospheric teleconnections. The GRUN dataset is available from the ETHZ Research Collection at https://doi.org/10.3929/ethz-b-000324386 (Ghiggi et al., 2019).

<p>Although river flow is the best-monitored variable of the terrestrial water cycle, the scarcity of available in situ observations in large portions of the world has until now hindered the development of consistent observational estimates with global coverage. Recently, fusing sparse in-situ river discharge observations with gridded precipitation and temperature using machine learning has shown great potential for developing global monthly runoff estimates (Ghiggi et al., 2019). However, the accuracy of the utilised gridded precipitation and temperature products is variable and the corresponding uncertainty in the resulting runoff and river flow estimates was not yet quantified.</p><p>Global-RUNoff ENSEMBLE (G-RUN ENSEMBLE) (Ghiggi et al., in review) provides a multi-forcing global reanalysis of monthly runoff rates at a 0.5° resolution, composed of up to 525 runoff simulations. The G-RUN ENSEMBLE is based on 21 different atmospheric forcing datasets, overall spanning the period 1901-2019. The reconstructions are benchmarked against a comprehensive set of global-scale hydrological models (GHMs) simulations, using a large database of river discharge observations that were not used for model training as a reference.</p><p>Overall, the G-RUN ENSEMBLE shows good accuracy compared to the set of GHMs evaluated, especially with respect to the reproduction of the dynamics and seasonality of monthly runoff rates. We found that the spread imposed by the atmospheric forcing data in the G-RUN ENSEMBLE is small compared to the spread observed within the ensemble of GHMs simulations driven with a subset of such forcings. This might occur because GHMs are more impacted by biases in the input meteorological forcing and are more susceptible to accumulate errors over the simulation time than the adopted machine learning approach.</p><p>In summary, the multi-forcing nature of the G-RUN ENSEMBLE allows to quantify the uncertainty associated with the currently available atmospheric forcings, thereby paving the way for more robust and reliable water resources assessments, climate change attribution studies, hydro-climatic process studies as well as evaluation, calibration and refinement of GHMs.</p><p><strong>R</strong><strong>eferences</strong></p><p>Ghiggi, G., Humphrey, V., Seneviratne, S. I., and Gudmundsson, L. 2019: GRUN: an observation-based global gridded runoff dataset from 1902 to 2014, Earth Syst. Sci. Data, 11, 1655–1674, https://doi.org/10.5194/essd-11-1655-2019.</p><p>Ghiggi, G., Humphrey, V., Seneviratne, S. I., and Gudmundsson, L.: G-RUN ENSEMBLE: A multi-forcing observation-based global runoff reanalysis, Water Res. Res., in review.</p>

No abstract

No abstract

Abstract. The use of meteorological radars to study snowfall microphysical properties and processes is well established, in particular through two techniques: the use of multi-frequency radar measurements and the analysis of radar Doppler spectra. We propose a novel approach to retrieve snowfall properties by combining both techniques, while relaxing some assumptions on e.g. beam matching and non-turbulent atmosphere. The method relies on a two-step deep-learning framework inspired from data compression techniques: an encoder model maps a high-dimensional signal to a lower-dimensional “latent” space, while the decoder reconstructs the original signal from this latent space. Here, Doppler spectrograms at two frequencies constitute the high-dimensional input, while the latent features are constrained to represent the snowfall properties of interest. The decoder network is first trained to emulate Doppler spectra from a set of microphysical variables, using simulations from the radiative transfer model PAMTRA as training data. In a second step, the encoder network learns the inverse mapping, from real measured dual-frequency spectrograms to the microphysical latent space; doing so, it leverages the spatial consistency of the measurements to mitigate the problem's ill-posedness. The method was implemented on X- and W-band data from the ICE GENESIS campaign that took place in the Swiss Jura in January 2021. An in-depth assessment of the retrieval’s accuracy was performed through comparisons with colocated aircraft in-situ measurements collected during 3 precipitation events. The agreement is overall good and opens up possibilities for acute characterization of snowfall microphysics on larger datasets. A discussion of the method's sensitivity and limitations is also conducted.  The main contribution of this work is on the one hand the theoretical framework itself, which can be applied to other remote sensing retrieval applications and is thus possibly of interest to a broad audience across atmospheric sciences. On the other hand, the retrieved seven microphysical descriptors provide relevant insights into snowfall processes.

No abstract