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&#176; 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&#8211;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>
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.
Snowfall information at the scale of individual particles is rare, difficult to gather, but fundamental for a better understanding of solid precipitation microphysics. In this article we present a dataset (with dedicated software) of in-situ measurements of snow particles in free fall. The dataset includes gray-scale (255 shades) images of snowflakes, co-located surface environmental measurements, a large number of geometrical and textural snowflake descriptors as well as the output of previously published retrieval algorithms. These include: hydrometeor classification, riming degree estimation, identification of melting particles, discrimination of wind-blown snow, as well as estimates of snow particle mass and volume. The measurements were collected in various locations of the Alps, Antarctica and Korea for a total of 2’555’091 snowflake images (or 851’697 image triplets). As the instrument used for data collection was a Multi-Angle Snowflake Camera (MASC), the dataset is named MASCDB. Given the large amount of snowflake images and associated descriptors, MASCDB can be exploited also by the computer vision community for the training and benchmarking of image processing systems.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
hi@scite.ai
10624 S. Eastern Ave., Ste. A-614
Henderson, NV 89052, USA
Copyright © 2024 scite LLC. All rights reserved.
Made with 💙 for researchers
Part of the Research Solutions Family.